Split (1)

train · 5.09k rows

pmcid stringlengths 6 6	title stringlengths 9 374	abstract stringlengths 2 4.62k ⌀	fulltext stringlengths 167 106k	file_path stringlengths 64 64
539060	fog-2 and the Evolution of Self-Fertile Hermaphroditism in Caenorhabditis	Somatic and germline sex determination pathways have diverged significantly in animals, making comparisons between taxa difficult. To overcome this difficulty, we compared the genes in the germline sex determination pathways of Caenorhabditis elegans and C. briggsae, two Caenorhabditis species with similar reproductive systems and sequenced genomes. We demonstrate that C. briggsae has orthologs of all known C. elegans sex determination genes with one exception: fog-2 . Hermaphroditic nematodes are essentially females that produce sperm early in life, which they use for self fertilization. In C. elegans, this brief period of spermatogenesis requires FOG-2 and the RNA-binding protein GLD-1, which together repress translation of the tra-2 mRNA. FOG-2 is part of a large C. elegans FOG-2-related protein family defined by the presence of an F-box and Duf38/FOG-2 homogy domain. A fog -2-related gene family is also present in C. briggsae, however, the branch containing fog-2 appears to have arisen relatively recently in C. elegans, post-speciation. The C-terminus of FOG-2 is rapidly evolving, is required for GLD-1 interaction, and is likely critical for the role of FOG-2 in sex determination. In addition, C. briggsae gld-1 appears to play the opposite role in sex determination (promoting the female fate) while maintaining conserved roles in meiotic progression during oogenesis. Our data indicate that the regulation of the hermaphrodite germline sex determination pathway at the level of FOG-2/GLD-1/ tra-2 mRNA is fundamentally different between C. elegans and C. briggsae, providing functional evidence in support of the independent evolution of self-fertile hermaphroditism. We speculate on the convergent evolution of hermaphroditism in Caenorhabditis based on the plasticity of the C. elegans germline sex determination cascade, in which multiple mutant paths yield self fertility.	Introduction Sex determination is an ancient and universal feature in metazoans. In spite of this, comparison of distantly related species such as Caenorhabditis elegans and Drosophila melanogaster has revealed little about the evolution of the complex pathways that mediate the sexual fate decision in the soma and germline [ 1 , 2 , 3 ]. This is likely due to the combination of gross morphological, functional, and behavioral dissimilarity and extensive sequence divergence. Thus, if we wish to clarify the etiology of diverged sex determination pathways, an alternative approach is required. One approach is to perform comparative analysis of sex determination genes in species separated by sufficient evolutionary time to allow for changes in pathway components yet retain comparable somatic and germline morphology and function. The clade containing C. elegans and C. briggsae represents an ideal case for this type of study, as the sex determination pathway has been well studied in C. elegans and an abundance of sequence information is available for both species [ 4 , 5 ]. C. elegans and C. briggsae, while sharing very similar germline and somatic morphology, are separated by approximately 100 million years and are members of a clade that employs multiple mating systems [ 5 , 6 , 7 , 8 , 9 , 10 ]. C. elegans and C. briggsae are self-fertile hermaphrodites that maintain males at a low frequency (androdioecious), whereas the morphologically similar C. remanei and C. sp. CB5161 are obligate female/male (gonochoristic) species [ 6 , 7 , 10 ]. Phylogenetic analysis of the four closely related Caenorhabditis species suggests that self-fertile hermaphroditism has evolved independently in C. elegans and C. briggsae from an ancestral male/female state [ 10 , 11 ]. Importantly, a transition in mating system from female/male to hermaphroditic (or hermaphroditic to male/female) requires that one or more changes in the sex determination pathway have occurred. C. elegans and C. briggsae, like many other animals, have two sexes specified by the ratio of X chromosomes to sets of autosomes [ 8 , 12 , 13 ]. In both species, XX animals are somatically female while the germline is hermaphroditic. Self fertility is achieved by a transient period of spermatogenesis beginning in the third larval (L3) stage before the organism switches to the production of oocytes in the L4 stage, which continues throughout adulthood [ 14 , 15 ]. In both species, XO males begin sperm production in the L3 stage and continue spermatogenesis throughout their reproductive lives [ 14 , 16 , 17 ]. A major determinant of germline sexual fate in C. elegans is the relative activity of two key regulators: tra-2, which promotes the female fate (oocyte), and fem-3, which promotes the male fate (sperm) [ 18 , 19 ] ( Figure 1 A). The activities of tra-2 and fem-3 must be regulated in both males and hermaphrodites to allow spermatogenesis to occur, however the mechanisms by which this regulation occurs differs between the two sexes. In males, her-1 represses tra-2 feminizing activity and raises the relative level of fem-3 activity so that spermatogenesis is continuous [ 20 , 21 ]. Since null mutations in her-1 have no effect on hermaphrodites and her-1 is not expressed in XX animals, a different mechanism is used to allow for the transient production of sperm [ 22 , 23 ]. Figure 1 The C. elegans XX Hermaphrodite Germline Sex Determination Pathway (A) Genetic pathway for gene activity, where arrows represent positive regulation and bars represent negative regulation. The key genes tra-2 and fem-3 and the upstream regulators of tra-2 that are the focus of this work, fog-2 and gld-1, are in large bold font. The upstream genes fog-2 and gld-1, which are key regulators of tra-2 and addressed in this work, are also in large bold font. The gene activities at each level in the hierarchy are indicated below as “ACTIVE” in bold or “inactive” in grey. In L3 and L4 hermaphrodites the activities of fog-2 and gld-1 are high, leading to repression of tra-2 activity (also see [B]) and the de-repression of fem-3, resulting in the onset of spermatogenesis. In L4 and adult hermaphrodites the activity of fog-2 and gld-1 are low, leading to high tra-2 activity and the repression of fem-3, resulting in oogenesis. The shift in tra-2 / fem-3 balance allows for the switch from spermatogenesis to oogenesis in an otherwise female somatic gonad in the hermaphrodite. (B) C. elegans FOG-2/GLD-1/ tra-2 mRNA ternary complex. Current data indicates that FOG-2 and GLD-1 are required for the translational repression of the tra-2 mRNA [ 25 ]. GLD-1 binds as a dimer to the tra-2 mRNA 3′UTR at two 28 nucleotide direct repeat elements (TGE/DRE, blocks) and FOG-2 makes contact with GLD-1 [ 32 , 34 ]. All three components are required for the proper specification of hermaphrodite spermatogenesis. Self fertility in C. elegans hermaphrodites is achieved by an early period of spermatogenesis followed by a later period of oogenesis ( Figure 1 A). The promotion of spermatogenesis during the L3 stage (early) is achieved by translational repression of the tra-2 mRNA mediated by gld-1 (“defective in germline development”) and fog-2 (“feminization of germline”)[ 24 , 25 ] ( Figure 1 A and 1 B). The transient reduction in the level of tra-2 feminizing activity raises the relative level of fem-3 masculinizing activity to promote spermatogenesis ( Figure 1 A). Later in L4 and adult animals, oogenesis is promoted by relieving the fog-2/gld-1 -mediated repression of tra-2 feminizing activity combined with repression of fem-3 masculinizing activity by mog-1 to mog-6, fbf-1 and fbf-2, and nos-1 to nos-3 [ 18 , 19 , 26 ]. Central to this work are the genes fog-2 and gld-1. fog-2 is required for hermaphrodite, but not male, spermatogenesis in C. elegans, as XX animals that lack fog-2 produce only oocytes, resulting in functional females, whereas XO males are unaffected [ 27 ]. Similarly, loss-of-function mutations in gld-1 result in the feminization of the hermaphrodite germline without affecting males [ 28 , 29 ]. Both fog-2 and gld-1 are germline-specific regulators of sexual fate, since they do not appear to be expressed in the soma, and null mutations in either gene do not affect somatic sexual fate [ 25 , 27 , 28 , 29 , 30 ]. C. elegans gld-1 is a germline-specific tumor suppressor that is indispensable for oogenesis [ 28 , 29 ] and encodes a conserved KH-type RNA-binding protein [ 30 ]. GLD-1 is a translational repressor that binds to multiple mRNA targets [ 31 ], including tra-2, where it binds as a dimer to each of two tra-2 and GLI elements (TGEs) present on the 3′ untranslated region (UTR) of the tra-2 mRNA [ 24 , 32 ] ( Figure 1 B). Deletion of the tra-2 TGEs results in a loss of GLD-1-mediated translational control and feminization of the germline, such that only oocytes are produced [ 20 , 25 , 33 , 34 ]. C. elegans FOG-2 was identified as a GLD-1-interacting protein with a structure similar to canonical F-box proteins; it has an N-terminal F-box and a C-terminal protein–protein interaction domain. In the case of FOG-2 the putative protein–protein interaction domain is referred to as Duf38 (Pfam in [ 35 ]) or FOG-2 homology domain (FTH) [ 25 ]. F-box proteins are often core components of the Skp1/Cullin/F-box-type E3 ubiquitin ligase complexes, and they serve to link specific substrates to the ubiquitin ligase machinery for subsequent proteolysis [ 36 ]. However, FOG-2 cannot target GLD-1 for degradation since both function to promote hermaphrodite spermatogenesis [ 25 ] ( Figure 1 A). Current data suggest that the formation of a FOG-2/GLD-1/ tra-2 mRNA ternary complex mediates translational repression of tra-2 and a corresponding reduction in feminizing activity to allow hermaphrodite spermatogenesis [ 24 , 25 ] ( Figure 1 B). The completion of the C. elegans genome sequence [ 4 ] and the 10X sequence (representing more than 98% coverage) of the closely related species C. briggsae [ 5 ] permits studies of the evolution of sex determination and the inception of hermaphrodite spermatogenesis in morphologically comparable species. Here, we pose the question, do C. elegans and C. briggsae specify male sexual fate in the hermaphrodite germline similarly? We find that 30 of 31 C. elegans sex determination genes have C. briggsae orthologs, indicating that there is extensive conservation of sex determination pathway components; the lone exception is fog-2 . We provide evidence that the essential role of FOG-2 in C. elegans hermaphrodite spermatogenesis evolved from post-speciation duplication and divergence of the fog-2 -related (FTR) gene family and that a fog-2 gene is not present in C. briggsae . Furthermore, double-stranded-RNA-mediated interference (RNAi) of the gld-1 ortholog in C. briggsae results in masculinization of the germline instead of the feminization of the germline phenotype observed in C. elegans . The lack of a potential C. briggsae fog-2 combined with the opposite sex determination function of GLD-1 in C. briggsae indicate that the control of hermaphrodite spermatogenesis, while using most of the same gene products, is fundamentally different between the species and is likely to have evolved independently. Results Components of Sex Determination Pathway Are Conserved between C. elegans and C. briggsae To survey conservation in the sex determination pathway between C. elegans and C. briggsae we used reciprocal best BLAST [ 37 , 38 , 39 ] to identify potential C. briggsae orthologs of 31 known C. elegans sex determination genes, some of which have been previously identified. The 31 genes included 16 that function only in germline sex determination, seven that function in both somatic and germline sex determination, two that function only in somatic sex determination, and six that coordinate sex determination and dosage compensation. We found that 30 of 31 genes have C. elegans –to– C. briggsae reciprocal best BLAST hits and alignments consistent with a high level of conservation ( Table 1 ). Using this method, putative orthologs of all known sex determination genes, including less conserved members, and previously identified genes were recovered [ 17 , 26 , 40 , 41 , 42 , 43 , 44 ], with the notable exception of fog-2 . Table 1 Comparative Analysis of Sex Determination Genes in C. elegans and C. briggsae WormPep ( C. elegans ) and BriggPep ( C. briggsae ) entries are protein identification numbers from Wormbase ( http://www.wormbase.org ). ID entries are C. elegans gene identifiers from Wormbase Reciprocal best BLAST hits are indicated by “yes” or “no,” and e-values are presented using WormPep release 112 and C. briggsae protein predictions (Wormbase). “Percent Length” is the extent of alignable sequence. All proteins with the exception of fog-2 returned reciprocal best BLAST hits in C. elegans and C. briggsae. Proteins that contain RNA-binding motifs or that function in RNA regulation are the following: ATX-2, FOX-1, FOG-1, FOG-2, GLD-1, GLD-3, NOS-1, NOS-2, NOS-3, FBF-1, FBF-2, MOG-1, MOG-4, MOG-5, and MOG-6 a C. elegans FBF-1 and FBF-2 share 90% amino acid identity and 95% amino acid similarity. BLAST searches using C. elegans FBF-1 or FBF-2 result in the same C. briggsae best hit (CBP14598). A partial FBF family phylogeny suggests recent duplications of a common FBF ancestor have occurred in both C. elegans and C. briggsae (data not shown) N/A, not applicable The functions of seven C. briggsae sex determination genes have been tested, and current data indicate that these genes exhibit similar and possibly identical functions in C. elegans and C. briggsae ( her-1 [ 43 ], tra-2 [ 21 ], fem-1 [A. Spence, personal communication], fem-2 [ 45 ], fem-3 [ 41 ], fog-3 [ 42 ], and tra-1 [ 17 ]). Importantly, the epistatic relationship and function of two key regulators of sex determination, tra-2 and fem-3, are essentially intact between the sister species in somatic sex determination [ 21 , 41 ] ( Figure 1 A). At first glance, given the conservation of 30/31 sex determination genes, similar or identical functions for 7/7 genes tested, and maintenance of a key epistatic relationship, it would appear that the sex determination pathway is generally conserved between C. elegans and C. briggsae . However, genetic and molecular studies will be required to determine whether the C. briggsae orthologs are functionally equivalent to their C. elegans counterparts. A single FOG-2 ortholog could not be resolved by reciprocal best BLAST or by using the reciprocal smallest distance algorithm [ 46 ], which uses global sequence alignment and maximum likelihood estimation of evolutionary distances, to infer putative orthologs (data not shown). This indicates that fog-2 is either highly diverged, present in an unsequenced portion (<2%) of the C. briggsae genome, or potentially a C. elegans –specific adaptation not present in C. briggsae . fog-2 Is a C. elegans –Specific Adaptation FOG-2 is part of a large, highly diverged F-box- and DUF38/FTH-containing protein family in C. elegans with more than 100 members referred to as FTR proteins [ 25 , 36 ]. The FTR family is also expanded in C. briggsae, making the identification of a single functionally equivalent ortholog from a large number of paralogs difficult. Therefore, to discern the relationships among C. elegans and C. briggsae FTR family members, 30 C. elegans and C. briggsae FTR proteins or protein predictions closely related to FOG-2 were used to generate a neighbor-joining phylogeny. The remaining, more diverged FTR members from either species were not included in the phylogeny to avoid long branch attraction [ 47 ]. The C. elegans and C. briggsae FTR phylogeny reveals that all of the C. elegans FOG-2 relatives form a single clade and all of the C. briggsae relatives a distinct clade. An unrooted radial phylogram illustrating C. elegans and C. briggsae FTR relationships is presented in Figure 2 , and a rectangular representation of the same phylogeny with bootstrap support information is shown in Figure S1 . If a closely related homolog of C. elegans FOG-2 were present in C. briggsae the expectation is that it would have clustered with the C. elegans proteins. Contrary to this, the phylogenetic separation of C. elegans and C. briggsae FTR family members into distinct lineages indicates that extensive expansion in the FTR family occurred post-speciation and that C. elegans and C. briggsae FTR genes do not have one-to-one orthologous relationships. Figure 2 The FTR Gene Family in C. elegans and C. briggsae A radial phylogram showing the relationships of 30 C. elegans and C. briggsae FTR genes closely related to FOG-2 was generated using neighbor-joining. C. elegans and C. briggsae protein predictions with complete F-box and Duf38/FTH (FTR proteins) were identified using BLAST and HMMs, aligned using CLUSTALW, trimmed, de-gapped, and realigned (see Materials and Methods ). A clear separation of C. elegans (below dashed line) and C. briggsae (above dashed line) FTR proteins is indicated by the phylogeny. The branch containing FOG-2 and FTR-1 is in bold. Tree is unrooted, and branch lengths are proportional to divergence (also see Figure S1 ). Bar represents 0.1 substitutions per site. FOG-2 and FTR-1, across their entire length, are more similar to each other than to any other gene in C. elegans . Comparison of the diverged approximately 40aa C-terminal region from both proteins to the closely related FTR genes in the FOG-2 cluster reveals 48% average pairwise identity between these FTRs and FTR-1 and 22% average pairwise identify between these FTRs and FOG-2 ( Figure S2 ). One interpretation of this greater similarity is that FTR-1 may be ancestral; however, it is not clear whether the slight increase in similarity over about 40aa is significant or whether selection rather than evolutionary history produced the sequence similarity observed. The above results could be misleading if a closely related C. briggsae fog-2 homolog were present in the less than 2% of the genome sequence that is not present in the final assembly or if the fog-2 ortholog diverged sufficiently such that the computational methods were not able to distinguish between orthologous and paralogous relationships. To address these possibilities we used low-stringency cross-species Southern blotting in an effort to identify closely related fog-2 -like sequences in unsequenced portions of the C. briggsae genome, and we used conserved synteny in an attempt to identify a diverged fog-2 ortholog that might reside in the same genomic location. Both approaches were used to effectively identify other diverged sex determination genes from C. briggsae ( tra-2, her-1, and fem-2 ) prior to the release of the C. briggsae genome sequence [ 40 , 43 , 44 ]. For low-stringency Southern blotting we used a C. elegans fog-2 probe and a fem-2 positive control probe against C. briggsae genomic DNA. Under conditions that detected cross-species hybridization with the C. elegans fem-2 probe against C. briggsae genomic DNA [ 40 ], no C. briggsae signal was observed with the C. elegans fog-2 probe ( Figure 3 A). This suggests either that a close fog-2 relative is not present in the less than 2% of the C. briggsae genome that is unsequenced or that it has diverged significantly beyond the level of fem-2 . Figure 3 fog-2 Is Likely Absent in C. briggsae Low-stringency Southern blotting (A) and conservation of synteny (B and C) were used in an attempt to identify a potential fog-2 gene in C. briggsae . (A) A total of 2–20 ug of digested genomic DNA was used in low-stringency Southern blotting. C. elegans fem-2 probe ( Ce _ fem-2 ) was able to detect fem-2 on both same-species and cross-species blots (first two panels). The C. elegans fog-2 probe ( Ce_fog-2 ), which detects both fog-2 and ftr-1 on the 5.8-kb XhoI fragment, produced a signal with C. elegans but not C. briggsae genomic DNA (next two panels). fog-2 cross-species blot integrity was verified by stripping and reprobing with same-species C. briggsae fem-2 (final panel). Same-species exposures were 4 h and cross-species were 4 d. The C. elegans fem-2 probe is 70% identical to the C. briggsae genomic sequence. (B) Scale diagram of the C. elegans Chromosome 5 region containing fog-2 . A 82.6-kb enlargement below, indicated by the dashed lines, shows the fog-2 cluster containing five canonical FTR genes, one FTR gene with divergent structure, and 16 non-FTR genes (also see Table S1 ). (C) C. briggsae contig from the genome assembly containing flanking regions with conserved synteny. A 194.4-kb enlargement below, indicated by the dashed lines, covers the C. briggsae region that is predicted to contain a putative fog-2 ortholog. The conserved genes used to identify the C. briggsae contig are indicated by the arrowheads, with the genes flanking fog-2 indicated by the large arrowheads. Each gene from the C. briggsae contig with an ortholog defined as a reciprocal best BLAST hit is present on both maps (B and C), and blocks of synteny defined by the C. elegans organization are in the same color. Only one (Y113G7B.11) of the 22 genes from the 82.6-kb fog-2 cluster was found to have a reciprocal best BLAST hit in C. briggsae (contig cb25.fpc0129, corresponding to the predicted gene CBG05618; Table S1 ). No FTR genes or genes related to those in the fog-2 cluster were found within 50-kb on either side of CBG05618, indicating that this region does not share conserved synteny with the fog-2 cluster. Instead, the potential C. briggsae ortholog of Y113G7B.11 is located on a C. briggsae contig region that shows extensive conserved synteny with a different portion of C. elegans Chromosome 5 not involving the fog-2 cluster ( Table S2 ). For analysis of conserved syntenic relationships, five conserved C. elegans genes surrounding fog-2 ( srg-34, sec-23, psa-1, Y113G7A.14, and Y113G7B.15) were used to query C. briggsae contigs. The genes srg-34, sec-23, and psa-1 are highly conserved across metazoans and have reciprocal best BLAST hits in C. briggsae ( Figure 3 B and 3 C, small arrow heads). The genes Y113G7A.14 and Y113G7B.15 flank the gene-dense C. elegans fog-2 region and also have reciprocal best BLAST hits in C. briggsae ( Figure 3 B and 3 C, large arrow heads). All five genes were found to be represented on a single C. briggsae contig, suggesting that the global syntenic relationships are conserved, but with detailed analysis revealing a number of differences in gene order ( Figure 3 B and 3 C). However, fog-2 , its four adjacent close FTR relatives, and 16 surrounding genes in an 82.6-kb region were absent from this C. briggsae contig, while the conserved genes on either side were present (Table S1 and S2 ). The closest relative of fog-2 is the gene ftr-1, which is part of a group of five closely related ftr genes that are colinear in C. elegans and not present in C. briggsae [ 25 ] ( Figures 2 and 3 ). If fog-2 and ftr-1 are the result of a “recent” post-speciation duplication within the C. elegans lineage, as suggested by the phylogeny, then we would expect that fewer synonymous substitutions (K s ) have occurred between fog-2 and ftr-1 relative to other C. elegans/C. briggsae best BLAST orthologs. Consistent with a recent duplication, the K s for fog-2/ftr-1 is not saturated (K s = 0.36) whereas the average K s for reciprocal best BLAST hits between C. elegans and C. briggse is saturated (K s = 1.72) [ 5 ]. The finding that fog-2 and ftr-1 arose from a relatively recent local duplication within C. elegans strongly supports the contention that fog-2 is not present in C. briggsae . These results imply that C. briggsae must regulate hermaphrodite spermatogenesis differently than C. elegans . The Diverged C-Terminal of FOG-2 Is Necessary for GLD-1 Binding Previous work has shown that FOG-2 is an integral part of the tra-2 3′ UTR translational repression complex. The RNA-binding protein GLD-1 makes direct contact with the tra-2 3′ UTR, and FOG-2 is recruited to the complex via its interaction with GLD-1 [ 24 , 25 ]. In spite of the high similarity between fog-2 and ftr-1 ( Figure 4 ), ftr-1 cannot compensate for fog-2 in the promotion of hermaphrodite spermatogenesis [ 25 ]. This indicates that fog-2 must contain unique sequences that allow it to function in sex determination. Figure 4 The Highly Diverged FOG-2 C-Terminal Region Is Responsible for GLD-1 Interaction in C. elegans (A) Dot plot of FOG-2/FTR-1, with the black diagonal line delimiting regions of greater than 70% identity based on a 10-aa sliding window. The dashed horizontal line at the C-terminus indicates a region of low identity. The arrow indicates the final exon 4 boundary. (B) Protein sequence alignment of FOG-2 and FTR-1 encoded by exon 4. Differences are shaded in black and illustrate the abrupt breakdown in sequence conservation. The dashed line marks the region required for GLD-1 interaction. (C) Nucleotide alignment of fog-2 and ftr-1 EST coding regions expanded from a portion of the protein sequence alignment, with vertical lines delimiting the reading frame relative to fog-2. Amino acid sequence for FOG-2 (above) and changes in FTR-1 (below) are below the alignment. Frame-shifting indels are indicated by the large open arrowheads. (D) The C-terminal FOG-2 region is required for GLD-1 interaction in the yeast two-hybrid system. Full-length FOG-2 (black) and FTR-1 (grey) constructs were tested for interaction with GLD-1. FOG-2 interacts with GLD-1 (++++) whereas FTR-1 does not (−). Progressive C-terminal deletions (black) in FOG-2 were generated to identify FOG-2 requirements for GLD-1 interaction. Binding to GLD-1 was completely eliminated with the removal of the C-terminal 64 aa of FOG-2 exon 4. Transfer of exon 4 to FTR-1 (grey/black chimera) resulted in the transfer of GLD-1 binding to FTR-1. Control interactions to test for the production of functional proteins were performed with the Skp1 homolog SKR-1, which binds to the F-box region (see Materials and Methods ). Searches for C. elegans and C. briggsae proteins with homology to the 64-aa FOG-2 region required for GLD-1 interaction (or FOG-2 exon 4) failed to identify any predicted proteins with significant homology (>35% or e-value = 0.01) other than FTR-1, which cannot bind GLD-1 and does not compensate for FOG-2 in sex determination. (E) Sliding-window (100-nt window, 25-nt shift) estimation of K a /K s ratio for fog-2/ftr-1 using full-length average K s . The K a /K s ratio is highest at the C-terminal end of the Duf38/FTH domain, reaching a peak of 2.2 in window 37. The position of the F-box and Duf38/FTH domain are indicated by grey shading. The bold horizontal line is at the K a /K s = 1 threshold. The dashed vertical line indicates the boundary between exon 3 and exon 4. Pairwise comparisons between FOG-2 and FTR-1 reveal a highly diverged C-terminal region encoded by the final exon (exon 4) ( Figure 4 A– 4 C). Before the C-terminal region of low similarity, the relative reading frames of fog-2 and ftr-1 are conserved with all insertions and deletions in three nucleotide multiples and an overall amino acid identity of 70%. Within the final exon, multiple amino acid substitutions, insertions, and deletions have occurred, resulting in a region of low nucleotide and amino acid identity ( Figure 4 B and 4 C). For example, an indel (deletion relative to fog-2 ) at nucleotide 805 shifts the reading frame of FOG-2 relative to FTR-1 and results in a region of low similarity between the proteins ( Figure 4 B). A second indel at position 819 restores the reading frame but additional substitutions result in a diverged amino acid sequence ( Figure 4 C). The dramatic differences between the FOG-2 and FTR-1 C-terminal regions suggested a connection between the unique functionality of FOG-2 in sex determination and the highly diverged C-terminal region. Since FOG-2 interacts with GLD-1 and both are required for the promotion of the male germ cell fate in the hermaphrodite, we determined whether the diverged FOG-2 C-terminal region was necessary for its interaction with GLD-1 ( Figure 4 ). Progressive C-terminal deletions of FOG-2 were tested for their ability to interact with GLD-1 in the yeast two-hybrid system ( Figure 4 D). Full-length FOG-2 interacts with GLD-1 [ 25 ]; however, C-terminal deletions of nine and 28 aa in FOG-2 reduced the interaction, and deletion of 64 and 76 aa (essentially all of exon 4) eliminated the interaction ( Figure 4 D), indicating that the highly divergent C-terminal region is necessary for GLD-1 binding. All full-length and deletion constructs were tested against the Skp1 homolog SKR-1 as a positive control for functionality in the two-hybrid system (see Materials and Methods ). To determine whether the C-terminal region of FOG-2 is sufficient to confer GLD-1 interaction, an FTR-1/FOG-2 exon 4 chimera was generated and assayed for its ability to interact with GLD-1. Normally FTR-1 lacks the ability to interact with GLD-1 [ 25 ] ( Figure 4 D). The replacement of exon 4 from ftr-1 with exon 4 from fog-2 allowed the chimera to interact with GLD-1 ( Figure 4 D). Thus, the C-terminal 74aa region of FOG-2, when in the context of the FTR-1 F-box and Duf38/FTH sequences, is sufficient to confer GLD-1 binding. FOG-2/GLD-1 Interaction Evolved Rapidly in C. elegans Gene duplication provides the raw material for the evolution of novel adaptations, having been implicated in the diversity of the host–pathogen immune response, rapid onset of insecticide resistance, and diversity of vertebrate body plans [ 48 ]. Rapidly evolving genes, or portions of genes, under positive selection can be identified by comparison of nucleotide alterations that result in amino acid changes (non-synonymous substitutions [K a ]) to alterations that do not change the amino acid (K s ) [ 49 , 50 ]. K a /K s ratios that are equal to or less than one are indicative of neutral or purifying selection, where substitutions that change amino acids offer no fitness advantage or result in lowered fitness. In contrast, K a /K s ratios greater than one, common in rapidly evolving genes, are indicative of positive selection, where non-synonymous changes offer some fitness advantage and are fixed at a higher rate than synonymous substitutions [ 51 ]. To determine the selection acting on the fog-2/ftr-1 duplication we compared K a /K s ratios between fog-2, ftr-1, and the five FTR genes closest to fog-2 in C. elegans . Pairwise comparisons of codon-delimited full-length coding sequences closely related to fog-2 suggest that purifying selection dominates along the fog-2 branch, as all comparisons produced K a /K s ratios less than one (mean = 0.46). However, while the overall K a /K s ratio for fog-2/ftr-1 is not indicative of positive selection (mean = 0.58), sliding-window K a /K s ratio estimates [ 52 ] for fog-2 and ftr-1 indicate that the highly diverged C-terminal region of FOG-2/FTR-1 contains residues under positive selection (K a /K s = 1.98 for nucleotides 777–987, windows 33–37) ( Figure 4 ). An alternate method using maximum likelihood estimation of K a /K s (PAML and codeml [ 53 ]) confirmed the presence of residues under positive selection within the C-terminal region (see Materials and Methods ). Thus, the primary differences between FOG-2 and FTR-1 are localized to the rapidly evolving C-terminus of FOG-2 that is required for GLD-1 binding and is under positive selection. The yeast two-hybrid data, together with the genetics of fog-2 [ 25 ], indicate that FOG-2 is unique among C. elegans FTR genes in functioning with GLD-1 in germline sex determination. Given the specificity of the FOG-2/GLD-1 interaction in C. elegans, phylogenetic analysis of FTR proteins (see Figure 2 ), and additional experiments (see Figures 3 and 4 ) that indicate that there are no close relatives of fog-2 among C. briggsae FTR genes, it is unlikely that any C. briggsae FTR protein functions with C. briggsae GLD-1 in sex determination. In contrast with FOG-2, a highly conserved GLD-1 ortholog is present in C. briggsae ( Table 1 ) and has a germline expression pattern essentially identical to that of C. elegans ( Figure 5 A, top right and middle right). In fact, C. elegans GLD-1 and C. briggsae GLD-1 share 81% amino acid identity overall and more than 90% in the maxi-KH RNA-binding region. Since FOG-2 and GLD-1 function together to promote the male germ cell fate in C. elegans hermaphrodites, this raised the question of what role, if any, C. briggsae GLD-1 plays in C. briggsae germline sex determination. Figure 5 GLD-1 Has the Opposite Sex Determination Function in C. elegans and C. briggsae For (A) and (B) the distal end of the gonad arm is indicated by the asterisk, and regions of the germline are delimited by dashed vertical lines as follows: M, mitotic zone; TZ, transition zone; P, pachytene; Pa, abnormal pachytene; and S, spermatocytes. For both (A) and (B) staining indicated is as follows: DAPI, blue, nuclear DNA; GLD-1, green; and MSP, red. (A) RNAi of C. briggsae gld-1 results in masculinization of the germline. Paired DAPI-stained (left) and GLD-1- and MSP-stained (right) images of dissected young adult hermphrodite germlines. Top four panels illustrate the similarity between C. elegans and C. briggsae germline morphology and polarity (DAPI, blue; GLD-1, green; MSP, red). In both species, sperm (“sperm” arrow) are produced first before switching to oogenesis (“oocytes” arrow), and the pattern of cytoplasmic GLD-1 accumulation (green) is identical. GFP-injected controls were identical to wild-type animals. C. briggsae gld-1 RNAi animals exhibit masculinization of the germline (lower panels). A vast excess of sperm extends to the loop region (“sperm” arrows), and spermatogenesis extends further distally (solid line). Masculinization is confirmed by a corresponding extension in MSP staining beyond the loop (compare lower right to controls above). (B) RNAi of gld-1 and fog-3 in C. elegans and C. briggsae results in a similar tumorous germline phenotype. C. elegans (top) and C. briggsae (bottom) have normal mitotic, transition, and entry into pachytene, but abnormal progression through pachytene, based on DAPI morphology. Both MSP and GLD-1 staining were below the level of detection in both cases. GLD-1 Has Distinct Functions in C. elegans and C. briggsae Germline Sex Determination To examine C. briggsae GLD-1 function in sex determination we performed RNAi [ 54 ] by injecting double-stranded C. briggsae gld-1 RNA into C. briggsae adult hermaphrodites followed by phenotypic analysis of F1 self progeny (see Materials and Methods ). From genetic analysis of C. elegans gld-1 [ 28 , 29 ] there are two functions relevant to this study. First, C. elegans GLD-1 has an essential function in meiotic prophase progression during oogenesis. In null mutant hermaphrodites oogenic germ cells progress to pachytene and then return to the mitotic cell cycle, giving rise to ectopic proliferation and a germline tumor [ 28 ]. For this function C. elegans GLD-1 acts to spatially restrict the translation of multiple target mRNAs during oogenesis. GLD-1 oogenic target mRNAs are repressed during early meiotic prophase, but then are translated during late meiotic prophase following the loss of GLD-1 at the end of pachytene [ 30 , 31 , 55 ]. Second, C. elegans GLD-1 is necessary for the specification of the male sexual fate in the hermaphrodite germline. This function is most simply revealed as a haplo-insufficient feminization of the hermaphrodite germline [ 28 , 29 ]. C. elegans gld-1 has no known essential functions in male meiotic prophase progression or in XO male germline sex determination as C. elegans null males are wild-type [ 28 , 29 ]. C. briggsae GLD-1 may still function as a translational repressor of C. briggsae tra-2 mRNA even in the absence of a FOG-2 ortholog. This is a possibility because FOG-2 is not required for C. elegans GLD-1 binding to the C. elegans tra-2 mRNA in vitro [ 25 ], and some conservation is preserved between the C. elegans and C. briggsae tra-2 3′ UTRs [ 34 ]. In this case, RNAi of GLD-1 in C. briggsae might feminize the germline given that C. briggsae tra-2 promotes female development in both the germline and soma [ 21 ]. Alternatively, C. briggsae GLD-1 might have no role in germline sex determination, in which case RNAi would not result in a sex determination phenotype. Surprisingly, C. briggsae gld-1 RNAi resulted in a masculinized germline ( Figure 5 A, bottom; Table 2 ), with no effect on the soma. Staining with 4′,6′-diamidino-2-phenylindole hydrochloride (DAPI) and anti–major sperm protein (MSP) (see Materials and Methods ) revealed continuous spermatogenesis leading to a vast excess of sperm at the expense of oogenesis. Anti-GLD-1 antibody staining of gld-1 RNAi F1 gonad arms indicated that the level of GLD-1 protein was reduced to below detectable limits ( Figure 5 A, bottom right). C. briggsae control hermaphrodites injected with double-stranded RNA for green fluorescent protein (GFP) had gonad morphology identical to wild-type ( Figure 5 A, top left and middle left). The masculinized phenotype of gld-1 RNAi in C. briggsae indicates that the wild-type function of GLD-1 in C. briggsae is to promote the female germ cell fate, likely by the translational repression of an mRNA that encodes a masculinizing gene product. This function is in direct contrast to that of C. elegans GLD-1, which promotes the male germ cell fate by translational repression of the feminizing tra-2 mRNA. Table 2 Summary of GLD-1 RNAi Germline Phenotype in C. elegans and C. briggsae a Results are from a single group of experiments. Similar results were obtained in other experiments b “Other” refers to masculinized arms with proximal proliferation GLD-1 Function in Meiotic Prophase Progression during Oogenesis Is Conserved Given the difference in sex determination function, it is possible that C. elegans and C. briggsae GLD-1 have few conserved functions in germline development. To investigate this we took advantage of well-defined activities of gld-1 in C. elegans such as its essential function in female meiotic prophase progression and in the translational repression of the evolutionarily conserved yolk receptor mRNA encoded by the rme-2 locus [ 28 , 31 ]. The gld-1 -null tumorous phenotype results from aberrant oogenic prophase progression and a return to mitosis [ 28 , 29 ]. This phenotype is dependent on germline sex because a tumor only occurs when germ cell fate is set to female [ 28 , 29 ]. The masculinized phenotype caused by gld-1 RNAi in C. briggsae is likely to preclude the detection of this function as the C. elegans gld-1 -null tumorous phenotype is suppressed by mutations that cause masculinization of the germline [ 29 ]. To overcome the masculinization we combined fog-3 RNAi with gld-1 RNAi in C. briggsae . Since C. elegans fog-3 functions near the end of the sex determination pathway and in C. briggsae fog-3 RNAi results in feminization of the germline [ 42 ], we predicted that C. briggsae fog-3 RNAi would be epistatic to the masculinization of the germline of C. briggsae gld-1 RNAi. Similar to the C. elegans gld-1 -null, RNAi of gld-1 or gld-1 and fog-3 in C. elegans and double RNAi of gld-1 and fog-3 in C. briggsae resulted in a robust proximal germline tumor ( Figure 5 B; Table 2 ). Control RNAi with fog-3 alone resulted in feminized germlines in both species [ 42 ]. Both the mitotic zone and transition zone appear to have roughly normal nuclear morphology, with more proximal nuclei having abnormal pachytene morphology ( Figure 5 B), suggesting that germ cells are entering meiosis but progressing aberrantly before returning to mitosis. The return-to-mitosis tumorous phenotype in each species was confirmed using phosphohistone H3 staining, a mitotic proliferation marker [ 56 ]. We cannot rule out the possibility that the C. briggsae phenotypes observed, masculinization of the germline with gld-1 RNAi alone and tumorous germline with gld-1 and fog-3 RNAi, are the result of incomplete knockdown leading to partial gld-1 loss of function. The rme-2 yolk receptor mRNA is a known target of GLD-1-mediated translational repression in C. elegans [ 31 ]. In C. elegans, GLD-1 and RME-2 have mutually exclusive expression patterns because rme-2 mRNA is translationally repressed in the transition zone and pachytene region, where GLD-1 levels are high, and translated in oocytes, where GLD-1 levels are low [ 31 ]. In C. elegans gld-1 -null germlines RME-2 is ectopically expressed in the transition zone and pachytene region owing to loss of GLD-1-mediated translational repression of the rme-2 mRNA [ 31 ]. A similar, mutually exclusive accumulation pattern in C. briggsae suggests that C. briggsae GLD-1 is a translational repressor of C. briggsae rme-2 mRNA ( Figure 6 ). To determine whether C. briggsae GLD-1 represses the rme-2 mRNA, double RNAi of gld-1 and fog-3 was performed in both species, and gonad arms were stained for RME-2 protein [ 57 ]. Reduction of GLD-1 and FOG-3 by RNAi results in the ectopic accumulation of RME-2 protein in both C. elegans and C. briggsae ( Figure 6 ), indicating that the role of GLD-1 in the translational repression of the rme-2 mRNA is conserved. Thus, despite the opposite roles of GLD-1 in sex determination, essential functions of GLD-1 in oogenesis are conserved between the species. Figure 6 GLD-1-Mediated Translational Repression of rme-2 mRNA in C. elegans and C. briggsae In both C. elegans and C. briggsae wild-type (WT) animals (left panels), GLD-1 (green) and RME-2 (red) have mutually exclusive accumulation patterns. In C. elegans (upper right), gld-1 and fog-3 RNAi results in a germline tumor with ectopic RME-2 accumulation (red expanded). In C. briggsae (lower right), RNAi of gld-1 and fog-3 also results in germline tumor with ectopic RME-2 accumulation (red expanded). The germline tumor and expansion of RME-2 expression due to ectopic translation are similar between the two species (compare right top and bottom, DAPI [blue]). The distal end of the gonad arm is indicated by the asterisk, and regions of the germline are delimited by dashed vertical lines. DAPI, blue, nuclear DNA; GLD-1, green; RME-2, red; M, mitotic zone; TZ, transition zone; P, pachytene; Pa, abnormal pachytene. Discussion Our results indicate that the control of hermaphrodite spermatogenesis is fundamentally different between the sister species C. elegans and C. briggsae at the level of FOG-2/GLD-1/ tra-2 mRNA regulation. While FOG-2 is essential for self-fertile hermaphroditism in C. elegans, a closely related homolog of FOG-2 could not be recovered in C. briggsae by reciprocal best BLAST, phylogenetic inference, low-stringency hybridization, or analysis of conserved synteny . Comparison of synonymous changes between fog-2 and its closest relative, ftr-1, indicates that fog-2 is the product of a recent expansion “specific” to C. elegans in the FTR gene family and implies that the evolution of FOG-2 and its incorporation into the sex determination pathway occurred post-speciation. Consistent with this, the C-terminal region of FOG-2 required for binding to GLD-1 was found to be highly diverged and “unique” to FOG-2 in C. elegans . Interestingly, GLD-1 was found to have a sex determination function in C. briggsae opposite that in C. elegans while retaining similar functions in female meiotic prophase progression and oogenesis. The absence of FOG-2, and the opposite sex determination function of GLD-1, provides evidence for the independent evolution of hermaphroditism in C. elegans and C. briggsae . General Conservation of the Sex Determination Pathway Reciprocal best BLAST indicates that C. elegans and C. briggsae have orthologs of 30 of 31 known sex determination pathway genes. Conserved functions for C. briggsae her-1, tra-2, fem-1, fem-2, fem-3, fog-3, and tra-1 have been demonstrated by transgene rescue of C. elegans mutations or similarity of RNAi loss-of-function phenotype [ 17 , 21 , 26 , 41 , 42 , 43 , 45 ]. The general conservation of genes that govern sex determination suggests that the underlying pathway remains largely intact between the species. RNAi and transgenic experiments have suggested that while fem-2 and fem-3 have conserved roles in the somatic sex determination of both species, they may play diminished roles in C. briggsae germline sex determination [ 41 , 45 ]. There are two possibilities that could explain these results. One is that there are inherent species-specific differences in susceptibility to RNAi or in the ability to reconstitute complete gene function by transgene rescue. The other is that differences in C. elegans and C. briggsae phenotypes reveal functional divergence in sex determination pathway components. Analysis of null mutations in C. briggsae orthologs of C. elegans sex determination genes will help to distinguish between these possibilities. While some functional differences may turn out to be valid, tra-2 (feminizing) and fem-3 (masculinizing) apparently play the same somatic roles in both species, and their epistatic relationship appears to be conserved [ 41 ]. fog-2 Is Unique to C. elegans Within the context of general conservation of sex determination pathway components and conserved key epistatic relationships, the absence of fog-2 in C. briggsae is intriguing. fog-2 arose as a consequence of recent C. elegans –specific gene duplication events, and none of the closely related C. elegans fog-2 paralogs can compensate for loss of fog-2 in sex determination [ 25 ]. Thus, it is unlikely that more distantly related C. briggsae FTRs are involved in GLD-1/ tra-2 -mRNA-mediated promotion of hermaphrodite spermatogenesis. Since fog-2 is essential for the promotion of spermatogenesis in C. elegans hermaphrodites and is not present in C. briggsae, the direct implication is that specification of the male germ cell fate in C. briggsae hermaphrodites is fundamentally different from that in C. elegans and that it evolved independently. The highly diverged C-terminus of FOG-2 is under positive selection and is necessary and sufficient for GLD-1 binding within the context of an F-box and FTH domain (see Figure 4 ). Acquiring the diverged C-terminus was crucial in FOG-2 becoming incorporated into the sex determination pathway. With respect to the C. elegans lineage, it is unclear whether fog-2 retains an ancestral function in sex determination and ftr-1 has changed/drifted away or, alternatively, whether ftr-1 represents the ancestral function and fog-2 has recently evolved a role in sex determination (also see Figure S2 ). The ftr-1 gene is expressed, though its function is currently unknown. RNAi of ftr-1 into the fog-2 null did not reveal any obvious phenotypes beyond feminization of the germline [ 25 ]. Conserved GLD-1 Functions in C. elegans and C. briggsae Meiotic Prophase during Oogenesis GLD-1 function in meiotic prophase progression and oogenesis shows substantial conservation between the species (see Figures 5 and 6 ), which is not surprising given the high level of sequence conservation between C. elegans and C. briggsae GLD-1. This is illustrated by the rme-2 yolk receptor mRNA being regulated similarly between the species ( Figure 6 ). Current data indicate that C. elegans GLD-1 binds to, and likely represses translation of, more than 100 mRNA targets [ 31 , 55 ] (M.-H. Lee, V. Reinke, and T. Schedl, unpublished data). The C. elegans gld-1 -null tumorous phenotype likely results from misregulation of multiple mRNA targets [ 31 ]. While the identity of the misregulated mRNA targets causing the gld-1 -null tumorous phenotype are currently unknown, the fact that C. briggsae gld-1 and fog-3 RNAi results in a similar tumorous phenotype suggests that a similar, if not identical, set of C. briggsae GLD-1 mRNA targets are misregulated. The absence of a FOG-2 ortholog in C. briggsae is unlikely to have a major effect on GLD-1-mediated translational control since FOG-2 appears to be required only as a cofactor for tra-2 repression [ 25 , 27 , 31 , 55 , 58 ]. Thus, it is possible that the majority of GLD-1 mRNA targets involved in prophase progression and oogenesis are regulated similarly between species. Divergent GLD-1 Function in C. elegans and C. briggsae Sex Determination Genetic analysis reveals that C. elegans and C. briggsae GLD-1 have opposite functions in germline sex determination; C. elegans GLD-1 promotes spermatogenesis while C. briggsae GLD-1 promotes oogenesis. This indicates that the major sex determination function of C. briggsae GLD-1 is not translational repression of tra-2 feminizing activity. C. elegans GLD-1 binds two 28 nucleotide direct repeat elements on the C. elegans tra-2 mRNA 3′ UTR to mediate translational repression [ 24 ]. Somatic reporter gene assays in C. elegans and C. briggsae have suggested that the tra-2 3′ UTRs of both species are able to function in translational repression [ 34 ], with the implication being that the C. elegans and C. briggsae 3′ UTRs are regulated similarly. However, these data are difficult to interpret in the context of germline sex determination, as GLD-1 and FOG-2 are not natively expressed in the soma and neither GLD-1 nor FOG-2 have essential functions in somatic sex determination [ 25 , 27 , 28 , 29 , 30 ]. One hypothesis to explain our results is that C. briggsae GLD-1 binds to the C. briggsae tra-2 mRNA but is necessary for translational activation instead of translational repression as in C. elegans . However, for all characterized C. elegans GLD-1 targets, and C. briggsae rme-2 mRNA, GLD-1 acts as a translational repressor [ 2 , 31 , 55 , 58 , 59 ]. We currently do not understand how FOG-2 acts with GLD-1 in translational repression of C. elegans tra-2 mRNA. In C. elegans, GLD-1 can bind the tra-2 mRNA in the absence of fog-2 in worm extracts but cannot properly repress its translation in vivo [ 25 ]. This suggests that the role of FOG-2 may be to recruit additional factors specific to the C. elegans tra-2 mRNA 3′ UTR that allow for efficient GLD-1 translational repression. Assuming C. briggsae GLD-1 binds C. briggsae tra-2 mRNA in vivo, given the absence of a FOG-2 ortholog, there may be no regulatory consequence of this binding. Another possibility is that C. briggsae GLD-1 binds and translationally represses an mRNA that promotes spermatogenesis. This could occur if a masculinizing sex determination gene, either present in both species or unique to C. briggsae, has come under GLD-1 control in C. briggsae . Given the conservation of GLD-1 and its regulation of at least some common targets (e.g., rme-2 ) it is unlikely that changes in GLD-1 are responsible for a new mRNA target in C. briggsae . Instead, it is more likely that one or more new target mRNAs have acquired sequences that direct GLD-1 binding and translational repression. The requirements for GLD-1 binding are only just being elucidated, with a hexanucleotide sequence being one important feature amid otherwise diverse GLD-1 binding regions [ 32 , 55 ]. Thus, small numbers of changes in UTRs are likely to be sufficient for new mRNAs to come under GLD-1-mediated regulation. Evolution of Self-Fertile Hermaphroditism Current phylogenetic data suggest that hermaphroditism evolved independently in Caenorhabditis and other lineages of Rhabditid nematodes from an ancestral female/male state [ 5 , 6 , 7 , 10 , 11 , 60 ]. This is consistent with our results showing that control of hermaphrodite spermatogenesis at the level of FOG-2/GLD-1/ tra-2 mRNA is fundamentally different between C. elegans and C. briggsae . This raises the question, how might the transition from the ancestral female/male to hermaphrodite/male system of reproduction have occurred multiple times within the Caenorhabditis clade? The anatomy and reproductive physiology of C. elegans allow both sperm that is introduced by mating and sperm that develops within the female gonad of the hermaphrodite to be effectively used in reproduction [ 14 , 61 , 62 ]. Either source of sperm generates a MSP-derived signal that is required for full-grown oocytes to undergo meiotic maturation, ovulation, and fertilization in the spermatheca [ 62 , 63 ]. Not only is the anatomy conserved but an MSP-derived sperm signal also appears to be utilized by both C. briggsae and C. remanei (a female/male species) to induce oocyte maturation and ovulation [ 63 , 64 ]. This conservation within Caenorhabditis indicates that major changes in anatomy and reproductive physiology are not necessary in the transition from female/male to hermaphrodite/male reproduction. The relative ease with which mutants and mutant combinations can alter the sex determination system in C. elegans has suggested that transitions between mating systems may not be difficult and that the overall sex determination pathway reflects selection for a particular mating system rather than a constant regulatory mechanism [ 65 ]. The hermaphrodite pattern of spermatogenesis first then oogenesis is achieved by high masculinizing/low feminizing activity in early larvae followed by low masculinizing/high feminizing activity in late larvae/adults (see Figure 1 ; reviewed in [ 18 , 19 , 26 , 66 ]). Lowering or eliminating germline masculinizing activity in XX animals can convert C. elegans from hermaphrodite/male to female/male reproduction ( Table 3 , and references therein [ 20 , 27 , 28 , 29 , 66 , 67 , 68 , 69 ]). For example, fog-2 -null mutations result in strains that reproduce as XX females and XO males. The mutant female/male strains can be converted back to hermaphrodite reproduction by introducing masculinizing mutations in certain genes (e.g., fog-2 -null; fem-3- gf; Table 3 ). The generality of high masculinizing/low feminizing activity early followed by low masculinizing/high feminizing activity late is borne out by other sets of mutually suppressing feminizing-plus-masculinizing combinations in which the double mutants are self-fertile while each single mutant is usually self-sterile (e.g., tra-1- gf; fem-3 -gf; Table 3 ). Thus, multiple genetic states can yield self-fertile hermaphrodite/male and male/female reproduction in C. elegans . Table 3 C. elegans Sex Determination Mutants That Yield Female/Male Reproduction and Mutually Suppressed Hermaphrodite Reproduction a gf, gain of function; Mog, allele(s) show a masculinization of the germline phenotype; Fog, allele(s) show a feminization of the germline phenotype; Fem, allele(s) show a feminization of the soma and germline phenotype; ts, temperature sensitive; lf, loss of function, in these cases non-null b All of the masculinizing and feminizing mutant combinations that show mutual suppression display the pattern of sperm first then oocytes as in wild-type. The opposite pattern, oocytes first then sperm, would not result in self fertility and thus would not be reproductively successful. The reason that these mutant combinations all display the wild-type pattern, instead of the oocyte then sperm pattern, is unclear and suggests that an additional level of sex determination pathway regulation remains to be uncovered c Mutually suppressing feminizing and masculinizing double-mutant hermaphrodites often have intersexual germ cells between the sperm and oocytes, unlike wild-type hermaphrodites d Embryos generated showed developmental arrest Given the conservation of anatomy and reproductive physiology, an initial conversion from an ancestral Caenorhabditis female/male species to a hermaphrodite/male mode of reproduction may only require a genetic event that results in a transient increase in germline masculinizing activity in early larvae to produce sperm. As long as this change does not interfere with the higher level of feminizing activity (oogenesis) in late larvae/adults, self fertility would be possible. After the establishment of self fertility, there would likely be strong selection for additional genetic events that would optimize self-fertile brood size [ 70 ] and result in a clean transition from sperm to oocyte development so that wasteful intersexual gametes are not formed ( Table 3 ). Thus, it is very likely that multiple genetic events now define the differences in the C. elegans and C. briggsae germline sex determination pathways. In C. elegans, the relative levels of TRA-2 feminizing to FEM-3 masculinizing activity appear to be the major regulatory point for the sperm-then-oocyte pattern. There is no a priori reason for TRA-2 or FEM-3 to be the major focus of regulation to achieve hermaphroditism in C. briggsae; if one of these is the focus, then at least some of the regulation must differ between C. elegans and C. briggsae, given the absence of fog-2 and the changed role of GLD-1. Since the last common ancestor of C. briggsae and C. elegans must have contained orthologs of 30 of 31 C. elegans sex determination genes, a change in the regulation of one or more of these genes might be responsible. Alternatively, since much of the regulation of C. elegans germline sex determination is by translational control, mutations in UTRs of mRNAs may result in new genes coming under the control of GLD-1 or another RNA sex determination gene regulator ( Table 1 ). Additionally, duplication and divergence, analogous to what we have found for FOG-2 in C. elegans, may have resulted in a new gene being incorporated into the germline sex determination pathway. To move beyond speculation, the forward genetic analysis currently in progress (R. Ellis and E. Haag, personal communication) will be important for the identification of C. briggsae –specific genes, analogous to fog-2, that are necessary for self-fertile hermaphroditism. Materials and Methods Sex determination pathway conservation Protein coding sequences of cloned C. elegans sex determination genes were obtained from Wormbase ( http://www.wormbase.org ; WormPep release 112). C. briggsae genomic sequence was obtained from The Sanger Institute (Cambridge, United Kingdom) or the Genome Sequencing Center (St. Louis, Missouri, United States), and protein sequences were obtained from either Wormbase or Ensemble ( http://www.ensembl.org/ ; version 17.25.1). Best BLAST orthologs of C. briggsae sex determination proteins were obtained using C. elegans sex determination protein sequences as queries against C. briggsae predicted proteins and six-frame translated C. briggsae genomic sequence. C. briggsae proteins obtained at an e-value cutoff of 1 × 10 −50 reciprocal best hits were recovered for 26 of 31 C. elegans proteins. NOS-1 and XOL-1 orthologs were identified at an e-value cutoff of 1 × 10 −20 and were also reciprocal best BLAST hits between species. In each case a single reciprocal best hit was identified for each component of the sex determination pathway with the exception of FBF-1 and FBF-2, which returned the same best BLAST hit, and FOG-2. Searches of the non-redundant National Center for Biotechnology Information protein database (GenBank CDS+PDB+SwissProt+PIR+WormPep) with full-length FOG-2 as query revealed only weak similarity to the F-box motif for non– C. elegans or – C. briggsae sequences. Using the highly diverged C-terminal end of FOG-2, including a portion of the Duf38/FTH, or the GLD-1 interaction region of FOG-2 as query did not reveal any hits below an e-value of 0.01 in C. elegans or C. briggsae other than FOG-2 and FTR-1. Identification of FTR family members FTR family members are defined by the presence of an N-terminal F-box and C-terminal Duf38/FTH domain (FTR) [ 25 ]. C. elegans FTR family members were identified using FOG-2 as a query against WormPep release 112. Each potential FTR was scanned for an N-terminal F-box motif and C-terminal Duf38/FTH domain using the hidden Markov models (HMMs) for each domain (HMMER 2.3.2) [ 35 ]. Similarly, C. briggsae FTR family members were identified using FOG-2 as a BLAST query and HMMs. In C. elegans, fog-2 (Y113G7B.5), ftr-1 (Y113G7B.4), CE35646 (Y113G7B.1), CE24144 (Y113G7B.3), CE23289 (Y113G7B.6), and CE23288 (Y113G7B.7) are closely related and tightly linked on Chromosome 5. CE35646 was not included in later analysis because of a divergent N-terminal structure. An FTR family also appears to be present and expanded in the obligate male/female species C. remanei based on the currently sequence assembly (Genome Sequencing Center, Washington University, St. Louis, Missouri, United States; 16 September 2004, BLASTn and tBLASTn; ftp://genome.wustl.edu/pub/seqmgr/remanei/plasmid_assembly ). Our preliminary analysis suggests that closest FOG-2 homologs from C. remanei have diverged from C. elegans approximately to the same level as the FTR genes in C. briggsae. A comprehensive phylogenetic analysis to resolve the relationships between C. elegans, C. briggsae, and C. remanei FTR family members will await accurate C. remanei protein predictions and a complete C. remanei assembly. Sequence alignments and analysis Alignments were generated using CLUSTALW, and conserved residues were identified with the Lasergene MEGALIGN (DNASTAR, Madison, Wisconsin, United States) package and Dialign [ 71 , 72 ], which was also used to identify conserved regions for subsequent phylogenetic analysis. The best BLAST C. briggsae hit to each C. elegans FTR protein used in the phylogeny was included in order to identify any potential one-to-one orthologous pairs along the FOG-2 branch. Non-homologous N- and C-terminal extensions were trimmed, and extremely distant family members unlikely to be functional FOG-2 orthologs were excluded to avoid long branch attraction [ 47 ]. Phylogenetic inference was performed using the neighbor-joining (neighbor) program in the PHYLIP package (Phylogeny Inference Package version 3.5c; Department of Genetics, University of Washington, Seattle, Washington, United States) using the BLOSUM45 distance matrix. Trees with and without gaps were generated, and comparison revealed some differences in branching order, but only within the species. For the tree presented here, positions with gaps were excluded and all non-homologous or highly divergent sequences trimmed. The topology of the tree structure was tested by bootstrapping with 1,000 replicates and by analysis of the alignment using protpars from the PHYLIP package (a maximum parsimony method), which produced a tree with a similar branching order. Trees were processed using TreeView [ 73 ]. Codon-restricted alignments for K a /K s calculation were generated using Se-Al (a sequence alignment editor by A. Rambaut, version 2; available at http://evolve.zoo.ox.ac.uk/software.html?id=seal ) to modify CLUSTALW-aligned cDNA or predicted cDNA sequences, and all gaps and frame-shifted regions were removed. Sliding-window K a and K s estimates [ 74 ] were generated using DNASP (version 3) [ 52 ], and codon-based analysis was performed using PAML (codeml) [ 53 ] (HKY substitution model) to confirm the presence of codons under positive selection (95% confidence) within the sliding windows. Worm culture and RNAi C. elegans (N2, Bristol, United Kingdom) and C. briggsae (AF16) were obtained from the Caenorhabditis Genetics Center University of Minnesota, Minneapolis, Minnesota, United States. Cultures of both were maintained on Escherichia coli OP50 on NGM plates at 20 °C as previously described [ 75 ]. RNAi was performed by injection in C. elegans and C. briggsae essentially as described previously [ 54 ]. Double-stranded RNAs for species-specific gld-1 and fog-3 were generated by PCR amplification of cDNA with SP6 (5′) and T7 (3′) linkers, gel purified, sequenced, and used in RNA synthesis reaction using the appropriate Ambion kit (MEGAscript SP6 or T7; Austin, Texas, United States). Double-stranded RNAs were injected at 0.5 mg/ml into young adult N2 animals and F1 progeny collected 12–48 h post injection and matured to 24 h post L4 stage before gonads were dissected, fixed, and stained to score for abnormal phenotypes. Staining Dissection, antibody, and DAPI staining of C. elegans and C. briggsae gonads were performed essentially as previously described with fixation in 3% formaldehyde, 80% methanol, and 100 mM dibasic potassium phosphate [ 29 , 30 ]. Affinity purified rabbit polyclonal anti-GLD-1 antibodies were used at 1:50, and MSP mouse monoclonal antibody was used at 1:2,000, both with overnight incubation at room temperature (anti-MSP antibody was the kind gift of M. Kosinski and D. Greenstein, Vanderbilt University School of Medicine, Nashville, Tennessee, United States). Texas Red or Alexa488 secondary antibodies were used to detect staining, and DAPI was used visualize DNA morphology. Epifluorescent images were captured with a Zeiss (Oberkochen, Germany) Axioskop coupled to a Hamamatsu Photonics (Hamamatsu City, Japan) digital CCD camera, and processed with Photoshop 7.0 (Adobe, San Jose, California, United States). All image post-processing (brightness, contrast, pseudo-color, unsharp mask) was performed identically for each image. Constructs and transformation GLD-1 and FOG-2 yeast two-hybrid binding assays were performed as previously described [ 25 ] with the inclusion of 20 mM 3-amino-triazole. Progressive C-terminal deletions in FOG-2 and FTR-1/FOG-2 chimeric constructs were generated using PCR amplification of the appropriate coding sequences (FOG-2 full-length [327 aa], 318 aa, 299 aa, 263 aa, or exon 4 [251aa], or FTR-1 full-length [318 aa]) and cloned by recombination in yeast. In each case GLD-1 was used as bait in the pAS1 vector (DNA binding) and FOG-2 deletion constructs in the pACTII vector (activation). FOG-2 was found to exhibit low levels of auto-activation in the pAS1 (DNA binding) vector, so binding assays were performed in only one direction to avoid background and using high levels of 3-amino-triazole. The constructs were sequenced, and the Skp1-related F-box-binding protein SKR-1 (in pAS1) was used as a positive control for interaction [ 76 , 77 ]. Supporting Information Figure S1 Phylogenetic Relationships of 30 C. elegans and C. briggsae FTR Genes Closely Related to FOG-2 Presented as a Rectangular Phylogram A clear separation of C. elegans and C. briggsae FTR genes ( C. briggsae is in grey shade) is suggested by the phylogeny. The branch containing FOG-2 and FTR-1 is in bold. Tree is unrooted, and branch lengths are proportional to divergence. Bar represents 0.1 substitutions per site. Bootstrap support for separation of C. elegans and C. briggsae sequences is indicated at the node (black dot) and at each node for the C. elegans FOG-2 branch. (34.1 MB TIF). Click here for additional data file. Figure S2 Alignments of FTR-1 and FOG-2 C-Terminal Regions to Other Closely related C. elegans FTR Family Members (A) FTR-1 and FTR family alignment. Residues identical to FTR-1 are shaded black, and residues identical between all FTR family members tested are shaded red. Average pairwise identity to FTR-1 is 48%. (B) FOG-2 and FTR family alignment. Residues identical to FOG-2 are shaded black, and residues identical between all FTR family members tested are shaded red. Average pairwise identity to FOG-2 is 22%. (15.6 MB TIF). Click here for additional data file. Table S1 Analysis of Genes in the fog-2 Cluster (59 KB PDF). Click here for additional data file. Table S2 Analysis of Genes Surrounding Y113G7B.11 in C. briggsae (59 KB PDF). Click here for additional data file.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC539060.xml
544869	Despite WT1 binding sites in the promoter region of human and mouse nucleoporin glycoprotein 210, WT1 does not influence expression of GP210	Background Glycoprotein 210 (GP210) is a transmembrane component of the nuclear pore complex of metazoans, with a short carboxyterminus protruding towards the cytoplasm. Its function is unknown, but it is considered to be a major structural component of metazoan nuclear pores. Yet, our previous findings showed pronounced differences in expression levels in embryonic mouse tissues and cell lines. In order to identify factors regulating GP210, the genomic organization of human GP210 was analyzed in silico . Results The human gene was mapped to chromosome 3 and consists of 40 exons spread over 102 kb. The deduced 1887 amino acid showed a high degree of alignment homology to previously reported orthologues. Experimentally we defined two transcription initiation sites, 18 and 29 bp upstream of the ATG start codon. The promoter region is characterized by a CpG island and several consensus binding motifs for gene regulatory transcription factors, including clustered sites associated with Sp1 and the Wilms' tumor suppressor gene zinc finger protein (WT1). In addition, distal to the translation start we found a (GT)n repetitive sequence, an element known for its ability to bind WT1. Homologies for these motifs could be identified in the corresponding mouse genomic region. However, experimental tetracycline dependent induction of WT1 in SAOS osteosarcoma cells did not influence GP210 transcription. Conclusion Although mouse GP210 was identified as an early response gene during induced metanephric kidney development, and WT1 binding sites were identified in the promoter region of the human GP210 gene, experimental modulation of WT1 expression did not influence expression of GP210. Therefore, WT1 is probably not regulating GP210 expression. Instead, we suggest that the identified Sp binding sites are involved.	Introduction Nuclear pore complexes (NPCs) provide the only known gateway for transport of RNAs to the cytoplasm and bidirectional transport of proteins between the nucleus and the cytoplasm. The NPC in vertebrates has an estimated mass of approximately 125 Mda. Structural studies suggest an octagonal rotational symmetry framework, from which 50–100-nm long fibrils extend into the nucleoplasm and cytoplasm. A comprehensive inventory of all NPC constituents has been made for yeast [ 1 ] and metazoans [ 2 ]. A polypeptide profile from purified rat liver NPCs revealed ~50 putative nucleoporins [ 3 ]. In the list of metazoan nucleoporins, there are only two integral membrane proteins, gp210 [ 4 - 6 ] and POM121 [ 7 , 8 ]. Both have been localized to the NPC structure, each with a distinct membrane topology and amino acid motifs. Primarily due to their location, both proteins are presumed to anchor NPCs by the nuclear envelope and to assemble nucleoporins postmitotically. No binding partners have so far been identified for either of these proteins. The 121-kDa pore membrane protein POM121 [ 7 , 8 ] is located in the pore membrane domain of the NPC with a short (29 residues) N-terminal tail protruding into the lumen of the nuclear envelope, with the C-terminus facing the cytoplasm [ 8 ]. POM121 contains a C-terminal tandem sequence repeat of a core XFXFG motif interrupted by hydrophilic spacers. These motifs typical for nucleoporins and have been shown to interact with components of the soluble transport machinery [ 3 , 9 ]. In contrast to POM121, gp210 has an inverted topology with its main bulk residing in the lumen of the NE and only a short 58 residue C-terminal portion facing the NPC [ 5 , 6 ]. The amino acid-sequence of gp210 lacks pentapeptide repeats indicating no direct interaction with the mobile receptors directing nucleocytoplasmic transport [ 5 , 10 ]. A 23-amino-acid hydrophobic peptide residing in the luminal part of gp210 has been predicted to be involved in formation of new pores acting as a nuclear membrane fusion agent [ 5 , 11 ]. It has also been experimentally shown that the C-terminus of gp210 is involved in nuclear pore dilation [ 11 ], even though this is not a conserved sequence in different species [ 12 ]. Remarkably, it has also been shown that gp210 is essential for viability of human HeLa cells and C. Elegans [ 13 ]. A fraction of the cellular pool of gp210 can form dimers that may constitute a lumenal submembranous protein skeleton [ 14 ]. The primary sequence of gp210 is known for rat [ 5 ] and mouse [ 10 ]. Interestingly, whereas several nucleoporins found in vertebrates have homologues in the completed yeast genome, no such similarities have so far been detected for POM121 or gp210. Possibly, this could be related to the fact that the yeast nuclear membrane does not break down during cell division, and assembly regulators are not needed. In a comprehensive analysis of a highly enriched NPC fraction, presumably containing all yeast NPC proteins [ 1 ], only three transmembrane nucleoporins were detected, but these have no resemblance with gp210 or POM121. Thus, if the role of POM121 or gp210 in metazoans is to anchor the NPC, different proteins or mechanisms should be involved in the anchorage of yeast NPCs. Mouse gp210 was initially identified as an early response gene to induction of metanephric kidney development and data from other embryonic tissues confirmed the differential distribution of its mRNA [ 10 ] and protein [ 15 ]. This suggested a novel cell-type specific regulation of gp210. It was thus of interest to characterize the promoter region of the human GP210 gene. In the current study we present the genomic structure of the human integral membrane glycoprotein 210 gene (GP210), the open reading frame sequence and a promoter region analysis. This was done in silico by taking advantage of the available human genomic sequence. Transcription start sites were determined experimentally by RNA ligase mediated rapid amplification of cDNA ends. Computer-assisted searches of the promoter sequence indicated putative consensus binding sites for transcription factors involved in tissue specific gene regulation. We also identified of shared putative cis -acting elements in the human promoter and its mouse counterpart. Several putative Wilms tumor suppressor binding 1 sites were found. Nevertheless, experimental overexpression of WT1 in SAOS osteosarcoma cells did not influence GP210 mRNA expression. Results Organization of the human GP210 gene We initially assumed that mouse gpP210 was a member of a yet undiscovered large family of tissue-specific nuclear pore membrane proteins, and initially named it POM210 [ 10 ] to emphasize the similar subcellular distribution with POM121 [ 7 ]. Since current data suggest a surprisingly low amount of pore membrane proteins both in vertebrates and yeast, renaming is unnecessary. A BLAST homology search was performed using the mouse gp210 cDNA sequence (POM210, accession AF113751) against the working draft sequences of the human genome. This identified a completed contig-component (clone RP11-220D14, accession AC090942.1) localized to chromosome 3, in a region defined by three genomic markers (stSG4499, Cda14e10 and WI-9637). These markers were cytogenetically positioned to 3p25.1. By comparing to the mouse cDNA sequence and taking advantage of the exon/intron prediction program provided by the Genscan web server, 40 exons covering 102551 bp were defined (see Table 1, additional file 1 ). The exons ranged between 63 (exon 24) and 251 (exon 36) bp. All exon-intron junctions conformed to the consensus splice donor (GT) and acceptor (AG) sites, except for the splice donor sites of intron 7 (AT) and10 (GC). The introns sizes were between 74 (intron 38) and 20198 bp (intron 1). Introns were classified relative to codon interruption, as follows: phase 0 (no codon interruption), phase 1 (interruption between first and second base) and phase 2 (after second base). Exons were interrupted by introns of all phases. Most introns were of phase 0 (55%). A number of efforts where made to identify alternatively spliced products using PCR with primer pairs directed to high probability putative splice variant exons. However, no such variants could be found. Manually, we could identify one single polyadenylation signal (AATAAA) 1423 nucleotides downstream of the translation stop codon (Fig. 1 ). The exons formed an ORF of 5664 bp including the stop codon. Figure 1 Genomic organization of GP210 and a model of the deduced amino acid chain of GP210 . Exons (black boxes) and intron sizes are scaled individually. In silico predictions of a signal peptide, the transmembrane region and 12 putative N- linked glycosylation sites (N). Primary structure of GP210 The deduced amino acid sequence of human GP210 contains 1887 residues, predicting a molecular mass of 205 198 Da and a pI of 6,41 of the non-processed protein. Alignment to the corresponding mouse and rat sequences displayed a high degree of homology (91,8% similarity and 88,9% identity compared to the primary structure of the rat protein). One insertion, an alanine at position 1858, makes the GP210 one residue longer compared to rat and mouse GP210. A signal peptide cleavage consensus site could be defined between residues 25 and 26 using the SignalP algorithm (Fig. 1 ). This signal sequence shows no resemblance to previously reported GP210 sequences. Hydrophobicity values along the deduced amino acid chain identified several putative membrane spanning regions. One of these (residues 1809 to 1828) corresponded to a domain mapped in the rat orthologe [ 5 ], leaving 59 residues facing the nuclear pore. Motifs found scanning the sequence through the ExPASy-Prosite database included 12 potential N-glycosylation sites (outlined in Fig. 1 ), and numerous putative consensus sites for various kinase related phosphorylations. Two cAMP- and cGMP dependent protein kinase sites (residues 1089, 1874), two tyrosine kinase sites (residues 227, 922), 30 protein kinase C phosphorylation sites and 26 casein kinase II phosphorylation sites were found. The sites associated to PKC and CK2 phosphorylation were evenly distributed throughout the sequence. Blast homology searches revealed a vast number of EST clones containing GP210 sequence and an 871 amino acid partial sequence of a hypothetical human protein KIAA09906 (accession Xp051621) identical to the C-terminal end of the full length translated GP210 open reading frame. Identification of the transcription start The sequence 1 kb upstream of the ATG start codon possesses neither TATA or CAAT boxes, but contains scattered initiator (Inr) elements (consensus Py Py C A N T/A Py Py) [ 16 ]. In order to determine the transcription start site (tss), we therefore performed RNA ligase mediated rapid amplification of cDNA ends [ 17 ]. By Northern blotting using a 935 bp cDNA probe (nt 4286–5220), a single transcript of 7,3 kb was seen in two different tumor cell lines (Fig 2 ). The size of the mRNA corresponds to the sum of the open reading frame and 3'-UTR including a potential poly A tail. Since expression was much more abundant in HeLa cells than in Wilms tumor cells, we used HeLa cell total RNA as a template. The nested PCR (see material and methods) gave a major specific cDNA product of approximately 80 nucleotides (fig. 3 , lane 2). Sequence analysis of 10 independently ligated PCR products obtained using nested adapter- and gene specific primers (outlined in fig. 3 ) revealed two different amplified, GP210 specific fragments of similar length, indicating two alternative tss (Fig. 4 ). Out of ten clones sequenced, eight ended at position -29 and two at position -18 upstream ATG. In an identical analysis using poly-A+ RNA from a human fetal myoblast cell line G6, 7 clones ended at position -29 and 3 at position -18 upstream ATG. These findings argue for a major tss at -29. In addition, to confirm our findings we performed sequence homology searches in the human EST clone database at NCBI and elsewhere. The results revealed no reported cDNA sequences located upstream of the experimentally determined start site. Figure 2 Northern blot analysis of HeLa cell and WCCS-1 mRNA . A 935 bp GP210 specific cDNA probe was hybridized to 10 μg of HeLa and WCCS-1. Loading and RNA quality was controlled by a methyleneblue staining of 18 and 28s rRNA. Figure 3 Determination of 5 prime end of GP210 mRNA . Transcription start site were determined analyzing nested PCR products generated with gene (fig. 5) and adapter specific primers. Lane 1, Marker. Lane 2, Nested PCR product obtained using two primer pairs specific to sequences within the adapter and the first exon. Lane 3, Negative water control. Lane 4, Positive control using bacterial adapter ligated cDNA and specific primers. Lane 5, Marker. Figure 4 Promoter analysis human GP210 and homology to its mouse counterpart . A 500 bp sequence upstream of translation start site was analyzed for the presence of consensus transcription factor binding sites. Upper lane is the human sequence, and lower sequence is mouse. Homology in the (GT)n repeat between human and mouse genomic sequences and its position relative to the translation start codon. The translation start site is in bold and numbered +1. Homology to the mouse sequence is marked in grey, Outer gene specific primer (ogsp) and inner gene specific primer (igsp) used for transcription start definition using RLM-RACE are underlined with arrows. The transcription initiation sites are positioned with empty arrows and the start nucleotide is in bold. Putative transcription binding motifs are underlined. Some elements for different transcription factors overlap. Only sense strand binding sites were considered. Legend: Sp1, Simian-virus-40-protein 1; EGR2, Early growth response gene 2; WT1, Wilms' tumor zink finger protein 1; PuF, c-myc purine-binding transcription factor. Promoter sequence analysis Analysis of the sequence surrounding the translation start site with the GRAIL program predicted a 1236 bp long CpG island, with a GC content of 75.3% starting 434 bp upstream of ATG, covering the first exon and extending into the first intron. We used a variety of promoter and transcription factor binding site algorithms to analyse the region upstream the start site for GP210, including TESS and Matinspector. By selecting for perfect matching and human consensus motifs a restricted number of putative transcription factor binding sites were found. Using these criteria, 22 motifs recognized by 14 different factors were defined on the sense strand, evenly distributed within 1000 bp proximal to the translation start codon (See Table 2, additional file 2 ). Seven Sp1 binding motifs were identified [ 18 ], five of them clustered in a region spanning 315 bases, starting eight nucleotides upstream of the major tss (Fig. 4 ). Four putative binding motifs for EGR1/WT1 (consensus GXGXGGGXG) were mapped within the same promoter region, starting at positions -47, -70, -76 and -283. Two of these matched completely with this consensus sequence, whereas two contained one mismatch in the 9bp-binding motif (positions -71 and -284 respectively). In addition, we found a WT1 binding site in the antisense strand (pos. -112). Only a few other upstream regulatory elements were defined within and proximate to the CpG rich region. We found one binding motifs associated to Ets-2 [ 19 ], one to the c-myc purine-binding transcription factor PuF [ 20 ] and one to the early growth response gene 2 (table 2, Fig. 4 ). A sequence of 9433 bp (c047302867. Contig1) was found in the Mouse Genome Sequencing Consortium (MGSC) database. This sequence mapped to mouse chromosome 6 and contained the first exon, part of the first intron, and 3 kb of the upstream promoter region of mouse GP210. Similar to human, a GPC island containing 644 bp and with a GC content of 74% was found starting 268 nt upstream of the start codon and extending into the gene. Pairwise ClustalW alignment of this genomic clone showed 54% homology to its human counterpart within the first 500 bp upstream of the translation start (fig. 4 ). In the same region as in human we found putative EGR1/WT1 binding sequences. The sequence at positions -40 to -32 matched completely with the consensus sequence. The sequences starting at -47, -76, and -283 had one, two and one mismatches, respectively. As in the human sequence, an additional putative WT1 binding site was located in the antisense strand starting at position -83, but it contained a single nucleotide insertion. An Ets-2 motif, identical to the motif starting at -500 in the human sequence, was found starting at position -390 in the mouse sequence. In addition, a 40 bp (GT)n repetitive sequence was located about 1700 bp upstream of the ATG both in the human and mouse promoter (Fig. 5 ). This repetitive sequence is known to exist redundantly in the genome [ 21 ], and has been reported to be a binding element for WT1 [ 22 , 23 ]. Figure 5 Alignment of mouse and human sequences demonstrating a conserved region about 1700 bp upstream of ATG. To determine whether the putative WT1 binding sites in the GP210 promoter might correspond to functional regulation of GP210 by WT1, we used a model system for the examination of WT1 target genes in which WT1- isoform A, devoid of a 17 amino acid insert and a KTS insert in the zinc finger region, was expressed upon removal of tetracycline from the growth media. Figure 6 shows in a triplicate experiment that the induction of WT1 (Lanes 4–6) in these cells did not alter GP210 transcription. An actin control showed comparable loading and integrity of mRNA. These data suggest that GP210 is not a target of WT1. Figure 6 WT1 isoform A does not regulate expression of GP210 . SAOS cells conditionally expressing WT1 isoform A were grown in the presence (Lanes 1–3) or absence (lane 4–6) of tetracycline to induce expression of WT1. Triplicate plates of cells were harvested for total RNA and subjected to sequential northern blot analysis with the GP210 probe, WT1probe and actin control. Discussion The present study describes the genomic structure of the human nucleoporin GP210 gene, including its exon and intron sizes, intron/exon junctions and the 5' UTR sequence. Transcription start sites were determined experimentally by RNA ligase mediated rapid amplification of cDNA ends. Analysis of the promoter sequence identified a number of putative binding motifs for factors involved in tissue- or cell -type specific gene regulation. Strikingly, we could identify five putative Wilms tumor 1 (WT1) suppressor protein binding sites, four Sp1 biding sites, and one ETS binding site in a range of 315 bases just upstream of the translation initiation site. Some of these were conserved in the mouse promoter region. Mouse GP210 was initially identified as an early response gene to induction of embryonic kidney tubule development [ 10 , 15 ], suggesting that transcription factors regulating conversion of mesenchyme to kidney tubules are involved in its activation. Transcription factors implicated in early kidney tubule development include members of the myc family, Pax-2, hox a11 and Hoxd11, lmx-1b, HNF-1a, Pod-1, and WT1 [ 24 - 29 ]. Except for the WT1 binding site, putative binding motifs for these factors were lacking in the promoter region of the GP210 gene. Experimentally we found that WT1 does not influence GP210 expression in human osteosarcoma cells. It is thus more likely that Sp1 and some member of the ETS transcription factor family are the positive regulators of GP210. WT1 is a zinc finger transcription factor known to exist in different isoforms due to alternative pre-mRNA splicing. DNA binding specificity is determined by insertion or removal of three amino acids between zinc finger III and IV (referred to as WT1(+KTS) and WT1(-KTS)). The -KTS isoform have been reported to repress or activate target genes containing variations of an EGR1 related, GC-rich motif (consensus GXGXGGGXG) in their promoter [ 24 ]. Other biological activities have been suggested for the +KTS isoform [ 26 ]. Mutations in the WT1 gene has been shown in a small proportion of nephroblastomas, an embryonic kidney tumor, as well as in other tumor types, such as leukemia, mesothelioma and desmoplastic small round cell tumor. The restricted expression pattern in the mouse embryonic kidney and the failure of kidney development in WT1 null mice shows that WT1 is important for mesenchyme-to-epithelial transition, especially for early organogenesis of kidney and gonads [ 29 ]. It is thus of considerable importance to identify downstream target genes for WT1. This could include the gene for GP210, but presumably not other nucleoporins. In the only previously reported nucleoporin promoter region, of mouse nup358, several binding sites for Sp1 but none for WT1 were detected [ 25 , 30 ]. Sp1 is a transcription factor included in a small protein family (Sp1, Sp2, Sp3, and Sp4), whose members are binding to cis -elements widely distributed in different types of transcription control regions [ 25 ]. Although traditionally considered as an activator for house keeping genes, it has become increasingly clear that Sp1 can act as a cell specific regulator of gene expression. Differential expression levels of Sp1 during nephrogenesis [ 31 ] and hematopoietic development [ 32 ] have been reported. Along with specific post-translational modifications, the substantial differences in the expression patterns of Sp1 suggest that Sp1 can induce specific gene expression in embryonic tissues, including GP210 in the kidney. We also found a putative ETS-2 binding site in the mouse and human promoter for gp/GP210. Ets-2 is a widely distributed member of the ETS family of transcription factors characterized by a unique winged helix-turn-helix domain, which specifically interacts with DNA sequences containing the purine-rich core motif, GGAA/T. Since several ETS family members binds to the same core motif it has been difficult to determine specific target genes for each member, gene targeting in mice implicates ETS-2 as an activator of metalloproteinases in placenta (MMP-3, MMP-9 and MMP-13) and a regulator of hair development [ 33 ]. Elevated ETS-2 expression can reverse ras dependent transformation in cell lines [ 34 ]. In contrast, a high expression of ETS-2 is needed to maintain the transformed state of human prostate cells [ 35 ]. These data suggests multiple roles for ETS-2 during development and cancer. Interestingly, a binding site for Pea3, a member of the ETS family, has been noted in the WT1 promoter, and Pea3 was found to transactivate the Wt1 promoter [ 36 ]. Our promoter region analyses, which identify WT1, SP1 and Ets-2 as putative transcription factors regulating GP210 expression are descriptive. GP210 expression in the developing kidney resembles that of E-cadherin, which has been show to be a bone fide WT1 target gene [ 24 ]. WT1 A isoform, which lacks a 17 amino acid insert and the KTS insert in the zinc finger region, did not influence GP210 expression in SAOS osterosarcoma cells, in a system using tetracycline-induced repression of expression. In vivo, expression of WT1 appears very early during nephrogenesis, and is downregulated when GP210 expression increases [ 10 , 15 , 28 , 29 ]. Based on these findings, it was difficult to predict whether WT1 is involved in the positive or negative regulation of GP210. Our data do not exclude the possibility that demonstrate that the WT1 A isoform in different setting can regulate GP210 expression. It is also possible that other isoforms of WT1 regulate GP210. The amino acid sequence of human GP210 revealed potential sites for phosphorylation and glycosylation. The role for phosphorylation of nuclear envelope associated proteins is not well understood, but is presumed to have a function in mitotic events [ 37 , 38 ]. Non-membrane nucleoporins 153, 214 and 358 are phosphorylated throughout the cell cycle, but hyperphosphorylated during cell division. In contrast, GP210, was in the same study specifically phosphorylated during mitosis and one single consensus Ser 1880 -Pro 1881 motif could be detected as a target for cyklin-B-p34 cdc2 kinase and MAP kinase in vitro [ 39 ]. A comparison of the cytoplasmic domain in mouse, rat and human reveals that this serine-proline dipeptide located seven amino acids downstream of the carboxyl terminus is conserved. Whether the many putative phosphorylation sites in GP210 are actively regulated has to be experimentally determined. Restricted to the lumenal region of GP210, there are 12 potential putative acceptor residues for N-linked oligosaccharides. This is one residue less than in the rat homologue [ 5 ], but the remaining 11 seem to be located at conserved locations. The binding of GP210 to the lectin ConA suggests presence of high mannose-type oligosaccharides in mature GP210 [ 4 ], but there are no reports on functional aspects of this posttranslational modification. Materials and methods Wilms tumor and HeLa cell cultures WCCS-1 Wilms tumor cells [ 40 ] and HeLa cells were cultured in Dulbecco's modified medium containing 10% heat- inactivated Fetal Calf Serum (FCS) and in the presence of 1% HEPES. RNA from WCCS-1 and HeLa cells was isolated from 5 × 10 7 cells using the RNAeasy midi kit (Qiagen) following the recommendations of the manufacturer, including a DNAse treatment step to avoid chromosomal DNA contaminations. Northern blotting of total RNA from cells was performed as described [ 41 ]. To visualize 18 and 28s rRNA, a control RNA filter lane were immersed in 5% acetic acid followed by colorization in 0,5 MNaAc (pH 5,2), 0,4% methyleneblue for 15 min. cDNA probes used: a 935 bp PCR generated GP210 probe (nt 4286–5220), a 1.8 kb human β-Actin control probe (Clontech) and a 1023 bp PCR generated human WT1 probe covering the 3' end of the ORF and 523 bp of the 3' untranslated region thereby hybridizing to all known splice variants [ 42 ]. Probes were labelled with [α 32 P] dCTP by random priming using Megaprime DNA labeling system kit (Pharmacia). Filters were hybridized in 20 mM Na 2 HPO 4 (pH 7.2), 7% SDS at 65°C for 18 hours. After washing in 20 mM Na 2 HPO 4 (pH 7.2), 5% SDS at 65°C for 2 × 60 min followed by 2 × 60 min in 20 mM Na 2 HPO 4 (pH 7.2), 1% SDS at 65°C the filters were exposed to Hyperfilm-MP films (Amersham) for 5 days at -70°C in the presence of intensifying screens. Band intensities were quantified using a PhosphoImager 400S (Molecular Dynamics, Sunnyvale, CA). Tetracycline-regulated expression of WT1 in human osteosarcoma cells WT1-A SAOS cells were constructed from a cell line harboring the tetracycline-repressor-VP16 fusion protein [ 43 ], transfecting the parental cell with a construct harboring WT1 isoform A (-17amino acids, -KTS) linked 3' to the CMV minimal promoter and tetracycline operators [ 44 ]. Conditional expression of WT1-A was demonstrated by immunoblotting. WT1-A-SAOS cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 1∞ penicillin/streptomycin, 0.3 mg/mL L-glutamine and 0.5 mg/mL G418. All cells were cultured at 37°C in a 5% CO 2 atmosphere. For the induction, cells (at 70% confluence) were washed twice with PBS and refed with fresh media in the absence or presence of 1 μg/mL of Tetracycline. After 18 hours, the cells were washed twice with PBS and total RNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, CA) according to the manufacture's instructions. 4 μg of the RNA were resolved on formaldehyde-containing agarose gels and transferred to Nytran membranes (Schleicher and Schuell, Keene, NH). Hybridization was performed in ULTRAhyb buffer (Ambion, Austin, TX) at 42°C. Briefly, filters were prehybridized in ULTRAhyb buffer for 6 hours followed by an overnight hybridization at 42°C with the 935 bp hGP210cDNA probe. A WT1 cDNA probe (exons 5 to 10) and a cDNA probe for actin were used as controls. Membranes were exposed to BIOMAX MS films (Kodak, Rochester, NY) at -80°C in the presence of intensifying screens. Probes were stripped by boiling the membranes in 0.1% sodium dodecyl sulfate/ standard saline citrate solution for 10 minutes. Oligo-capping To determine transcription start, the RNA Ligase Mediated Rapid Amplification of cDNA Ends (RML-RACE) kit (Ambion) was used according to the instructions of the manufacturer's. Briefly 5 μg of Hela cell RNA or 2–5 μg of poly-A+ RNA from partially differentiated human G6 satellite cells [ 45 ] was treated with calf intestinal phosphatase (CIP) at 37°C for 60 min. RNA from G6 cells was kindly provided by Donald Gullberg at ICM, Uppsala University, Sweden. The mixture was phenol:chloroform (1:1) extracted followed by ethanol precipitation. The RNA was subsequently incubated with Tobacco Acid Phosphatase (TAP) at 37°C for 60 min. A 45 nt adapter RNA oligonucleotides (5'GCUGAUGGCGAUGAAUGAACACUGCGUUUGCUGGCUUUGAUGAAA3') was ligated to the CIP/TAP treated 5' RNA end using T4 ligase. cDNA was generated using random decamers and MMLV reverse transcriptase at 42°C for one hour. The nested PCR were performed using the advantaq 2 polymerase system (Clontech) and the following primers: in the first PCR reaction the outer adapter (5'GCTGATGGCGATGAATGAACACTG3') was combined with the gene specific outer primer (5'CAGCAGCACTTTGGGGATGTTGAG3'), in the second PCR the inner adapter primer (5'CGCGGATCCGAACACTGCGTTTGCTGGCTTTGATG-3') was used in combination with the gene specific inner primer (5'CCCGCCGCCAACAGCACCGACAGC3'). The conditions for both PCR reactions were as follows: 94°C for one min (hot start) followed by 95°C for 20 s, 68°C for 60 s, repeated 35 cycles. After the final cycle, the reactions were extended for an additional 5 min at 68°C. PCR products were analysed on a 2% agarose gel, ligated into the pCRII vector (Invitrogen) and sequenced using M13 primers. Sequencing Sequencing was performed using ABI PRISM and the Dye Terminator cycle sequencing kit according to the manufacturer's directions (Perkin Elmer) and analysed by an automated ABI-310 fluorescent-dy sequencer (Applied Biosystems). Bioinformatics and sequence analyses Exon-intron boundary predictions were done manually and using the Genscan web server at MIT . Open reading frame finding and all sequence analyses were done using the MacVector 6.5.3 sequence analysing software (Oxford Molecular Group). The BLAST (Basic Local Alignment Search Tool) server of the National Center of Biotechnology Information (NCBI) and The Mouse Genome Sequencing Consortium (MGSC) database was used for sequence alignments. Signal peptide prediction was done using the SignalP v1.1 at WWW Prediction Server (Center for Biological Sequence Analysis, and membrane topology predictions at the HMMTOP server using the TopPred2 program [ 46 ]. Transcription factor mapping in the 5' untranslated region of hGP210 was analysed in TESS and Matinspector [ 47 ] search programs. Additional amino acid sequence patterns and domains were analysed using the PROSITE database . Prediction of GPC islands and detection of repeats in sequences analysed were done with the GRAIL program [ 48 ]. Supplementary Material Additional File 1 Table 1. Organization of the human GP210 gene including exon/intron sizes, splice acceptor consensus sequences and intron phases. Click here for file Additional File 2 Table 2. Predicted cis -acting elements of the human GP210 promoter. The start sites of the elements are indicated as 5' end of consensus sequence and relative to translation start as +1. The compilation was made using the TESS and MatInspector analyze programs. Results are restricted to human species and perfect match except for four WT1 binding sites indicated as a boxed nucleotide. Some abbreviations: EGR2, Early growth response gene 2; WT1, Wilms' tumor zinc finger protein 1; PuF, c-myc purine-binding transcription factor; IL-6. RE-BP, IL-6 Response element-Binding protein; NF-1, Nuclear factor 1; Ets-2, proto-oncoprotein; USF2, upstream stimulating factor. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544869.xml
545593	A comprehensive transcript index of the human genome generated using microarrays and computational approaches	A combination of microarray data with extensive genome annotations resulted in a set of 28,456 experimentally supported transcripts, providing the first experiment-driven annotation of the human genome.	Background The completion of the sequencing of the human, mouse and other genomes has enabled efforts to extensively annotate these genomes using a combination of computational and experimental approaches. Generating a comprehensive list of transcripts coupled with basic information on where the different transcripts are expressed is an important first step towards annotating a genome once it has been fully sequenced. The task of identifying the transcribed regions of a sequenced genome is complicated by the fact that transcripts are composed of multiple short exons that are distributed over much larger regions of genomic DNA. This challenge is underscored by the widely divergent predictions of the number of genes in the human genome. For example, direct clustering of human expressed sequence tag (EST) sequences has predicted as many as 120,000 genes [ 1 ], whereas sampling and sequence-similarity-based methods have predicted far lower numbers, ranging from 28,000 to 35,000 genes [ 2 - 5 ], and a hybrid approach has suggested an intermediate number [ 6 ]. Furthermore, the availability of a completed draft sequence of the human genome has yielded neither a proven method for gene identification nor a definitive count of human genes. Two initial analyses of the human genome sequence that used strikingly different methods both suggested the human genome contains 30,000 to 40,000 genes [ 2 , 3 ]. However, a direct comparison of the predicted genes revealed agreement in the identification of well-characterized genes but little overlap of the novel predictions. Specifically, 84% of the RefSeq transcripts agreed with fewer than 20% of the predicted transcripts matching between the two analyses. This result suggests that, individually, these datasets are incomplete and that the human genome potentially contains substantially more unidentified genes [ 7 ]. Several recent studies have highlighted the limitations of relying solely on computational approaches to identify genes in the draft of the human genome [ 8 - 13 ]. Furthermore, substantial experimental data from direct assays of gene expression provide evidence for many genes that would not have been recognized in the analyses just mentioned. Saha and colleagues used a new LongSAGE technology to provide strong evidence that there are thousands of genes left to be discovered in the human genome [ 9 ]. Specifically, they sequenced over 27,000 tags from a human colorectal cell line that collapsed down to 5,641 unique groups. Interestingly, only 61% (3,419) of the tags matched known or predicted genes, whereas 10% (575) matched novel internal exons and 14% (803) appear to represent completely novel genes [ 9 ]. They extrapolate from these data to predict as many as 7,500 exons from previously unrecognized genes. A recent analysis by Camargo et al . [ 8 ] also indicates that we are far from defining a complete catalog of human genes based on the analysis of 700,000 ORESTES (Open Reading Frame ESTs) that were recently released into GenBank. Finally, Kapranov and colleagues recently constructed genome-tiling arrays for human chromosomes 21 and 22 to comprehensively query transcription activity over 11 human tissues and cell lines [ 10 ]. They detected significant, widespread expression activity over a substantial proportion of these chromosomes outside of all known and predicted gene regions. Most current methods in widespread use for identifying novel genes in genomic sequence depend on sequence similarity to expressed sequence and protein data. For example, ab initio prediction programs operate by recognizing coding potential in stretches of genomic sequence, where the recognition capability of these programs depends on a training set of known coding regions [ 14 ]. Therefore, genes identified by ab initio prediction programs or assembled from EST data are also inaccurate or incomplete much of the time [ 10 - 12 ]. While ab initio prediction programs perform well at identifying known genes, predictions that do not use existing expressed sequence and protein data often miss exons, incorrectly identify exon boundaries, and fail to accurately detect the 3' and 5' untranslated regions UTRs [ 14 ]. Similarly, EST data may be biased towards the 3' or 5' UTR [ 13 ]. These deficiencies are addressed in full-length gene cloning strategies [ 13 ], but cloning is still a laborious process which could be accelerated if we were able to start from a more accurate view of a putative gene [ 13 ]. Recently, several groups have used microarrays to test computational gene predictions experimentally and to tile across genomic sequence to discover the transcribed regions in the human and other genomes [ 10 - 12 , 15 - 17 ]. These array-based approaches detected widespread transcriptional activity outside of the annotated gene regions in the human, Arabidopsis thaliana and Escherichia coli genomes. The recent sequencing and analysis of the mouse genome indicates extensive homology between intergenic regions of the human and mouse genomes, further highlighting the potential for other classes of transcribed regions [ 18 ]. Interestingly, recent tiling data suggests that many of these conserved intergenic regions are transcribed [ 15 , 16 ]. In the study reported here, we describe hybridization results generated from two large microarray-based gene-expression experiments involving predicted transcript arrays spanning the entire human genome and a comprehensive set of genomic tiling arrays for human chromosomes 20 and 22. mRNA samples collected from a diversity of conditions were amplified using a strand-specific labeling protocol that was optimized to generate full-length copies of the transcripts. Analyses of the resulting hybridization data from both sets of arrays revealed widespread transcriptional activity in both known or high-confidence predicted genes, as well as regions outside current annotations. The results from this analysis are summarized with respect to published genes on chromosomes 20 and 22 in addition to our own extensive set of genome alignments and gene predictions. Combining computational and experimental approaches has allowed us to generate a comprehensive transcript index for the human genome, which has been a valuable resource for guiding our array design and full-length cloning efforts. In addition, the expression data from the 60 conditions provides a comprehensive atlas of human gene expression over a unique set of gene predictions [ 19 ]. Results Generating a comprehensive transcript index of the human genome Figure 1 illustrates the process we used to generate a comprehensive transcript index (CTI) for the human genome that represents just over 28,000 known and predicted transcripts with some level of experimental validation. The first step in this process was to generate a 'primary transcript index' (PTI) by mapping a comprehensive set of computationally and experimentally derived annotations onto the genomic sequence. The computational predictions include the output of gene-finding algorithms and protein similarities, while the experimentally derived alignments are based on ESTs, serial analysis of gene expression (SAGE), and full-length cDNAs. The resulting list of transcripts in the PTI can be loosely ranked or classified into different categories, ranging from high confidence to low confidence, on the basis of the level of underlying experimental support. The advantages of a PTI are that the computations can be performed on a genome-wide scale and it incorporates the massive amounts of publicly available EST, SAGE and cDNA sequence data. However, the resulting transcript index has two significant limitations. First, the ab initio gene-finding algorithms tend to have a high false-positive rate when applied at a low-stringency setting to cast as broad a discovery net as possible. Second, gene-finding algorithms are trained on known protein-coding genes, which may limit their ability to detect truly novel classes of transcribed sequences. The second step towards the CTI is the use of two different types of microarrays to address these limitations (Figure 1 ). First, predicted transcript arrays (PTA) were used to determine experimentally which of the lower-confidence predictions in the PTI were likely to represent real transcripts. Second, genomic tiling arrays were used to survey transcriptional activity in a completely unbiased and comprehensive fashion. As shown in Figure 1 , the CTI plays a central part in the subsequent design of screening arrays. These are used to monitor RNA levels for all the transcripts across a large number of diverse conditions to begin the process of assigning biological functions to novel genes based on co-regulation with known genes [ 20 ]. The CTI is also used to design exon/junction arrays that can be used to discover and monitor alternative splicing across different tissues and stages of development [ 21 ]. Generating a PTI To generate the PTI, three distinct computational analysis steps were executed in parallel: predictions based on similarity to expressed sequences from human and mouse; predictions based on similarity to all known proteins; and ab initio gene predictions. The process resulted in mapping 91% of the well characterized genes found in the RefSeq database [ 22 ], a percentage consistent with initial genome annotation results [ 2 , 3 ]. The mapping results were generated by collapsing overlapping gene models and regions of similarity to define locus projections, which comprise the distinct transcribed regions making up our PTI. While the reliance on gene predictions and protein alignments biases the PTI towards protein-coding genes, the alignment of all expressed sequences should represent many of the non-coding genes reported to date. A comprehensive index of non-coding genes would require tiling arrays, as described later. All locus projections were classified into one of eight categories on the basis of the level of underlying evidence from expressed sequence similarity, protein similarity and ab initio predictions. The categories, in decreasing order of support, are as follows: (1) known genes, taken as the set of 11,214 human genes represented in the RefSeq database when the arrays were designed; (2) ab initio gene models with expressed sequence and protein support; (3) ab initio gene models with expressed sequence support; (4) ab initio gene models with protein support; (5) alignments of expressed sequence and protein data; (6) alignments of expressed sequence data, requiring at least two overlapping expressed sequences; (7) ab initio gene models with no expressed sequence or protein support; and (8) alignments of protein data. Because of the limitations discussed in the previous section, we considered predictions with a single line of evidence (categories 6-8) as low confidence. Table 1 provides summaries resulting from a comparison between our PTI and the published Sanger Institute data for chromosomes 20 and 22 [ 23 , 24 ]. Our locus projections overlap 1,177 of 1,297 (91%) Sanger genes on chromosome 20 and 854 of 936 (91%) Sanger genes on chromosome 22, and our predicted exons overlap 7,306 of 7,556 (97%) and 4,819 of 5,014 (96%) total Sanger chromosome 20 and 22 exons, respectively. This comparison highlights the fact that our annotations result in the detection of both genes and exons in genomic sequence with high sensitivity. Predicted transcript arrays We previously described a high-throughput, experimental procedure to validate predicted exons and assemble exons into genes by using co-regulated expression over a diversity of conditions [ 11 ]. Here we employ a similar strategy over the entire genome by hybridizing RNA from 60 diverse tissue and cell-line samples to a set of arrays designed from the PTI. For a complete list of the transcripts represented on the predicted transcript arrays and 60 tissues and cell lines hybridized to these arrays (see Additional data files 1 and 2). We designed two probes per exon, where possible, for exons containing the highest-scoring probes as described in the methods from each transcript in our PTI set (on average, a total of four probes per transcript). This was done to balance the poor specificity of ab initio gene-finding algorithms [ 14 , 25 , 26 ] against the significant microarray costs associated with large-scale gene-expression experiments. The resulting hybridization data provides experimental validation of those low-confidence predicted genes that are either unsupported or minimally supported by existing EST data, thereby providing a means of determining which transcripts are included in the CTI. Summary of predicted transcript validation on chromosomes 20 and 22 We used an enhanced version of a previously described gene-detection algorithm to analyze the predicted transcript array dataset [ 11 ]. Basically, the hybridization data from probes each transcript from the PTI were examined to identify those transcripts with probes that appear to be more highly correlated over the 60 diverse conditions. Transcripts with probes that behaved similarly over the different conditions tested were considered to be expression-validated genes (EVGs). Unlike our original algorithm that used Pearson correlations to group similarly behaving probes, our enhanced algorithm incorporated a probe-specific model to assess the most likely set of probes making up a transcriptional unit [ 27 ] (see Materials and methods for details). We used the extensive publicly available annotations on chromosomes 20 and 22 to assess the sensitivity and specificity of our array-based detection procedure. The sensitivity of our procedure was assessed by computing the EVG detection rate for those Sanger genes that overlap predictions (locus projections) represented in our PTI (Table 2 ). The average detection rate for our locus projections on chromosomes 20 and 22 is approximately 70% for those overlapping Sanger genes and just over 80% for those locus projections derived from RefSeq alignments (locus category = known) that represent Sanger genes. A true positive in this instance was defined as an expression-verified gene containing at least two probes, where at least one of the probes was contained within the exon of a Sanger or RefSeq gene. This 20% false-negative rate is the result of a complex mixture of issues, including limitations in our EVG-detection algorithm, limitations in the probe design step, lack of expression in the conditions profiled, and/or alternative splicing events. While the EVG-detection algorithm provides an efficient method to assemble probes into transcript units, the detection capabilities of this model could be expected to improve as the number of samples and the number of probes targeting any given transcript increases. The use of four probes per predicted transcript was determined to be sufficient for detection of most transcripts, as supported by the overall detection rate of known genes, although in many cases the probe design step was limited by our ability to find four high-quality probes per transcript. For many transcripts, there were not four nonoverlapping probes predicted to have good hybridization characteristics for the microarray experiment carried out here. The 60 samples were chosen to represent a broad array of tissue types, as an exhaustive list of human tissues is impossible to obtain. Because no replicate tissues/cell lines were run for any of the 60 chosen samples, we relied on the replication inherent in monitoring the same transcripts over 60 different conditions. In this case, genes expressed in multiple samples provide the replication necessary to increase our confidence in the detections. However, there are clear limitations in not replicating tissues/cell lines, as genes may be expressed in only a single condition or may be switched on only under certain physiological conditions or only during a certain stages of development. In such cases, we would have reduced power to detect these genes. Genes in the lower-confidence categories of our PTI annotations, which are not typically considered genes by Sanger, were detected at a significantly reduced rate. Interestingly, of the 337 (188 +149) higher-confidence transcripts on chromosomes 20 and 22 that did not intersect with Sanger genes, 47 (or 14%) were detected as EVGs (Table 2 ). These transcripts represent potential novel transcripts on these two highly characterized chromosomes. However, before we can make claims to the discovery potential for this method over the entire genome, we need to assess the false-positive detection rates. To this end, we defined as false positives all detections made in regions with support by only a single gene model that fell outside Sanger-annotated genes on chromosomes 20 and 22. Applying this definition over all transcripts in our PTI leads to a false-positive rate of 3% (11 out of 406). Because we cannot exclude the possibility that some of the transcripts supported by a single gene model represent real genes, we consider this false-detection rate as an upper bound on the actual false-positive rate. Accepting that the Sanger annotations represent the gold standard for chromosome 22, we detected 70% of all Sanger-annotated genes, while only 4% of the chromosome 22 locus projections that did not intersect Sanger genes were detected by our procedure, highlighting the sensitivity and specificity of this approach. In addition, the enrichment for EVG detections in Sanger genes versus the non-Sanger PTI on chromosomes 20 and 22 was extremely significant with a p -value effectively equal to 0 when using the chi-square test for independence ( χ 2 = 3,093, with 1 degree of freedom (df)). Summarizing EVG data over the entire genome and assessing the discovery potential. The last column of Table 2 provides the number of expression verified genes detected over the entire genome for locus projections in our PTI. This represents the most comprehensive direct experimental screening of ab initio gene predictions ever undertaken. We can use the false-positive and negative rates derived above to assess the discovery potential on that part of the genome that has not been as extensively characterized as chromosomes 20 and 22. First, we note that our detection rates over the genome were similar to that given for chromosomes 20 and 22. That is, 75% of the category 1 genes (RefSeq genes) were detected over the entire genome, compared to 80% for chromosomes 20 and 22. In total, 15,642 genes in the PTI were experimentally validated using this array-based approach. Assuming the false-positive rate of 3% defined above and a conservative false-negative rate of 30%, defined as the percentage of Sanger genes we failed to detect on chromosomes 20 and 22, these data suggest there are close to 21,675 potential coding genes represented in our PTI set. Because our PTI misses close to 10% of the Sanger genes, we corrected this number for those genes not represented in this set and provide an estimate of the total number of protein-coding genes in the human genome supported by our data to be approximately 25,000. This number is consistent with estimates given in the current release (22.34d.1) of the Ensembl database [ 28 , 29 ]. However, we caution that the estimate provided is based solely on the data described here, and that orthogonal sources of data [ 30 ] continue to suggest that the actual number of genes will be known only after the transcriptome has been completely characterized. From Table 2 we note that 2,093 (1,428 + 555 + 110) of the transcripts that were detected as EVGs had only one line of evidence (EST alignment, protein alignment or ab initio prediction). These 2,093 transcripts represent a rich source of potential discoveries in our PTI. To assess the potential biological functions of this novel gene set, we annotated translations of this set by searching the domains represented in the Protein Families database (Pfam) [ 31 ]. The search results were used to assign each of the translations to Gene Ontology (GO) [ 32 ] codes as described in the methods. Figure 2 graphically depicts the breakdown of the most common GO codes for two of the three major GO categories. These data suggest there may still be a significant number of protein-coding genes with important biological functions, given that domains/motifs represented in these predicted genes are similar to those found in known genes. The 339 predictions that were validated as EVGs and that had protein domains of biological interest would be natural candidates for full-length cloning, over the 24,532 (7,170 + 16,822 + 540 from Table 2 ) other lower-confidence predictions in our set. EVG data as an expression index Because multiple probes in each of the approximate 50,000 predicted genes in the human genome have been monitored over 60 different tissues and cell lines, the EVG data represent a significant atlas of human gene expression that is now publicly available [ 19 ]. For each transcript, the intensity information from the corresponding probes was optimally combined as described by Johnson et al. [ 21 ] to provide a quantitative measure of the relative abundance across the panel of 60 conditions, as shown in Figure 3 . Tiling arrays for chromosomes 20 and 22 To complement the use of PTI arrays, we constructed a set of genome tiling arrays comprised of 60 mer oligonucleotide probes tiled in 30 base-pair steps through both strands of human chromosomes 20 and 22. Repetitive sequences identified by RepeatMasker were ignored for probe design. These genome tiling arrays allow for an unbiased view of the transcriptional activity outside of known and predicted genes on these two chromosomes. mRNA from six (chromosome 20) or eight (chromosome 22) conditions was amplified and hybridized to the tiling arrays (see [ 19 ] and Additional data files 3 and 4). As with the PTI arrays, the amplification protocol generated strand-specific cDNA copies of the transcripts, which were full-length. Using a two-step procedure, the resulting data were analyzed to detect sequences expressed in at least one condition [ 33 ]. First, we examined probe behavior over conditions in overlapping windows of size 15,000 bp to identify windows that probably contained transcribed sequences, using a robust principal component analysis (PCA) method [ 33 ]. Second, for regions identified as likely to contain transcribed sequences, we attempted to discriminate between probes corresponding to expressed sequences (expressed 'exons') and probes corresponding to untranscribed sequences ('introns' or intergenic sequence) using a clustering procedure on variables derived from the PCA procedure [ 33 ]. All analysis results derived from this procedure were interpreted in the light of the Sanger annotations and our custom PTI set described above. Figure 4 provides two representative examples of tiling data for two known Sanger genes, KDELR3 and EWRS1 . In the first case (Figure 4a ), the tiling data almost perfectly correspond to the RefSeq annotation of KDELR3 , with just two potential false positives out of the 178 intron probes. The KDELR3 gene is annotated as having two alternative transcripts in the RefSeq database, given by the RefSeq accession numbers NM_006855 and NM_016657. The NCBI Acembly alternative splicing predictions further suggest the presence of additional isoforms of this gene (see Figure 4 ). One of the alternative forms, KDELR3 .e, depicted in Figure 4a , includes a novel 5' exon. The presence of this exon is supported by the EST with GenBank accession number BM921831. The tiling data for the KDELR3 gene in two conditions clearly show expression of NM_006855 but not NM_016657, thereby reliably detecting distinct splice forms. Further, there is a significant signal 5' to exon 2 in both transcripts that seems to suggest a novel exon, as opposed to a true false positive. This putative exon exactly matches the location of the first exon given in the Acembly prediction track noted in Figure 4a ( KDELR3 .e). Figure 4b shows the tiling data for the EWSR1 gene. In contrast to the first example, this gene has intense transcriptional activity outside of the annotated exons. Specifically, the EWSR1 gene has 43 potentially false-positive calls out of 203 intron probes. However, the EST data and alternative splicing predictions strongly suggest that these probes represent biologically relevant transcriptional activity. As with the KDELR3 gene, EWRS1 is annotated by RefSeq as having two transcripts: NM_005243 and NM_013986. The Acembly predictions identify four additional alternative splice forms; most noteworthy among these are EWSR1 .b and EWSR .g, shown in Figure 4b . These predictions indicate that alternative transcripts may exist for the EWSR1 gene that essentially divide the largest transcript into two transcripts, suggesting that multiple promoter and transcription-stop signals are present in this gene. The tiling data depicted in Figure 4b shows that all exons from both RefSeq splice forms were detected. In addition, there is a region to the right of probe position 400 in Figure 4b that indicates significant transcription activity but where there are no RefSeq exons annotated. However, the green bars indicate exons that are supported by EST data as well as the EWSR .b and EWSR .g predicted alternative splice forms, providing experimental support that these predictions represent actual isoforms of this gene. In fact, these data may provide a more accurate representation of the putative structure of this gene, as they support multiple alternatively spliced transcripts in this gene, beyond what has already been annotated in the RefSeq database. In all, 5% of the probes detected as expressed in intronic sequence mapped to predicted alternative splice forms. Given the extent of alternative splicing that is yet to be characterized [ 21 ], we believe a significant proportion of the 'intron' transcriptional activity in our data may represent alternative splicing. Summarizing the tiling results Our genome tiling arrays consisted of 2,119,794 and 1,201,632 probes for chromosomes 20 and 22, respectively. Of these, 1,615,034 probes fell into Sanger gene regions, with 239,542 probes actually overlapping Sanger exons. Under stringent criteria 64,241 probes were detected as expressed, with 34,245 of these falling within Sanger exons, 18,551 falling within Sanger introns, and 15,835 probes falling completely outside all Sanger annotations. This widespread transcriptional activity outside annotated regions of the human genome is consistent with other reports from multiple species [ 10 , 12 , 15 , 16 ]. Overall, at least one exon in each of 876 Sanger genes was detected as expressed out of 1,703 total genes covered by probes (excluding annotated pseudogenes), leading to an overall gene detection rate of 52%. The bias of probes identified as exon probes that actually fall in exons is striking, given that exons comprise roughly 2% of the genomic sequence (the p -value for this enrichment using the Fisher exact test is less than 10 -15 ). To estimate the upper bound of false-positive calls, we counted as false-positive events each probe identified as expressed by the detection process, but falling within an annotated intron of the RefSeq genes we detected as expressed. This resulted in an estimated false-positive rate of 1.3%. As indicated in Figure 4 , a percentage of these false-positive calls will be due to unannotated isoforms of genes. Others still will be due to cross-hybridization of the intron probes to genes in other parts of the genome. We consider cross-hybridization as made up of two components: specific cross-hybridization resulting from transcripts with similar, usually homologous, sequences; and nonspecific cross-hybridization resulting from the base composition of the probe sequence (J.C. and G.C., unpublished work). Of the intron probes detected as expressed, 23% had sequence similarities to known transcripts considered to render them susceptible to specific cross-hybridization, and 17% contained sequence features associated with nonspecific cross-hybridization. Accounting for probes that were positive for both specific and nonspecific cross-hybridization, we are left with 55% of the probes detected as expressed in the introns of Sanger genes that cannot easily be explained as alternative splicing or cross-hybridization. These data support recent observations that significant levels of transcription exist within the introns of known genes [ 15 , 16 ]. For those probes falling outside all Sanger genes, we again made use of our custom genome annotations to help interpret the extent of transcriptional activity in these regions. Table 3 summarizes the detections made for each of the categories described above. Filtering probes using the same cross-hybridization predictors described above suggests that 65% of those probes falling outside all annotations are not likely to be the result of cross-hybridization. Furthermore, for those detections that overlap low-confidence locus projections in our PTI, we used the classification procedure discussed above to assign GO codes to these transcripts. Only seven of the 297 transcribed regions detected outside of all Sanger genes via the tiling results (see Table 3 ) contained GO protein domains. This suggests that a large fraction of the transcriptional activity detected using tiling arrays is non-coding and of unknown biological function [ 15 , 34 ]. Tiling data help classify conserved sequences between species One further advantage of the tiling data is that they can be used to discriminate between transcribed and non-transcribed sequences conserved between human and mouse, or between any other pair of species. Figure 4c highlights tiling data under one condition for the beta-actin gene, a gene that is constitutively expressed in all tissues and often serves as a positive control in mRNA and protein expression experiments. The genomic region containing the complete beta-actin mRNA and 10 kilobases (kb) of genomic sequence upstream of the transcription start, was obtained from the mouse and human genomes, aligned using the AVID program [ 35 ] and then fed into the rVista program [ 36 ]. From this, we identified the conserved regions in this gene between mouse and human, including several relevant transcription factor binding domains that are key to the transcriptional regulation of this gene [ 37 - 39 ]. As can be seen directly from the figure, the exons are all detected as highly expressed, but none of the conserved transcription factor regions shows activity. This combination of expressed sequence in close proximity to conserved regions that are not expressed (as determined by the tiling data), offers another powerful advantage of the tiling data in discriminating among the possible roles of conserved sequences. Discussion A complete understanding of the human genome will only come after all genes have been identified and the functions of those genes have been determined. There has been much recent progress in defining the human transcriptome with ab initio methods, sequencing of EST libraries, full-length gene cloning projects, and comparative analyses between fully sequenced genomes of different species. However, we are still a long way from having a comprehensive set of annotations for the human and other genomes. There is need for new high-throughput experimental approaches to accelerate the process of annotating sequenced genomes in a comprehensive and accurate fashion. Toward this goal, we have used two microarray-based experimental approaches to provide evidence of widespread transcription activity outside of any known or predicted genes in the human genome. We have also provided experimental support for many ab initio predicted genes that have no other or minimal experimental sequence support, suggesting a small but significant class of genes that have evaded all other forms of experimental detection. Similar identifications have been made recently in the first extensive comparative analysis between mouse and human genomes [ 18 ]. Despite the extent of novel discovery, our data suggest there are only 25,000-30,000 protein-coding genes in the human genome, with perhaps an equal number of non-coding transcripts that may serve important regulatory roles [ 34 , 40 ]. Finally, our data indicate widespread alternative splicing across known genes, providing a glimpse into the extent of transcript complexity that may exist in mammalian genomes. We have used the expression data for the approximate 50,000 predicted transcripts hybridized to 60 diverse conditions in combination with genomic tiling data to generate a CTI containing 28,456 experimentally supported transcripts. The transcripts represented in the CTI include all computational predictions with two or more lines of evidence from our PTI (independent of microarray validation), in addition to the overlapping set of 15,642 transcripts detected as EVGs. This resulting comprehensive list of known and predicted transcripts provides the starting point for large-scale systematic studies to determine the biological function of genes in both normal and disease states. The primary goal of the CTI is to allow researchers to focus experimental efforts on a comprehensive set of genes that are likely to be real. It is of note that between the time the predicted transcript arrays were designed and annotated using the custom genome annotations described above, and the time this work was published, more than 6,000 genes were added to the RefSeq collection. These newer RefSeq genes were represented by 5,100 locus projections in our original PTI that were not classified in the RefSeq category. Interestingly, 4,212 were detected as EVGs in the present analysis and had already been included in our CTI, a validation rate slightly greater than 82%. Only 19% of the non-RefSeq genes in our PTI had been detected as EVGs (see Table 2 ), yet more than 82% of the new RefSeq genes coming from this set were detected as EVGs. This result speaks to the utility of the microarray-based approach to gene validation described here (see Additional data file 5 for a complete breakdown of validation rates by category). While using microarrays to test computational gene predictions experimentally has the advantage of being economically feasible across the whole genome, the tiling data represent a more comprehensive and unbiased view of transcription. Our data indicate widespread transcriptional activity in the introns of annotated genes and in intergenic regions, where a significant proportion of this activity can be explained by nonspecific and specific cross-hybridization. The transcriptional activity noted for our low-onfidence transcripts in the PTI indicates that 25% of the activity we observe may be coding for proteins that are at least somewhat related to existing protein data. Much of the transcription activity we have noted in the introns of genes may also represent uncharacterized alternative splicing, and potentially novel genes, in addition to specific and nonspecific cross-hybridization. Conclusions At present, predicted transcript arrays allow for the discovery of most protein-coding genes genome wide when many different conditions are considered. Until the discovery and characterization of these protein-coding genes is completed, this method will continue to be a cost-effective solution to drive such discovery. In contrast, genomic tiling represents a completely unbiased method for monitoring transcriptional activity in genomes, but due to cost will probably be limited to screening a smaller number of conditions. However, as novel transcription regions are identified from the tiling data, these regions can be represented on predicted transcript arrays that are hybridized over many more conditions, as described in Figure 1 . As the microarray technologies have evolved, tiling the entire human genome is now possible, with such efforts presently being supported by the ENCODE (Encyclopedia of DNA Elements) project of the National Human Genome Research Institute (NHGRI) [ 41 ]. We believe the steps taken here are necessary for querying all potential transcription activity in the genome, for the purpose of identifying novel genes, more completely characterizing existing genes, and identifying a more comprehensive set of probes for these genes that can be used to monitor transcription abundances in more standard gene expression studies. Not all uses of microarrays demand an exhaustive representation of probes to all genes in the genome under study. However, experiments that seek to identify key drivers of pathways [ 42 ] or that seek to discriminate between alternative splice forms of genes within a given tissue [ 21 ] require a more comprehensive set of arrays to ensure success. These data provide an essential first step to generating a comprehensive set of arrays that are based on experimental support combined with computational annotation, instead of relying solely on the latter. These comprehensive arrays will be invaluable as we seek to better understand mechanisms of action for existing and novel drug targets and elucidate pathways underlying complex diseases. In addition, further study of the extensive noncoding RNA identified via the methods described here and elsewhere [ 10 , 12 , 15 , 16 ] is likely to open new fields of biology as the functional roles for these entities are determined. Materials and methods Data preparation The NCBI 8/2001 assembly of the human genome was the input data for this analysis. The 4/21/1999 release of RepeatMasker [ 43 ] was used to mask for human repeats. An internal database of RNA genes and bacterial and vector sequences was aligned to the genome with BLASTN. Genomic sequences with 95% or higher identity over at least 50 bases were masked. No probes were designed from masked regions. Gene index production To predict genes on the basis of expressed sequence similarity, we first clustered and aligned all expressed human and mouse sequences in GenBank to create a human gene index (HGI) and a mouse gene index (MGI). Clustering and alignment were performed with the DoubleTwist Clustering and Alignment Tools (CAT) [ 44 ]. Input data included all mouse and human RefSeq mRNA sequences, and EST and mRNA sequences from GenBank release 124, masked as described above for repeats and contaminating sequences. For each species, the CAT software first clustered sequences and then defined subclusters on the basis of a multiple sequence alignment. The subclusters represent candidate alternatively spliced gene transcripts. The 644,168 human and 291,656 mouse sequences that were singleton ESTs or completely masked were excluded from the HGI and MGI. Expressed sequence mapping Human and mouse UniGene and RefSeq, MGI, and HGI sequences were aligned with the genome first by BLASTN 2.2.1 [ 45 ], followed by refinement of intron/exon boundaries by the sim4 algorithm (12/17/2000 release) [ 46 ]. Only the representative sequences (Hs.seq.uniq) for each UniGene cluster designated in the 3 August 2001/build 138 version of the UniGene database were used in this analysis. We only refined BLAST alignments with an E-value of less than 10 -20 for human sequences and 10 -8 for mouse sequences. For human UniGene and HGI, we refined only those BLAST hits where the target sequence showed greater than or equal to 92% identity to the genomic sequence over 75 bp. For human RefSeq, we refined hits with greater than or equal to 95% identity, and for MGI, RefSeq, and UniGene, we refined hits with greater than or equal to 80% identity. These thresholds were empirically determined to provide good sensitivity in aligning most sequences to the genome while limiting multiple alignments past those expected from paralogs present in the human genome. In all cases percent identity was measured over 75 bp. Individual sim4 exons of questionable confidence were then removed on the basis of percent identity and length thresholds. All sequence databases were downloaded from GenBank August, 2001. Protein sequence mapping The GenBank nonredundant protein database (downloaded 25 August 2001) was aligned to the genomic sequence with BLASTX 2.2.1 [ 45 ] using an E-score threshold of 10 -5 . Adjacent protein alignments from a single protein were grouped together as a prediction whenever the protein sequence coordinates of the alignments were consistent in direction and did not significantly overlap. Ab initio gene prediction GrailEXP 4.0 [ 47 ], GENSCAN 1.0 [ 48 ], FGENESH [ 49 ], and FGENESH+ [ 49 ] ab initio gene-prediction algorithms were run independently across the entire genome assembly to augment alignment-based gene identification methods. GrailEXP 4.0, GENSCAN 1.0, and FGENESH version 1.c were run with default parameters for human sequence. GrailEXP used expressed sequence evidence from RefSeq, UniGene and DoubleTwist HGI to refine gene predictions. FGENESH+ was run with protein sequences from BLASTX with E-score lower than 10 -5 . When multiple protein alignments overlapped, all overlapping protein sequences were clustered with BLASTClust [ 50 ] and the lowest E-score hit was used by FGENESH+. Synthesis and analysis Locus projections contained the union of all exons from all overlapping predictions in a contiguous region of the chromosome that were derived from sequence alignments or gene-finding algorithms. Predictions to a given strand of the genomic sequence that overlapped by even a single nucleotide were grouped into a single locus projection (antisense transcripts were not considered in defining the locus projections). The criteria for grouping predictions were intentionally kept loose, given that the intent was to include as many potential exons as possible in a given genomic region, and then use the experimental microarray-based approach to elucidate the actual gene structure. These merged overlapping predictions defined the 5' and 3' ends of the locus projections. Overlapping predicted exons were merged to form an exon prediction of maximal extent. Low-quality predicted exons from sim4 alignments that contained a high percentage of A or T were removed. We also removed sim4-predicted exons that overlapped two or more predicted exons from another sim4 alignment. Additionally, 3' sim4 and 3' or 5' FGENESH+ predicted exons that were short and/or distant from internal predicted exons were removed. Finally, locus projections that contained mRNAs from RefSeq were split at the 5' end of the RefSeq sequence. Locus projections supported by expressed sequences alone could be portions of 3' or 5' UTRs of genes included in the other gene-prediction categories described in the text. To minimize the consequences of this potential artifact, we used a UTR filter to exclude locus projections from the expressed sequence alone category that were within 20 kb of a locus projection supported by an ab initio gene model. All data were loaded into a relational database to count and categorize locus projections. At least one type of evidence was assigned to each predicted exon for each locus projection. Multiple types of evidence were assigned to a merged predicted exon if there was overlap between predicted exons of different types for at least 1% of the length of the merged exon prediction. One of the eight evidence categories discussed in the text was assigned to each exon on the basis of the combination of types of evidence. Locus projections inherited the highest-ranking evidence category of their constituent exons. Evidence categories were ranked in the following order: Refseq (highest); expressed sequence + protein + ab initio ; expressed sequence + ab initio ; protein + ab initio ; expressed sequence + protein; ab initio alone; protein alone; expressed sequence alone. FGENESH+ predictions were counted as protein + ab initio . For the ab initio category, predictions from at least two of FGENESH, GENSCAN and GrailEXP were required to overlap in at least one exon to be merged. Probe selection for the genome tiling and predicted transcript arrays Input sequences for probe selection were masked for vector, interspersed repeats, simple repeats, poly(A) tails, Escherichia coli contamination and human non-coding RNA and mitochrondrial DNA contamination using Scylla (Paracel). For genomic tiling arrays, 60 mer probes were then selected from unmasked regions of both forward and reverse complement strands at uniform 30-base intervals. For predicted transcript arrays, up to four oligonucleotide probes were selected from the unmasked regions of each transcript using a multistep process. The first step in the probe-selection process was the generation of a pool of candidate probes 60 nucleotides long (60 mers), where each probe was required to fall entirely within an exon from the set of exons under consideration. If there were fewer than four 60 mers then all 50 mers were considered as well. If there were fewer than four 50 mers or 60 mers then all 40 mers were considered, and so on. Stilts composed of sequence from Saccharomyces cerevisiae were added to the 3' ends of probes shorter than 60 nucleotides so that they had a total length of 60 bases when printed onto the arrays. The second step in the probe-selection process was the classification and reduction of the probe pool on the basis of base composition and related filters. Probes were sorted into four classes on the basis of several criteria, including A, G, C and T content, GC content, the length of the longest homopolymeric run and the number of A residues at the 5' end. For example, a probe had to have GC content between 35 and 45% to be in class 1, between 15 and 55% to be in class 2, and between 10 and 60% to be in class 3. After all classifications were made, probes from lower-quality classes were discarded, keeping the number of probes per gene greater than 15. In cases where a pair of probes was overlapping by more than 50 bases, only a single probe was chosen. The final step in the probe-selection process identified probes with minimal overlap, and predicted cross-hybridization and desirable positions in the transcript sequence. Cross-hybridization prediction was based on BLAST searching of the full collection of transcript sequences [ 51 ]. Probes with perfect matches to transcript sequences for genes other than the one undergoing design were discarded unless they were the only probes available. Otherwise the probes with the weakest predicted cross-hybridization interactions were preferred. Probes were also selected to have as little overlap as possible, and probes located in the last 500 bp of each transcript were discarded where possible to reduce the effects of impaired amplification in this region [ 52 ]. All arrays included a set of standard control probes which were used for image processing and quality control. Each array also included 30 randomly distributed copies of each of 51 negative-control probes. These probes were selected for their low intensities in previous human hybridizations. The negative controls local to each experimental probe were used for background correction. Non-control probes were added to each array such that all probes for a given input sequence were grouped together and ordered by their position on the sequence. Preparation of labeled cDNA and array hybridization Hybridization material was generated through a random-priming amplification procedure using primers with a random sequence at the 3' end and fixed motif at the 5' end. This amplification procedure has been fully described [ 52 ] and has been optimized to generate strand-specific cDNA copies of the mRNA transcripts that are full-length. The 60 mRNA samples from the human tissues described in Additional data files 2 and 3 were purchased from Clontech. The 60 mRNA samples hybridized to the predicted transcript set of arrays were done in duplicate with fluor reversal to systematically correct for dye bias. For tiling hybridizations, six samples were used for chromosome 20 arrays and eight samples for chromosome 22. The mRNA samples hybridized to the set of tiling arrays were not done in duplicate as the analysis carried out on these data was intensity based, and our preliminary data demonstrated reasonable results without performing the tiling experiments in fluor-reverse pairs (data not shown). Additional data files 2-4 contain the full list of samples used for each set of arrays. Array images were processed as described [ 53 ] to obtain background noise, single channel intensity and associated measurement error estimates. Expression changes between two samples were quantified as log 10 (expression ratio) where the expression ratio was taken to be the ratio between normalized, background-corrected intensity values for the two channels (red and green) for each spot on the predicted transcript arrays. An independent normalization routine was carried out on the tiling data as described [ 33 ] to correct for dye biases, given the lack of technical replicates for these data. Analysis of predicted transcript array data Probes from each computationally determined locus were analyzed for coordinated expression over 60 tissues by adapting an additive, probe-specific model initially developed to estimate gene expression indices [ 27 ]. The model for a single probe in a single sample pair is given by y ij = μ + φ j + θ i + ε j , where the y ij represent the mlratio measurements for sample pair i and probe j in the current transcriptional model, μ is the grand mean term, φ i is the probe-specific term for probe j in the model, θ i is the sample-specific term for sample i , and ε j is the probe-specific error term, which is taken to be normally distributed with mean 0 and variance . Given the above representation for an observed mlratio value, the likelihood for a single probe over N condition pairs is simply From this, the likelihood for a given transcriptional model, where a transcriptional model in this context is defined as a set of probes that are adjacent to one another in the genomic sequence and that co-regulate over a number of conditions, is easily seen to be the product of the individual probe likelihoods defined above over the M probes comprising the current model: The maximum likelihood estimates for the parameters of this model are obtained using standard optimization techniques. With the likelihood model described above, probe groups making up a transcriptional model were formed by iteratively considering whether neighboring probes (within a PTI member based on genomic location) of a given probe improved the fit of the model just described. This was determined by examining the likelihood ratio statistics between the current, best transcriptional model with or without an additional probe included in the model. Thresholds for the likelihood ratio test statistic and the different model parameters were empirically determined to minimize false-positive and false-negative rates. False positives were estimated by the detection of PTI members supported by only a single ab initio prediction that fell outside annotated Sanger genes on chromosomes 20 and 22. False negatives were defined as Sanger genes on chromosome 20 and 22 that were not detected. Probe sets with a maximum likelihood statistic greater than 100 and an r 2 value for fit of data to the model greater than 0.8 were considered high-confidence candidates for EVGs. For each high-confidence EVG candidate, probes were further assessed by considering the number of conditions in which the absolute intensity of the probe was seen to be significantly above background, and the number of times the probe was seen significantly differentially expressed. Candidate EVGs with at least one probe that was: significantly above background ( p -value < 0.01) in at least 10% of the samples; or significantly differentially expressed ( p -value < 0.01) in at least 10% of the condition pairs, were considered validated. Analysis of tiling array data The analysis of the tiling data has been described in detail by Ying et al. [ 33 ]. Briefly, probes were separated into 15 kb windows along the genome with 7.5 kb overlap between the windows. For each window, a robust principal component analysis was applied to the between-sample correlation matrix for probes in the window. Windows containing transcriptional activity were characterized by comparing the distribution of the Mahalanobis distances for the probes in the window (the Mahalanobis distance for each probe was calculated from the probe location to the center of the data in the first dimensions of the principal component score (PCS)) space with the reference distribution calculated based on known intron probes. Individual probes were then classified as belonging to the transcribed unit or not on the basis of the log of the Mahalanobis distance and an approximation of the diagonal distance (slope) of the probe from the minimum first PCS and median second PCS. Using these measures for distance, the probes were clustered using standard clustering techniques as described [ 33 ]. The procedure for estimating cross-hybridization of the probes is the subject of a separate manuscript. For the analyses described in this paper, the nonspecific cross-hybridization was estimated by the presence of motifs within the probe sequence that were enriched in probes observed to have a high level of nonspecific cross-hybridization. These probes were observed to have significant intensity when hybridized to human mRNA samples despite having no EST support and falling in introns of well characterized genes on chromosomes 20 and 22. Specific cross-hybridization was estimated by the minimum predicted ΔG value for hybridization of the probe to all genes other than the intended target in the UniGene database (build 157). Annotation of EVG and tiling data Hidden Markov model Pfam (HMMPfam) domain predictions were run on six-frame translations of the PTIs using the HFRAME software from Paracel with an E-value cutoff of 0.01 and frameshift penalty of -12. Information on Pfam [ 31 ] domains is available [ 54 , 55 ]. GO terms [ 32 ] were then assigned to each locus projection using the full set of Pfam to GO mappings available from EBI FTP site [ 56 ]. The domain-to-ontology mapping is a part of the larger InterPro effort on annotating the proteome [ 57 , 58 ]. Multiple GO categories can be assigned to a single element of the PTI. Additional data files The following additional data is available with the online version of this paper and at [ 19 ]. Additional data file 1 gives a complete list of 48,614 transcripts in the PTI that were represented on the set of predicted transcript arrays. Additional data file 2 gives a complete list of 60 tissues and cell lines hybridized to the predicted transcript arrays. Additional data file 3 gives a list of six tissues and cell lines hybridized to the chromosome 20 genomic tiling arrays. Additional data file 4 lists the eight tissues and cell lines hybridized to the chromosome 22 genomic tiling arrays. Additional data file 5 contains a comparison of EVG predictions with RefSeq sequences (March 2004). Also available on our website [ 19 ] are: ratio data and body atlas data along with the EVG status, and full sequences for the locus projections in fasta format. All probe sequences and expression data are available from the GEO database [ 59 ]. The series accession numbers for the tiling and predicted transcript arrays are GSE1097 and GSE918 respectively. Supplementary Material Additional data file 1 A complete list of 48,614 transcripts in the PTI that were represented on the set of predicted transcript arrays Click here for additional data file Additional data file 2 A complete list of 60 tissues and cell lines hybridized to the predicted transcript arrays Click here for additional data file Additional data file 3 A list of six tissues and cell lines hybridized to the chromosome 20 genomic tiling arrays Click here for additional data file Additional data file 4 The eight tissues and cell lines hybridized to the chromosome 22 genomic tiling arrays Click here for additional data file Additional data file 5 A comparison of EVG predictions with RefSeq sequencesP Click here for additional data file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545593.xml
534801	Genetic diversity of Plasmodium vivax isolates from Azerbaijan	Background Plasmodium vivax , although causing a less serious disease than Plasmodium falciparum , is the most widespread of the four human malarial species. Further to the recent recrudescence of P. vivax cases in the Newly Independent States (NIS) of central Asia, a survey on the genetic diversity and dissemination in Azerbaijan was undertaken. Azerbaijan is at the crossroads of Asia and, as such, could see a rise in the number of cases, although an effective malaria control programme has been established in the country. Methods Thirty-six P. vivax isolates from Central Azerbaijan were characterized by analysing the genetic polymorphism of the circumsporozoite protein (CSP) and the merozoite surface protein 1 (MSP-1) genes, using PCR amplifications and amplicons sequencing. Results Analysis of CSP sequences showed that all the processed isolates belong to the VK 210 type, with variations in the alternation of alanine residue (A) or aspartic acid residue (D) in the repeat motif GDRA(A/D)GQPA along the sequence. As far as MSP-1 genotyping is concerned, it was found that the majority of isolates analysed belong to Belem and Sal I types. Five recombinant isolates were also identified. Combined analysis with the two genetic markers allowed the identification of 19 plasmodial sub-types. Conclusion The results obtained in the present study indicate that there are several P. vivax clones circulating in Azerbaijan and, consequently, a careful malaria surveillance could be of paramount importance to identify, at early stage, the occurrence of possible P. vivax malaria outbreaks.	Introduction Plasmodium vivax is the most widely distributed human parasite, with an estimated burden of 70–80 million cases annually [ 1 ]. In some parts of the world (Asia, South America), it is the most prevalent form of the four human malarial parasites. Although it causes a less severe disease than Plasmodium falciparum , being rarely lethal, P. vivax affects the working capacity of the population and the lack of efficient drug distribution favors the onset of drug resistant strains [ 2 , 3 ]. Imported malaria is an increasing health problem in Western Europe, where about 6,500 cases are reported annually in Germany, France, Italy and the United Kingdom [ 4 ]. Although P. falciparum infections account for the majority of cases (64%), P. vivax is responsible for an additional 23% [ 4 ]. Presence in this area of residual anopheline populations susceptible to P. vivax infection represents a permanent risk for the occurrence of P. vivax indigenous malaria cases, as recently occurred in central Italy [ 5 , 6 ]. Since 1970, malaria had been eradicated in central Asia, except for some residual foci in two countries belonging to the Newly Independent States (NIS), i. e. Azerbaijan and Tajikistan (WHO, Regional Office for Europe, unpublished document). At the beginning of the 1990s, the situation changed dramatically due to the re-emergence of malaria in the NIS area and especially in Tajikistan, where at present an epidemic is still in progress [ 7 , 8 ]. In these countries, the existing state of the primary health care system is extremely precarious, especially in rural areas and in small villages. Malaria is a common disease, which can easily re-establish itself when a lack of control occurs. In comparison with P. falciparum , molecular studies of the genetic diversity and dissemination of P. vivax are scanty. Recently, 33 polymorphic tandem repeats (TRs) of P. vivax and a P. vivax polymorphic microsatellite have been identified and shown to be useful in population studies [ 9 , 10 ]. The merozoite surface protein 3α (MSP3-α) gene also seems to be a good candidate for studying the genetic diversity of P. vivax populations, since PCR-RFLP products indicate the presence of up to 13 alleles [ 11 , 12 ]. However, the circumsporozoite protein (CSP) and merozoite surface protein 1 (MSP-1) genes still remain the most studied molecular markers in genetic epidemiological surveys carried out in P. vivax endemic areas. In the frame of a malaria research project funded by the European Commission, a molecular study was undertaken in Azerbaijan, aimed at collecting information on the genetic make-up of P. vivax natural populations present in this endemic country. For this purpose the extent of polymorphism of CSP and MSP-1 genes were analysed in parasite isolates from five localities of central Azerbaijan by using PCR amplification and sequencing. Materials and Methods Study area and samples collection Azerbaijan covers an area of 29.540 Km 2 , with a populations of approximatively 2,5 millions. The climate is typically continental with an average temperature between 12 and 15°C and a rainfall between 200 and 600 mm per year. Climatic and agro-ecological conditions of this area make the environment favourable to mosquito vectors breeding. The major malaria vector is Anopheles sacharovi that breeds preferably in lakes, swamps, irrigation canals and pools. Although it prefers well-oxygenated water, it is known to tolerate moderate salty water. Other anopheline species found in this area are A. maculipennis , A. subalpinus , A. superpictus and A. hyrcanus , all of which are considered secondary vectors of malaria transmission [ 13 ]. Malaria transmission occurs in Azerbaijan mainly from June to October. In the last years, number of malaria cases showed a negative trend, accounting for 610 cases in 2000, 418 cases in 2001, 203 cases in 2002. During summer 2002, a malaria epidemiological survey was performed in central Azerbaijan, in the frame of a Malaria Surveillance Programme launched in year 2001 by the Ministry of Health in collaboration with WHO. Active case detection was carried out in five districts included in previously identified sentinel sites, namely Mingaçevir, Beylagan, Imisli, Saatli and Sabirabad. A map of Azerbaijan with the study area is shown in Figure 1 . Figure 1 Map of central Azerbaijan showing localities (underlined names on the map) included in the present study. All individuals who visited the district health centres or were found in villages with a history of recent fever and no history of travel in the past few months were considered. In this context, a total of 36 infected individuals with positives blood smears at the microscopic examination, collected between August and September 2002, were selected for the genetic study. The age of patients ranged from 6 to 78 years and parasitaemia varied from 288 to 12,800 parasites/μl. For the molecular analysis, a blood sample of about 1 ml was taken from by venipuncture before drug treatment was given. Patients or the guardians of children were informed about the study. According to the international rules for research involving human subjects, any information which would identify a participant was removed in order to keep each sample processed anonymous. Number of samples for each district is shown in Table 1 . Table 1 Geographic origin of Azerbaijan isolates with the corresponding MSP-1 and CSP characteristics identified in the present study. ISOLATE NAMES MSP-1 CSP Genotype No. polyQ Sub-type Genotype Sub-type Bey1 Belem 21 G VK210 4 Bey2 Belem 21 G " 4 Bey4 Sal1 - O " 5 Bey7 Belem 21 G " 8 Bey14 recombinant 19 S " 2 Im3 Belem 21 G " 4 Im5 Belem 21 G " 4 Im8 Sal1 - H " 1 Im9 Belem 21 G " 4 Im10 Belem 21 F " 4 Im11 Belem 21 G " 4 Im12 Belem 21 C " 4 Im14 Belem 21 D " 4 Im15 Belem 21 G " 4 Min1 Belem 21 G " 4 Min3 Belem 21 G " 4 Min6 Sal1 - L " 3 Min7 recombinant 19 S " 2 Min8 Sal1 - I " 2 Min9 Belem 21 G " 4 Min10 Sal1 - P " 6 Sat2 Belem 21 G " 4 Sat3 Belem 21 F " 4 Sat5 recombinant 18 U " 5 Sat7 recombinant 18 U " 5 Sat11 recombinant 12 T " 5 Sab1 Belem 21 G " 7 Sab2 Belem 21 A " 4 Sab4 Belem 21 B " 4 Sab6 Sal1 - M " 4 Sab7 Belem 21 G " 4 Sab8 Belem 21 G " 4 Sab10 Sal1 - M " 4 Sab12 Belem 21 G " 4 Sab13 Sal1 - R " 4 Sab15 Sal1 - Q " 4 DNA preparation Plasmodial DNA was extracted from 200 μl of each infected blood sample using QIAamp DNA blood kit following the manufacturer's instructions (Qiagen, CA). Circumsporozoite (CSP) marker analysis The CSP gene was amplified for the most part of samples using PV5 and PV6 primers [ 14 ]. Samples that did not provide good PCR products with this set of primers were processed a second time by using CSP-A2 [ 15 ] as forward primer and PV6bis (5'-CACAGGTTACACTGCATGGAGT-3') as original reverse primer. PCR amplification was performed in a reaction mixture of 50 μl containing the parasite DNA, 1x reaction buffer, 2.5 mM MgC1 2 , 80 μM of each deoxynucleotide triphosphate, 6 pmol of each primer and 1.3 U of Taq polymerase (Promega, Madison, USA). The PCR programme was: denaturation at 94°C for five minutes; 34 cycles of one minute at 94°C, one minute at 54°C and two minutes at 72°C. The PCR products were separated using electrophoresis on a 1.5 % NuSieve gel and the band of interest was cut out and purified using the QIAquick PCR purification Kit (Qiagen). The purified product was sequenced in both directions using an ABI-PRISM 373 sequencer. Nucleotide or amino acid sequences were aligned first using the CLUSTAL X programme [ 16 ] with manual editing and adjustments made using the MUST package [ 17 ]. The ExPASy Molecular Biology Server was used to convert nucleotide sequences into amino acid sequences. The GenBank accession numbers of the eight sub-types of VK210 type are from AY792359 to AY 792366. Merozoite surface protein 1 (MSP-1) marker analyses A portion of the MSP-1 gene (the region encompassing the interspecies conserved blocks ICB5 and ICB6) was amplified using a nested PCR with, respectively, the two outer primers A5 and A6 [ 18 ] and the two inner primers MSP1N1 forward and MSP1N2 reverse [ 6 ]. The first round of amplification was performed in a reaction mixture of 50 μl containing parasite DNA, 1x reaction buffer, 2.5 mM MgC1 2 , 200 μM of each deoxynucleotide triphosphate, 30 pmol of each primer and 2.5 U of Taq polymerase (Promega, Madison, USA). For the second round, 10 μl of the first amplification product was added to a fresh PCR mixture with 30 pmoles of each inner primer. The thermal profile was: denaturation at 94°C for five minutes; 35 cycles of 94°C for one minute, 60°C for one minute and 75°C for three minutes. All nested-PCR products were purified by Microcon-PCR (Millipore), following the manufacturer's instructions and sequenced in both directions at the MWG Biotech. The results were analysed by means of Omiga 2.0 (ACCELRYS, Cambrige) and Mega 2 (S. Kumar, K. Tamura, M. Nei and Pennsylvania State University) computer programmes. The GenBank accession numbers of the 36 nucleotide sequences from P. vivax isolates are from AY789657 to AY789692. Distance analyses The aligned nucleotide sequences of CSP were converted to a distance matrix (% of differences) using the Net algorithm of the MUST package [ 17 ]. The dendrogram was generated using the neighbour-joining method [ 19 ]. Bootstrap proportions were used to assess the robustness of the tree with 1,000 bootstrap replications [ 20 ]. MSP-1 and CSP data were analysed using the Cavalli-Sforza distance [ 21 ] from Genetics v.4.01 package. The dendrogramme was generated using the neighbour-joining method [ 19 ] Results CSP marker CSP sequences obtained from 36 Azerbaijan P. vivax isolates were found to belong to the VK210 type [ 22 ]. The isolates tested displayed variations in the peptide repeat motifs GDRA(A/D)GQPA with different alternations of non-synonymous codons GCT or GAT, respectively, coding for alanine (A) and aspartic acid (D) (Figure 2 ). All our sequence types had the same three repeat units (GDRAAGQPA) at the 3' end, identical to that of the VK210 type. Furthermore four non-synonymous mutations were found, one being the RDRADGQPA variant (sequence named in the present study as sub-type 1), already described in North Korean and Chinese isolates [ 23 ]. In summary, eight different sub-types of VK 210 were observed (Figure 2 and Figure 4 ). Among all 36 azeri isolates analysed, 24 isolates were found to have identical sequence (sub-type 4, Table 1 and Figure 4 ). In particular, the Beylagan (n = 5) and Mingaçevir (n = 7) isolates appeared the most diversified since they displayed four and five different sub-types respectively. The Imishli (n = 9), Saatli (n = 5) and Sabirabad (n = 10) isolates only showed two different sub-types each-one. Figure 4 clearly shows that the genetic diversity of CSP is relatively small inside the Azerbaijan isolates when compared to the South Korean and Chinese isolates. Figure 2 Amino acid sequence alignment of eight CSP sub-types found from 36 Azerbaijan P. vivax isolates with that of VK210 type (Accession No. M28745). a Imi 8; b Bey 14, Min 7, 8; c Min 6; d Bey 1, 2, Imi 3, 5, 9, 10, 11, 12, 14, 15, Min 1, 3, 9, Sat 2, 3 Sab 2, 4, 6, 7, 8, 10, 12, 13, 15; e Bey 4, Sat 5, 7, 11; f Min 10; g Sab 1; h Bey 7. Figure 4 Distance tree (built with the neighbor-joining method) inferred from 443 nucleotide positions and 264 variable sites of CSP gene. Numbers on the branches indicate bootstrap proportions (1000 replicates); only bootstrap values above 70 % are displayed on the tree. MSP-1 marker The majority of Azerbaijan isolates (Table 1 ) belong to either the Belem (22 isolates, all with the same poly-Q region of 21 repeats) or the Sal I (9 isolates) types already described [ 24 ]. Only 5 P. vivax isolates were identified as recombinant types. Isolate Satl1 (sequence named in the present study as sub-type T) could be ascribed to the type 3a (accession no. D85252, [ 25 ]), while isolates Bey 14 and Min7 (sub-type S) and isolates Sat5 and Sat7 (sub-type U) seem to be the result of recombinant events between the recombinant type 3a and Sal I. All the three recombinant types showed a different number of poly-Q repeats (Table 1 ). In addition to these sources of diversity, nucleotide substitutions could be observed, allowing the identification of 17 sub-types (Table 1 and Figure 3 ). The Imisli and Sabirabad districts appeared to be less diversified, accounting for five different genotypes for 9 isolates and six genotypes for 10 isolates, respectively. Finally, Saatli district was found to have the greatest variability, with four different genotypes for 5 isolates. Figure 3 Amino acid sequence alignment of seventeen MSP-1 sub-types found from 36 Azerbaijan P. vivax isolates compared with that of MSPlBelem (Accession No. M60807), MSPlSal1 (Accession No. M75674) and recombinant type 3a (D85252). Classification of Azeri isolates according to the different types is shown in Table 1. Combined analysis between the two markers By combining the results of genotyping obtained by CSP and MSP-1, 19 P. vivax sub-types (Figure 5 ) were identified as circulating in the central region of Azerbaijan. The sub-type named G/4 with the greatest representation (n = 14 isolates), was detected in all districts investigated. Genotypes identified as M/4 and U/5 were observed twice in the districts of Sabirabad and Saatli, respectively. Genotype F/4 was detected once in Imisli and Saatli districts, as was for genotype S/2, detected once in Beylagan and Mingacevir districts. Figure 5 Neighbour-joining tree from the MSP-1 and CSP data (results in parenthesis) reflecting the relationships between the Azerbaijan P. vivax isolates. Discussion and conclusions For CSP, the main variations already reported in the literature consist of two variant sequences, VK210 and VK247 that show a variable number of repeat units, GDRA(A/D)GQPAA and ANGAGNQPG, respectively, with some variant positions within the sequence [ 22 , 23 ]. These two variants have a worldwide distribution, and locally their distribution have been also correlated with climatic gradients or with the Anophelines vector specificity [ 26 ]. Studies carried out in some Asian endemic countries, i.e. South and North Korea, China, the Philippines, the Solomon Islands and Thailand [ 27 , 15 , 28 , 12 ] suggest that CSP has a limited value as a molecular marker for genetic variability when used alone. In the present study, all the analysed P. vivax isolates from Azerbaijan were found to belong to the type VK210 and they constitute a group of eight CSP sub-types that closely linked in the dendrogramme shown in Figure 4 . Differently from what reported by other Authors [ 28 ] who showed that the CSP sequence analysis allows detecting the geographic origin of plasmodial isolates, our results did not support its use for tracking the geographic origin of Azeri isolates since we dealt with a limited number of samples studied. No genotype association with particular sampling districts was observed since, for example, the most common sub-type identified (G/4) was present in all five districts. Our results confirmed that MSP-1 is a good polymorphic marker. In particular, the region of the gene known to be highly polymorphic and discriminative between the Belem and Sal I types was analysed [ 23 ]. A variable poly-glutamine (poly-Q) region is characteristic of the Belem type and represents the principal source of genetic diversity of this marker. Moreover, a poli-Q region is the recombination site between the two types Belem and Sal I and, as shown in the literature, interallelic recombinations between the two types are frequent [ 25 ]. The total genetic diversity observed when including the nucleotide substitutions is relatively important taking into account the low endemicity of studied area. Similar results were observed in Southeastern Iran and in Thailand [ 29 , 12 ], low endemic countries for P. vivax malaria as well, where the authors detected the two types Belem and Sal I, together with several recombinant types. In particular, in the study carried out in Iran by Zakeri et al. , the analysis of MSP1 genetic diversity on a total of 16 plasmodial isolates leaded to the identification of 14 genetic sub-types. It is worth noting that such a high level of diversity is probably due to the small sample size. Our results show 17 genetic sub-types on a total of 36 isolates analysed and a quite high MSP-1 polymorphism also in Azerbaijan. As suggested in other studies [ 12 ] and also reported by Zakeri et al. , it is possible to speculate that the observed genetic diversity could be also explained considering the studied area, i.e. central Azerbaijan, as transit road of the country and also of neighboring Asian country, where the circulating P. vivax populations show considerable MSP-1 genetic diversity. However, further studies aimed at collecting more information about people moving within the whole country and to closer countries are needed to verify the above hypothesis. The combined analysis of CSP and MSP1 sequence polymorphism has led to the identification of a total of 19 P. vivax sub-types, confirming that the simultaneous use of more than one genetic marker in this kind of study enhances the knowledge of genetic diversity existing in the parasite populations. The results of the current study show the circulation of multiple plasmodial clones in the studied area thus leading to the conclusion that malaria surveillance activities must be maintained in Azerbaijan in order to avoid serious disease outbreaks in the future. The understanding of the polymorphism extent in surface antigens as CSP and MSP-1 and the resulting genetic diversity in P. vivax field populations could help in implementing malaria control activities being a crucial step for the development of a malaria vaccine. Authors' contributions S. Mammadov, N. Aliyev, E. Gasimov were involved in field collection of blood samples and microscopy examinations. M.C. Leclerc and A. Cligny did the CSP sequence analysis and M. Menegon did the MSP-1 sequence analysis. M.C. Leclerc and J.L. Noyer did the distance analyses. M.C. Leclerc wrote the report with major contributions of C Severini, M Menegon and G. Majori. G. Majori coordinated the field activities carried out in Azerbaijan. C. Severini, as scientific coordinator of the VIVAXNIS project mentioned below, got the financial support.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC534801.xml
555555	Orientation determination by wavelets matching for 3D reconstruction of very noisy electron microscopic virus images	Background In order to perform a 3D reconstruction of electron microscopic images of viruses, it is necessary to determine the orientation (Euler angels) of the 2D projections of the virus. The projections containing high resolution information are usually very noisy. This paper proposes a new method, based on weighted-projection matching in wavelet space for virus orientation determination. In order to speed the retrieval of the best match between projections from a model and real virus particle, a hierarchical correlation matching method is also proposed. Results A data set of 600 HSV-1 capsid particle images in different orientations was used to test the proposed method. An initial model of about 40 Å resolutions was used to generate projections of an HSV-1 capsid. Results show that a significant improvement, in terms of accuracy and speed, is obtained for the initial orientation estimates of noisy herpes virus images. For the bacteriophage (P22), the proposed method gave the correct reconstruction compared to the model, while the classical method failed to resolve the correct orientations of the smooth spherical P22 viruses. Conclusion This paper introduces a new method for orientation determination of low contrast images and highly noisy virus particles. This method is based on weighted projection matching in wavelet space, which increases the accuracy of the orientations. A hierarchical implementation of this method increases the speed of orientation determination. The estimated number of particles needed for a higher resolution reconstruction increased exponentially. For a 6 Å resolution reconstruction of the HSV virus, 50,000 particles are necessary. The results show that the proposed method reduces the amount of data needed in a reconstruction by at least 50 %. This may result in savings 2 to 3 man-years invested in acquiring images from the microscope and data processing. Furthermore, the proposed method is able to determine orientations for some difficult particles like P22 with accuracy and consistency. Recently a low PH sindbis capsid was determined with the proposed method, where other methods based on the common line fail.	Background Three-dimensional (3D) reconstruction of virus particles like SARS (Severe Acute Respiratory Syndrome) and HSV (Herpes Simplex Virus) using electron microscopy yields crucial information for understanding the assembly and infectivity mechanism. The structural determination begins with acquisition of projection images in an electron-microscope. A major part of data processing is aimed at determining the direction of projection for each particle image (2D projection of virus) so that a 3D reconstruction can be computed. The first step in a virus reconstruction is the detection and selection of the individual particle images from a large area of an electron micrograph. There are different criteria to determine the particle orientation. One criterion is based on the computational search of the common lines in the computed Fourier Transforms of individual or multiple particle images [ 1 ]. An improvement of the Fourier Common Line algorithm [ 2 ] has been proposed, but a significant amount of the low contrast particle images are still discarded, partly because of the impossibility of obtaining a reliable estimate of their orientations. Another criterion for the particle orientation estimate is to find the correlation match between the raw images with many projections from a 3D model [ 3 ]. Regardless of the criterion used, finding the orientation determination for a particle image such as that in Fig. 1a is difficult because of its extremely low contrast. One approach is to take two consecutive pictures of the same particles one close-to-focus (Fig. 1-a ) and another farther from focus with a higher contrast (Fig. 1-c ) from which the initial orientations are easily determined [ 4 , 5 ]. The initial orientations are then assigned to the corresponding particles in the close-to-focus images for structural refinement (henceforth, called focal pair method). In a high resolution structure determination, one would require over 6000 particles of data for 8.5 Å resolutions [ 6 ]. If a focal pair is required, one would need over 12,000 particles and hence it is a labor-intensive process of data recording, digitization and archiving. In this paper, we propose a method for determining the initial orientations of the particles from low contrast (close-to-focus) images without necessity for a second set of highly defocused images. In this technique, we use the wavelet transformation in a multi-resolution analysis [ 7 , 8 ] to enhance the contrast of the image and the hierarchical weighted projection matching to accelerate the processing. The wavelet-transformed images have the same size as the original images. Wavelet decomposition separates the low-resolution information, called "approximation", from the high resolution information, called "details". This method computationally generates an image equivalent to the far-from-focus picture taken by the microscope and separates images containing details and noise. The technique proposed here is a model-based approach in wavelet space, which we call Hierarchical Wavelet Projection Matching (HWPM). Results A data set of 600 HSV-1 capsid particle images in different orientations was used to test the HWPM method. The defocus range of herpes particles was chosen to be close to focus between 1.7 μm and 0.4 μm. An initial model of about a 40 Å resolution [ 4 ] was used to generate projections uniformly covering the asymmetric triangle of the icosahedrally symmetric HSV-1 capsid particle [ 4 , 5 ]. A grid sampling of 0.5° in each direction of the asymmetric triangle of icosahedral particles was used. The number of projections obtained with this grid was relatively high (2616 projections). First, the 2616 projections were grouped into 200 classes, each class containing about 13 projections. A match of the particle into the best 3 of the 200 classes was obtained using the wavelet correlation coefficient (wccf) criterion. Next, the particle was compared to the 39 projections of the best three classes, and the correct orientation was that of the projection giving the highest wccf. The hierarchical implementation wavelet projection matching reduced the time at least by a factor of 10 compared with the classical projection matching method. In the example of 600 particles, by using HWPM it took approximately 3 hours to determine the orientations, instead the 33 hours it took with the classical matching algorithm. Both algorithms were running on the SGI Origin-2000 supercomputer using 10 processors. At this point, each particle had been assigned the orientation of the closest projection. A quality factor was assigned to each orientation, which was the wavelet correlation coefficient. Particles having high wccf coefficients were selected for reconstruction of a first 3D model of the virus. Refinement of initial orientations obtained by HWPM was realized by the same iterative refinement process used in focal pair method [ 5 ]. This refinement process uses both local and global refinement. Local-refinement refines orientations against a set of projections from the 3D density map. In global refinement, all the raw particle orientations are refined against each other, without using projections from the 3D model. A potential merit of global refinement is the absence of possible bias arising from the 3D model. In order to assess the accuracy of the orientations obtained with the HWPM. A comparison with the focal pair method (Fig. 5.a ), which is currently the most appropriate method for low contrast virus images was accomplished. The following steps were executed. First the initial orientations of the far-from focus particles were determined by using the cross-common line method between real particles and a set of projections obtained from the low resolution model. Next, a global refinement process was realized in order to determine the initial orientation. The same software as in [ 4 ] was used with the same initial parameters. The parameters used in this software were the minimum radius and maximum radius limiting the resolution and the sampling step size of 4.67Å/pixel. The minimum valid radius ensured that the minimal radius was computationally accurate when the two common lines angles were close and also to avoid the biasing of the orientations of particles by the very low frequency components. This parameter choice for herpes at the above sampling step was 5 pixels. The maximum radius was chosen to limit the maximum resolution expected from the reconstruction, here in the initial orientation the maximum radius corresponding to a resolution about 40 Å. Then an assignment of the particle orientations from the far to focus to the close to focus is realized. Next, an iterative refinement process to the close-to-focus data was accomplished as described in [ 5 ]. A 3D reconstruction using the best 300 HSV-1 particle orientations was performed for each method. Fig. 5.a and Fig. 5.b show surfaces density contour, displayed at one standard deviation above the mean density [ 9 ], obtained respectively from focal pair method and HWPM method. Both structures show a similar visual resemblance. In order to assess the reliability of the 3D density maps and the quality of particles orientations obtained from each method, the Fourier Shell Correlation (FSC) criterion, which is the most robust criterion [ 10 , 11 ], was employed. The FSC was calculated between 2 independent reconstructions from the same set of orientations for each method. The effective resolution assessment of the 3D structure obtained from each method is estimated at FSC correlation value of 0.5, which correspond to 45° phase difference. Fig. 6 shows three different plots. The green curve shows a resolution of 32 Å of the reconstruction using the best 300 particles with orientations obtained from the focal pair method. The blue curve shows a resolution of 24 Å of the reconstruction using 300 particles with orientations obtained from the HWPM method. This result shows that the resolution of the structure obtained from the HWPM is higher than the one using the orientations from the focal pair method. Therefore, the orientations obtained from HWPM method are more accurate. Furthermore, HWPM method uses only one set of close-to-focus data instead of the two sets used by the focal pair method. The purple curve shows a resolution of about 14.5 Å of the reconstruction using 500 particles with orientations assigned by HWPM. The red curve plots twice the expected FSC for Gaussian noise. A less stringent criterion to assess the resolution as the intersection between the FSC curve and the curve plotting the 2 times expected Gaussian noise. HWPM was tested on a P22 empty shell capsid which was circular and whose shell is very thin (~40 Angstrom). Twenty micrographs of the P22 empty shell capsid with defocus range [0.5 to 2 μm] were used for testing purpose. The total number of particles is 1340, each image has a size of 300 × 300 pixels, and the dimension of each pixel is 2.8 Å. Concerning the initial orientations determinations using HWPM method, an initial model of around 20 Å resolutions was used to generate projections which uniformly covered the asymmetric triangle of the icosahedrally symmetric model. A grid sampling of 2° in each direction of the asymmetric triangle of icosahedral symmetry was used to obtain an initial orientation, targeting a structure of 30 Å. The number of projections obtained with this grid was about 200 projections. A match of the particle with the projections was obtained by using the wccf criterion. The correct orientation was selected as the one of the projection giving the highest wccf. The better half of the orientations projections (650) according to wccf criterion was chosen for final reconstruction. The initial orientations for the same set of data were determined using the Improved Common Line (ICL) method, with the same input parameters for the software described in [ 2 ]. ICL use one single micrograph and does not use focal pair technique. The best half of the particles orientations (650) was chosen, according to the phase residual criterion, in the 3D reconstruction of the P22. Fig. 7 shows three surface views of the P22 empty shell capsid. The Top image shows the original surface [ 12 , 13 ]. The lower right image shows the surface obtained by HWPM, which shows a very similar view to the original structure. The resolution assessment of the structure, by Fourier shell correlation criterion, gives a resolution of 14.5 Å. The lower left surface shows the result obtained by ICL method. The surface view of the reconstruction obtained from the ICL of the P22 empty shell capsid is different from the original P22 capsid. Fig. 7 proves the inaccuracy of some of the initial orientations obtained from the ICL method for such a smooth virus. Discussion During the last thirty years the common lines methods were a great method to resolve icoshedral particles up to 7–8 Å [ 6 ]. Recently, a method using polar transformation and projection matching were used for the purpose of orientation determination [ 3 ], but this last method is not suitable for the high resolution of large virus because the resulting transformed images, could be double the size of the original image. The proposed method combines the projection matching of wavelet denoising for an initial determination of particle orientation, with the common lines method for refinement to a higher resolution. It is clear that HWPM method works only if the initial low resolution model of the particle is already known. This method is very interesting if we need to add more particles to an existing intermediate resolution reconstruction in order to increase the resolution. Particles having high resolution information are very noisy [ 9 , 14 ]. The best that we can get using the ICL method is less than 40 % of good orientations, for defocus values between 1.9 μm and 1.2 μm, for the P22 capsid [ 2 ]. Usually, very high resolutions use defocus values which go much lower than 1.2 μm as in the HSV data, or the current P22 data which goes to 0.5 μm. The 40% rate of correct orientations would certainly become smaller if we used data at closer defocus. The study accomplished on high resolution for HSV reconstruction showed that using a close-to-focus single micrograph with CL method was not effective, because a small number of orientations were found to be correct [ 9 ], for this reason a focal pair method was used for 8.5 Å structure[ 9 ]. At high resolution reconstructions, the number of particles needed increases drastically, and the data with a signal-to-noise ratio valid up to the targeted resolution, tend to be very noisy. For an 8.5 Å structure of HSV-1 it took about 6000 particles for a final reconstruction. For a 6.5 Å structure resolution, the estimated value was about 50,000 particles using the same electron microscope [ 14 ]. To further increase the resolution of the HSV virus to 6.5 Å or higher (4 Å), the focal pair method would be impracticable. The focal pair method, for intermediate resolution up to (8 Å) for big viruses like HSV, works well for orientations determinations. The number of particles selected for the final reconstruction about 40% of the original number of particles (taking into account the far-focus and close-focus micrographs). It is necessary to emphasize that results from both methods are very similar in terms of visual resemblance. But, there are two advantages of HWPM over the focal pair method. First, focal pair method uses as much as double the data used for the HWPM. Second, the quality of the density maps shows that HWPM gives a better resolution for the same number of particles (figure 6 ). This proves a better accuracy of orientations determinations obtained by the HWPM. One of the more obvious advantages of the HWPM for orientation accuracy appears in two examples of real reconstructions. The first is for the P22 capsid, the ICL method does not give a good initial orientation, and the refinement of the orientations does not help to converge toward the right orientations. The probable reasons why the ICL method did not work properly for the P22 capsid are: first the P22 capsid has a smooth surface (the thickness of the shell is about 40 Å); second most of the data are very close-to-focus with defocus range of 0.5 μm to 1.3 μm. The data was noisy and had a very low contrast. The ICL method was able to give 40% of good orientations for the defocus range between 1.9 and 1.2 μm, here the data was closer to focus, which reduced the percentage of good orientations to less then 22%. The application of the HWPM to the P22 empty shell capsid gave the expected structure (Fig. 7 ). The wavelet denoising in the HWPM not only helped in reducing the noise and enhancing the contrast of the particles, but also used the entire information from the image (instead of using several lines) which is enhanced accuracy for highly noisy particles. Another example of real data reconstruction is the VP5-VP19C recombinant. After long investigation using CL and ICL algorithms, the classical projection matching scheme was also tested in order to determine the orientations, but unfortunately all those methods failed. The wavelet filtering and matching was used during the classification step of the recombinant particle VP5-VP19C [ 15 , 16 ], which significantly improved the quality of the class averages [ 16 - 18 ] and enabled the determination of the structure of that particle. A study [ 16 ] shows the superiority of the wavelet projection matching over the Gaussian filtered projection matching. The third examples for low PH sindbis: Three years of investigation using CL and ICL methods failed to obtain the correct density map of the low PH sindbis capsid which is subject to conformational changes and an alteration of the symmetry. Recently the proposed method (HWPM) was tested on low PH sindbis and the correct structure was finally observed and analyzed [ 19 ]. Wavelet multi-resolution analysis and processing improves particle detections [ 8 ], classification [ 15 , 16 ], and orientation determination on a variety of electron microscopy images which are highly noisy and have an extremely low contrast. This prove that wavelet techniques are adequate in the 3 main steps of 3D virus reconstruction and in the classification step of single particle reconstruction [ 16 , 17 ]. Conclusion This paper describes the development and implementation of a new method for orientation determination for low contrast images of virus particles. This method is based on wavelet filtering, which enhances the contrast of the particles and reduces the noise, and on weighted projection matching in wavelet space. A hierarchical implementation of this method increases the speed of orientation determination. Results show that, HWPM have been able to determine accurately more than 85% of the orientations of low-contrast particles. Compared to the focal pair method (for orientation determination from low contrast data) the HWPM reduced the amount of data required in a reconstruction by at least 50 %. In addition the accuracy of the orientations obtained by the proposed method is higher than those obtained by focal pair method [ 9 ]. This improved accuracy is shown clearly by the resolution assessment in Fig. 6 . The estimated number of particles needed for a 6.5 Å reconstruction of the HSV-1 capsid was about 50,000 [ 14 ]. By using the HWPM method, only half as much data was necessary. The proposed method could save 2 to 3 man-years invested in acquiring images from the microscope and data processing. Another advantage of this method is the ability to give accurate orientations for some particles having conformational changes or alteration of symmetry as seen for VP5-VP19C recombinant and recently with the low PH sindbis capsid. Methods Choice of wavelet Base The choice of wavelet filter bases depends on the signal. Signals coming from different sources have different characteristics. For audio, speech, image and video signals the best choices of wavelet bases are known. The best choice for electron microscopic images is not clear. The problem is to represent typical signals with a small number of convenient computable functions. An investigation to choose the best wavelet bases for electron microscopic images was performed here. During this study, simulated and real electron microscopy images were used. The majority of the wavelets basis existing in Matlab-5 software [ 20 - 24 ] was tested. The criterion used to determine the best wavelet base was the one which optimizes the signal to noise ratio in a broad spectrum of spatial frequencies. The bi-orthogonal wavelets basis [ 25 - 27 ] especially the 3.5 basis in Matlab-5 yielded the best average signal to noise ratio in the range of the spatial frequency (1/100 - 1/8 Å -1 ) relevant to data analysis. Wavelet Projection Matching (WPM) Principle The principle of the wavelet decomposition is to transform the original raw particle image into several components: one low-resolution component called "approximation" [ 21 ], which is mainly used in this method, and the other components called "details" (Fig. 2 ). The approximation component is obtained after applying a bi-orthogonal low-pass wavelet filter in each direction (horizontal and vertical) followed by a sub-sampling of each image by a factor of 2 for each dimension. The details are obtained with the application of a low-pass filter in one direction and a high-pass filter in the other, or a high-pass filter in both directions. The noise is mainly present in the detail components. A higher level of decomposition is obtained by repeating the same filtering operations on the approximation. The wavelet correlation coefficient between two wavelet-transformed images, for a given level, is : Where W 1 to Wp are weights given for each components of the wavelet correlation, p is the number of components of wavelet decomposition. A 1 , A 2 are the approximations. ⊗ denote the correlation between two components images. D 1i , D 2i are the details (Fig. 2 ). This implementation starts first by a wavelet filtering which is performed by thresholding [ 21 , 28 , 29 ] of the details components in order to reduce the noise effects in the correlation matching. Higher weight is given to the approximation component to further reduce the noise effect in the decision. The weights given in this implementation are 0.75 for the approximation and 0.25 for the details. Orientation determination with Hierarchical WPM (HWPM) Initial orientation determination is based on model-based projection matching approach [ 3 ]. The level of wavelet decomposition depends on the dimension of the virus and the sampling rate. For herpes simplex virus type-1 (HSV-1) B-capsid, which has a diameter of 1250 Å with a sampling of 2.1 Å/pixel, a level two of wavelet decomposition (Fig. 2 ) is appropriate for the initial orientation estimate, because of the contrast enhancement and the consideration of computational speed. The method starts by generating the wavelet decomposition at level two for each projection and raw image. In order to have accurate orientation estimation a small angular grid (figure 3 ) to generate projections from the initial model is needed, and this results in a large number of projections. The classical projection matching, which consists of comparing the wavelet-transformed raw images with every projection, is very slow even when using multiple processors on a parallel computer. In order to significantly increase the speed of processing, a hierarchical implementation is performed. This consists of grouping projections into classes of similar orientations [ 30 ]. Fig. 3 shows the classification scheme applied for the icosahedral viruses, only an asymmetric triangle representing the possible orientations for icosahedrally symmetric object [ 4 ] is considered. The choice of the number of classes is optimized to give the best tradeoff between speed and accuracy. The classification gives a uniform distribution of projections into the classes. The next step is to compare each wavelet-transformed raw image with the closest projection to the center of each class, and then rank the classes in terms of wccf (Fig. 4 ). The final step is to compare the raw image with all the projections of the three classes given the highest wccf coefficients. Next, the orientation of the projection yielding the highest wccf will be assigned to the raw image as the correct orientation. The software is written in C++ (a parallel version of the software has been written to run on the SGI Origin-2000 supercomputer).	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555555.xml
543451	Vitamin A deficiency and inflammatory markers among preschool children in the Republic of the Marshall Islands	Background The exclusion of individuals with elevated acute phase proteins has been advocated in order to improve prevalence estimates of vitamin A deficiency in surveys, but it is unclear whether this will lead to sampling bias. The purpose of the study was to determine whether the exclusion of individuals with elevated acute phase proteins is associated with sampling bias and to characterize inflammation in children with night blindness. Methods In a survey in the Republic of the Marshall Islands involving 281 children, aged 1–5 years, serum retinol, C-reactive protein (CRP), and α 1 -acid glycoprotein (AGP) were measured. Results Of 281 children, 24 (8.5%) had night blindness and 165 (58.7%) had serum retinol <0.70 μmol/L. Of 248 children with AGP and CRP measurements, 123 (49.6%) had elevated acute phase proteins (CRP >5 mg/L and/or AGP >1000 mg/L). Among children with and without night blindness, the proportion with serum retinol <0.70 μmol/L was 79.2% and 56.8% ( P = 0.03) and with anemia was 58.3% and 35.7% ( P = 0.029), respectively. The proportion of children with serum retinol <0.70 μmol/L was 52.0% after excluding children with elevated acute phase proteins. Among children with and without elevated acute phase proteins, mean age was 2.8 vs 3.2 years ( P = 0.016), the proportion of boys was 43.1% vs. 54.3% ( P = 0.075), with no hospitalizations in the last year was 11.0% vs 23.6% ( P = 0.024), and with anemia was 43.8% vs 31.7% ( P = 0.05), respectively. Conclusions Exclusion of children with inflammation in this survey of vitamin A deficiency does not improve prevalence estimates for vitamin A deficiency and instead leads to sampling bias for variables such as age, gender, anemia, and hospitalization history.	Vitamin A deficiency is a major cause of morbidity and mortality among preschool children in developing countries [ 1 ]. Vitamin A, or all- trans retinol, is available as preformed vitamin A in foods such as eggs and dairy products and as provitamin A carotenoids in foods such as dark green leafy vegetables, pumpkin, and papaya. Vitamin A is essential for normal immune function, hematopoiesis, growth, and vision [ 1 ]. Among preschool children, risk factors for vitamin A deficiency include age, such as the period that follows weaning when vitamin A intake is low and risk of precipitating infections is high [ 2 , 3 ], recent infections such as diarrheal disease [ 4 ], and low socioeconomic status [ 5 ]. The syndrome of vitamin A deficiency is characterized by increased susceptibility to infections, anemia, and elevated acute phase proteins [ 1 , 6 , 7 ]. The identification of populations at risk for vitamin A deficiency is important in order to target groups that would benefit from interventions to improve vitamin A status. Xerophthalmia, such as night blindness and Bitot spots, is used to determine the prevalence of clinical vitamin A deficiency in populations [ 1 ]. Other surveys that are conducted to determine whether vitamin A deficiency is a public health problem often rely upon the measurement of serum or plasma retinol concentrations [ 8 ]. However, plasma retinol is a negative acute phase reactant that decreases during infections [ 9 ], and it has been thought that the measurement of plasma retinol concentrations may not accurately reflect the vitamin A status of populations with a high prevalence of subclinical infections, as some portion of low retinol concentrations may be due to an acute phase response [ 10 ]. Recently, Thurnham and colleagues proposed that in surveys that rely upon plasma or serum retinol concentrations to estimate the prevalence of vitamin A deficiency, one method to improve the accuracy of the prevalence estimates is to exclude all individuals with elevated acute phase proteins [ 11 ]. The underlying assumption using this method is that subclinical infection is randomly distributed in a population, and that the exclusion of individuals with elevated acute phase proteins will not lead to sampling bias. We hypothesized that exclusion of individuals with elevated acute phase proteins, and selection of only apparently healthy children would lead to lead to sampling bias in a survey of vitamin A deficiency. A secondary hypothesis was that children with night blindness would have elevated acute phase proteins. Vitamin A deficiency has been recognized as a major public health problem in many areas of the South Pacific region [ 12 ], including the Republic of the Marshall Islands [ 13 , 14 ]. In order to address these hypotheses, we characterized serum retinol concentrations, night blindness, and markers of inflammation in a population-based survey among preschool children in the Republic of the Marshall Islands. Subjects and Methods A population-based survey, the Republic of the Marshall Islands Vitamin A Deficiency Study, was conducted between November 1994 and March 1995 to assess the prevalence of vitamin A deficiency among children, aged 1–5, in the Republic of the Marshall Islands. The Republic of the Marshall Islands (population 68,000) is a complex of 28 coral atolls and 5 small (non-atoll) islands, with a total of 1136 islands spreading over a very small land area of 134 km 2 . The sampling strategy for the study was based on the 1988 census of the Republic of the Marshall Islands, which provided data on the average number of children of the target age group within each household, determined by dividing the number of children in a locality by the number of households in the same location. This number was then divided into the number of children to be sampled to obtain the number of households to be visited. Households to be visited were chosen by systematic sampling of every fifth household. When available, the birth dates of the children were ascertained from the children's health cards; otherwise, the birth dates were obtained by asking the parent or guardian. The survey teams consisted of at least one Marshallese-speaking health care worker, a phlebotomist, and a medical doctor. Of the children in the survey, a systematic subsample of every fifth child was obtained in which a standardized questionnaire was used to collect information on various risk factors such as night blindness, demographics, breastfeeding, visits to the hospital, presence of a vegetable garden, and food consumption. Parents were asked if the child currently had night blindness. A history of night blindness has been shown to be a reliable indicator of vitamin A deficiency [ 15 ]. Oral informed consent was obtained from a parent or guardian prior to participation in the survey as considered appropriate by the institutional review board for this setting. Blood samples were obtained by venipuncture. Hemoglobin was measured using a HemoCue instrument (HemoCue, Inc., Mission Viejo, CA). Venous blood samples were collected in red-top serum separator tubes and immediately wrapped in aluminum foil, continuously shielded from light, and stored at 4°C until centrifugation (200 g × 10 min, room temperature). Aliquots of serum were made, kept in liquid nitrogen or at -70°C, and shipped by overnight air express from the Republic of the Marshall Islands to New York City. Retinol remains stable at -70°C for 15 y or more [ 16 ]. Serum retinol was measured using reverse-phase high performance liquid chromatography, as described elsewhere [ 13 ]. Serum C-reactive protein (CRP) was determined using a commercial enzyme-linked immunosorbent assay kit (Hemagen Diagnostics, Waltham, MA). Serum α 1 -acid glycoprotein (AGP) was measured using radial immunodiffusion assay (Bindarid, The Binding Site, Birmingham, UK). Interleukin-6 (IL-6) was analyzed using a high sensitivity commercial enzyme-linked immunosorbent assay (R&D systems, Inc, Minneapolis, MN). Serum retinol, CRP, AGP, and IL-6 values were available for 281, 250, 248, and 176 of 281 pre-school children, respectively. Missing values were due to inadequate volume of sera. Controls provided by the manufacturer were used to measure intra- and inter-assay coefficients of variation (CV) in laboratory analysis. For serum retinol, the within-assay and between-assay CVs were 3% and 8%, respectively. For serum CRP, AGP, and IL-6, the within-assay and between-assay CVs were 4.1 and 5.0, 3.9 and 1.6, and 3.2 and 2.9, respectively. The study protocol was approved by the institutional review board of the Pacific Health Research Institute of Hawaii and the Ministry of Health and Environment of the Republic of the Marshall Islands. The study was conducted in accordance with the Helsinki Declaration. Groups were compared using Student's t -test or analysis of variance for continuous variables where appropriate, and categorical variables were compared using χ 2 or exact tests. Vitamin A deficiency was defined as serum retinol <0.70 μmol/L [ 1 ]. Inflammation was defined as CRP >5 mg/L and/or AGP >1000 mg/L [ 6 , 7 , 11 ]. Anemia was defined as hemoglobin <110 g/L [ 14 ]. Nonparametric Wilcoxon rank sum test was used on continuous variables in which the distribution was not normal. The proportion of children with serum retinol <0.70 μmol/L was "adjusted" using two methods: (1) excluding all children with inflammation and then presenting the proportion of children who had serum retinol <0.70 μmol/L for the remaining children without inflammation, and (2) calculating geometric mean serum retinol concentrations in children with and without inflammation, multiplying the serum retinol concentrations in the group with inflammation by a constant that makes the geometric mean serum retinol in the group with inflammation equivalent to the group without inflammation, and then reporting the proportion of children who had serum retinol <0.70 μmol/L [ 11 ]. One sample test of proportions was used to compare proportions of children with serum retinol <0.70 μmol/L between "adjusted" and unadjusted groups. Spearman correlation was used to examine correlation between AGP, CRP, IL-6, retinol, and hemoglobin. The level of significance used in this study was P < 0.05. Results A total of 919 Marshallese children, age 1–5, participated in the survey, and the systematic subsample of every fifth child involved 281 children from the following atolls (n): Ailuk (16), Arno (26), Enejelar (3), Enewetak (14), Kwajalein (73), Majuro (77), Namu (32), Utrik (25), and Wotje (15). There were 141 boys and 140 girls, and the mean (±standard deviation) age of the children was 2.93 ± 1.38 years. Due to limited sample volume, serum CRP, AGP, and IL-6 were not measured in 31, 33, and 105 pre-school children, respectively. Of the 281 children, there were 24 (8.5%) with night blindness and 165 (58.7%) with serum retinol <0.70 μmol/L. Of 248 children who had both AGP and CRP measured, there were 123 (49.6)% with inflammation (CRP >5 mg/L and/or AGP >1000 mg/L). The characteristics of children with and without inflammation are shown in Table 1 . Children with inflammation were older ( P = 0.016), were more likely to have been hospitalized in the last year ( P = 0.024), and were more likely to be anemic ( P = 0.05) compared with children without inflammation. Mean retinol concentration was lower ( P = 0.0005) and the proportion of children with serum retinol <0.70 μmol/L was higher ( P = 0.013) among children with inflammation compared to children without inflammation. The findings suggest that children with inflammation were more likely to be girls ( P = 0.075) compared to children without inflammation. There were no significant differences in the proportion with night blindness, the mean number of people living in the house with the child, current breastfeeding status, and the presence of a vegetable garden at home between children with and without inflammation. Table 1 Relationship between inflammation and demographic characteristics of children, age 1–5 years, in the Republic of the Marshall Islands Inflammation 1 Characteristic 2 No ( n = 125) Yes ( n = 123) P Mean age (years) 3.2 (3.0, 3.4) 2.8 (2.5, 3.0) 0.016 Sex (% male) 54.3 43.1 0.075 Night blind (%) 9.5 8.9 0.83 Hospitalized in last year (%) 0 23.6 11.0 0.024 1–2 44.7 43.2 3–4 18.7 31.4 5+ 13.0 14.4 Mean number of people living in house 11.4 (10.1, 12.7) 11.3 (10.1, 12.4) 0.62 Currently breastfeeding (%) 14.3 18.9 0.33 Presence of a vegetable garden at home (%) 32.2 32.8 0.93 Mean hemoglobin (g/L) 119 (108, 112) 109 (106, 110) 0.19 Hemoglobin <110 g/L (%) 31.7 43.8 0.05 Mean retinol (μmol/L) 0.66 (0.61, 0.72) 0.52 (0.47, 0.58) 0.0005 Retinol <0.70 μmol/L (%) 52.0 67.5 0.013 1 Defined as CRP >5 mg/L and/or AGP >1000 mg/L. 2 Geometric mean; 95% CI in parentheses. Of the 281 children 58.7% had serum retinol <0.70 μmol/L. In contrast, if children with inflammation are excluded as proposed by Thurnham and colleagues [ 11 ], the proportion with serum retinol <0.70 μmol/L is 52.0% (Table 1 ). Alternatively, it has been proposed that serum retinol concentrations are "adjusted using acute phase proteins" based upon retinol values among children with and without inflammation [ 11 ]. When this alternative method is used, the proportion of children with serum retinol <0.70 μmol/L is 47% (95% C.I. 0.43–0.60), compared with the unadjusted proportion of 58.7% ( P = 0.08). The characteristics of children with and without night blindness are shown in Table 2 . Children with night blindness were slightly older ( P = 0.03), had lower serum retinol ( P = 0.007), were more likely to have serum retinol <0.70 μmol/L ( P = 0.03), and to be anemic ( P = 0.029) compared to children without night blindness. The findings were suggestive that there may be a higher proportion of boys than girls who are night blind ( P = 0.09), and with lower mean hemoglobin concentrations ( P = 0.09). There were no significant differences in geometric mean AGP, CRP, IL-6, or inflammation between children with and without night blindness. Table 2 Characteristics of children, aged 1–5 years, with and without night blindness in the Republic of the Marshall Islands Night blindness Characteristic 1 Yes ( n = 24) No ( n = 257) P Mean age (years) 3.5 (3.0, 4.0) 2.9 (2.7, 3.1) 0.03 Sex (% male) 66.7 48.6 0.09 Geometric mean AGP (mg/L) 925 (819, 1044) 924 (881, 970) 0.99 AGP >1000 mg/L (%) 43.5 45.8 0.83 Geometric mean CRP (mg/L) 1.3 (0.2, 11.8) 1.7 (0.1, 324.3) 0.36 CRP >5 mg/L (%) 8.7 23.1 0.11 With inflammation (%) 2 47.8 49.3 0.89 Geometric mean IL-6 (pg/mL) 3.9 (0.9, 14.1) 3.9 (0.7, 16.2) 0.96 Mean retinol (μmol/L) 0.40 (0.04, 1.00) 0.62 (0.02, 1.90) 0.007 Retinol <0.70 μmol/L (%) 79.2 56.8 0.03 Mean hemoglobin (g/L) 106 (101, 110) 109 (108, 110) 0.09 Hemoglobin <110 g/L (%) 58.3 35.7 0.029 1 For continuous variables (95% CI). 2 Inflammation defined as AGP >1000 mg/L and/or CRP >5 mg/L. Spearman correlations of serum AGP, CRP, IL-6, retinol, and hemoglobin are shown in Table 3 . Serum AGP was positively correlated with serum IL-6 and CRP ( P < 0.0001 for both). Serum AGP and CRP individually were inversely correlated with serum hemoglobin ( P < 0.0001) and retinol ( P = 0.0004). IL-6 was positively associated with CRP ( P < 0.0001), and inversely associated with retinol ( P < 0.0001). Hemoglobin had low correlation with retinol that was of borderline significance ( P = 0.09) and low inverse correlation with IL-6 that was also of borderline significance ( P = 0.08). Table 3 Spearman correlation between AGP, CRP, IL-6, retinol, and hemoglobin in preschool children in the Republic of the Marshall Islands Hemoglobin Retinol IL-6 CRP AGP -0.19 -0.25 0.48 0.60 P = 0.0026 P < 0.0001 P < 0.0001 P < 0.0001 CRP -0.13 -0.22 0.44 P < 0.0001 P = 0.0004 P < 0.0001 IL-6 -0.12 -0.28 P = 0.08 P < 0.0001 Retinol 0.10 P = 0.09 Discussion In this population of children from the Republic of the Marshall Islands, more than half had serum retinol concentrations <0.70 μmol/L. According to criteria of the World Health Organization, vitamin A deficiency is considered a public health problem if more than 15% of the population has serum or plasma retinol concentrations <0.70 μmol/L [ 17 ]. Vitamin A deficiency is also considered a public health problem if >1% of children less than six years old have night blindness [ 1 ], and in this survey, the prevalence of night blindness was more than eight times higher than this criterion. Thus, vitamin A deficiency was certainly a public health problem among children, aged 1–5 years, in the Republic of the Marshall Islands. Since the time of this survey, the Republic of the Marshall Islands has implemented a countrywide vitamin A capsule distribution program. Nearly half of the children in this study had inflammation, as indicated by elevated AGP and/or CRP. These findings suggest that the prevalence of subclinical infection is high in this population that has a high prevalence of vitamin A deficiency. These findings are consistent with the concept that the syndrome of vitamin A deficiency is associated with depressed immunity and increased infections [ 6 , 7 ]. Children with subclinical vitamin A deficiency may have pathological alterations in T and B cell function and mucosal immunity that make them more susceptible to subclinical infections, such as diarrheal disease [ 18 ]. Although the exclusion of individuals with elevated acute phase proteins has been advocated to improve prevalence of vitamin A deficiency in surveys that rely upon plasma or serum retinol concentrations [ 11 ], this study shows that exclusion of these individuals leads to sampling bias. The remaining subjects in the sample without inflammation are different from those excluded from the sample of the study, as there was selection bias in regard to age, gender, anemia, and morbidity history. Thus, by excluding those with inflammation, the prevalence estimates of vitamin A deficiency are based upon a biased sample that may be healthier. Subclinical infections are more likely to occur among malnourished children and among children from poorer families [ 19 , 20 ]. Studies among adults suggest that elevated acute phase proteins are more common among those with lower socioeconomic status [ 21 , 22 ]. The present study is limited in that data on maternal education, socioeconomic status, and other demographic indicators was not collected. Similar analyses could be conducted with other large existing data sets to corroborate and characterize the extent of sampling bias in vitamin A surveys when individuals with elevated acute phase proteins are excluded. The alternative method of having serum retinol concentrations "adjusted using acute phase proteins" [ 11 ] involves the same problem of sampling bias, as the group without inflammation that is used for "adjusting" the serum retinol concentrations of the group with inflammation is different as discussed above. Children with night blindness had significantly lower serum retinol concentrations compared with children without night blindness, a finding that is consistent with previous studies [ 6 , 7 , 15 ]. The ability to see at night depends on the visual pigment, rhodopsin, in rod photoreceptors of the retina. Synthesis of rhodopsin depends in part upon the availability of circulating retinol. The prevalence of anemia was higher among children with night blindness than children without night blindness. Vitamin A deficiency is associated with anemia, and there may be several mechanisms by which vitamin A deficiency could cause anemia, including impairment of iron metabolism, and immune dysfunction and associated anemia of infection [ 23 ]. In the present study, children with night blindness were not more likely to have elevated acute phase proteins than children without night blindness. However, the number of children with night blindness in this study was small, and the study had limited statistical power to address this secondary hypothesis. Other studies among pregnant women in Nepal [ 6 ] and preschool children in Indonesia [ 7 ] show that individuals with night blindness are more likely to have elevated acute phase proteins. The relationship between IL-6 and elevated acute phase proteins has not well been characterized in epidemiological studies of vitamin A deficiency. IL-6 is a proinflammatory cytokine that plays a role in the upregulation of CRP [ 24 ] and AGP [ 25 ]. In the present study, IL-6 concentrations had a moderate correlation with both CRP and AGP. CRP is one component of a first line of innate host defense against infectious diseases [ 24 ]. The biological function of AGP has not been well characterized [ 25 ]. The inverse correlation of hemoglobin with AGP, CRP, and IL-6 suggests the role of proinflammatory cytokines and inflammation in the anemia of chronic infection [ 26 ]. In summary, the method of excluding individuals with elevated acute phase proteins from this survey of vitamin A deficiency resulted in sampling bias and a prevalence estimate that was based upon about half of the original sample. The remaining sample was healthier, older, less likely to have been hospitalized, and with a higher proportion of boys and a lower proportion with anemia. The method of excluding individuals with elevated acute phase proteins may lead to underestimation of the prevalence of vitamin A deficiency. Vitamin A deficiency remains a major problem in many developing countries worldwide, and further studies are needed to develop unbiased epidemiological methods for the estimation of the prevalence of vitamin A deficiency in populations.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC543451.xml
548504	Factors associated with reporting multiple causes of death	Background There is analytical potential for multiple cause of death data collected from death certificates. This study examines relationships of multiple causes of death as a function of factors available on the death certificate (demographics of decedent, place of death, type of certifier, disposal method, whether an autopsy was performed, and year of death). Methods Data from 326,332 Minnesota death certificates from 1990–1998 are examined. Underlying and non-underlying causes of death are examined (based on record axis codes) as well as demographic and death-related covariates. Associations between covariates and prevalence of multiple causes of death and conditional probability of underlying compared to non-underlying causes of death are examined. The occurrence of ischemic heart disease or diabetes as underlying causes are specifically examined. Results Both the probability of multiple causes of death and the proportion of underlying cause compared to non-underlying cause of death are associated with demographic characteristics of the deceased and other non-medical conditions related to filing death certificate such as place of death. Conclusions Multiple cause of death data provide a potentially useful way of looking for inaccuracies in reporting of causes of death. Differences across demographics in the proportion of time a cause is selected as underlying compared to non-underlying exist and can potentially provide useful information about the overall impact of causes of death in different populations.	Background In their 1986 paper Israel, Rosenberg, and Curtin [ 1 ] gave a sort of rallying call for researchers to consider the analytical potential for multiple cause of death data collected by the United States National Center for Health Statistics (NCHS). Beginning with the implementation of the Eighth revision of the ICD in 1968, the NCHS developed and employed several computer systems to automatically select the underlying cause for each death certificate and to produce multiple cause of death data [ 2 ]. The resulting multiple cause of death datasets by year are made publically available through the NCHS website. Acknowledgment of the potential for multiple cause of death data analysis is increasing in other countries as well [ 3 , 4 ]. For example, the Australian Bureau of statistics point out that using multiple cause of death data allows researchers to: more comprehensively understanding and track death due to chronic disease which do not often appear as the underlying cause of death (e.g. Alzheimer's, diabetes, pneumonia), to provide better documentation on multi-morbid associations and the strength of associations between conditions which led to death (for example by examining the frequency of associations between diseases such as diabetes and ischaemic heart disease), and to assist in identifying problems with the process of recording and coding cause of death information [ 4 ]. Multiple cause of death data has been used to look at trends in certain diseases, e.g. HIV [ 5 , 6 ] and lung disease [ 7 ], but despite its availability, surprisingly few studies have looked at it broadly. Indeed there is no annual standard summary tabulation report of the multiple cause of death data put out by NCHS. This may be due in part to the overwhelming amount of information that arises when combinations of causes of death are considered. There are an enormous number of complex combinations which could be summarized and perhaps it is not clear what tables may be of general interest. The purpose of this article is to examine some straightforward relationships of multiple causes of death as a function of factors available on the death certificate (demographics of decedent, place of death, type of certifier, disposal method, whether an autopsy was performed, and year of death). Using all death certificates issued by the state of Minnesota between 1990 and 1998 (326,332 deaths), the current study documents the relationship between these factors and the associated frequency of reporting of multiple causes of death as well as the associated frequency that a cause is considered underlying (after data processing) given that it is mentioned on the death certificate. The implication being that differences found are either due to actual differences in causes of death in these groups or due to systematic biases in the reporting of causes of death, or a combination of both. The study will not be able to discern which is the cause but hopes to contribute at the very least by providing an example of the potential relationships which can be examined with the rich multiple cause of death data. Methods Data source The data used are from the Minnesota Department of Health Mortality Database and include entries from 326,332 individual death certificates, which represent all deaths occurring in Minnesota during the period of 1990–1998. Record axis codes (those codes which have been completely data processed) are used for all analyses in this paper rather than entity axis codes. A brief description of the entity and record axis coding is given here. The translation of causes of death listed on the death certificate (see Figures 1 and 2 for the actual certificate) to the codes used for statistical analysis goes through many steps. As seen in Figures 1 and 2 , the medical information which focuses on the sequence of medical conditions that resulted in death is provided in a two-part format. Part I is for the conditions which directly lead to death, and Part II is for other conditions which contribute to death but are not directly related to the immediate cause of death [ 1 ]. The underlying cause of death is defined as the "(a) the disease or injury which initiated the train of events leading directly to death, or (b) the circumstances of the accident or violence which produced the fatal injury" [ 8 ]. The entity axis codes represent what is actually written on the death certificate by the certifier expressed in terms of ICD codes including an indicator of which line the code came from and which position on the line it came from (if more than one code was listed per line). While the conditions listed in Part I should form a causal sequence initiated by the underlying cause listed on the lowest line, errors in properly completing the form occur regularly and a reselection of the underlying cause of death is done nationally 30–40% of the time. The decision to reselect an underlying cause other than that listed on the lowest used line in Part I is governed by a set of rules developed by WHO as part of the periodic revision of the International Classification of Disease [ 9 ] and is incorporated, along with a complex set of decision tables, into the Automated Classification of Medical Entities (ACME) software. The record axis codes represent a further processing of the entity axis codes to be consistent with the underlying cause data and more amenable to statistical tabulation and analysis. The record axis codes distinguish the ICD code selected as the underlying cause of death and lists all additional causes of death mentioned but does not distinguish them in terms of their ordering or original location on the death certificate. For more detail on entity and record access codes, see [ 10 ]. Figure 1 US standard certificate of death. Line 27 Part I and Part II are where the causes of death are listed. Figure 2 This figure displays the backside of the certificate of death . Details are given for filling out specific lines. Using the record axis codes, we have for each death record: one underlying cause of death and up to 14 non-underlying causes of death (with no distinction of importance given amongst the non-underlying). When we refer to a cause of death and do not want to distinguish if it is underlying or non-underlying we will refer to it as a "mentioned" cause of death. In addition to listing one underlying and up to 14 non-underlying causes of death, each death certificate also contains information about the demographics of the deceased, including age, gender, race, marital status, and educational attainment. Also, other conditions related to the death are recorded – place and time of death, who completed the death certificate, if autopsy has been performed and type of body disposal. Minnesota Laws and guidelines govern the process for who and how a death is certified under different circumstances in Minnesota. For example when an unattended death occurs (e.g. at a persons residence) a medical examiner's investigator must arrive at the scene. The medical examiner will contact the last attending physician asking about past medical history of the decedent and most likely cause of death. When an attending physician has seen the decedent within 90 days and the death is natural, jurisdiction is usually given to the physician to certify the death. Sudden or unexpected deaths due in part to any factor other than natural disease must be referred to the medical examiner's office. Autopsies are performed at the discretion of the medical examiner but can also be performed for any death at the request of the immediate family. The underlying and non-underlying causes of death derived from the death certificate, in this study, are coded according to the 9 th revision of International Classification of Disease (ICD). The specific ICD9 codes are grouped into standard reporting of cause of death categories resulting in a total of 107 different causes of death. In this study, individuals are dichotomized as having multiple causes of death (i.e. at least one non-underlying cause) or not having multiple causes of death (i.e. only having an underlying). In addition, because heart disease is the leading cause of death and diabetes is a good example of a disease which often shows up as a non-underlying cause of death, this research investigates two sub -populations: 1) Individuals that have ischemic heart disease (ICD-9: 410 – 414) mentioned as a cause of death (n = 79,833), and 2) Individuals that have diabetes mellitus (ICD-9: 250) mentioned as a cause of death (n = 27,181). For both sub-populations, a dichotomous variable is created to indicate whether the mentioned disease is coded as the underlying cause of death or not. Data analysis Descriptive statistics including total numbers and proportion of all deaths (n = 326,332) in each of the covariate categories are reported as well as proportions of people within each covariate category who have multiple causes of death. In order to examine the association between each covariate and the dichotomous outcome of multiple causes of death, logistic regression is used to mutually adjust each factor for the others. 95% confidence intervals of odds ratios are reported. Trends in multiple cause of death reporting across time are investigated graphically. Similarly, descriptive statistics including total numbers and proportions will be presented for the two sub-populations with ischemic heart disease (n = 79,833) or diabetes mellitus (n = 27,181) mentioned either as underlying or non-underlying cause of death. Logistic regression is used and 95% confidence intervals are reported to examine factors that are associated with each of these diseases being reported as underlying cause of death rather than non-underlying. Results Overall, 68.9% of the 326,332 deaths from 1990–1998 had at least one non-underlying cause of death in addition to the underlying cause (i.e. have multiple causes). There was a noticeable decreasing trend of reporting multiple causes of death over the 9 year period from 1990 to 1998 with 74.0% in 1990 consistently dropping down to 64.8% in 1996 and remaining around 66% until 1998. Table 1 presents the marginal percentage of individuals in each demographic and death related category as well as the proportions and adjusted odds ratios of having multiple causes of death by each of the covariates. The youngest (<25) and oldest (85+) age groups had the lowest and highest percent of multiple causes of death (61.7% and 71.8%, respectively). Interestingly, the age group from 45–64 did not have higher odds of having multiple causes than the young (<25) group. The percentage of men with multiple causes of death reported was slightly higher (1%) than that of women. Individuals over 25 years old with less education had a higher percentage of multiple causes of death (71.3%) compared to those with higher education (66.6%). The most pronounced difference with respect to demographics was found in race categories, with Native American having the highest percentage of multiple cause of death (74.5%), compared to 68.9% of white. Table 1 Percent of all deaths (n = 326,332) by each covariate. Probability of reporting multiple causes of death given covariate, marginal percent by category, and adjusted odds ratios of reporting multiple causes of death given covariates. % of death by categories % with multiple COD by categories Odds Ratio (95% CI) 1 DEMOGRAPHIC Age 0–24 3.0 61.7 1 25–44 4.6 69.0 1.38(1.31–1.45) 45–64 13.3 62.1 1.01(0.97,1.06) 65–84 47.8 69.4 1.41(1.35,1.47) 85+ 31.3 71.8 1.58(1.51,1.65) Sex Female 50.3 68.6 1 Male 49.7 69.3 1.11(1.09,1.13) Race White 96.4 68.9 1 Black 1.7 67.5 1.15(1.078,1.22) Native American 0.9 74.5 1.54(1.41,1.69) Asian 0.5 65.0 1.04(0.94,1.16) Hispanic 0.5 66.6 1.13(1.01,1.26) Education 2 High school 32.4 68.1 1 Below High School 44.3 71.3 1.03(1.01,1.05) Above High School 23.3 66.6 0.94(0.92,0.96) Marital Status Married 41.0 67.4 1 Single 12.7 68.0 1.08(1.05,1.11) Widowed 38.5 71.1 1.09(1.07,1.11) Divorced 7.8 68.2 1.12(1.08,1.15) DEATH RELATED Autopsy No 77.8 69.0 1 Yes 9.4 74.4 1.57(1.52,1.62) Unspecified 12.8 64.7 0.85(0.83,0.87) Certifier Physician 82.0 69.2 1 Coroner 12.6 70.2 1.23(1.19,1.26) Osteopath 1.4 75.3 1.27(1.183,1.37) Other/Unknown 4.1 57.8 0.61(0.59,0.63) Disposal method Burial 76.3 70.1 1 Donation 0.3 69.7 1.02(0.89,1.17) Removal 1.5 47.1 0.35(0.33,0.37) Cremation 20.5 66.8 0.94(0.93,0.96) Unknown 1.4 Death place Hospital Inpatient 35.2 73.4 1 Residential 19.1 58.7 0.48(0.47,0.49) Nursing home 35.5 70.6 0.77(0.76,0.79) ER 5.5 65.5 0.64(0.61,0.66) Unknown 4.7 1 Odds ratios are odds of having multiple causes of death given particular category as compared to odds in reference category. Odds ratios are obtained from logistic regression and are mutually adjusted for all other variables. 2 Education is listed only for population where age>25. For places of death, hospital in-patient (73.4%) and nursing home (70.6%) had the highest probability of reporting multiple causes of death, and residence had the lowest percentage (58.7%) (Table 1 ). Between different types of body disposal methods, "removal", which refers to moving the body outside of the US, had the lowest percentage (47.1%) of reporting multiple cause of death. For deaths that had autopsy performed, there was an increased odds of 1.57 that multiple causes of death would be reported. In terms of different types of medical examiners, the difference was less than 1% (69.2% vs. 70.1%) marginally between physician and coroners, the two most frequently seen types of examiners, but examining this difference across time (Figure 3 ) found an interesting interaction effect in the trend. The physicians showed a decrease in multiple cause of death reporting while the coroners stayed constant or slightly increased over the decade. Table 2 provides reference for the 25 leading underlying causes of death and leading mentioned causes of death in this dataset. It also lists the leading causes of death which occur on death certificates only reporting an underlying cause of death with no non-underlying. The top four causes based on only one cause certificates are the same as the overall top four causes. But it is interesting that the fifth leading cause in this category is "Symptoms and ill-defined conditions" which typically are assigned as the underlying cause only if the sole cause listed. Figure 3 Interaction between certifier and year. Table 2 Based on Minnesota death records (n = 326,332) from 1990–1998. Top 25 causes of death ranked by underlying and any mention cause of death. Top 25 causes of death for only those deaths where only one cause was listed (i.e. n = 101,423 deaths). ranked by underlying cause number of deaths with underlying cause % all deaths rankeded by any mention number of deaths with cause mentioned % all deaths ranked by cause when only one cause mentioned number of deaths with cause as only one cause % of deaths with only one cause mentioned 1 Ischemic Heart Disease 61540 0.1886 Other diseases of the Heart 88502 0.2712 Ischemic Heart Disease 13753 0.1356 2 Cerebrovascular Disease 26413 0.0809 Ischemic Heart Disease 79833 0.2446 Other diseases of the Heart 8862 0.0874 3 Other diseases of the Heart 22566 0.0692 Cerebrovascular Disease 43885 0.1345 MN of Trachea, Bronchus & Lung 8157 0.0804 4 MN of Trachea, Bronchus & Lung 18476 0.0566 Symptoms & ill-defined conditions 42196 0.1293 Cerebrovascular Disease 7602 0.0750 5 Pneumonia – except newborn 12465 0.0382 Other mental disorders 38565 0.1182 Symptoms & ill-defined conditions 6459 0.0637 6 Other COPD 11114 0.0341 Pneumonia – except newborn 31312 0.0960 Other mental disorders 3494 0.0345 7 Other mental disorders 10120 0.0310 Diabetes mellitus 27181 0.0833 MN of Breast 3360 0.0331 8 Diabetes mellitus 7959 0.0244 Hypertension without heart disease 26036 0.0798 MN of Intestine, not rectum 3218 0.0317 9 MN of Intestine, not rectum 7388 0.0226 Other COPD 25872 0.0793 Pneumonia – except newborn 3131 0.0309 10 Diseases of the arteries, veins & pulmonary circulation 7181 0.0220 MN of Trachea, Bronchus & Lung 19896 0.0610 MN of Other & unspecified sites 3042 0.0300 11 MN of Breast 6646 0.0204 MN of Other & unspecified sites 19386 0.0594 MN of Prostate 2380 0.0235 12 Symptoms & ill-defined conditions 6488 0.0199 Diseases of the arteries, veins & pulmonary circulation 19180 0.0588 Other COPD 2332 0.0230 13 Transportation accidents – Motor Vehicle 5809 0.0178 Other diseases of the digestive system 17943 0.0550 MN of Pancreas 2313 0.0228 14 Other diseases of the digestive system 5777 0.0177 Pneumoconiosis & other resp. Diseases 17728 0.0543 Diseases of arteries, veins & pulmonary circulation 1758 0.0173 15 Other disease of the Nervous System and Sense Organs 5714 0.0175 Other disease of the Nervous System & Sense Organs 16799 0.0515 Other Neoplasms of lymphatic & Hematopoietic tissue 1701 0.0168 16 MN of Prostate 5669 0.0174 Chronic and Unspec. Nephritis & renal failure & renal sclerosis 16065 0.0492 Alzeimer Disease 1536 0.0151 17 MN of Other & unspecified sites 5504 0.0169 Arteriosclerosis 13879 0.0425 Other disease of the Nervous System & Sense Organs 1477 0.0146 18 Suicide 4435 0.0136 Transportation accidents – Motor Vehicle 10367 0.0318 Residual, Undefined 1300 0.0128 19 MN of Pancreas 4136 0.0127 Medical complications & misadventures 10299 0.0316 MN of Brain, other nervous 1292 0.0127 20 Residual, Undefined 4135 0.0127 MN of Intestine, not rectum 9116 0.0279 Leukemia, & Aleukemia 1271 0.0125 21 Pneumoconiosis & other resp. Diseases 4028 0.0123 Septicemia 8790 0.0269 Perinatal conditions 1206 0.0119 22 Accidental falls 4010 0.0123 Suicide 8740 0.0268 MN of Ovary, Fallopian tube, Broad ligament 1124 0.0111 23 Alzeimer Disease 3757 0.0115 MN of Breast 8704 0.0267 Other diseases of the digestive system 1079 0.0106 24 Other Neoplasms of lymphatic & Hematopoietic tissue 3584 0.0110 Other genito-urinary disease 8382 0.0257 Suicide 990 0.0098 25 Leukemia, & Aleukemia 3440 0.0105 MN of Prostate 8369 0.0256 MN of Kidney 955 0.0094 The results presented so far explored how covariates may be correlated with multiple causes of death being reported. The following results pertain to the conditional probability that a particular cause of death (ischemic hear disease or diabetes) is selected as underlying given that it is mentioned. Results for the subpopulations with ischemic heart disease or diabetes mentioned are shown in Table 3 and Table 4 , respectively. Table 3 Population with Ischemic Heart Disease mentioned on death certificate (N = 79833). Marginal percent by category, conditional percent with ischemic heart disease as underlying given that it is mentioned by category and odds ratios of ischemic heart disease being reported as underlying cause when it is mentioned given covariates. % of deaths by category % with heart disease as underlying Odds Ratio (95% CI) 1 DEMOGRAPHIC Age 0–44 1.7 80.0 1 45–64 12.4 81.8 1.2 (1.03,1.36) 65–84 53.0 76.1 0.84 (0.74,0.95) 85+ 32.8 76.8 0.87 (0.77,1.00) Sex Female 45.2 76.3 1 Male 54.9 77.8 0.95(0.91,0.99) Race White 97.9 77.2 1 Black 0.8 71.8 0.78(0.66,0.94) Native American 0.7 69.7 0.55(0.45,0.66) Asian 0.3 77.2 1.12(0.81,1.56) Hispanic 0.3 72.2 0.79(0.59,1.06) Education 2 High school 29.1 76.3 1 Below High School 49.1 77.9 1.07(1.03,1.11) Above High School 21.8 76.5 1.02(0.97,1.07) Marital Married 44.9 77.5 1 Single 8.2 79.7 1.25(1.16,1.33) Widowed 40.0 75.9 1.04(0.99,1.09) Divorced 6.9 77.9 1.13(1.05,1.22) DEATH RELATED Autopsy NO 77.8 76.9 1 Yes 10.5 78.0 0.91(0.86,0.97) Unspecified 11.7 77.8 1.12(1.06,1.18) Certifier Physician 79.5 75.5 1 Coroner 15.2 84.3 1.34(1.25,1.42) Osteopath 1.5 80.9 1.25(1.07,1.46) Other/Unknown 3.8 81.0 1.23(1.13,1.35) Disposal method Burial 78.9 77.3 1 Donation 0.3 75.3 0.74(0.56,0.98) Removal 1.6 85.3 1.74(1.48,2.05) Cremation 18.0 66.8 0.95(0.91,0.99) Unknown 1.2 Death place Hospital Inpatient 36.1 72.9 1 Residential 20.1 90.5 1.83(1.74,1.93) Nursing home 27.9 71.2 0.83(0.79,0.87) ER 11.4 83.2 1.67(1.39,1.97) Unknown 4.5 1 Odds ratios are odds of having ischemic heart disease as underlying rather than contributing given particular category as compared to odds in reference category. Odds ratios are obtained from logistic regression and are mutually adjusted for all other variables. 2 Education is listed only for population where age>25. Table 4 Population with Diabetes mentioned on death certificate (N = 27181). Marginal percent by category, conditional percent with diabetes as underlying given that it is mentioned and odds ratios of diabetes being reported as underlying cause when it is mentioned given covariates. % of deaths by category % with diabetes as underlying given that it was mentioned Odds Ratio (95% CI) 1 DEMOGRAPHIC Age 0–44 2.3 51.7 1 45–64 13.2 34.5 0.51(0.43,0.60) 65–84 59.5 27.7 0.37(0.32,0.43) 85+ 24.9 28.4 0.39(0.33,0.45) Sex Female 51.7 30.2 1 Male 48.3 28.3 0.92(0.86,0.97) Race White 95.4 28.9 1 Black 2.0 36.5 1.19(0.98,1.43) Native American 1.5 41.0 1.42(1.136,1.76) Asian 0.5 27.2 0.88(0.59,1.30) Hispanic 0.7 36.6 1.33(0.978,1.82) Education 2 High school 31.2 30.1 1 Below High School 47.8 28.0 0.99(0.93,1.06) Above High School 21.0 30.8 1.03(0.96,1.11) Marital Married 43.6 27.8 1 Single 8.5 34.3 1.25(1.13,1.38) Widowed 40.2 29.0 1.09(1.01,1.16) Divorced 7.7 33.5 1.11(0.99,1.23) DEATH RELATED Autopsy No 83.3 28.3 1 Yes 5.5 22.8 0.70(0.62,0.80) Unspecified 11.3 39.4 1.49(1.37,1.62) Certifier Physician 85.0 29.7 1 Coroner 10.0 24.2 0.69(0.62,0.77) Osteopath 1.5 30.4 1.03(0.81,1.29) Other/Unknown 3.5 34.3 1.23(1.08,1.42) Disposal method Burial 79.2 28.6 1 Donation 0.3 25.0 0.84(0.51,1.39) Removal 1.1 37.0 1.46(1.143,1.87) Cremation 18.3 31.4 1.06(0.99,1.14) Unknown 1.1 Death place Hospital Inpatient 34.8 25.2 1 Residential 18.8 30.9 1.29(1.19,1.41) Nursing home 37.4 32.7 1.57(1.47,1.68) ER 6.4 28.3 1.15(1.02,1.29) Unknown 2.6 1 Odds ratios are odds of having diabetes as underlying rather than contributing given particular category as compared to odds in reference category. Odds ratios are obtained from logistic regression and are mutually adjusted for all other variables. 2 Education is listed only for population where age>25. Table 3 gives the odds ratio of ischemic heart disease being selected as underlying cause of death when it was mentioned as a cause, given the covariates main effect. Overall 77.1% of the time that heart disease was mentioned as a cause, it was selected as the underlying cause of death. The 45–65 year age group had the highest probability of heart disease being codes as underlying when it was mentioned (81.8%). Males had a slightly lower probability than females to have heart disease as underlying cause of death when it was mentioned on the death certificate. Furthermore, Blacks and Native Americans were less likely to have heart disease coded as underlying cause of death when it was present on the certificate as compared to Whites. Individuals that had an autopsy performed were less likely (0.91 odds ratio) to have ischemic heart disease selected as underlying when it was mentioned. If a physician is the death certifier, the probability of selecting heart disease as underlying cause of death is relatively the lowest when compared to coroner, osteopath and other and unknown certifiers. Amongst body disposal methods, the probability for heart disease to be reported as underlying cause was the lowest if bodies were donated (OR = .7 with "burial" as baseline category), and highest if bodies were removed (OR = 1.7). Finally, for those individuals who had heart disease mentioned on their death certificate, patients who died at a residence (not a nursing home) were most likely to have ischemic heart disease selected as the underlying cause of death (90.5% or an OR= 1.8 compare to hospital in-patient). Unlike Ischemic heart disease, only 29.3% of deaths with diabetes mentioned on the certificate had it selected as the underlying cause of death. While only 2.3% of deaths with diabetes mentioned occurred in the youngest age group (0–44 years), (Table 4 ) this group has a much larger probability of having diabetes be the underlying cause compared to non-underlying (51.7% reported as underlying). Men were less likely to have diabetes selected as underlying when it is mentioned on the certificate than women. Blacks and Native Americans both have significantly higher odds (OR = 1.2 and 1.4, respectively) of diabetes being the underlying cause given that it was mentioned as compared to Whites. The role of autopsy is that it was less likely diabetes was reported as underlying (OR = 0.7) when one was performed than if one was not. Moreover, if a coroner was the death certifier, diabetes was less likely to be reported as underlying. An increase in the reporting of diabetes as underlying was found for deaths that were removed. Finally, deaths occurring outside of the hospital inpatient setting all show increased odds of diabetes being selected as the underlying cause of death when it has been mentioned. We also examined what other leading causes of death showed up as underlying when ischemic heart disease or diabetes was mentioned on the certificate. As mentioned above, 77.1% of the individual with ischemic heart disease mentioned on their death certificate had it reported as the underlying. The second most common underlying cause of death when ischemic heart disease was mentioned was, in fact, diabetes (3.4% of the time underlying), followed by cerebrovascular disease (2.7% of the time underlying), then pneumonia (1.5% of the time underlying). When we focus on the population that has diabetes mentioned on the death certificate, as mentioned above 29.3% of the time diabetes is selected as the underlying, and the second most common underlying cause selected is ischemic hear disease at 25.4%, followed by cerebrovascular disease at 7.75% then followed by Other diseases of the heart 2.9%. Discussion Distinct differences in the frequency of multiple causes of death were found across time, age, race, disposal method and place of death. Definitive explanations for the differences cannot be given based on this study, but it is of interest to consider plausible explanations which may motivate further investigation. The increased reporting of non-underlying causes of death as the age of the decedent increases is likely due to actual increases in co-morbidity with age, hence would be explained by actual differences in the causes of death. The differences found in reporting of multiple causes of death for the other factors may be partly due to systematic reporting biases. According to the NCHS All Mortality Altas [[ 11 ], p. 3], the quality of cause of death determination in the US is affected by the accuracy and completeness of information – from medical diagnosis to final coding and processing of underlying cause of death. Although since 1968 the automated selection of the underlying cause of death has helped to reduce coding and processing errors, the completeness and accuracy of the information supplied on the certificate and the decedent's medical diagnosis remain as potential sources of error. If the certifier enters only one underlying cause and no other causes, then that cause will have to be selected as the underlying and there will not be multiple causes of deaths for that record. It is interesting to note that "Symptoms and ill-defined conditions" is the 5 th most commonly reported cause of death to be the only cause of death listed on the certificate. This reporting of it as the only cause of death pushes it up to be the overall 12 th leading cause of death. If almost any other cause would be listed simultaneously on the death certificate, this code would not end up as underlying. The decreasing trend in reporting multiple causes over the decade may be reflective of a gradual change in the procedures of death certification. It would be of interest to consider this trend across different states and longer periods of time including shifts from one ICD coding system to the next. Previous literature offers various plausible explanations to what contributes to the inaccuracy of reporting causes of death. The cause of death reported on the death certificates depends on a person's disease history that leads to death [ 12 ]. If a person dies after a long, well-characterized illness, the cause of death on the certificate is likely to be more accurate than a sudden or unobserved death. Also, when lack of adequate information on the decedent's disease history, the more narrowly characterized the cause of death on the certificate, the more likely it is to be in error. If we assume that reporting multiple causes on the death certificate can be considerd a proxy for level of familiarity of the death certifier with the patient, we would expect that a death which occurs in a hospital or nursing home would be more likely to have multiple causes reported, possibly due to a better documentation of disease history. On the other hand, death at the ER and in particular at the person's residence, which is conceivably often sudden should show a much lower percentage of multiple cause of death reporting. Analysis results from this current study match such speculations, supporting the argument that a good understanding of disease history is crucial. Still another result that supports this conclusion is the fact that performing autopsy, which gain better understanding of the disease condition, increased the probability of reporting multiple cause of death. Gender and race can also play a role in the accuracy of reporting. Lloyd [ 13 ] showed that positive predictive value of the death certificate tended to be lower in women than in men. Although no large differences were seen between men and women with respect to frequency of multiple causes, there was a higher percentage of multiple causes reported for Native Americans. It is conceivable that the high percentage of multiple cause of death observed for Native Americans might be associated with the geographic factors of concentrated residence and the unique practices of local clinics. Moreover, results (not shown) indicate differences exist across counties of Minnesota in the reporting of multiple causes of death ranging from 50% to 80%. These results support suggestions for better standardized training for physicians and coroners. Similar to the case of reporting multiple cause of death, the selection of ischemic heart disease and diabetes as underlying compared to non-underlying differs across the several factors considered. The implications of these differences across demographics are that mortality rates would be differentially affected when underlying cause of death is used compared to any mention cause of death. For example, for diabetes we might conclude that diabetes is being under-reported in Whites compared to Blacks, Native Americans and Hispanics if only underlying cause of death were considered since the proportion of diabetes as underlying to mentioned is substantially lower in Whites. This is not to say there is any inaccuracy in the way it is being coded but it points out where multiple cause of death reporting will provide a different perspective than underlying. Nevertheless, studies have shown the sensitivity and positive predictive value of the death certificate are particularly poor with regard to stroke and diabetes [ 14 ]. Furthermore, Lloyd [ 13 ] concluded that physicians may use coronary heart disease as a "default" cause when facing some unknown cause of death cases. The fact that individuals with autopsy performed have lower probability of having heart disease selected as underlying when it is mentioned might suggest that heart disease is often over-assigned as the default disease when no further medical details are available. This is further demonstrated by the very high ratio of ischemic heart disease being coded as the underlying compared to non-underlying cause of death when the death occurred at the person's residence. As mentioned in the introduction, one limitation of this research is the fact that there is no outside panel of experts who decide independently what the true causes of death are for each decedent, thus whether the associations we found are due to actual differences or reporting bias cannot be discerned. Therefore, this study cannot provide sensitivity or specificity per se, but it aims to identify factors that are associated with variability in reporting multiple cause of death and that perhaps contribute to inaccuracy in reporting underlying cause. Conclusions There is much to be learned from multiple cause of death data. It provides ways of looking at mortality data that go well beyond the typical examination of underlying cause of death. Future research is needed to understand further what the greatest concerns are about the accuracy of reporting causes of death. Multiple cause of death data have the potential to help point out potential concerns in the accuracy as well as provide a more complete picture of mortality for causes which are frequently not recorded as the underlying cause of death. Competing interests The author(s) declare that they have no competing interests. Authors' contributions MW conceived of the study and wrote most of the manuscript. JH performed the statistical analysis and wrote part of the manuscript. JO provided access to the data and collaboration regarding processes underlying data collection. DD provided details about cause of death reporting and collaboration regarding processes underlying data collection. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC548504.xml
548276	Tissue classification for the epidemiological assessment of surgical transmission of sporadic Creutzfeldt-Jakob disease. A proposal on hypothetical risk levels	Background Epidemiological studies on the potential role of surgery in Creutzfeldt-Jakob Disease transmission have disclosed associations with history of specific surgical interventions or reported negative results. Methods Within the context of a case-control study designed to address surgical risk of sporadic Creutzfeldt-Jakob Disease in Nordic European countries (EUROSURGYCJD Project), a strategy was adopted to categorise reported surgical procedures in terms of potential risk of Creutzfeldt-Jakob Disease acquisition. We took into account elements of biological plausibility, either clinically or experimentally demonstrated, such as tissue infectivity, PrP expression content or successful route of infection. Results We propose a classification of exposed tissues and anatomic structures, drawn up on the basis of their specific putative role as entry site for prion transmission through contact with surgical instruments that are not fully decontaminated. Conclusions This classification can serve as a reference, both in our study and in further epidemiological research, for categorisation of surgical procedures in terms of risk level of Creutzfeldt-Jakob Disease acquisition.	Background Case-control research on the association between surgery and sporadic Creutzfeldt-Jakob Disease (CJD) encompasses six studies and one meta-analysis [ 1 - 8 ], which frequently yielded diverging, partly inconsistent, positive [ 5 , 6 ] or negative [ 7 , 8 ] results, potentially attributable to methodological difficulties. This paper constitutes the first part of methods development for a case-control study summarily described in the Methods/Design section. "Development" in this sense refers to theoretical development built on scientific evidence and will focus on classification of surgical procedures, using technical and biological criteria appropriate for identifying characteristics of the intervention reflecting the putative risk level of sporadic CJD transmission. Methods In this section, we summarily describe the on-going study, review reports that possibly shed light on the biological plausibility of surgery-related CJD transmission, and propose principles for re-classification of surgical procedures. Study design The on-going case-control study, entitled "Surgery and risk of Creutzfeldt-Jakob Disease" (EUROSURGYCJD), constitutes a Concerted Action funded by the EU Research Commission, contract QLG3-CT-2002-81223. The main objective of this study is to quantify a putative excess risk of CJD associated with surgery. A secondary objective is to establish a basis for the design of preventive strategies. The most relevant methodological characteristics are: case-control design; exposure measurement prior to disease onset, registered as codes for surgical procedures; matched 5:1, randomly chosen population controls and random sample of population controls. The population base is the resident population in Denmark, Finland and Sweden, covered by the respective hospital in-patient registers. Cases are individuals with diagnoses corresponding to ICD-9 codes 046.1 and 331.5 and ICD-10 code A81.0 at death or at hospital discharge for the period 1987–2002, identified from the respective national hospital in-patient registers and corresponding national surveillance units. A questionnaire will be mailed to the heads of the registered hospital department or surveillance unit, and a copy of the medical record will be obtained for diagnosis validation. Approximately 300 patients will fulfil criteria for definite or probable sporadic CJD, and constitute the study cases. Population controls are: 1) 5 × 1 controls (approximately 1500), matched to the corresponding case by age, sex, county of residence of the case (at first discharge from hospital with CJD diagnosis or death if never hospitalised with CJD diagnosis); and 2) a 20/million sample of the 1987–2002 resident population aged >40 years, randomly selected from the corresponding national population registers. Individual person-numbers will be used for each resident case or control. Diagnoses and surgical procedures at hospital discharge of the corresponding CJD case at any registered time before date of death will be obtained from the three national hospital discharge registers in Sweden, Denmark and Finland. Perusal of surgical records might be undertaken for selected associations in order to understand transmission mechanisms and interpret results. Open-care surgery or dentistry will not be studied. Surgical procedures coded in registers as per national or NOMESCO classifications will be re-classified in accordance with putative levels of transmission risk based on scientific evidence/plausibility. The quality of CJD diagnoses will be assessed by a review of medical records. The accuracy of surgical history given by the registers will be assessed by comparison with that of a sample of controls and surrogate respondents obtained by interview. Centralised data analyses will be conducted by the Spanish team. In specific instances, risk due to blood transfusion, whether or not performed during surgical procedures, might also be studied. Review of reports and proposal Surgery may be a pathway for patient-to-patient transmission of sporadic Creutzfeldt-Jakob Disease (CJD). In many invasive surgical procedures, non-disposable surgical instruments come in contact with tissues that are known to be infective in CJD patients. These same instruments may retain a considerable level of infectivity after routine sterilisation, and in successive patients can come into contact with tissues that may act as entry sites for CJD transmission. Among the almost 300 recorded cases of iatrogenic transmission of CJD, 5 cases have been attributed to surgical instruments employed in neurosurgical procedures, whilst 2 additional cases were caused by the use of a contaminated intracerebral EEG electrode [ 9 ]. To date, proven surgical transmission of CJD has only been shown to have taken place through instruments contaminated with high-infectivity tissues (brain). However, stainless steel instruments exposed to infective tissue can acquire a maximum load of infectivity in a considerably short period of time (5 minutes) and are highly efficient in transmitting disease even after thorough washing [ 10 ]. The possibility of prion transmission through surgical interventions involving nervous or peripheral tissue has raised concern about decontaminating procedures, particularly after the emergence of variant CJD in the United Kingdom and several other countries [ 11 ]. In the above-mentioned studies [ 1 - 7 ], surgical procedures have been grouped and analysed according to gross anatomical regions (e.g., thyroid, gallbladder, prostate, etc.), which limits an interpretation of results based on biological inference. For a surgical instrument to act as a vehicle of prion transmission, it should come into contact with infective tissue during surgery of the "donor" (contaminating procedure), should maintain any adhered infectivity after being washed and sterilised, and, finally, should make contact with receptive tissues in the "recipient" patient (transmitting procedure). Different surgical interventions on the same organ may result in direct exposure of different tissues to surgical instruments, and may consequently involve a different risk of prion transmission. Within the context of a case-control study designed to address surgical risk in sporadic CJD in Nordic European countries (EUROSURGYCJD Project), we adopted the strategy of categorising all reported surgical procedures (putative transmitting procedures) in terms of potential risk of CJD acquisition. For this purpose, a classification of exposed tissues and anatomic structures has been drawn up on the basis of their specific putative role as entry site for prion transmission through surgical instruments. This classification can serve, both in our study and in further epidemiological studies, as a reference for a categorisation of surgical procedures in terms of risk of CJD acquisition. According to the "protein only" hypothesis [ 12 ], pathogenic prion protein (PrP Sc ) is a conformational isoform of PrP C , a normal host protein present in neurons and other cell types. In sporadic and familial transmissible spongiform encephalopathies (TSEs), "spontaneous" conversion of PrP C into PrP Sc is the key pathogenic event, followed by the accumulation, deposition and further conversion of PrP Sc in tissues, together with its propagation along specific neural pathways. In the case of transmitted TSEs -when these are not due to direct inoculation into the CNS- a peripheral phase of neuroinvasion by PrP Sc is followed by a subsequent phase of prion replication and propagation along the peripheral nervous system, with final access to the central nervous system [ 13 ]. In scrapie, bovine spongiform encephalopathy (BSE) and variant CJD, neuroinvasion follows widespread deposition of PrP Sc in mucosae-associated lymphoid tissue. For the purpose of classifying tissues susceptible to prion inoculation, the following considerations can be derived from this pathogenic model: i) a tissue can act as entry site for prion transmission if it normally expresses PrP C ; ii) the level of PrP Sc expression of an infected tissue correlates positively with the risk of prion acquisition by that tissue; and, iii) all tissues involved in the propagation chain of infection from peripheral tissues to the central nervous system can act as entry sites for prion transmission. The recently published WHO classification of tissue infectivity in TSEs [ 14 ], though aimed at public health issues radically different from those addressed in our study, may nonetheless serve as a conceptual framework for a tissue classification in terms of risk level of prion acquisition. This approach is based on our above-mentioned assumption (ii). The WHO classification groups tissues in three levels (high, lower and no detected infectivity) on the basis of bioassay infectivity data and/or detection of PrP Sc by Western blot. This three-level classification correlates quite closely with the distribution and levels of PrP C expression in normal nervous and non-nervous tissues in mammals.[ 15 ] The WHO tissue classification presents data on vCJD, other human TSEs, BSE and scrapie. Since no vCJD cases have been registered in Nordic countries, our epidemiological study must be limited to sporadic CJD. Consequently, our working classification excludes all tissues where positive data on infectivity or PrP Sc detection have been obtained exclusively in animal TSEs and/or vCJD. This is the case of the small bowel, large bowel (including enteric nerve plexuses), adrenal tissue, pancreas and bone marrow. Further relevant data for tissue classification derive from iatrogenic CJD cases and from experimental transmission of prion diseases to animals. In roughly half of iCJD cases the entry site has been the CNS or the eye (dura mater transplants, neurosurgical instruments or devices, corneal transplants), whilst in the other half, injection of pituitary hormones means that a peripheral route of entry has to be assumed [ 9 ]. Experimental efficiency of prion disease transmission to animals depends on various factors, such as the inoculum dose, the species barrier between the species of origin of the inoculum and the host, and the route of administration, among others. Under the same experimental conditions, different routes of administration show different efficacy of disease transmission, in terms of length of incubation period and percentage of infected animals [ 16 , 17 ]. While the most efficient route of transmission is intracerebral administration, other routes, such as intraperitoneal, intraneural, intraocular, intravenous, subcutaneous and intramuscular administration, have been used successfully in bioassays and other experimental models [ 18 ]. Still other routes, such as oral administration and conjunctival instillation [ 19 ], have shown a lower efficiency of transmission. Accordingly, clinical and experimental evidence includes several routes of prion transmission that cannot be easily reduced to a simple tissue classification involving tissues of known infectivity in CJD and/or expression of PrP C under normal conditions. This is the case of anterior ophthalmic tissues, skeletal muscle, peritoneum, and subcutaneous tissue rich in sensitive nerve fibres. These anatomical structures have therefore been independently added to our classification as putative routes of entry, with a lower level of risk compared to the high level represented by the central nervous system, sensitive ganglia and posterior eye tissues. The fact that PrP Sc has been recently found in 1/4 skeletal muscle samples of sCJD cases[ 20 ] prompted us to classify it as a tissue for potential entry rather than a route. A final classification of entry sites for putative surgical transmission of CJD contains tissues, including all those showing positive results for sporadic and familial CJD in the WHO classification [ 14 ], with minor additions (tonsil and thymus), and maintains the three risk levels of the original classification along with several putative routes of entry, based on clinical and experimental evidence (see Table 1 ). Table 1 Proposed classification of entry sites for putative surgical transmission of CJD by risk level. Risk level Tissues Anatomical structures / routes High Brain Spinal cord Retina, optic nerve Spinal ganglia Trigeminal ganglia Pituitary gland Dura mater a Lower Peripheral nerves b Spleen Lymph nodes c Tonsil d Thymus d Placenta Lung Liver Kidney Blood vessels e Olfactory mucosa CSF Skeletal muscle Anterior ophthalmic Peritoneum Subcutaneous (high density of sensitive nerve terminals) f Lowest Other Other a Dura mater does not contain pathological PrP in CJD patients and its infectivity has not been tested. It is included among high-infectivity tissues in the WHO classification because of evidence of iatrogenic transmission through dura mater grafts. 11 The same rationale has been applied to its inclusion in the present table. b Only surgical procedures on peripheral nerves (e.g., amputation, vagotomy, etc.) will be classified according to this tissue. c Surgical procedures that include this tissue as putative risk are those involving direct manipulation of lymph node chains, e.g., lymph node excision, oncological lymphadenectomy, and intra-abdominal procedures with extensive section of lymph node chains, such as cholecystectomy, gastrectomy and diverse types of bowel resection. d Although tonsil and thymic tissue have yielded negative results for infectivity and presence of pathological PrP in sporadic CJD tissue (tonsillar tissue has not yet been tested for infectivity), they are included in the table in the lower level group, together with the spleen and lymph nodes for biological reasons, in order to assess the role of peripheral lymphoid tissue in surgical transmission. e Only procedures involving direct surgery on blood vessels will be classified according to this tissue. f Hand and facial subcutaneous tissue will be included under this heading. Discussion We propose a list intended to be used to generate attributes of single, well-defined surgical procedures and criteria for their classification in terms of putative risk level of transmission. The two main attributes will be: 1) use of non-disposable surgical instruments; and, 2) exposure during surgery of tissues included in the preceding list. High- and lower-risk surgical procedures will be defined by exposure of tissue corresponding to the respective risk level during surgery. In addition, the lowest risk level is represented by those surgical procedures where disposable surgical instruments are not employed or where no listed tissue or anatomical structure is exposed to surgical instruments. A categorisation of surgical procedures based on the above attributes is inevitably prone to some degree of subjectivity, something that should be minimised by adequate methodological assessment. However, this drawback is more than offset by the possible benefits of identifying specific surgical procedures that pose a significant risk of CJD transmission, in terms of increased study power and control of misclassification of exposure by the removal of surgical procedures, which are probably irrelevant for CJD transmission, from gross anatomical classifications of surgery. We are also aware of the fact that whereas the same surgical instruments are commonly employed in an homogeneous group of procedures in some countries, as is the case of Nordic countries, the same instruments may circulate through a much wider range of procedures and putative risk levels in other countries. Final risk for disease transmission in each surgical procedure combines postulated risk related to tissues exposed in that procedure with the highest risk level of tissues to which instruments have been previously exposed. Accordingly, results should always be interpreted in the light of a known or assumed pattern of instrument circulation between surgical procedure groups. Finally, it is worth stressing that the aim of the approach presented here is exclusively to produce a useful tool for epidemiological research in CJD transmission. Under the present state of knowledge, no consequences for the possible adoption of any practical recommendation concerning surgery or further preventive measures are to be drawn from this approach. Competing interests The author(s) declare that they have no competing interests. Authors' contributions AR assumed the basic task of reviewing literature, proposing tissues and structures, and drafting the first manuscript version. JPC indicated the subject domain, suggested differences between disease transmission and disease acquisition as seen from experimental and observational epidemiological research, generated first paragraphs relating to epidemiological aspects. KM suggested changes in epidemiological aspects. ÅS provided some criticisms. MC gave diverse comments, particularly on biological plausibility. HL clarified the need for methodological refinement in future work. All authors read and accepted the final version. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC548276.xml
548262	A genomic island present along the bacterial chromosome of the Parachlamydiaceae UWE25, an obligate amoebal endosymbiont, encodes a potentially functional F-like conjugative DNA transfer system	Background The genome of Protochlamydia amoebophila UWE25, a Parachlamydia -related endosymbiont of free-living amoebae, was recently published, providing the opportunity to search for genomic islands (GIs). Results On the residual cumulative G+C content curve, a G+C-rich 19-kb region was observed. This sequence is part of a 100-kb chromosome region, containing 100 highly co-oriented ORFs, flanked by two 17-bp direct repeats. Two identical gly-tRNA genes in tandem are present at the proximal end of this genetic element. Several mobility genes encoding transposases and bacteriophage-related proteins are located within this chromosome region. Thus, this region largely fulfills the criteria of GIs. The G+C content analysis shows that several modules compose this GI. Surprisingly, one of them encodes all genes essential for F-like conjugative DNA transfer ( traF , traG , traH , traN , traU , traW , and trbC ), involved in sex pilus retraction and mating pair stabilization, strongly suggesting that, similarly to the other F-like operons, the parachlamydial tra unit is devoted to DNA transfer. A close relatedness of this tra unit to F-like tra operons involved in conjugative transfer is confirmed by phylogenetic analyses performed on concatenated genes and gene order conservation. These analyses and that of gly-tRNA distribution in 140 GIs suggest a proteobacterial origin of the parachlamydial tra unit. Conclusions A GI of the UWE25 chromosome encodes a potentially functional F-like DNA conjugative system. This is the first hint of a putative conjugative system in chlamydiae. Conjugation most probably occurs within free-living amoebae, that may contain hundreds of Parachlamydia bacteria tightly packed in vacuoles. Such a conjugative system might be involved in DNA transfer between internalized bacteria. Since this system is absent from the sequenced genomes of Chlamydiaceae , we hypothesize that it was acquired after the divergence between Parachlamydiaceae and Chlamydiaceae , when the Parachlamydia -related symbiont was an intracellular bacteria. It suggests that this heterologous DNA was acquired from a phylogenetically-distant bacteria sharing an amoebal vacuole. Since Parachlamydiaceae are emerging agents of pneumonia, this GI might be involved in pathogenicity. In future, conjugative systems might be developed as genetic tools for Chlamydiales .	Background First described in 1997, Parachlamydia acanthamoebae is an obligate intracellular bacterium naturally infecting free-living amoebae [ 1 , 2 ]. It was isolated from Acanthamoeba spp. recovered from the nasal mucosa of healthy volunteers [ 1 ]. Later, additional strains of Parachlamydiaceae have been found within about 5% of Acanthamoeba spp. and once within Hartmanella vermiformis [ 2 , 3 ]. The 16S rRNA sequences of these Parachlamydiaceae are about 14% different from those of both genera Chlamydophila and Chlamydia [ 2 , 3 ]. Since the 16S rRNA sequence difference between Chlamydophila sp. and Chlamydia sp. is 6% only, it clearly appears that the speciation between the two latter occurred after the divergence between Parachlamydiaceae and Chlamydiaceae . Like other Chlamydiales , Parachlamydiaceae can present two developmental stages: the reticulate body, a metabolically active dividing form, and the elementary body, an infective stage; the crescent body is another infective form, not observed in Chlamydiaceae [ 4 ]. Differentiation of the infective stages in reticulate bodies and multiplication of the latter were recently shown to occur within amoebal vacuoles, that may contain hundreds of bacteria [ 4 ]. Depending on the symbiotic/pathogenic relationships prevailing between both organisms, the escape of the bacteria from the amoeba may occur either by the release of secreted vesicles or by the lysis of the host [ 4 ]. There is a growing evidence of the human pathogenicity of Parachlamydiaceae [ 2 ]. For instance, positive Parachlamydia serologies were shown to be associated with a febrile epidemic [ 5 ], community-acquired pneumonia [ 6 ], and inhalation pneumonia [ 7 ]. The role of Parachlamydia -related bacteria as agents of inhalation pneumonia is further suggested by the temperature-dependent release of the bacteria from their amoebal reservoir [ 8 ]. PCR amplification of parachlamydial DNA from monocytes, sputa and bronchoalveolar lavages collected from patients suffering of bronchitis or pneumonia also supports the pathogenic potential of Parachlamydia [ 9 - 12 ]. The survival of these Chlamydia -like organisms within human macrophages [ 13 ] is an additional hint of parachlamydial pathogenicity. Horn et al . [ 14 ], by sequencing and annotating the whole genome of the Parachlamydia -related UWE25 contributed much to the understanding of the evolution of chlamydiae. Indeed, they demonstrated that major virulence mechanisms of Chlamydiaceae such as the Type Three Secretion System (TTSS) and the Chlamydial Protease-like Activity Factor (CPAF) are also encoded by the chromosome of the evolutionary early-branching Parachlamydiaceae UWE25. Genome analysis of the parachlamydial endosymbiont also identified Open Reading Frames (ORFs) homologous to Type Four Secretion Systems (TFSS) and characterized by a high G+C content, suggesting that they result from an horizontal transfer. Based on their annotation revealing the apparent absence of genes necessary for DNA transfer, Horn et al . [ 14 ] proposed that this TFSS was involved in protein export but not in DNA transfer. To date, numerous genomic islands (GIs) were already identified along whole chromosomal sequences of various bacterial species. For instance, 140 GIs are described in the Islander database, including GIs of proteobacteria, firmicutes, actinobacteria and cyanobacteria [ 15 ]. Thus, we wondered whether any GIs were located along the bacterial chromosome of the amoebal endosymbiont UWE25. GIs are genetic elements which length vary from 10 to 200 kb and are inserted in a chromosome after a lateral transfer occurring, in some instances, between phylogenetically-distant microorganisms. Their heterologous origins are generally evidenced by a G+C content different from that of the remaining bacterial chromosome and by the presence of various mobility genes (i.e. involved in transposition, transduction or conjugative transfer), that are occasionally source of GI instability [ 16 , 17 ]. They are often flanked by particular DNA sequences, such as direct repeats or insertion sequences. Moreover, tRNA loci are generally used as insertion sites by GIs for their chromosomal integration [ 16 - 18 ]. Since no genetic tools are available for the study of this obligate intracellular bacteria, a bioinformatic approach was chosen to locate putative GIs. Results A genomic island is present in the genome of UWE25 Using standard G+C content analyses of Parachlamydia -related UWE25 chromosome, we observed a G+C-rich region (Figure 1A and 1B ), similar to that shown by Horn et al . [ 14 ]. Using the residual cumulative G+C content analysis adapted from the GC profile of Zhang and Zhang [ 19 ], we were able to precisely define a 19-kb region (Figure 1C ). The presence of 17-bp direct repeats flanking a 100-kb chromosome region (1648 to 1748 kb, Table 1 ) that encompasses the 19-kb DNA sequence enabled us to define a new region composed of 100 ORFs (See additional data file 1 for the description of these genes and their location on the chromosome of UWE25). Interestingly, this 100-kb region is characterized by a higher level of local gene coorientation (75/100) than that characterizing the remaining of the genome (1015/1931, 52.6%, p < 0.001) and by a particular signature in the cumulative GC skew analysis. Two identical gly-tRNA genes in tandem are located at the proximal end of this 100-kb genetic element (Figure 1A,1C and Table 1 ). Several mobility genes (eight putative transposases, one recombinase and seven bacteriophage related-proteins) are encoded within the 100-kb region (Figure 1C , Table 1 ). Thus, this region largely fulfills the accepted criteria of GIs [ 16 - 18 ]. We termed this newly described GI "Pam100G" ( Protochlamydia amoebophila , 100-kb, Gly-tRNA) according to the nomenclature used in the Islander database [ 15 ]. Mosaicism of the 100-kb genomic island Interestingly, this GI can be divided into clearly distinct regions, according to their G+C content (Figure 1B , Table 2 ). The residual cumulative G+C content analysis highlights a modular structure with different slopes, each linear segment indicates that genes of this unit present a rather constant local G+C content (Figure 1C ). A positive or a negative slope would indicate that each block of genes presents a G+C content higher or lower that of the UWE25 chromosome, respectively. The first module begins with a direct repeat and two identical gly-tRNAs in tandem. Composed of 28 ORFs, this unit exhibits a G+C content (36.4%) similar to that of the remaining of the genome of UWE25 (1931 ORFs, 36.1%). Sixteen homologs to these 28 genes (57%) were found in databases, 12 of them (75%) exhibiting a best score in BLAST analyses with a Chlamydiaceae ORF (See additional data file 1 ). Interestingly, no gene of the other modules of the 100-kb GI exhibited a best hit in similarity analyses with any Chlamydiaceae counterpart. Some of the genes present in the first module, such as sctN and sctQ , are part of a TTSS also present in Chlamydiaceae . The other TTSS genes are disseminated along the chromosome of UWE25. The presence of some TTSS genes in the first module and of a gene encoding a putative transposase at the distal end of the first module of this 100-kb GI suggests that this first unit was acquired by chromosomal rearrangements. (See additional data file 1 for the results of BLAST analyses). Characterized by a low G+C content (34.1%), the 2 nd module encodes 18 ORFs. Only five are similar to known protein sequences (28%), four of them being identified as mobility genes (three bacteriophage-related genes and one putative transposase encoding gene). The 3 rd module (19 kb), exhibiting the second highest G+C content of the UWE25 genome (40.9%), comprises 21 ORFs. Some of these genes were identified as tra genes by Horn et al . [ 14 ]. Using BLAST analyses and alignment tools, we re-annotated the whole module (see below) and, if we except two transposase genes and one ORF of unknown function, we unveiled that all ORFs of this module belong to a genetic unit similar to the tra operons encoding the TFSS previously described in proteobacterial genomes (Figure 2 , See additional data file 1 for the re-annotations of this module). Presenting a low G+C content (33.4%), the 4 th module (10 kb) is composed of 13 ORFs. All these ORFs were previously annotated by Horn et al. [ 14 ] as encoding hypothetical proteins or without homolog. Our BLAST analysis identified one ORF homologous to genes encoding bacteriophage-related proteins and two genes of proteins involved in DNA metabolism (Table 1 and 2 , see also the table in the additional data file 1 ). Interestingly, a direct repeat is located between the 9th and 10th genes of the module. This 17-bp direct repeat, that presents 3 mismatches is similar to those present at the proximal and distal ends of the GI, exhibiting the same 14 conserved nucleotides. It may reflect a complex evolutionary history of the GI, possibly enabling it to be mobile as 25-kb, 75-kb or 100-kb DNA segments. A single large protein is encoded along the 5 th module (6 kb). Its G+C content is one of the highest of the UWE25 chromosome (41.8%). By BLAST analysis, this protein exhibits the strongest similarity with the human Nod3 protein. The 6 th module (12 kb) is characterized by a low G+C content (33.3%). This unit is composed of 13 ORFs, the first ORF encoding a product similar to the Death on cure (Doc) protein of P1 bacteriophage. Two ORFs code for proteins involved in DNA metabolism and an additional ORF encodes a putative transposase. The 7 th module is short (2 kb) and present a G+C-rich unit (38.7%). Five of the six ORFs of this unit encode a probable resolvase, three putative transposases and a phage-related Doc protein. The final direct repeat is located at the end of this module. With the only exception of the phage-related protein, all other ORFs of the 7 th module appear to be similar to gamma-proteobacterial proteins, possibly explaining the observed different signal in the G+C content analysis. Role of the type IV secretion system encoded by the 100-kb genomic island The functions of genes encoded by GIs may be related, among others, to pathogenicity such as the ability to exploit the host intracellular environment. Since no genetic system has been described for any obligate intracellular chlamydiae, we investigated the putative functions of this GI by bioinformatics. We focused our analyses on the TFSS, for which a previous annotation of the tra genes showed a genetic unit unable to transfer DNA [ 14 ]. Using different protein comparison methods described in the additional data file 1 , we identified supplementary tra genes, and compared the general organization of this tra unit with other genetic elements encoding TFSS genes [ 20 ]. The UWE25 tra unit displays a striking colinearity with tra operons encoding F-like conjugative DNA transfer system, especially to those of the F and pNL1 plasmids of Escherichia coli and Novosphingobium aromaticivorans , respectively (Figure 2 ). All homologous genes essential for DNA transfer in plasmid F ( traF , traG , traH , traN , traU , traW , and trbC ) and involved in sex pilus retraction and mating pair stabilization [ 20 ] are present, strongly suggesting that, similarly to the other F-like TFSSs, the gene products encoded by the UWE25 tra unit are devoted to DNA transfer. With the only exception of traG , these genes are not present on P-like and I-like plasmids, reinforcing the close relationship prevailing between the UWE25 tra unit and their F-like plasmids counterparts. Figure 3 shows that the UWE25 tra unit clusters within F-like TFSSs, confirming that it may function as a F-like conjugative system. Drawn as an UPGMA tree (Figure 3A ), the comparison of the genetic organization of all tra units was performed as a gene order breakpoint analysis developed for the study of the mitochondrial genome evolution [ 21 ]. This analysis clearly shows that the closest relatives of the UWE25 tra units are the tra operons of the F-like conjugative plasmids. The Fitch-Margoliash- and the minimum evolution comparisons performed on the same dataset presented the same tree topologies, confirming the former UPGMA results (data not shown). An omit test performed on this tree confirms that the results are robust: with one exception (involving the deep branching of one cluster on one tree), all 11 trees were congruent in all their nodes. Figure 3B shows an UPGMA tree comparing the Kimura corrected p -distances (the proportion p of nucleotide sites at which two sequences are different, taking into account the proportion of transversion- and transition-substitution rates) of nucleotide sequences of the concatenated traA , traK , traB , traV , and traC genes. A similar topology is observed with (i) neighbor-joining- and minimum evolution trees inferred using the Kimura-corrected p -distances and (ii) UPGMA, neighbor-joining- and minimum evolution trees performed on p -distance of the whole coding sequences of the concatenated tra genes (See additional data file 2 for these trees). Neighbor-joining- and minimum evolution methods comparing Kimura-corrected p -distances of the complete coding sequences confirmed that the tra unit of UWE25 is phylogenetically closely related to the tra operons of the F-like plasmids: the bootstrap values of 94% and 91% respectively, support the node separating the concatenated tra genes of the chromosomal UWE25 and the R27 plasmid, a gamma-proteobacterial F-like conjugative plasmid, from those of all other plasmids (See the additional data file 2 for these trees). In neighbor-joining and minimum evolution analyses of p -distances, the tra unit of UWE25 also clusters with the tra operons of gamma-proteobacterial F-like plasmids: the bootstrap of 96% and 92%, respectively, support the node separating the concatenated tra genes of UWE25 and RTS1, SXT, R391, three gamma-proteobacterial F-like conjugative plasmid, from their closest relative, R27 plasmid (See additional data file 2 for these trees). Taken together, all these data strongly suggest that the UWE25 tra unit is closely related to F-like conjugative tra operons. Origin of the genomic island and of its type four secretion system Our BLAST analyses [ 22 ] reveal that a majority (24/43) of genes not presenting a best hit for chlamydial genes but having homologs in other taxa are more related to proteobacterial genes (see Table 2 and the additional data file 1 for similarity analyses indicating for each parachlamydial tra gene the most similar gene and its taxonomical background). Moreover, the BLAST analyses of the 21 ORFs of the third module, encoding the tra genes, show that most ORFs of this unit (15/21) are of proteobacterial origin. However, since six of them present the highest similarity to alpha-proteobacterial genes and six others to gamma-proteobacterial genes, a more precise origin of the parachlamydial tra unit could not be precisely defined by this first approach. The presence of gly-tRNA at the proximal end of the GI of UWE25 is consistent with a close relatedness between this GI and proteobacteria: out of 14 GIs described in the Islander database of Mantri et al . [ 15 , 22 ] inserted along a chromosome by a gly-tRNA (14/140), 12 of them (86%) were sequenced in a proteobacterial genome. No GI of Gram-positives described in the Islander database are inserted in a chromosome within a gly-tRNA gene. Again, a precise proteobacterial origin could not be proposed, because the distribution of gly-tRNA genes in alpha- (4/22) and gamma-proteobacterial (8/72) GIs is not significantly different: by including only the non-redundant GIs, the distribution of gly-tRNA genes in alpha and gamma-proteobacterial GIs is 2/20 and 7/71, respectively. Comparison of gene order between all tra units also failed in assigning a precise origin to the UWE25 tra unit since it branched near the alpha- and gamma-proteobacterial tra operons (Figure 3A ). The only first hint of a possible gamma-proteobacterial origin for the UWE25 tra unit was brought by the phylogenetic analyses (Figure 3B and additional files 1 & 2 ). Thus, bootstraps values of 94, 91, 96 and 92% supported the node separating the concatenated tra genes of UWE25 and several tra operons of gamma-proteobacterial F-like plasmids from the F-plasmids of an alpha-proteobacteria and of other gamma-proteobacteria. (See above, and additional data file 2 for these trees). Discussion We showed that the Parachlamydia -related endosymbiont UWE25 presents a 100-kb region largely fullfilling the criteria of GIs [ 16 - 18 ]. Indeed, this DNA region characterized by a high level of gene co-orientation presents a G+C content different from that of the remainder of the genome. The presence of direct repeats flanking this chromosome region enabled us to focus on 100 ORFs. Two identical gly-tRNA genes in tandem are present at the proximal end of this genetic element. Moreover, several mobility genes encoding transposases and bacteriophage related-proteins are located within this chromosome region. The cumulative residual G+C content analysis shows that this GI is composed of seven modules. Such a chimeric organization was already described in other GIs [ 23 , 24 ]. The first module contains chlamydiae genes probably brought by chromosome rearrangements. Some of these genes, homologous to TTSS genes of Chlamydiaceae , might provide selective advantages to strains that retained the GI. The 2 nd , 4 th and 6 th modules are mainly composed of bacteriophage-related protein genes, that could reflect a putative phage implication in GI formation. The 3 rd module codes for a TFSS similar to tra operons. We propose that this tra unit is devoted to DNA transfer, based (i) on similarity analyses demonstrating the presence of all genes encoding proteins used during a DNA transfer, (ii) on phylogenetic analyses of tra unit genes and, (iii) on comparison of gene order. These analyses clearly demonstrate that the UWE25 tra unit is strongly more related to F-like conjugative system than to P-like and I-like secretion systems. The significant bootstraps of all trees obtained by standard gene phylogeny and their congruent topologies with others obtained by the gene order breakpoint analysis not biased by codon usage homing, strongly support the validity of these analyses confirming the F-like conjugative nature of the parachlamydial tra unit. Thus, our model significantly differs from the other proposed by Horn et al . [ 14 ], who did not identify traA , traL , traK , traV , and concluded that the UWE25 tra unit is involved in protein export, and not in DNA transfer. The 5 th module presents a nucleotide composition similar to the tra unit and is composed of a single high G+C 6-kb gene, whose product is similar to the human Nod3 protein. The Nod (Nucleotide-binding oligomerization domain) proteins are members of a family that also includes the apoptosis regulator Apaf1 (Apoptotic protease activating factor 1) and plant disease-resistance gene products [ 25 ]. The function of the human Nod3 is still unknown. Like Nod1 and Nod2, Nod3 might be involved in the recognition of conserved motifs present at the surface of bacteria, such as peptidoglycan. The nucleotide G+C composition of the 2 nd , 4 th , and 6 th modules are similar, explaining the observed similar negative slope of the residual G+C curves. Moreover, these three modules encode phage-related proteins and proteins involved in DNA metabolism. These modules probably involved in mobility might have a common origin, the ancestral single phage module being currently separated in three pieces by the presence of the tra unit and of the Nod3-like protein encoding gene. The positive slope in the G+C analysis of the 7 th module echoes those of the tra unit (3 rd module) and of the Nod3-like protein (5 th module). The 7 th module encodes a transposition resolvase and three transposases similar to gamma-proteobacterial homologs. With the only exception of the phage-related Doc protein, that has an homolog at the beginning of the sixth module, and that might be located there after transposition, the 7 th module appears thus to have a different origin than the 2 nd , 4 th and 6 th modules, though also encoding mobility genes. The presence of a F-like tra unit along the sequences of UWE25 is the first evidence of a putative conjugative system in chlamydiae. If conjugation occurs, it most probably takes place within free-living amoebae, that may contain several hundreds of Parachlamydia bacteria tightly packed in their vacuoles [ 4 ]. Such a conjugation system would be a mechanism to transfer DNA between internalized bacteria sharing an amoebal vacuole. Moreover, it may provide molecular genetic tools for obligate intracellular bacteria. The presence of tra units/operons in the parachlamydial UWE25 and in proteobacteria could be explained by an emergence of this unit in a common ancestor of both clades, and by its subsequent loss in Chlamydiaceae . Another evolutionary scenario is that the tra unit was acquired from a proteobacteria by a Parachlamydiaceae in a common amoebal vacuole. Since the tra unit is absent from all sequenced Chlamydiaceae genomes, this transfer would have occurred after the divergence of Parachlamydiaceae and Chlamydiaceae , at a time when Parachlamydia was already an intracellular bacteria. An intra-amoebal transfer of this GI is supported by the permissivity of free-living amoebae to proteobacteria [ 26 ], and by several hints suggesting its proteobacterial origin. Though phylogenetic analyses suggested a gamma-proteobacterial origin of the F-like parachlamydial tra , further analyses have to confirm whether this GI module was acquired from an alpha-, beta-, or gamma-proteobacteria unit. We hypothesize that the F-like parachlamydial tra unit has been brought by a lateral transfer from a proteobacterial genome. This hypothesis is strongly supported by the cumulative GC skew analysis [ 27 - 30 ] producing a signal of the GI differing from that of the remaining of the genome (Figure 1A and 1B ). The value of nucleotide skew analyses as good taxonomical markers is supported by (i) routine analyses on prokaryotic genome by cumulative TA-skews [ 30 ] and (ii) comparison of intragenic nucleotide skews of small subunit ribosomal RNA of the whole living world [ 31 ]. The genometric approach appeared to be able to identify GIs of Chlamydiales . Sequencing additional genomes of environmental chlamydiae, that present a large biodiversity [ 3 ], will provide major insights on bacterial evolution and hopefully a better comprehension of the emergence of this parachlamydial GI. Conclusions We showed that a GI present on the UWE25 chromosome encodes a potentially functional F-like DNA conjugative system. This is the first hint of a putative conjugative system in chlamydiae. Conjugation most probably occurs within free-living amoebae, that may contain hundreds of Parachlamydiaceae bacteria tightly packed in vacuoles. Such a conjugative system might be involved in DNA transfer between internalized bacteria. Since this system is absent from the sequenced genomes of Chlamydiaceae , we hypothesize that it was acquired after the divergence between Parachlamydiaceae and Chlamydiaceae , when the Parachlamydia -related symbiont was an intracellular bacteria. It suggests that this heterologous DNA was acquired by a Parachlamydiaceae from phylogenetically-distant bacteria sharing an amoebal vacuole. Since Parachlamydiaceae are emerging agents of pneumonia [ 2 ] and since many GIs are also considered as pathogenicity islands [ 17 ], the Pam100G GI might be involved in pathogenicity. In future, conjugative systems might be developed as genetic tools for studying Chlamydiales . Methods Sequence The genome sequence of UWE25 [ 14 ] (Accession number: NC_005861) is available at the NCBI website [ 32 , 33 ]. In this contribution, the acronym UWE25 refers only to the Parachlamydia -related endosymbiont UWE25, and thus not to the Acanthamoeba sp. strain UWE25 from which the parachlamydial endosymbiont UWE25 was recovered [ 3 ]. Horn et al . recently proposed UWE25 as the type strain of a new bacterial species: Protochlamydia amoebophila [ 34 ]. BLAST analyses BLAST analyses were performed with BLASTP 2.2.9 [ 35 ] available on the NCBI website [ 36 ] using the BLOSUM62 matrix, and gap penalties of 11 and 1. Each ORF was compared against all genes of non-redundant databases available at the NCBI website. An e-value of 0.001 was selected as a standard cut-off. To further identify possible homologous ORFs, we also BLASTed each tra gene of F plasmid versus all genes of the full genome of Parachlamydia and conversely, each ORF of the putative parachlamydial tra unit versus counterparts of the different F-like plasmids. CLUSTALW was used to detect the best relatedness of a given parachlamydial Tra protein with its possible homologs encoded by the F and pNL1 plasmids. Residual cumulative GC content The residual cumulative G+C content, a slightly modified version of the cumulative GC profile defined by Zhang and Zhang [ 19 ], was used to reveal local variations of G+C content of a genome, without using sliding windows of arbitrary size. First, a G+C content analysis was performed on 100-bp windows of the selected chromosome sequence, as for a cumulative GC skew analysis. The cumulative G+C content GC n of the n th window is obtained by cumulating the G+C contents from the first to the n th window: where, in the window i , G i and C i are the numbers of Gs and Cs, respectively, and N i is the total number of nucleotides. To visualize genomic regions differing from the average G+C content, a linear regression y defined by a slope k is performed on the cumulative curve using the least square methods: y ( n ) = kn where n is the position of the center of the n th window. The residual cumulative G+C content curve GC ' can then be drawn as a function of the position of each window center: GC ' n = GC n - kn Zhang and Zhang [ 19 ] recently demonstrated that, in some instances, abrupt changes in the residual cumulative G+C content curve correspond to genomic islands. Repeats identification The perfect tandem repeats identification was first performed using the EQUICKTANDEM software (Richard Durbin, Sanger Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge, CB10 1SA, UK) [ 37 ] on a 200-kb DNA sequence (UWE25 genome position: 1.6 to 1.8 Mb) encompassing the tra genes previously identified by Horn et al . [ 14 ]. The duplicated genes and the ORFs containing internal repeats were removed. For each pair of direct repeats, potential unperfect matches of flanking nucleotides were scanned using DNA strider 1.2.1 [ 38 ], with the following settings: a minimal size of 11 bp and 3 mismatches. Furthermore, sequences similar to direct repeats were searched along the whole chromosome, and sequences also found outside the selected 200-kb region were discarded from our analysis. Finally, the direct repeats positions were compared to the G+C content analysis, the cumulative GC skew curve, and to tRNA genes locations. Phylogenetic analyses Since Horn et al . [ 14 ] did not identified traA , traL , traK , traV , re-annotation of the UWE25 tra unit was necessary for phylogenetic analyses. We used i) the genes of F-like plasmids encoding the following tra genes, i.e. traA , traK , traB , traV , traC , and ii) the corresponding ORFs of P- and I-like plasmids [ 20 ], i.e. trbC/VirB2 , trbG/VirB9 , trbI/VirB10 , trbH/VirB7 , trbE/VirB4 of P-plasmids and traX , traN , traO , traI , traU of I-plasmids, respectively. The genes were concatenated to obtain a single nucleotide sequence and aligned with CLUSTALW ([ 39 ] as it was already performed for genes of ribosomal proteins [ 40 ]. Using this alignment and the MEGA 2.1 software [ 41 ], we inferred phylogenetic relationships by drawing trees using p -distances (the proportion p of nucleotide sites at which two sequences compared are different) and Kimura corrected p -distance (correction for the rates of transition and transversion) with Unweighted Pair Group Method with Arithmetic Mean (UPGMA), neighbor-joining, and minimum evolution methods. To prevent alignment biases, trees were drawn using the complete deletion option implemented on MEGA 2.1. Gene order breakpoint analyses To quantify the inversion and transposition events leading to the current organization of tra operons, the gene order breakpoint analysis developed for small genomes (mitochondria) by Blanchette et al . [ 21 ] was used to estimate the similarity of gene order existing between the tra unit of UWE25 and the tra operons reviewed by Lawley et al . [ 20 ]. The distance calculated for two given operons O i and O j containing homologous genes proposed by Blanchette et al . [ 21 ] was slightly modified to take into account the variation of gene numbers of tra operons: instead of counting the number of minimal breakpoints existing between two tra operons, a distance was estimated by measuring the proportion of conserved gene pairs between both genomic entities. Next, a comparison matrix is established by calculating the distance for each pairwise comparison. Finally, a dissimilarity matrix is obtained by subtracting each distance from 1. For instance, if the operon O i encodes sequentially four genes ( a , b , c , and d ) and the operon O j , six genes ( a , b , e , -d , -c , and -f ; genes labeled by a minus sign are encoded by the complementary strand), the gene order breakpoint analysis reveals that two gene pairs are conserved: ab and cd . The dissimilarity distances existing between the operons i) O i and O j , and ii) O j and O i would be: 1-(2/3) = 1/3 and 1-(2/5) = 3/5, respectively. From the square dissimilarity matrix, phylogenetic trees were drawn. Three different distance-matrix analyses were used: the UPGMA, the Fitch-Margoliash- and the minimum evolution methods. To assess the robustness of the tree, an omit test [ 42 ] was performed on 11 UPGMA trees, in each one organism is missing. Authors' contributions GG and CAR initiated the project. GG and FC reannotated the parachlamydial tra unit and performed all BLAST analyses. FC and LG delimited the GI by direct repeat analyses. GG drew the phylogenetic trees of the concatenated tra genes. After developing the residual cumulative G+C content analysis used for a software development, LG performed the G+C content analyses and the gene order comparison of tra units by the gene order breakpoint analysis. CAR established the correlation existing between the tRNA genes and the GIs according to taxonomy and coordinated the team work. GG wrote the first draft of the paper. All authors improved the manuscript and approved its final version. Supplementary Material Additional File 1 Supplementary table. Results of BLAST [ 35 , 36 ] analyses of 100 ORFs present in the 100-kb region. BLAST analyses were performed using BLOSUM62 matrix and gap penalties of 11 and 1. Chromosome location of each ORF, its G+C content, coding strand, and the presence of at least one homolog in Chlamydiaceae are presented. Direct repeats (DR), gly-tRNA genes and limits of each modules of the GI are highlighted. Click here for file Additional File 2 Supplementary figure. Phylogenetic analyses suggest that the UWE25 tra unit is phylogenetically closely related to F-like DNA conjugative tra operons: (A) UPGMA-, (B) Neighbor-joining-, and (C) minimum evolution-trees comparing p-distances and Kimura corrected p -distances of nucleotide sequences of the concatenated traA , traK , traB , traV , and traC genes (the UPGMA tree comparing the Kimura corrected p -distances of tra genes, shown in Figure 3B , is presented here to facilitate comparison with the other trees). Interestingly, in neighbor-joining and minimum evolution analyses of the p -distances, the tra unit of UWE25 is clustered with tra operons of gamma-proteobacterial F-like plasmids: bootstrap values of 96% and 92%, respectively, support the node separating the concatenated tra genes of UWE25 and RTS1, SXT, R391, three gamma-proteobacterial F-like conjugative plasmids, from their closest relative R27 plasmid. Similarly, in neighbor-joining and minimum evolution analyses of the Kimura corrected p -distances, bootstrap values of 94% and 91%, respectively, support the node separating the concatenated tra genes of the chromosomal UWE25 and the R27 plasmid, another gamma-proteobacterial F-like conjugative plasmid, from those of all other plasmids. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC548262.xml
529313	Adaptive Amplification and Point Mutation Are Independent Mechanisms: Evidence for Various Stress-Inducible Mutation Mechanisms	“Adaptive mutation” denotes a collection of processes in which cells respond to growth-limiting environments by producing compensatory mutants that grow well, apparently violating fundamental principles of evolution. In a well-studied model, starvation of stationary-phase lac − Escherichia coli cells on lactose medium induces Lac + revertants at higher frequencies than predicted by usual mutation models. These revertants carry either a compensatory frameshift mutation or a greater than 20-fold amplification of the leaky lac allele. A crucial distinction between alternative hypotheses for the mechanisms of adaptive mutation hinges on whether these amplification and frameshift mutation events are distinct, or whether amplification is a molecular intermediate, producing an intermediate cell type, in colonies on a pathway to frameshift mutation. The latter model allows the evolutionarily conservative idea of increased mutations (per cell) without increased mutation rate (by virtue of extra gene copies per cell), whereas the former requires an increase in mutation rate, potentially accelerating evolution. To resolve these models, we probed early events leading to rare adaptive mutations and report several results that show that amplification is not the precursor to frameshift mutation but rather is an independent adaptive outcome. (i) Using new high-resolution selection methods and stringent analysis of all cells in very young (micro)colonies (500–10,000 cells), we find that most mutant colonies contain no detectable lac -amplified cells, in contrast with previous reports. (ii) Analysis of nascent colonies, as young as the two-cell stage, revealed mutant Lac + cells with no lac- amplified cells present. (iii) Stringent colony-fate experiments show that microcolonies of lac -amplified cells grow to form visible colonies of lac -amplified, not mutant, cells. (iv) Mutant cells do not overgrow lac -amplified cells in microcolonies fast enough to mask the lac -amplified cells. (v) lac -amplified cells are not SOS-induced, as was proposed to explain elevated mutation in a sequential model. (vi) Amplification, and not frameshift mutation, requires DNA polymerase I, demonstrating that mutation is separable from amplification, and also illuminating the amplification mechanism. We conclude that amplification and mutation are independent outcomes of adaptive genetic change. We suggest that the availability of alternative pathways for genetic/evolutionary adaptation and clonal expansion under stress may be exploited during processes ranging from the evolution of drug resistance to cancer progression.	Introduction Adaptive mutation was first brought to wide attention by Cairns et al. (1988) , and encompasses a collection of processes whereby cells adapt genetically in response to growth-limiting environments. The phenomenon is reported in several different microbial systems, and appears to occur via different molecular mechanisms (reviewed by Rosenberg 2001 ; Hersh et al. 2004 ). The Lac frameshift assay system of Escherichia coli ( Cairns and Foster 1991 ) is the best understood in terms of mutation mechanism (reviewed by Foster 1999 ; Rosenberg 2001 ; Hersh et al. 2004 ). In this system, cells carrying a +1 frameshift mutation in lac genes on an F′ conjugative plasmid are starved on solid medium with lactose as the sole carbon source, selecting Lac + revertants. Lac + clones appear over time during selection in a population of cells that shows no net growth (i.e., is in stationary phase). In most clones, Lac + is conferred by a compensatory −1 frameshift mutation in the lac gene (Lac + point mutants). Lac + point mutation during stationary phase differs from mutation during rapid growth in that it requires the recombination proteins RecA, RecBC, RuvA, RuvB, and RuvC ( Harris et al. 1994 , 1996 ; Foster et al. 1996 ), and induction of the SOS DNA damage response regulon ( Cairns and Foster 1991 ; McKenzie et al. 2000 ), particularly error-prone DNA polymerase (Pol) IV (DinB) ( McKenzie et al. 2001 , 2003 ; Wolff et al. 2004 ). The lac frameshift allele is leaky, conferring 1%–2% of the wild-type level of β-galactosidase ( Foster 1994 ; Andersson et al. 1998 ), and Lac + colonies can also occur by amplification of the lac locus into a tandem array of 20–100 copies ( Foster 1994 ; Andersson et al. 1998 ; Hastings et al. 2000 ). Adaptive amplification requires RpoS, the transcriptional activator of approximately 50 stationary-phase/starvation- and general-stress-response-specific genes, which is also required for point mutation ( Lombardo et al. 2004 ). Amplification does not require DinB or induction of other SOS proteins ( McKenzie et al. 2001 ), and whether or not it requires recombination proteins has not been determined unambiguously (Tlsty et al. 1984; Hastings and Rosenberg 2002 ). Consistent with Darwinian principles of evolution by selection of undirected and random genetic changes, adaptive point mutation in the Lac system is accompanied by a high level of mutation in genes unrelated to lactose catabolism (secondary mutation). Lac + point mutants show about 50-fold more secondary mutations than do cells starved on the same plate that did not become Lac + ( Torkelson et al. 1997 ; Rosche and Foster 1999 ; Godoy et al. 2000 ). This indicates that a subpopulation of the starved cells experiences a transient episode of hypermutation (adaptive point mutants are not mutator mutants; Torkelson et al. 1997 ; Rosenberg et al. 1998 ; Rosche and Foster 1999 ). In keeping with their independence of DinB and SOS, the amplified clones do not show hypermutation of unrelated genes ( Hastings et al. 2000 ). These differences, together with the finding that cells carrying lac amplification (“ lac -amplified cells”) do not readily yield Lac + point mutants when resubjected to selection on lactose medium, led us to propose that amplification reflects a pathway of genetic change wholly or partly separate from point mutation (we suggested an alternative outcome of a branched pathway; Hastings et al. 2000 ; Rosenberg 2001 ; Hastings and Rosenberg 2002 ). Two general models for the adaptive point mutation mechanism in this system are currently at odds, each focused on the role of DNA amplification, and each representing a different view of evolution (e.g., Rosenberg and Hastings 2004a , 2004b ; Roth and Andersson 2004a , 2004b ). In “hypermutation models,” the observed high frequency of Lac + point mutants is proposed to result from transient hypermutation as part of a stress response, one consequence of which is acceleration of genetic change (and so potentially of evolution) (e.g., see Rosenberg 2001 , and below, where specific molecular mechanisms are suggested). In this view, amplification is an adaptive pathway alternative to point mutation that allows growth of cells, relieving the stress of starvation, and so deflects them from the point mutation route. This view is compatible with Darwinism, which allows for changing rates of generation of heritable variations ( Darwin 1859 ). In an alternative model, called “amplification–mutagenesis,” amplification is an intermediate in the formation of Lac + point mutants, and its most important role is to provide more lac copies per cell so that point mutation can occur without increase in mutation rate per copy of lac ( Andersson et al. 1998 ; Hendrickson et al. 2002 ). This adheres to conservative neo-Darwinist ideas of constant and gradual evolutionary change (e.g., Mayr 1982 ). Here, we report several experiments that demonstrate that amplification and point mutation are separate adaptive outcomes, compatible with hypermutation models, and not part of a sequential pathway leading to mutation (as in the amplification–mutagenesis model). We also report the first genetic requirement specific to the adaptive amplification pathway, Pol I. Results/Discussion Colony Phenotypes Experiments described below make use of the different colony color phenotypes of point-mutant and lac -amplified clones grown on rich medium with the dye 5-bromo-4-chloro-3-indoyl β-D-galactoside (X-gal) (e.g., Tlsty et al. 1984 ; Andersson et al. 1998 ; Hastings et al. 2000 ) ( Figure 1 ). We use the word “sectored” to describe the appearance of a colony of cells on X-gal medium (e.g., Figure 1 B and 1 C). The word “unstable” is used to denote continued sectoring when cells from sectored colonies are replated. This instability has been shown in many examples to reflect tandem amplification of the lac region ( Tlsty et al. 1984 ; Foster 1994 ; Andersson et al. 1998 ; Hastings et al. 2000 ). Sectoring that is not unstable ( Figure 1 C) may reflect incidental juxtaposition of blue and white colonies on a plate (see below). “Stable” describes the absence of visible sectoring that indicates Lac + point mutation ( Tlsty et al. 1984 ; Hastings et al. 2000 ; see Figure 1 A). Figure 1 Colony Morphologies (A) Point-mutant Lac + colony showing solid blue color (the pale colonies are derived from Lac − cells). (B) lac -amplified colonies showing sectoring caused by the instability of the amplified array. Cells from these colonies grow either into sectored blue colonies or, if they have lost the amplification, into white colonies, a phenotype that we call “unstable.” (C) A sectored colony that is not unstable in that it was found to contain only stable blue and stable white cfu upon retesting. (D) An example of a microcolony of the sort used in this work. The visible colony on the lower edge of the field has a diameter of 1.4 mm (>10 8 cells). (E and F) Phase contrast (E) and green fluorescence (F) of the same field, showing two of 30 SMR6039 cells fluorescing. Most Microcolonies Are Pure, Not Mixed If most adaptive point mutants arise by mutation in young colonies carrying lac amplification ( Andersson et al. 1998 ; Hendrickson et al. 2002 ), then young colonies should contain both the lac -amplified progenitor cells and their point-mutant descendants, as reported by Hendrickson et al. (2002) and reexamined here. To determine whether a microcolony carries point-mutant or lac- amplified cells or both, in principle, one picks the microcolony from the lactose-minimal selection medium and replates its cells onto rich X-gal medium to observe whether the resulting colonies are pure blue (point mutant), unstable sectoring ( lac- amplified), or both. A significant problem in interpreting data from such analyses is that the picked microcolony carries with it many Lac − background cells from the selection plate ( Figure 2C ) and may contain cells from unrelated microcolonies, because, of necessity, these microcolonies cannot be streaked to purify them before analysis. Figure 2 Stringent Analysis of Whole Microcolonies: Analyzing All Cells in a Microcolony and Reducing Contamination by Unrelated Neighboring Bacteria To analyze all cells in a microcolony, very young microcolonies were harvested (10 3 –10 4 cfu/microcolony; see Figure 1 D; Table 1 ). To reduce contamination with neighboring Lac − bacteria and unrelated Lac + microcolonies from the minimal-lactose selection plate, first, a minority of the Lac − cells plated carried an antibiotic-resistance marker, and only resistant microcolonies were analyzed, and, second, sectored colonies observed were retested for instability (to eliminate sectored colonies that are not unstable, shown in Figure 1 C, which may result from accidental overlap of blue and white cfu). Procedures described in Materials and Methods , “whole microcolony analysis.” Cam R , chloramphenicol resistant; Kan R , kanamycin resistant; Tet R , tetracycline resistant. Streptomycin resistance (not shown) was also used. Table 1 Composition of Whole Microcolonies Microcolonies were harvested from five adaptive mutation experiments and analyzed as described in Figure 2 and Materials and Methods a Cells derived from these two cfu were Lac − (see text) 10.1371/journal.pbio.0020399.t001 We have improved the sensitivity of microcolony analysis substantially, first, by analyzing all cells in a microcolony (rather than a sample) and, second, by imposing stringent methods to reduce contamination of microcolonies with neighboring cells and microcolonies from the selection plate ( Materials and Methods ; Figure 2 ). First, frameshift-bearing (tester) cells were mixed with 3% each of three or four different derivative testers that carry antibiotic resistance markers ( Figure 2 A), and only microcolonies that carried one of these minority antibiotic resistance markers were analyzed to search for sectored colonies on medium carrying the antibiotic. This reduces, but does not eliminate, accidental overlap with Lac − background cells and with unrelated Lac + microcolonies accidentally harvested in the same sample (shown below). Second, because some sectored colonies are not unstable (e.g., mixtures of stable, white Lac − background cells, and blue colony formers that overlapped), all sectored colonies found were retested by plating a sample on the same X-gal antibiotic medium to determine whether there was continued instability indicative of amplification ( Figure 2 E). Third, because the sequential model predicts that the younger the microcolony is, the greater the proportion of lac -amplified cells (colony-forming units [cfu]) it should contain ( Hendrickson et al. 2002 ), and because one might miss the putative lac- amplified cfu by not examining enough cfu per microcolony, we analyzed very young microcolonies of 10 2 –10 4 cfu/microcolony (as compared with approximately 10 5 cfu/microcolony used by Hendrickson et al. [2002] ) and analyzed all cfu in each of them (rather than a sample) ( Figure 2 Table 1 ). The results of analyzing every cell in 18 whole microcolonies of stable cells, isolated on days 3 through 6 of adaptive mutation experiments, are shown in Table 1 . Although we found sectored colonies at frequencies of 10 −2 –10 −3 in almost all microcolonies (as reported by Hendrickson et al. 2002 ), these sectored colonies, when replated, were almost all found to consist of cells giving rise to stable white and stable blue colonies only, indicating that these were not from lac -amplified clones (also demonstrated below). In general, we looked at 50 to 100 cells from these replated sectored colonies, but, for three of them, we screened 3,000–4,000 cells with the same result: we could find no unstable cfu within them. In contrast, cells from lac -amplified colonies, upon replating, again give rise to sectored colonies. Therefore, most of the sectored colonies that we found give no evidence of carrying amplification of the lac locus. Sectoring in these colonies may occur by chance juxtaposition of Lac + point-mutant cells with non-mutated background cells (which are numerous) during the plating ( Figure 2 ). Other explanations might also be possible; however, these colonies are not lac- amplification bearers by this test and also as demonstrated in the following section. An example of such a colony is shown in Figure 1 C. Such colonies may account for the report of Hendrickson et al. (2002) that more than 10 −3 of cells in every Lac + microcolony are sectored colony formers. These sectored, non- lac -amplified colonies have also been reported by another laboratory ( Poteete et al. 2002 ). Of the 18 whole microcolonies examined ( Table 1 ), 16 were pure stable. One day-6 microcolony was found to contain seven unstable cfu. (Another, found on day 5, contained two cfu showing sectors of an extremely pale blue color along with sectors of the same color as the Lac − parental cells.) Those unstable cfu might be clonally related to the rest of the microcolony, or might result from chance juxtaposition of two different clones carrying the same antibiotic resistance marker. To determine the frequency at which such chance juxtapositions are expected, we analyzed the cells carried on the agar plugs on which microcolonies were harvested. We did this by plating the microcolony suspensions on medium containing an antibiotic to which cells of the microcolony were sensitive, to ascertain the frequency of marked Lac + cells that are unrelated to a microcolony but get harvested with it. The microcolony plugs contained from 900 to 7,000 antibiotic-resistant lac − background cells each. Of eleven such plugs analyzed, one contained three unrelated (different antibiotic resistance) point-mutant Lac + cfu, one contained one unrelated unstable cfu, and three contained unrelated cfu of the very pale sectored phenotype described above. The presence on agar plugs of cells giving rise to sectored colonies that are unrelated to the microcolony, but occur at a frequency comparable to the frequency of unstables in a microcolony of the same antibiotic-resistance genotype ( Table 1 , last line), implies that the presence of unstable cfu was independent of the microcolony, and thus not relevant to the origin of the microcolony. The very pale sectored colonies that we found were examined further and were shown to be unstable upon retesting but contained no cells capable of forming colonies on lactose medium under the conditions of an adaptive mutation experiment. Presumably they contain too few copies of the lac region to allow growth on lactose medium. They might carry lac gene duplication, or tandem amplification at a very low level. In support of this possibility, we note that their pale appearance on X-gal medium is similar to known lac gene duplications of this frameshift allele that we have constructed (A. Slack and P. J. Hastings, unpublished data). Another possibility is that they contain multiple F′ plasmids ( Foster and Rosche 1999 ). Their failure to form adaptive point-mutant or lac -amplified colonies shows that they do not advance to a Lac + phenotype efficiently (so are not obvious precursors to point mutation). Thus, colonies composed of a mixture of lac -amplified and point-mutant cells, as reported by Andersson et al. (1998) and Hendrickson et al. (2002) , are not common when stringent methods of scoring are imposed, and such mixed colonies are certainly not every microcolony in which a mutation has occurred. Those apparent mixed colonies that do occur are likely to be unrelated mixtures and not relevant to the origin of the Lac + point mutants in the colony. Selection for lac- Amplified Cells in Point-Mutant Microcolonies As an independent test for the presence of rare lac- amplified cells in point-mutant microcolonies, we devised a genetic selection capable of revealing single lac -amplified cells among numerous non-amplified cells. The selection depends on the quantitative relationship between the number of copies of the chloramphenicol acetyl transferase gene (cat) and the concentration of the antibiotic chloramphenicol that cells resist ( Petit et al. 1992 ). We inserted the cat gene into codA, about 5 kb from lac. Control experiments established that strains with lac and cat amplification (as shown by unstable phenotypes and Southern blot analyses) could grow on rich Luria-Bertani-Herskowitz medium (LBH) containing 100 μg/ml chloramphenicol, whereas Lac + point-mutant cells, and those amplified at lac but not at cat, did not form colonies at this concentration. We determined that 61% or more of all lac -amplified cells are detected by this selection ( Materials and Methods ). The data are shown in Table 2 . Among 75 Lac + point-mutant microcolonies isolated from days 4, 5, and 6 of an adaptive mutation experiment, three (4%) were found to contain a few chloramphenicol-resistant (CamR) cfu ( Table 2 ). All of these showed the lac- amplified colony phenotype. Correction for those that would not be detected by this selection (above; Materials and Methods ) shows that, at most, 6.6% of point-mutant microcolonies contain any lac -amplified cells (and probably fewer; see below) . As discussed above, this contrasts with the report of Hendrickson et al. (2002) that all Lac + microcolonies contain lac -amplified cells. Also, these data verify independently that the high levels of sectored, but not unstable, cfu found among microcolonies (see Table 1 ) were not lac -amplification bearers, because most lac- amplified cfu would have scored positively in this selection. Table 2 Selection for Rare Amplified Cells in Colonies of Point-Mutant Cells Results of selection for CamR cells in 152 microcolonies (10 3 to 2 × 10 4 cells each) and 70 control samples with no visible microcolony. The data show that amplified cells are no more common in microcolonies of point-mutant cells than they are in the negative control carrying only background cells (no microcolony). Numbers in parentheses are the numbers of CamR cfu found in each case. For unstable microcolonies, all cfu in any microcolony were either CamR or chloramphenocol sensitive 10.1371/journal.pbio.0020399.t002 A negative control was performed by plating samples of cells from agar plugs that did not carry a microcolony visible on a dissecting microscope. Five of 70 such samples yielded one or a few cells that were lac -amplified and showed CamR. The similarity of this proportion to that for lac -amplified cells in point-mutant microcolonies suggests that none of the mixtures detected represents an occurrence of lac -amplified cells that arose in the same cell clone as the point mutation. Thus, we conclude that few, if any, point-mutant microcolonies include lac -amplified cells. Nascent Lac + Point-Mutant Colonies Do Not Contain lac -Amplified Cells The selection method described above allowed us to detect amplification in samples of stressed cells that did not include a visible microcolony. This suggested the possibility of analyzing point mutation and amplification in nascent Lac + colonies as young as the two-cell stage and up to about 40 cells. The sequential amplification–mutagenesis model demands that all colonies as young as 2–40 cells must consist only of lac- amplified cells, because these are proposed not to generate Lac + point mutations until the number of lac copies reaches 10 7 or more (e.g., a colony of 10 5 cells with 100 lac copies per cell caused by amplification; Hendrickson et al. 2002 ). Cells from 600 sample plugs, harvested on days 3 to 5 in two experiments, that carried no microcolony visible at 20-fold magnification were analyzed by selecting for CamR in half the sample, and Lac + phenotype in the other half. The samples were found to contain 1 × 10 4 to 5 × 10 4 total cells. We detected 22 events by either Lac + or CamR selection (3.7% of the samples). This frequency of events is compatible with the observation of 30–60 Lac + colonies per 10 8 cells per day, which we see routinely in adaptive mutation experiments (e.g., Hastings et al. 2000 ). These events are listed in Table 3 . Table 3 Nascent Lac + Point-Mutant Colonies at the Few-Cell Stage Do Not Include Amplified Cells Number of cfu per sample appearing in two selections, half of each sample applied to each selection. Data for 600 samples of stressed cells that did not include a microcolony 10.1371/journal.pbio.0020399.t003 Twelve samples gave lac- amplified cells (cfu) detected as CamR. The cell numbers ranged from one to 16 cells (from half the sample). Two amplification events gave Lac + cells that were chloramphenicol sensitive, detected on lactose-minimal medium plates. Many more amplification events were detected by CamR selection than by growth on lactose medium (even though all were lac- amplified because of their being selected on lactose medium), implying that the CamR test is more sensitive, and capable of detecting levels of amplification that are too low to allow colony formation on lactose medium. Eight samples gave stable Lac + cells (cfu), ranging in number from one to 20 cells from half the sample (implying colonies ranging in size from 2–40 cells). None of these eight contained amplified cells, as determined by CamR, or by rescoring the lactose plates after 7 d, by which time any pre-existing lac- amplified cells would have formed colonies ( Hastings et al. 2000 ). Thus, we see that, even at the 2- to 40-cell stage of colony growth, we are unable to detect a correlation between mutation and amplification events, and we conclude that these are independent processes. These results are incompatible with the idea that Lac + adaptive point mutants result from point mutation events in young colonies of lac -amplified cells ( Andersson et al. 1998 ; Hendrickson et al. 2002 ). lac- Amplified Microcolonies Arise Later than Point-Mutant Microcolonies If point-mutant colonies arose from lac -amplified microcolonies, the majority of microcolonies of sufficiently small size would be expected to consist of lac -amplified cells. For example, in the sequential model of Hendrickson et al. (2002) , Lac + point mutation is expected to occur when a colony of lac -amplified cells grows to 10 5 cells (with a mutation rate of 10 −8 mutations per cell per generation that is enhanced 35-fold by SOS induction and acts on an additional 30 copies of lac; 35 × 30 × 10 5 × 10 −8 = 1 mutation per microcolony of 10 5 cells). This model predicts that lac -amplified microcolonies of approximately 10 5 cells should precede point-mutant colonies. As shown in Figure 3 , the microcolonies harvested in the course of these experiments, all much smaller than 10 5 cells, show a very distinct pattern that conflicts with this expectation. Only 5% of microcolonies isolated on day 3 consisted of lac -amplified cells (the rest were stable point mutant). The proportion of unstable microcolonies increases on later days. This is the pattern of arising of lac -amplified visible colonies, which are rare in the first few days of an experiment and increase in frequency later ( Hastings et al. 2000 ; Figure 3 ). The proportion of unstable microcolonies appears to reflect the proportion of visible colonies that would consist of lac -amplified cells 1 to 4 d later ( Figure 3 ), as expected if lac -amplified microcolonies produce lac -amplified (not point-mutant) visible colonies. Figure 3 Temporal Distribution of lac -Amplified Microcolonies Data from six adaptive mutation microcolony experiments were pooled to give the distribution of point-mutant and lac -amplified phenotypes. Squares, the mean percentage of Lac + microcolonies with unstable phenotype (± SEM; n = 3 experiments on day 3 and n = 4 on days 4, 5, and 6); diamonds, the percentage of lac -amplified visible colonies on the same days included for comparison (data from Hastings et al. 2000 ). Note that the observed proportion of unstable microcolonies is higher than the actual proportion because lac -amplified clones are slower growing than point mutants ( Hastings et al. 2000 ), and so spend more time as microcolonies before becoming visible colonies. Thus, the lack of unstable microcolonies on early days is even more severe than the data show. lac -Amplified Microcolonies Produce lac -Amplified, Not Point-Mutant, Visible Colonies The sequential amplification–mutagenesis model states that mutation occurs in a cell in a microcolony of lac -amplified cells ( Hendrickson et al. 2002 ). Growth of the point-mutant cell clone is proposed to overtake the lac -amplified cells to give a visible colony consisting mainly of point-mutant cells. We looked for this change from lac -amplified to point-mutant phenotype by removing a few cells from microcolonies to determine which microcolonies consisted of lac -amplified cells, and sampling them again after they had grown to form visible colonies ( Materials and Methods ). The data in Figure 4 show that we did not see this change occur. lac -amplified microcolonies grew into lac -amplified visible colonies, and point-mutant microcolonies remained point mutant. We conclude that microcolonies of lac -amplified cells do not become visible colonies of point-mutant cells. Figure 4 Fate of Microcolonies Developing In Situ Microcolonies on days 4 and 5 of separate adaptive mutation experiments were sampled by touching with a needle, then the sample was spread on rich X-gal medium and scored for point-mutant or sectored Lac + phenotype (see Materials and Methods ). The microcolonies were then left to grow into visible colonies (∼10 7 cells), and the colonies sampled again and samples plated to score for sectoring (hatched bars, day 4 microcolony samples; black bars, day 5 microcolony samples). The numbers of each type of colony observed are shown next to the bars. One single stable microcolony produced a mixed visible colony, which may have been caused by overlap with another colony or microcolony. Hendrickson et al. (2002) reported a seemingly similar experiment that gave the opposite result. They scored small (presumed young) and large (old) colonies and found many more unstable cfu in small colonies than in larger colonies. We note that they did not follow the same small colonies to score later, when those small colonies had grown larger, as we have done. Their result is probably caused by the slow growth rate of lac -amplified cells, which causes microcolonies of lac- amplified cells to remain at the microcolony size for longer than do microcolonies of point-mutant cells, such that many more small colonies are lac -amplified. We have found that point-mutant microcolonies stay small for a much briefer period of time than do lac -amplified ones (data not shown). We overcame this problem by scoring the same colonies as microcolonies and found that these small lac -amplified colonies do not grow to become point mutant ( Figure 4 ). An additional explanation for why the small colonies of Hendrickson et al. (2002) contained more lac- amplified cfu is that, in most of their experiments, these authors did not employ the stringent methods that we devised to avoid scoring unrelated mixtures of cells in microcolony samples. These include the antibiotic-resistant-minority-scoring method outlined in Figure 2 . Thus, when they picked small colonies in a fixed volume of agar, a greater fraction of the cells in their sample would be unrelated background cells than when they picked larger colonies (same number of background cells each time, but fewer or more cells from the colony they intended to sample). This, too, should result in the appearance of more mixtures of lac- amplified and point-mutant cfu in the small colony samples and more pure point mutant in the large colony samples; but the mixtures in the small colony samples would be unrelated. We conclude that when stringent colony-fate experiments are performed, lac- amplified microcolonies are seen to produce lac -amplified and not point-mutant visible colonies ( Figure 4 ). Point Mutants Do Not Overgrow lac -Amplified Cells in Competition Experiments We tested experimentally whether point-mutant cells occurring in a lac -amplified microcolony would be able to overgrow and mask the lac -amplified cells on the timescale of adaptive mutation experiments, as is postulated in sequential models ( Hendrickson et al. 2002 ). Cultures of lac -amplified cells were seeded with Lac + point-mutant cells in various proportions and grown for a known number of cell generations, as spots on lactose plates, to form pseudocolonies. The proportion of lac -amplified and point-mutant cells was determined after 3 d, by which time the pseudocolony had reached 8 × 10 7 to 1.6 × 10 8 cells ( Table 4 ). This is three to four more generations than are needed to form a visible colony (approximately 0.5–1 × 10 7 cells; Hastings et al. 2000 ). These experiments were performed with two different lac -amplified isolates that have 2.0 and 4.2 times the generation time on lactose medium of the point-mutant Lac + strain used. That is, these strains can produce 2 and 4.2 times fewer generations than point mutants, and so might, in principle, be overtaken by point mutants during colony growth. However, this was not observed under the real-life conditions of growth in colonies under selective conditions. Only when point-mutant cells were introduced at a ratio of one to ten was the proportion of lac -amplified cells reduced to below 1% (the level reported in mixed colonies by Hendrickson et al. [2002] ), and then only with the slower growing lac -amplified isolate ( Table 4 ). We did not see the point-mutant cells take over when the point mutant started at 10 −5 of the cells, the predicted time of occurrence of a mutation ( Hendrickson et al. 2002 ). Similar results were obtained when the experiment was performed in liquid culture (data not shown). Table 4 Competition between Two lac -Amplified Isolates and a Point-Mutant Lac + Strain Lac + point-mutant cells were seeded into suspensions of lac -amplified cells of two different isolates with different growth rates, at dilutions from 10 −5 to 10 −1 . Cell numbers ranging from 10 2 to 10 5 cells were spotted as 0.5-μl spots on M9 lactose plates to form pseudocolonies. After 3 d of growth, the pseudocolonies were suspended, and dilutions of cells plated to determine cell number and the proportion of Lac + point-mutant and lac -amplified cells. Each number is the mean ± SEM of three parallel cultures with two pseudocolonies per culture. This experiment was performed three times on plates (as above) and twice in liquid culture, giving comparable results each time 10.1371/journal.pbio.0020399.t004 We conclude that overgrowth of colonies of lac -amplified cells by point-mutant cells does not occur as postulated in the sequential, amplification–mutagenesis model (in which one point-mutant cell is proposed to overgrow 10 5 lac -amplified cells; Hendrickson et al. 2002 ). The only circumstance under which overgrowth might be seen would be if the mutation occurred in the first ten cells of a microcolony. With only approximately 300 lac copies present at the ten-cell stage, this mutation frequency, approximately 3 × 10 −3 mutations per lac copy, would be untenably high in any current mutation model. Amplification and Not Point Mutation Requires DNA Pol I If adaptive amplification were a prerequisite for point mutation, then all genetic requirements for amplification should also be required for point mutation. We report the first genetic requirement found for amplification that is not also required for point mutation: Pol I. Pol I, encoded by polA, is required for most amplification ( Figure 5 A), but adaptive mutation is increased in a polA mutant ( Figure 5 C). We used the temperature-sensitive polA12 (polA[Ts]) allele incubated at semi-permissive temperature, because polA null mutation causes inviability in this strain background (Harris 1997). In repeat experiments, we saw a 2- to 6-fold reduction in amplification rate. In the same experiments, adaptive mutation was unchanged or increased by up to 8-fold. Figure 5 Pol I Is Required for Adaptive Amplification and Not Point Mutation Strains were plated on lactose-minimal medium and Lac + colonies counted daily (see Materials and Methods ). The plots are cumulative, showing the mean of 3–4 cultures with one SEM. Strains used: FC40 dinB + polA + fadAB ::Tn 10 Kan , SMR3490 (squares); FC40 dinB + polA12 (Ts) fadAB ::Tn 10 Kan , SMR3491 (diamonds); FC40 dinB10 polA12 (Ts) fadAB ::Tn 10 Kan , PJH308 and PJH309 (triangles, inverted triangles); and FC40 dinB10 pol + fadAB ::Tn 10 Kan , PJH310 (circles). All cultures were grown at 30 °C and the experiments conducted at 37 °C. (A) An example of the effect of polA (Ts) on the yield of lac- amplified colonies, showing a partial requirement for polA at a semi-permissive temperature. (B) and (C) show the effect of the dinB10 mutation on adaptive lac- amplification and point mutation, respectively. The dinB polA (Ts) cells display the decreased lac amplification of the polA mutant (B), and the decreased point mutation characteristic of the dinB mutant (C), demonstrating that the decrease in lac amplification rate does not result from channeling of lac -amplified cells into a point mutation pathway. These data (C) also show that the absence of Pol I increases point mutation (reported previously, Harris 1997 ) in a completely DinB-dependent manner. This could occur via the absence of Pol I leading to SOS induction ( Bates et al. 1989 ) and more DinB/Pol IV, or via relief of a competition between high-fidelity Pol I and error-prone Pol IV at the replisome. Neither of these ways should affect lac amplification, which is Pol IV–independent, as observed (B). The possibility that amplification occurs in the Pol I–deficient strain but that the Lac + colonies are unable to grow to visibility in the time-span of an adaptive mutation experiment is ruled out by reconstruction experiments, in which three independent lac- amplified isolates carrying polA (Ts) or polA + formed visible colonies in 5.25 ± 0.17 d and 5.39 ± 0.24 d, respectively (mean of three strains ± standard error of the mean [SEM]). Another possibility that we considered is that the polA mutation, which causes an increase in SOS induction ( Bates et al. 1989 ), might channel amplified DNA into the point mutation pathway by increasing SOS-induced mutagenesis, so that there is an increase in mutation concomitant with a decrease in amplification. We were unable to test this directly by preventing SOS induction, because the non-inducible allele of lexA confers low viability in combination with polA (Ts) (data not shown) . However we tested this possibility using dinB -defective cells. DinB/Pol IV accounts for most, if not all, of the requirement of adaptive mutation for SOS induction ( McKenzie et al. 2001 ). The results are shown in Figure 5 Two independent isolates of the double mutant dinB polA (Ts) show the reduction in point mutation ( Figure 5 C) expected of dinB ( McKenzie et al. 2001 ) and the reduction in amplification ( Figure 5 B) reported above for polA (Ts) . This demonstrates that the decrease in lac- amplified colonies in polA (Ts) strains is not due to conversion/channeling of lac -amplified cells into point mutants, but rather is caused by reduction of the number of lac -amplified cells formed. We conclude that Pol I is required specifically for adaptive amplification and not point mutation. These data are incompatible with models in which amplification is a prerequisite for point mutation, and support models in which point mutation is an independent mechanism. Amplification Does Not Induce SOS A final aspect of the sequential model of Hendrickson et al. (2002) was tested. To explain the high frequency of unselected secondary mutations observed in Lac + adaptive point mutants ( Torkelson et al. 1997 ; Rosche and Foster 1999 ; Godoy et al. 2000 ), Hendrickson et al. proposed that amplified DNA itself is sufficient to induce a chronic SOS response leading to increased mutation, and that, therefore, the SOS regulon should be induced in all or most cells carrying amplification. By contrast, hypermutation models (e.g., Rosenberg 2001 ; Hersh et al. 2004 ; Rosenberg and Hastings 2004a ) suggest that a hypermutable state, possibly an SOS response, is induced only transiently in a subpopulation of starved cells, and that this condition will not persist for long in cells that are able to utilize lactose because they no longer experience the stress of starvation. To determine whether the involvement of SOS in adaptive point mutation ( Cairns and Foster 1991 ; McKenzie et al. 2000 ) is a consequence of amplification, we constructed a strain in which the green fluorescent protein (GFP) gene is expressed under the control of the SOS-regulated sulA promoter (see Figure 1 E and 1 F). sulA is one of about 40 SOS (DNA damage inducible) genes in E. coli ( Courcelle et al. 2001 ). Figure 6 A shows that about 2% of cells in the growing cultures showed spontaneous SOS induction, as revealed by GFP fluorescence. There are no detectable differences in this proportion between Lac + point-mutant and lac -amplified cultures, or between cultures grown with (lactose medium) or without (glycerol medium) selection for lactose utilization. Control experiments showed that 3 h after exposure to 20 J/m 2 of ultraviolet light, all cells were induced for SOS (data not shown). Thus, amplification per se does not induce an SOS response. Figure 6 DNA Amplification Does Not Induce the SOS Response (A) Known lac -amplified and point-mutant (control) derivatives of SMR6039 were grown in liquid M9 medium containing either lactose or glycerol. Mid-logarithmic phase cells were harvested and scored microscopically for GFP fluorescence, using an Olympus BL51 microscope mounted with a mercury lamp UV source and a High Q Endow GFP emission fluorescence filter cube. Some 1,000–2,000 cells from each of 4–10 fields were scored per determination. Error bars indicate one SEM for 8–13 cultures as indicated. (B) Microcolonies were harvested and suspended in 500 μl of buffer, 50 μl of which was spread on LBH X-gal rif solid medium to determine sectoring in resulting colonies. The remainders were concentrated and examined microscopically for GFP fluorescence, counting 60–400 cells per microcolony. These experiments were plated at very low cell density to avoid significant numbers of background Lac − cells being harvested with the microcolonies. Open bars, fraction of stable Lac + isolates (32 microcolonies); black bars, fraction of sectored isolates (11 microcolonies). The two distributions do not differ ( p = 0.8 by Student's t -test). These experiments were repeated, giving similar results. We also studied the persistence of SOS induction in microcolonies during an adaptive mutation experiment. We isolated microcolonies from lactose plates, spread a sample on LBH (rich) X-gal rifampicin (rif) plates to determine which colonies were point mutant and which lac -amplified, and screened other cells from the microcolonies for GFP production. The data presented in Figure 6 B show that there is no difference between point-mutant and lac -amplified microcolonies in the distribution of cells producing GFP ( p = 0.8 by Student's t -test) and that neither shows persistent SOS induction. Thus, even under adaptive mutation experimental conditions, amplification is not sufficient to induce the SOS regulon. The results also show that SOS induction does not persist for many generations after stress has been released. The data are compatible with hypermutation models for adaptive mutation ( Rosenberg 2001 ; Figure 7 ), wherein the SOS response is induced transiently during starvation, and are incompatible with the postulate that chronic SOS induction is caused by amplified DNA. Figure 7 Stress-Response Models for Adaptive Amplification and Point Mutation Modified from Lombardo et al. (2004) and Rosenberg and Hastings (2004a) . DSBR, DSB repair; hypermutation, increased mutation at lac and elsewhere. The origins of DSEs during starvation on lactose medium in this assay system are unknown, and possibilities are reviewed elsewhere ( Rosenberg 2001 ). On the F′ plasmid carrying the lac gene, DSEs may be frequent and may be derived from chronic single-strand nicks at the origin of transfer. However chromosomal mutation during starvation on lactose medium in these cells also requires DSB repair proteins ( Bull et al. 2001 ), and so probably results from (lower levels of) DSEs generated in the chromosome, independently of the transfer origin, or perhaps by transient integration of the F′ into the chromosome (Hfr formation). Both adaptive point mutation and amplification are proposed to be outcomes of RpoS-dependent stress response, which both mechanisms require ( Lombardo et al. 2004 ), and to result from alternative error-prone ways of repairing DSBs. Further Discussion Amplification is a separate genetic endpoint that allows growth under selection The experiments reported here were designed to determine whether DNA amplification is an intermediate in the formation of adaptive point mutations in the Lac system, or whether the two are independent end points of genetic change under stress, each of which adapts cells to their environment. All of the data presented support the idea that lac amplification and point mutation are alternative end points, either of which allows growth on lactose. These data are reviewed conveniently by contrast with the sequential, amplification–mutagenesis model, as follows. Significance and predictions of amplification–mutagenesis The experiments reported test the amplification–mutagenesis model for the mechanism of adaptive mutation in the E. coli Lac system ( Andersson et al. 1998 ; Hendrickson et al. 2002 ). The model's significance lies in its apparent explanation of adaptive mutation phenomenology without invoking an environmentally induced increase in mutation rate, in adherence to conservative neo-Darwinist ideas about constant, gradual evolutionary change (e.g., Mayr 1982 ). In the model, spontaneous amplification of lac allows growth of cells into microcolonies on lactose medium; replication of the multiple lac copies generates a point mutant at about the 10 5 -cell stage; and the point-mutant cell overgrows the lac -amplified cells, leading to colonies consisting mostly of point-mutant cells with a few (10 −2 –10 −3 ) lac -amplified cells. This model derives most of its support from the observation of sectored-colony-forming cells in colonies of point-mutant cells ( Andersson et al. 1998 ; Hendrickson et al. 2002 ). Independent point mutation and amplification mechanisms supported By contrast, we found, first, that when stringent scoring methods are applied, and all cells in very young microcolonies analyzed, most point-mutant Lac + colonies are not mixtures of stable point-mutant cells with 10 −2 –10 −3 unstable (amplification bearing) cells, but rather are pure (see Table 1 ). This was verified with sensitive genetic selection methods that can detect single lac- amplified cells among microcolonies (see Table 2 ), and was found to be true even at the two- to 40-cell stage of colony growth (see Table 3 ). We conclude that adaptive point mutants arise directly from lac − single cells without an intermediate stage involving colonies of lac- amplified cells, and that the lac- amplified cells are a separate adaptive outcome. Previous observations of mixed colonies ( Andersson et al. 1998 ; Hendrickson et al. 2002 ) seem likely to have arisen from accidental overlap of point-mutant and Lac − colonies that were not related, which would cause sectored, but not unstable colonies. We observed these, recapitulating the previous results (see Table 1 ; Figure 1 C; Poteete et al. 2002 ), only when we failed to apply stringent criteria for identification of lac -amplified cells. Such colonies might also have arisen from overlap of stable with unrelated unstable microcolonies, which were not screened out stringently ( Hendrickson et al. 2002 ), as done here (see Figure 2 ; Table 1 ). Second, we found that microcolonies smaller than 10 5 cells are not mostly lac -amplified, as the amplification–mutagenesis model demands, but rather that microcolonies carrying amplification are rare initially and frequent later, with kinetics that anticipate the arising of lac -amplified visible colonies (See Figure 3 ). Third, when sampled early and late, young microcolonies with lac amplification give rise to lac -amplified, not point-mutant, visible colonies (see Figure 4 ). Fourth, artificial mixtures of point-mutant with lac -amplified cells do not show the predicted takeover of the amplification carriers by the point mutants (one point mutant overtaking 10 5 lac -amplified, as predicted Hendrickson et al. [2002] and Roth et al. [2003] ) unless the point mutants comprise a very large fraction (10%) of the total (see Table 4 ). Fifth, we show that DNA Pol I is required for lac amplification and not point mutation (see Figure 5 ), demonstrating that lac amplification is not a prerequisite for point mutation. We conclude that amplification is an independent outcome that adapts cells, not a transient intermediate en route to point mutation. Amplification does not induce SOS To explain the genome-wide hypermutation observed in Lac + point mutants ( Torkelson et al. 1997 ; Rosche and Foster 1999 ; Godoy et al. 2000 ) with a sequential model, Hendrickson et al. (2002) suggested that amplified DNA per se induces an SOS response and increased mutability. Using an SOS-inducible gfp reporter gene, we find that amplified DNA does not induce SOS (see Figure 6 ). This conclusion was also supported previously by the observation that cells carrying amplification do not display general hypermutation ( Hastings et al. 2000 ). Apparent support for sequential model compatible with independent pathways An additional result previously interpreted as support for sequential models was that when a gene counter-selectable at more than one copy was placed next to lac, point mutation was decreased ( Hendrickson et al. 2002 ). However, we note that this is predicted by our non-sequential hypermutation model for point mutation (below) as well, in which point mutation is proposed to result from error-prone DNA double-strand-break (DSB) repair, because more than one copy of lac must be present to allow repair by homologous recombination. This copy could be on another DNA molecule (a sister chromosome) or the same one (a duplication), but either way it is required for error-prone repair leading to mutation, and its counter-selection would decrease point mutation, even though amplification is not an intermediate in the process. Branched- or independent-pathway, stress-response models for adaptive amplification and point mutation In Figure 7 , we outline a model supported by previous and current data. In this model, amplification and point mutation result from alternative modes of error-prone DNA DSB repair caused as stress responses to starvation. Adaptive point mutation requires homologous recombination proteins, including those specific to DSB repair, implicating DSBs or double-strand ends (DSEs) as a molecular intermediate ( Harris et al. 1994 ). We suggest that adaptive point mutations form by homologous recombinational DSB/DBE repair that is error prone due to use of the SOS-inducible ( Kim et al. 1997 ) and starvation-inducible ( Layton and Foster 2003 ) error-prone DNA polymerase, Pol IV/DinB. The SOS response ( Cairns and Foster 1991 ; McKenzie et al. 2000 ) and DinB are required for most (85%) adaptive point mutation and not for lac amplification in E. coli ( McKenzie et al. 2001 ). Point mutants, but not lac -amplified cells ( Hastings et al. 2000 ), carry high levels of mutation of genes throughout their genome ( Torkelson et al. 1997 ; Rosche and Foster 1999 ; Godoy et al. 2000 ). Because these levels of mutation are not found in most Lac − cells starved on the same plates, we infer that only a subpopulation of the starving cells experiences this transient hypermutation ( Torkelson et al. 1997 ), perhaps those that are induced for the SOS response ( McKenzie et al. 2000 ) or the SOS and RpoS stress response (which is required, discussed below) (but see Foster 2004 ; Rosenberg and Hastings 2004a ). Mismatch repair, which corrects replication errors, also becomes limiting specifically during adaptive mutation, leading to increased mutagenesis ( Harris et al. 1997 ); this might occur by titration of mismatch repair proteins by excess DNA polymerase errors ( Harris et al. 1997 ) made by Pol IV ( McKenzie et al. 2001 ). This model, which includes elevated rates of general (not lac -specific) genetic change in response to stress, is compatible with Darwinism, which also encompasses changing rates of heritable variations ( Darwin 1859 ), but not with the conservative neo-Darwinist constraint of gradual evolutionary change (e.g., Mayr 1982 ), a major impetus for the sequential, amplification–mutagenesis model ( Andersson et al. 1998 ; Hendrickson et al. 2002 ). Amplification mechanism and the role of Pol I Amplification, on the other hand, could be formed by non-homologous repair of the DSEs ( Figure 7 ), leading to a gene duplication or rolling-circle replication, either of which might amplify rapidly to produce the 20–100 copies of lac that allow growth without a point mutation ( Hastings and Rosenberg 2002 ). Alternatively, homologous DSE repair might produce replication forks that lack some of the control of origin-initiated replication and stall in the nucleotide-depleted milieu of starvation. We suggest that template switching can occur at a stalled fork and produce a duplicated-DNA segment or a rolling-circle intermediate from which amplification results. This is similar to template-switching recombination models ( Bzymek and Lovett 2001 ). This template-switching event might occur preferentially with Pol I, which is capable of template switching in vitro ( Kornberg and Baker 1992 ). The partial requirement for Pol I might result from the use of semi-permissive conditions for a conditional mutant. Acquisition of lac amplification, and thus the ability to grow on lactose, alleviates the stress and deflects lac -amplified cells from the point mutation pathway, making it a (more or less) stable, alternative outcome ( Figure 7 ). There is no appearance of directed mutation in the lac system Unlike the hypermutable state model of Hall (1990) , our model ( Figure 7 ) does not specify that cells that do not acquire an adaptive point mutation or amplification die. That feature would be necessary for explaining a lack of mutations in genes other than lac (“directed” mutation; Cairns et al. 1988 ). However, in the E. coli Lac assay, exposure to selection conditions generates (non-adaptive) mutations in other genes efficiently—and by the same recombination-protein- and DinB-dependent mechanism as Lac + adaptive point mutation—both in the F′ that carries lac ( Foster 1997 ) and the bacterial chromosome ( Bull et al. 2001 ). A report to the contrary used a different bacterium (Salmonella) with many different genetic features that might not parallel the E. coli Lac assay ( Slechta et al. 2002 ). Adaptive point mutation and amplification in the lac system are stress responses That both adaptive point mutation and amplification are outcomes of a stress response is supported by their requirements for the general-stress-response and stationary-phase transcription factor RpoS (σ S ) ( Lombardo et al. 2004 ). RpoS upregulates approximately 50 genes in response to starvation, oxidation, pH, heat, and osmotic stresses. RpoS upregulates error-prone Pol IV ( Layton and Foster 2003 ), which might explain its role in the Pol IV–dependent process of adaptive point mutation ( McKenzie et al. 2001 ). However RpoS must play some other role in adaptive amplification, which is Pol IV–independent ( McKenzie et al. 2001 ). RpoS is also required for a Pol IV–independent, starvation-induced hypermutation mechanism in a natural isolate of E. coli ( Bjedov et al. 2003 ) and in other adaptive mutation pathways ( Gomez-Gomez et al. 1997 ; Saumaa et al. 2002 ), implying that this is a general feature of starvation-induced mutability. Thus, several mechanisms of adaptive genetic change appear to require a differentiated, stress-response condition of the cells, most probably stationary-phase cell physiology. This further supports the idea of stress-induced increases in mutation rate generally. Mutation rates are increased in starving/stressed cells generally Although it has been argued that increasing mutation generally in response to stress is implausible ( Roth et al. 2003 ), a recent study found that starvation-induced general mutability is the norm in natural isolates of E. coli ( Bjedov et al. 2003 ). The authors found that the vast majority of 787 natural isolates from diverse habitats worldwide produced random, unselected mutations in response to starvation. Multiple mechanisms of mutation in the natural isolates are implied, but at least one of them has similarities to that of the Lac system, and to a different stationary-phase-mutation model system (reviewed by Rosenberg and Hastings 2003 ), in that these mutation responses require RecA, an SOS-regulated DNA polymerase, and RpoS ( Bjedov et al. 2003 ). Moreover, in the eukaryote Caenorhabditis elegans, a new study of mutation ( Denver et al. 2004 ) suggests that cellular stress responses might provoke hypermutation generally, and also lead to a mismatch-repair-compromised transient state ( Rosenberg and Hastings 2004c ) similar to that suggested here. These systems support the idea that evolution might be hastened during stress. They promise to reveal mutation mechanisms that are likely to pertain to cancer formation and progression, acquisition of drug resistance in pathogens and tumors, and many processes in which clonal expansion under stress or growth limitation follows from an adaptive genetic change. Materials and Methods Strains E. coli strains used are isogenic with FC40 ( Cairns and Foster 1991 ) (also SMR4562, an independent construction of FC40; McKenzie et al. 2000 ), the lac -frameshift-bearing strain in which adaptive mutation and amplification are assayed, and FC29, a non-revertible lac deletion strain used to scavenge other carbon sources from the lactose plates ( Cairns and Foster 1991 ). Antibiotic-resistant derivatives SMR4481 (FC40 malB ::Tn 10 dTet), SMR3598 (FC40 zff3139 ::Tn 10 Kan), SMR533 (FC40 malB ::Tn 9 ), and PJH223, a spontaneous streptomycin-resistant mutant of SMR4562, are resistant to 10 μg/ml tetracycline, 30 μg/ml kanamycin, 25 μg/ml chloramphenicol, and 100 μg/ml streptomycin, respectively. SMR4481 was made by transposition of Tn 10 dTet from lambda NK1323 ( Kleckner et al. 1991 ). SMR533 and SMR3598 were constructed by transduction of FC40 with phage P1 grown on FS2055 (F.W. Stahl, University of Oregon) carrying malB ::Tn 9, and STL1605 (S.T. Lovett, Brandeis University) carrying zff3139 ::Tn 10 Kan, respectively. SMR4562 codA :: cat (PJH220) was constructed by linear replacement into SMR4562[pKD46] (plasmid from Datsenko and Wanner 2000 ) using the cat gene from pKRP10 ( Reece and Phillips 1995 ). The isogenic strain pair FC40 fadAB3165 ::Tn 10 Kan (SMR3490) and FC40 polA12 (Ts) fadAB3165 ::Tn 10 Kan (SMR3491) were constructed by P1 transduction of FC36 ( Cairns and Foster 1991 ) with a P1 lysate carrying fadAB3165 ::Tn 10 Kan from CAG18495 ( Singer et al. 1989 ), or with P1 grown on a strain with fadAB3165 ::Tn 10 Kan from CAG18495 linked with polA12 (Ts) ( Monk and Kinross 1972 , via F.W. Stahl, University of Oregon), respectively, followed by transfer of F′128 from FC40 by conjugation. PJH308 and PJH309 were made by P1 transduction of SMR5380 ( McKenzie et al. 2001 ) with lysate from SMR3491 and screened for sensitivity to ultra-violet light (UV) to give two separate UV-sensitive isolates of FC40 polA (Ts) fadAB3165 ::Tn 10 Kan dinB10 and the UV-resistant isogenic control strain PJH310 (FC40 fadAB3165 ::Tn 10 Kan dinB10 ). polA mutant strains were grown at 30 °C, and experiments were carried out at 37 °C (a semi-permissive temperature). SMR6039 is a derivative of SMR4562 carrying Δ attλ ::P sulAΩgfp-mut2: the gfp-mut2 reporter gene ( Cormack et al. 1996 ) under the control of the LexA-regulated sulA promoter replacing the phage attλ site in the E. coli chromosome (as per Gumbiner-Russo et al. 2001 ; see below). Construction of an SOS-GFP reporter The gfp - mut2 gene ( Cormack et al. 1996 ) was inserted into plasmid pACYC184 (New England Biolabs, Beverly, Massachusetts, United States), at the BamHI and HindIII restriction sites, to create pJP21. To fuse the sulA promoter to the gfp gene, two 5′ phosphorylated oligonucleotides corresponding to the top and bottom DNA fragments of the sulA − 1 to −67 region were synthesized: BspHI-pSulA-BamHI-top, 5′-P-CATGAGGGTTGATCTTTGTTGTCACTGGATGTACTGTACATCCATACAGTAACTCACAGGGGCTGGATTGATTG-3′ and BspHI-pSulA-BamHI-bot, 5′-P-GATCCAATCAATCCAGCCCCTGTGAGTTACTGTATGGATGTACAGTACATCCAGTGACAACAAAGATCAACCCCT-3′. When these are annealed, a BspHI-complementary end is reconstituted at one end and a BamHI-complementary end at the other. This fragment was ligated into BspHI /BamHI-digested pJP21 to create pJP23. A SphI fragment from pJP23, containing the P sulA Ω gfp-mut2 fusion, was isolated and ligated into SphI-digested pTGV-Light ( Gumbiner-Russo et al. 2001 ) to create pJP30. pJP30 was digested with AscI to generate a linear fragment for chromosomal integration replacing the att λ site ( Gumbiner-Russo et al. 2001 ) with the P sulA Ω gfp-mut2 fusion, Δ att λ::P sulA Ω gfp-mut2 . Phage P1 grown on this strain was used to construct SMR6039, an SMR4562 derivative carrying Δ att λ::P sulA Ω gfp-mut2 , by transduction into SMR5084, genotype SMR4562(λ xis1 c I ts857 Ind), as per ( Gumbiner-Russo et al. 2001 ). Adaptive mutation experiments Adaptive mutation experiments were performed as per Harris et al. (1996) , with the following variations for microcolony experiments: lactose plates were prescavenged by plating 5 × 10 9 FC29 “scavenger” cells on them and incubating at 37 °C for 24 h, then washing the cells off with M9 buffer; plating of frameshift-bearing cells was not in top agar, but on the plate's surface; growth of Lac − cells throughout the experiments was monitored either by sampling the lawn as described ( Cairns and Foster 1991 ; Harris et al. 1994 , 1996 ) or by microscopic examination. Microscopic examination used 50-fold magnification on a dissecting microscope. The lawn was seen to be smooth at this magnification when less than about 2-fold growth had occurred (number of viable cells relative to starting viable cell measurement). Plates on which growth occurred acquired a punctate appearance. This method was validated by using the two methods in parallel in two experiments, the largest of which involved 24 plates from two cultures. The correlation between growth detected as a doubling in cell number obtained from an agar plug, and the appearance of puncti was complete. No growth (≤ 2-fold) was detected during the relevant times in the experiments reported. For quantitative measure of mutation and amplification, three or four parallel cultures of each genotype were scored for Lac + colonies each day. Cells from 20 colonies from each culture, or 40 colonies in the case of SMR3491, were plated on LBH rif X-gal to determine the proportion amplified. Whole microcolony analysis Frameshift-bearing (tester) cells were mixed with 3% each of three or four different derivative testers that carried antibiotic-resistance markers (see Figure 2 ). Microcolonies were found by use of a dissecting microscope, and harvested on a plug of agar with a capillary tube (see Figure 2 ). Working at 25× magnification on a dissecting microscope, we were able to find and harvest very young colonies (see Figure 1 D), down to a size of 10 2 cells. The cells were suspended in 500 μl M9 buffer. One percent of the suspension was plated on LBH plates ( Torkelson et al. 1997 ) containing 40 μg/ml X-gal and 100 μg/ml rif (to prevent growth of scavenger cells). The remainder of the suspension was diluted with glycerol (to 12.5%) and stored at −80 °C. Many of the cells in any plating of a harvested microcolony were background (white colony forming) Lac − tester cells that were on the agar plug (see Figure 2 ). The cells from the microcolony either formed solid blue colonies on LBH X-gal rif plates (see Figure 1 A), indicating a point-mutant colony, or sectored blue colonies (see Figure 1 B), indicating an unstable lac -amplified colony ( Hastings et al. 2000 ). The colonies were scored for sectoring, counted, and patched from the LBH X-gal rif plates onto plates containing antibiotics. Those microcolonies that showed a stable Lac + point-mutant phenotype, had a suitable cell number, and carried one of the minority antibiotic-resistance markers were then plated in toto from the frozen suspension onto LBH X-gal rif medium with the appropriate antibiotic, to search for sectored colonies. Because some sectored colonies are not unstable (e.g., mixtures of stable white and blue colonies that overlap; see Results/Discussion and Table 1 ), all sectored colonies found were retested by plating a sample on the same LBH X-gal antibiotic medium to determine whether there was continued instability indicative of amplification. Harvesting the microcolony on a plug of agar with a 25-μl capillary pipette also picks up many Lac − tester cells. The number of these that form colonies during the analysis is reduced by use of antibiotics (as described above) because most cells are not resistant, but the background of tester cells may still be equal in number to those in the microcolony (see Figure 2 C). Use of a smaller caliber tube was found to risk losing microcolony cells on the wall of the tube. Selection for lac -amplified cells Microcolonies of PJH220 were harvested as above and suspended in 200 μl of buffer. One microliter (0.5%) of the suspension was spread on LBH X-gal rif medium to determine cell number and which microcolonies were point mutant. Point-mutant microcolonies were then plated to select CamR on LBH plates containing X-gal, rif, and 100 μg/ml chloramphenicol. Plates were scored after 15 to 20 h at 37 °C. To analyze background cells without microcolonies, plugs of agar were taken from areas of the plates where no microcolony was visible at 20× magnification and the cells suspended in 200 μl of buffer. Half of each suspension was plated on chloramphenicol-containing medium as above, and the other half was plated with scavenger cells onto lactose-minimal medium. Chloramphenicol plates were scored between 15 and 18 h, and lactose plates were scored at 2 d, and again at 7 d. Colonies appearing on lactose medium were scored for amplification by plating a sample of the cells on LBH X-gal rif medium. Colonies appearing on chloramphenicol-containing medium were scored for lac amplification (blue and white sectoring) directly on that plate. All CamR colonies were seen to be lac -amplified. The fraction of lac -amplified cells detectable by CamR selection was determined as follows. There are two reasons why lac amplification might fail to be detected with the CamR selection: first, because some lac amplifications have not co-amplified the neighboring codA :: cat gene, and, second, because some cat- amplified microcolonies obtained from the lactose plates (grown selectively for Lac + and not for CamR) might have too few cat copies for growth on 100 μg/ml chloramphenicol. We determined each of these separately. We found that 59 of 77 lac -amplified microcolonies were CamR (at least 77% had co-amplified cat ), and that 84% of cells from CamR ( cat -amplified) isolates grown on lactose without chloramphenicol were then able to form colonies on chloramphenicol medium. The cells that did not form colonies on chloramphenicol presumably had lost or reduced their amplification of cat and lac. This would indicate that 77% of 84% (65%) of all lac -amplified cells should be detected by the CamR selection. An independent determination of the number of cat -amplified cells that grow on chloramphenicol after growth on lactose was obtained in a reconstruction experiment in which a few CamR cells were plated with a 10 4 -fold excess of Lac + point-mutant cells, mimicking more closely the conditions of the experiments in which microcolonies are obtained. A total of 79% of the cat- amplified cells formed colonies on chloramphenicol medium, so 77% of 79% (61%) of all lac -amplified cells should be detected. The selection for CamR is actually a more sensitive assay for lac amplification than is selection for Lac + , as shown by the fact that only 60% of cells from the cultures that showed CamR were able to form colonies on lactose medium under the conditions of an adaptive mutation experiment, though all carried lac amplification (indicating that 40% had too few lac copies for growth on lactose), whereas 79% were detected on chloramphenicol medium. Colony-fate experiments Microcolonies were touched with the point of a needle, and cells were transferred to LBH X-gal rif plates, where they were spread by streaking for scoring. Control experiments (not shown) showed that this procedure removed, on average, 5% of the microcolony from the plate. About 50 separate colonies from each microcolony were scored for sectoring. The remainder of the colony was left for 1 to 4 d to form a visible colony in situ. The visible colony was then sampled, plated, and scored for sectoring (see “whole microcolony analysis” above). Competition experiments M9 lactose liquid cultures of Lac + point-mutant and lac -amplified cells were grown (separately) to saturation, then mixed in various proportions, and 0.5 μl of the mixed cell suspensions was put onto the surface of prescavenged lactose plates and grown into visible colonies. These “pseudocolonies” were suspended, and cells were spread on LBH X-gal rif plates at 100 to 200 cells per plate and scored for sectoring blue phenotypes after 2 d growth. Supporting Information Accession Numbers The Swiss-Prot ( http://www.ebi.ac.uk/swissprot/ ) and Ecogene ( http://bmb.med.miami.edu/EcoGene/EcoWeb/ ) accession numbers for the genes and gene products discussed in this paper are cat (Swiss-Prot P62577), codA (Swiss-Prot P25524), DNA Pol I (Swiss-Prot P00582), DNA Pol IV/ dinB (Swiss-Prot Q47155), fadAB (Swiss-Prot P21177), malB ( Swiss-Prot P02943), gfp-mut2 (Swiss-Prot P42212), lac (Swiss-Prot P00722 and P03023), recA (Swiss-Prot P03017), recBC (Swiss-Prot P08394 and EcoGene EG10825), rpoS (Swiss-Prot P13445), ruvA (Ecogene EG10923), ruvB (Ecogene EG10924), ruvC (Ecogene EG10925), and sulA (Swiss-Prot P08846).	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC529313.xml
548510	Inhibition of early steps in the lentiviral replication cycle by cathelicidin host defense peptides	Background The antibacterial activity of host defense peptides (HDP) is largely mediated by permeabilization of bacterial membranes. The lipid membrane of enveloped viruses might also be a target of antimicrobial peptides. Therefore, we screened a panel of naturally occurring HDPs representing different classes for inhibition of early, Env-independent steps in the HIV replication cycle. A lentiviral vector-based screening assay was used to determine the inhibitory effect of HDPs on early steps in the replication cycle and on cell metabolism. Results Human LL37 and porcine Protegrin-1 specifically reduced lentiviral vector infectivity, whereas the reduction of luciferase activities observed at high concentrations of the other HDPs is primarily due to modulation of cellular activity and/ or cytotoxicity rather than antiviral activity. A retroviral vector was inhibited by LL37 and Protegrin-1 to similar extent, while no specific inhibition of adenoviral vector mediated gene transfer was observed. Specific inhibitory effects of Protegrin-1 were confirmed for wild type HIV-1. Conclusion Although Protegrin-1 apparently inhibits an early step in the HIV-replication cycle, cytotoxic effects might limit its use as an antiviral agent unless the specificity for the virus can be improved.	Background As a barrier and immune organ, the gastrointestinal tract, lung and skin play a key role in protecting the body from a hostile environment [ 1 ]. The low incidence of infection at normal epithelial surfaces reflects the presence of innate, broad-spectrum antimicrobial defense mechanisms [ 2 ]. Host defense peptides (HDPs) of the innate immune response play an important role in the protective barrier function of the epithelia [ 3 ]. Host defense peptides have been isolated from diverse organisms, including plants, insects, bacteria and vertebrates [ 4 ]. Several classes of mammalian peptide antibiotics have been ascribed pivotal roles in innate immunity [ 5 ]. Among these are various cysteine-rich peptides such as defensins [ 6 , 7 ] and the more structurally diverse cathelicidins [ 8 ]. Produced as precursors, they require proteolytic processing to liberate the mature functional antimicrobial peptide. Cathelicidins contain a conserved N-terminal cathelin domain, and a structurally diverse C-terminal domain that possesses the peptide's antimicrobial activity. Rabbit CAP18 was the first cathelicidin precursor described, and its mature peptide has broad-spectrum antimicrobial activity [ 9 ]. Cathelicidins have since been identified in many other species including hCAP18/LL37 in humans [ 10 ], protegrins in swine [ 11 - 13 ], CRAMP in mice [ 14 , 15 ] and SMAP29 in sheep [ 16 ]. Many of these peptides demonstrate extremely broad-spectrum antimicrobial activity, including Gram positive and Gram negative bacteria and fungi [ 4 , 15 ]. In addition, they achieve bacterial killing much more rapidly than any commercially available antibiotic [ 17 ]. Recently, a new family of synthetic, α-helical HDPs called "ovispirins" was described [ 18 - 20 ]. Although some of these modified peptides had similar antimicrobial activity of naturally occurring peptides, they manifested appreciable cytotoxicity. We have demonstrated recently that variants of Ovispirin, the so called Novispirin peptides, displayed more favorable toxic/ therapeutic ratios in vitro and broad spectrum activity in infected rat burn model [ 21 , 22 ]. Some of these peptides are induced at epithelial surfaces in response to invading organisms [ 23 - 25 ]. Many HDPs kill microorganisms by causing membrane permeabilization, although not necessarily as their sole mode of action [ 26 ]. Some HDPs also direct chemotaxis, promote wound healing, angiogenesis and contribute to adaptive immunity by mobilizing memory T cells and immature dendritic cells [ 25 , 27 ]. Recent studies have also demonstrated antitumor activity after treatment with HDPs [ 28 ]. In addition, several antiviral activities were reported. Recently it has been demonstrated that rabbit neutrophil peptide alpha-defensin NP1 protects cells from infection with HSV-1 and 2 [ 29 ]. Other studies revealed that human neutrophil peptide HNP1 to 3 and Theta-defensins also inhibit HSV infection although by different mechanisms [ 30 - 32 ]. The ancestral human theta-defensins retrocyclin blocked HSV attachment [ 33 ]. The inhibition of adenovirus replication by the antimicrobial peptide awaits identification of a mechanism of action [ 30 ]. Anti-HIV activity of defensins were first reported 1993 by Nakashima and coworkers [ 34 ]. The alpha-defensins exhibited anti-HIV activity on at least two levels: directly inactivating virus particles; and affecting the ability of target CD4 cells to replicate the virus [ 35 - 37 ]. Binding to gp120 of HIV-1 and inhibition of HIV entry has also been identified as the mechanism of inhibition of HIV infection by theta defensins [ 38 ]. Due to their inhibitory effect on HIV-1 replication and due to an association of a single-nucleotide polymorphism in a beta defensin gene human beta-defensins might also play an important role in host defense against HIV-1 [ 39 ]. The antibacterial activity of HDPs is largely mediated by pore formation leading to permeablization of the bacterial membrane. Although some selectivity for bacterial membranes has been described, the lipid membrane of enveloped viruses might also be a target of antimicrobial peptides [ 32 , 40 ]. This might allow development of antiviral effector molecules for topical application against a broad spectrum of enveloped viruses. Targeting host cell-derived membrane components might be a particularly interesting approach to inhibit viruses that rapidly develop resistance to compounds directed against viral proteins such as HIV. Therefore, we screened a panel of different natural occurring and designer HDPs for Env-independent inhibition of HIV infection at an early step in the viral replication cycle. Results Given the potential cytotoxicity of HDPs, it was important to discriminate between modulation of cell metabolism and direct antiviral effects. We were concerned that HDPs could modulate host cell metabolism without affecting cell viability as assayed for example by standard MTT assays. We therefore decided to use immunodeficiency virus-based vectors transferring the luciferase gene to determine both, the HDP antiviral activity and modulation of cell metabolism. The luciferase activity of target cells, stably transduced with the lentiviral vector prior to HDP treatment should reveal any effect of HDPs on cellular transcriptional and translational efficacy. Treating cells with HDPs during infection with the lentiviral vector and comparison with results obtained with the stably transduced cells should then allow identifying HDPs that inhibit early steps in the viral replication cycle. To validate the luciferase-based assay for modulation of cell metabolism, 293 cells were stably transduced with a lentiviral vector transferring the luciferase gene. Incubation of transduced cells with increasing concentrations of Protegrin-1 and LL37 led to a dose-dependent inhibition of luciferase activity (Fig. 1 ). Comparison to the MTT test revealed a similar dose response curve, although the MTT test might be less sensitive at lower concentrations of the HDPs. Testing cell proliferation by a BrdU incorporation assay revealed a threshold level above which proliferation is strongly reduced. To allow side by side evaluation of cytotoxic and antiviral effects of HDPs with the same read-out the luciferase-based assay was used in most subsequent experiments. Figure 1 Comparison of different cytotoxicity assays. 293A target cells stably transduced with the luciferase gene were incubated for 48 hours in the indicated concentrations of LL37 or Protegrin-1. Viability, cell proliferation and cell metabolism of parallel cultures were assessed by a standard MTT assay, Brd-U incorporation and the luciferase assay, respectively. Values are expressed as percentage of the values obtained from cultures without HDPs. The mean and the standard deviation of triplicates are given. A panel of HDPs containing members of the major classes of antimicrobial peptides were analyzed for inhibitory effects against lentiviral vectors. Given the variability of the viral envelope protein, we focused on identifying Env-independent inhibitory activities by using VSV-G pseudotyped lentiviral vectors. All antimicrobial peptides used in this study, human cathelicidin LL37, recombinant human β-Defensin-2, porcine Protegrin-1 (PG-1), fungal Plectasin and Novispirin G10, inhibited gram positive and gram negative bacteria revealing antimicrobial activity in the expected range (Table 1 ) and confirmed bioactivity of the peptides used. Table 1 Summary of the radial diffusion assay results comparing host defense peptides with a clinically used antibiotic (Ampicillin) M E C (μg/ ml) Organisms Protegrin-1 LL–37 HBD 2 Plectasin Novi- G10 Ampicillin S. aureus 0,91 ± 0,04 4,74 ± 0,1 9,13 ± 0,5 2,42 ± 0,4 4,2 ± 0,4 10,50 ± 0,2 S. epidermidis 4,48 ± 0,2 13,09 ± 0,6 8,96 ± 0,2 8,20 ± 0,5 2,8 ± 0,02 ND E. faecalis 4,46 ± 0,3 13,61 ± 0,7 - 10,60 ± 0,7 4,3 ± 0,3 28,85 ± 0,6 P. aeruginosa 3,3 ± 1,1 12, 28 ± 0,5 6,87 ± 0,8 > 128 1,56 ± 0,4 ND E. coli 2 ± 0,1 11,66 ± 1,5 3,66 ± 0,3 79,10 ± 0,3 1,63 ± 0,3 18,9 ± 0,9 A. baumanii 5,01 ± 0,2 13,13 ± 0,2 3,19 ± 0,5 > 128 4,5 ± 0,02 ND All clinical isolates from human wounds showed significant sensitivity to HDPs. ND: not determined. MEC: minimal effectory concentration. MEC > 128 means that the tested bacteria is not susceptible to the drug tested. Depicted data consists of 3 individual experiments; each condition was performed in quadruplicates The inhibitory effect against the lentiviral vector was determined by preincubation of the vector with human cathelicidin LL37, recombinant human β-Defensin-2, porcine PG-1, fungal Plectasin and Novispirin G10 for 30 minutes in increasing concentrations of peptide prior to the addition of vectors with antimicrobial peptide to the target cells. Stably transduced target cells were incubated in parallel with the HDPs to detect effects on cell metabolism. Luciferase activities were determined 2 days after infection. All HDPs led to a reduction in luciferase activity of cells transduced with the lentiviral vector (Fig. 2A–E ), but most of them also reduced the luciferase activity of stably transduced target cells. Specific inhibition of early steps of infection was only seen for the cathelicidin LL37 and PG-1. As an additional control for the specificity of inhibition, a non-enveloped adenoviral vector transferring the luciferase gene was also incubated with the same panel of HDPs. None of the HDPs led to a dose-dependent inhibition of early steps of the adenoviral replication cycle (Fig. 2F–J ). Two-fold serial dilutions were used to determine 50% inhibitory concentrations (IC 50 ) of LL37 and PG-1, resulting in IC 50 s of approximately 30 μg/ml and 16.8 μg/ml, respectively (Fig. 3 ). Figure 2 Inhibitory activity of HDPs against lentiviral and adenoviral vectors. Percent luciferase activity of 293A target cells transduced in the presence of the indicated amounts of HDP with the VLΔBH lentiviral vector (A to E) or an adenoviral vector (F to J) both transferring the luciferase gene is shown. Modulation of cell metabolism was investigated in parallel by incubating 293A target cells stably transduced with a luciferase gene with the indicated amounts of HDP. The luciferase activity is expressed as percentage of the luciferase activity of cells cultured in the absence of HDP. The mean and the standard deviation of triplicates are given. Figure 3 50% inhibitory concentrations of LL37 and Protegrin-1. Two-fold serial dilutions were used to determine the IC 50 s of LL37 (A) and Protegrin-1 for the VLΔBH vector (B). Modulation of cell metabolism was investigated in parallel by incubating 293A target cells stably transduced with a luciferase gene with the indicated amounts of HDP. The luciferase activity is expressed as percentage of the luciferase activity of cells cultured in the absence of HDP. The mean and the standard deviation of triplicates are given. In initial attempts to characterize the mechanism of inhibition of lentiviral vector infectivity, LL37 and PG-1 were added at different time points during infection: both HDPs were either preincubated with the vector preparation for 30 minutes prior to addition to the cells or the HDPs and the vector were added simultaneously to the cells (Fig. 4A+B ). In addition, cells were first infected with the lentiviral vector for two hours prior to addition of the HDPs. LL37 and PG-1 exerted the strongest inhibition after preincubation of HDPs and the lentiviral vectors, indicating a direct effect on infectivity of the vector particles. However, adding LL37 and PG-1 two hours after incubation of cells with the lentiviral vector also led to a stronger reduction of luciferase activity than observed after incubation of cells stably transduced with the luciferase gene suggesting a second target in the infection cycle that is affected by LL37 and PG-1. To further discriminate between direct inhibitory effects on vector particles and effects mediated by potential HDP cell interactions lentiviral vector particles were first incubated with LL37 and PG-1 for 30 minutes and then added either undiluted or at a 1:10 dilution to the target cells. Due to a 10-fold lower HDP concentration in the latter cultures, vector infectivity should be reduced to a lesser extent, if inhibitory effects are mediated by cellular targets. Comparable dose-dependent inhibition curves (Fig 4C,D ) of diluted and undiluted vectors demonstrate that the inhibitory effects of LL37 and PG-1 depend on the HDP concentration during preincubation of the vector particles and not on the HDP concentration during subsequent cell culture. Figure 4 Time and concentration dependent inhibition of lentiviral vectors by LL37 and Protegrin-1. The lentiviral vector VLΔBH transferring the luciferase gene was either preincubated with 200 μg/ml of LL37 (A) or 50 μg/ml of Protegrin-1 (B) for 30 minutes (-30) or added simultaneously (0) with LL37 and Protegrin-1 to 293A target cells. Target cells were also preincubated for 120 minutes with the lentiviral vector prior to addition of LL37 and Protegrin-1. Two days after infection luciferase activities were determined as percentage of luciferase activities of cells cultured in the absence of HDPs. Cells stably transduced with the luciferase gene were also cultured in the presence and absence of LL37 and PG-1 to determine the effect of HDPs on the cell metabolism. The mean and the standard deviation of triplicates is given. A lentiviral vector transferring the GFP gene (HIV-CSCG) was incubated for 30 minutes at the indicated concentrations of LL37 (C) or Protegrin-1 (D). The vector was then added directly to 293A target cells (undiluted) or after a 1:10 dilution in medium lacking the HDPs. The number of GFP positive cells at each HDP concentration is given as percentage of GFP-positive cells of cultures transduced with diluted and undiluted vectors in the absence of HDPs. The mean and standard deviation of triplicates are shown. The lentiviral vector used in this study had been pseudotyped with the G protein of vesicular stomatitis virus (VSV-G). To evaluate whether LL37 and PG-1 directly target VSV-G or a lentiviral protein, the inhibitory effect of these HDP against the lentiviral vector was compared side by side to their effect on a retroviral vector containing the amphotropic MLV Env for entry into target cells. The dose dependent reduction in the luciferase activity in the target cells was very similar in cells transduced with the lentiviral or the MLV vector (Fig. 5 ). Figure 5 Comparative analysis of inhibition of lentiviral and retroviral vector infectivity. Percent luciferase activity of 293A target cells transduced in the presence of the indicated amounts of HDP with a lentiviral vector (VLΔBH) or a retroviral vector (pRV-172) both transferring the luciferase gene is shown. Modulation of cell metabolism was investigated in parallel by incubating 293A target cells stably transduced with a luciferase gene with the indicated amounts of HDP. The luciferase activity is expressed as percentage of the luciferase activity of cells cultured in the absence of HDP. The mean and the standard deviation of triplicates of two independent experiments are shown. The inhibition of HIV vectors containing the HIV-1 envelope by LL37 and PG-1 were studied on P4CCR5 cells expressing CD4 and coreceptors. IC 50 s of 25 μg/ml and 14 μg/ml were observed for LL37 and PG-1, respectively (Fig 6A,B ), while only minimal inhibitory effects on cell proliferation were detected at these concentrations. Inhibition of early steps of wild type HIV-1 infection by LL37 and PG-1 was also evaluated on P4CCR5 indicator cells, which produce beta-Galactosidase upon expression of the viral tat gene after infection. The BrdU incorporation assay was used to evaluate modulation of cell function. A dose dependent reduction of the titer of HIV-1 on P4CCR5 cells was observed for both HDPs. However, the IC 50 of LL37 was approximately 3-fold higher than the IC 50 previously determined for the lentiviral vectors resulting in a narrow gap between antiviral and antiproliferative effects of LL37. In contrast, the IC 50 of PG-1 was below 10 μg/ml, while inhibitory effects on cell proliferation were not observed up to concentrations of 50 μg/ml. Figure 6 Inhibitory effects of Protegrin-1 and LL37 on HIV-1 Env mediated vector entry (A, B) and HIV-1 infection (C,D). A lentiviral vector transferring the GFP gene (VGΔBH-SIN) was incubated at increasing concentrations of LL37 (A) or Protegrin-1 (B) prior to transduction of P4CCR5 cells. The vector titer is given as percentage of the titer of the vector incubated in the absence of HDPs. Wild type HIV-1 was incubated with increasing concentrations of LL37 (C) and Protegrin-1 (D). The virus titer was subsequently determined on P4CCR5 indicator cells and is expressed as percentage of the titer of the untreated HIV-1 virus stock. The toxicity of LL37 and Protegrin-1 was determined in parallel using the BrdU incorporation assay. Discussion From the panel of five HDP studied, the cathelicidin LL37 and PG-1 were found to specifically inhibit lentiviral and retroviral vector, but not adenoviral vector infectivity. The strongest inhibition was seen if the lentiviral vectors were preincubated with LL37 and PG-1. This suggests that these HDPs directly interacted with the vector particles, which is consistent with our observation that inhibition was dependent on the HDP concentration during preincubation of the vectors with HDPs, but not on the HDP concentration during infection of the cells. Since lentiviral vectors and retroviral vectors were inhibited to a similar degree although they do not share any viral protein, the target for the HDP on the particles is probably cell-derived. This could either be the lipid membrane derived from the cell, which surrounds the vector particles or cellular membrane proteins that are frequently incorporated in lentiviral and retroviral particles during budding [ 41 ]. A permeabilizing effect of LL37 and PG-1 on the viral particles would be consistent with our data, but other mechanisms of inhibition cannot be excluded. While the inhibitory effect of PG-1 was also detected with wild type HIV-1 on P4CCR5 cells, LL37 inhibited HIV-1 to lesser degree then the lentiviral vectors. Due an IC 50 of 88 μg/ml against wild type HIV-1, it is questionable whether LL37 concentrations are sufficiently high at mucosal membranes to play a role in host defense against HIV-1. Conclusions Modulation of cell metabolism was generally seen at concentrations of HDPs exceeding 50 μg/ml, while the MEC of the antibacterial activity ranged from 1 to 10 μg/ml. This might leave a sufficient window for therapeutic intervention of bacterial infection. However, for the treatment of HIV-1, the therapeutic window of LL37 and PG-1 is rather narrow. It should also be noted that the HDP-induced modulation of cell metabolism and cytotoxicity can be cell type dependent. Therefore, increasing the selectivity of HDPs for early steps in the viral replication cycle seems to be necessary for further development of the human cathelicidin LL37 and the porcine Protegrin-1 as antiviral agents for systemic or topical applications. Methods Preparation of vectors transferring the luciferase or GFP genes To generate lentiviral vector particles transferring the luciferase gene, a codon-optimized (Geneart GmbH, Regensburg, Germany) HIV-1 gag-pol expression plasmid (Hgp syn ) [ 42 ] and a VSV-G expression plasmid (pHIT-G) [ 43 ] were used to package the SIV-based vector VLΔBH. This vector contains the luciferase gene replacing the GFP gene of VGΔBH [ 44 ].5 μg of Hgp syn , 2 μg of pHIT-G and 5 μg of VLΔBH were transiently cotransfected by the CaPO 4 coprecipitation method into 293T cells as previously described [ 45 ]. An HIV vector construct containing the GFP reporter gene (HIV-CSCG) [ 46 ] was also used to prepare lentiviral vector particles by cotransfection with Hgp syn , pcTat [ 47 ], pcRev [ 47 ] and pHIT-G or pSVIIIenv3-2, an HIV-1 envelope expression plasmid [ 48 ]. The MLV vectors were prepared by cotransfection of pHIT-456, pHIT-60 and pRV-172 [ 49 ]. Two days after transfection, the supernatants were cleared from cellular debris by low speed centrifugation (10 minutes, 1000 × g) and filtration through 0.2 μm filters from Roth (Karlsruhe, Germany). Aliquots were stored at -80°C. Construction and Production of Ad.OW126 Vector Beginning with a first generation E1- and E3-deleted adenoviral vector, we generated a replication-competent adenoviral vector Ad.OW126, which harbors in the E1 region the firefly luciferase cDNA (subcloned from pGEM-Luc (Promega, Madison, WI)), a IRES element [ 50 , 51 ], and an Ad5 E1A ΔE1B-55K gene. The entire expression cassette is driven by the human CMV-IE promoter in parallel to the transcriptional orientation of the adenovirus E1 gene products and terminated by the bovine growth hormone polyadenylation site. The expression cassette was flanked upstream by the Ad5 packaging sequence and downstream by the Ad5 pIX. The Ad.OW126 vector was generated by in vitro ligation [ 52 ] to H5 dl 327 (kindly provided by T. Shenk, Princeton University, Princeton, NJ), utilizing the unique Bst 1107 I restriction site. The vector was propagated in 293 cells and purified by two rounds of CsCl density centrifugation [ 53 ], dialyzed (Slide-A-Lyzer, Pierce, Rockford, IL) against 1500 ml of PBS with 1 mM MgCl 2 and 10% glycerol four-times (1 hour each) at 4°C, and stored at -80°C until use. The concentration of the vector was determined by measuring absorbency at 260 nm [ 54 ], and the infectious titer was determined by plaque assay on 293 cells [ 55 ]. The ratio of infectious to non-infectious virus particles was approximately 1:80. Generation of 293A cells stably transduced with a luciferase gene To stably transduce the luciferase gene into 293A cells, a self-inactivating version of VLΔBH, VLΔBH-SIN similar to VGΔBH-SIN [ 44 ] was packaged by cotransfection with Hgp syn and pHIT-G. 293A cells were plated in 24 well plates at a density of 50.000 cells / well and transduced with 200 μl of VLΔBH-SIN vector for two hours. Two days after plating cells were transferred to one well of a six well plate and transduced again with 1 ml of VLΔBH-SIN vector. Cells were subsequently expanded resulting in 293-Luc cells. Luciferase assay The supernatant of infected 293A or 293-Luc cells, cultured in 96 well plates was removed and cells were lysed in 50 μl of cell lysis buffer (Promega, Pittsburgh, PA). 20 μl of the cell lysates were used in the firefly luciferase assay system of Promega as described by the manufacturer. Each single value of the triplicates was expressed as percent of the mean of triplicates of control cultures infected with the same vector in the absence of HDPs and the mean and the standard deviation of the percent values was calculated for each triplicate. Host defense peptides The antimicrobial peptides (human LL37, porcine PG1-1, mutants from the ovine SAP29: Novispirin G10 and fungal Plectasin) used in this study were prepared by solid phase synthesis and purified by RP-HPLC. Recombinant human β-Defensin-2 was produced by a molecular farming approach in transgenic potato tubers and purified by perfusion chromatography (data not shown). The peptides (≥ 98% pure) were dissolved in 0.01% acetic acid and used for all in vitro and in vivo studies. Potential endotoxin contamination was monitored with the chromogenic Limulus amoebocyte lysate assay (BioWhittaker, Walkersville, MD) using Escherichia coli endotoxin (supplied with the kit) as the standard. Endotoxin levels for the peptides were not detectable. Bacteria The following strains were used in this study: Gram-negative strains: Acinetobacter baumannii (ATCC 19606), Escherichia coli (ATCC 25922) and Pseudomonas aeruginosa (ATCC 27853) Gram-positive strains: Staphylococcus aureus (ATCC 25923), Staphylococcus epidermidis (ATCC 12228) and Enterococcus faecalis (ATCC 29212). All bacterial strains were analyzed with API test strips (BioMerieux, Hazelwood, MO) to confirm identity and aliquots were stored frozen in 50 % skim milk at -80°C. Bacteria were grown overnight in trypticase soy broth (Becton Dickinson, Franklin Lakes, NJ) at 275 rpm and 37°C. An aliquot of the resulting stationary phase cultures was then transferred to 20 ml of trypticase soy broth and incubated at 37°C for 2.5 hours to reach log phase. This subculture was transferred to a 50 ml conical polystyrene tube and centrifuged for 10 min at 4°C at 880 g. The bacterial pellet was washed once with chilled phosphate buffered saline, pH 7.4, and resuspended in 5 ml of the same cold buffer. One milliliter was removed to measure its optical density at 620 nm. The bacterial concentration was calculated from the following formula: CFU/ml = OD 620 × 2.5 × 10 8 . Growth Inhibition Assay To monitor bacterial growth inhibition in vitro a radial diffusion assay was performed as previously described [ 56 ]. Briefly, the underlay agar consisted of 1% agarose (A-6013, Sigma Chemical, St. Louis, MO) and 0.3 mg/ml trypticase soy broth (TSB) powder in 10 mM sodium phosphate with 100 mM NaCl (normal salt medium), pH 7.4. Bacteria (approximately 5 × 10 6 CFU) were mixed with 10 ml of underlay gel (43°C) and immediately poured into square 9 × 9 cm petri dishes. A series of 3 mm wells was punched after the agarose solidified. After appropriate serial dilutions were done, 5 μl of HDP, vancomycin (Abbott Labs, Chicago, IL), gentamicin, ciprofloxacin, or fluconazole (Sigma-Aldrich, St. Louis, MO) were added to the designated wells. Plates were incubated at 37°C for 3 hours. The bacteria-containing layer was covered with a 10 ml overlay of the nutrient rich agar. The overlay agar consisted of 6% (w/v) TSB and 1% agarose in PBS for all assays. After 18 h of incubation at 37°C, the plates were stained with 0.001% Coomassie blue for 10 h. The clear zones (bacterial growth inhibition) around the punched wells indicated antibacterial activity. The diameters of the clear zones were converted into units by subtracting the well diameter and multiplying the difference by 10. Results were plotted using a semi log scale and correlation coefficients and X-intercepts obtained from linear regression analysis. The minimal effective concentration (MEC) corresponded to the X-intercept value. All assays were performed in triplicates and repeated at least once. Cytotoxicity and proliferation Assay 293-Luc cells were plated in 96 well plates at a density of 2 × 10 3 cells / well. After 48 hours, 50 μl of MTT-solution (3 mg/ml) was added and incubated at 37° C under 5% CO 2 for 1 hour. After this time medium was removed and 100 μl 0.04 N HCL + 10% SDS was used to dissolve the resulting blue formazan crystals in living cells. The optical density was determined at 550 nm. Each single value of the triplicates was expressed as percent of the mean of triplicates of control cultures infected with the same vector in the absence of HDPs and the mean and the standard deviation of the percent values was calculated for each triplicate. In addition the BrdU Cell proliferation ELISA with chemiluminescence detection (Roche Diagnostics GmbH, Mannheim, Germany) was performed. After 293-Luc cells or P4CCR5 cells were plated in 96 well plates at a density of 2 × 10 3 cells/well BrdU (5-bromo-2'-deoxyuridine) was added to the cells with a resulting concentration of 10 μM for the last 22 h of the incubation period. After removing the culture medium, the cells were fixed and DNA was denatured in one step with Fixdenat. Thereafter the cells were incubated with Anti-BrdU-POD for 1 h at room temperature. The chemiluminescence detection was measured after automatic injection of substrate solution with a microplate-luminometer (Orion, Berthold detection systems, Pforzheim, Germany). Inhibition of vector infectivity To determine the effect of HDPs on vector infectivity and cell metabolism, 293A target cells were plated in triplicates into 96-well plates at 2 × 10 3 cells / well. After overnight incubation, the supernatant of the wells were removed and replaced by 25 μl vector preparation and 25 μl HDP at twice the final concentration indicated. 50 μl fresh medium with HDP at the final concentration indicated was added after two hours. One (adenoviral vector) or two (lentiviral and retroviral vector) days after infection, the supernatant was removed and 50 μl of cell lysis buffer (Promega) was added. Lysates were stored at -80°C until determination of the luciferase activity of the extracts. The affect of HDPs on cell metabolism was determined in parallel by plating 293-Luc cells exactly the same way as 293A cells and the luciferase activity was determined after one (as a control to inhibition of adenoviral vector infectivity) or two (as a control of lentiviral infectivity) days of incubation with HDPs. For GFP-expressing vectors, vector titers were calculated from the number of GFP positive cells per well as previously described [ 45 ] and the BrdU incorporation assay was used to monitor cytotoxic effects. Inhibition of HIV-1 Stocks of HIV-1 were generated by transient transfection of 293T cells with the molecular clone pNL4-3. 25 μl of the virus stock were incubated for 30 minutes at room temperature with 25 μl of LL37 or PG-1 adjusted to twice the final concentration indicated. The mixture was added to P4CCR5 cells plated the day before at a density of 2 × 10 3 cells / well of 96 well plate. 50 μl fresh medium with HDP at the final concentration indicated was added after two hours. Two days after infection, the supernatant was removed and cells were stained by X-Gal. The number of infected cells per well were counted in the microscope. Competing interest The author(s) declare that they have no competing interests. Authors' contributions LS, BT and JM performed most of the experiments. LS, OW, ML, EL, HH, HS and KU participated in the experimental design, data interpretation and writing of the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC548510.xml
534815	Tracking Orangutans from the Sky	null	After apprenticing at her mother's side for some eight years—the first three clinging to her body—an orangutan is ready to make her own way in the forest canopy. The only great ape specializing in arboreal living, orangutans forage the treetops mostly for fruit, nuts, insects, leaves, and tree bark. They can recognize hundreds of species of edible fruit from trees and woody climbers and remember their location and fruiting season. Malay legend says orangutans (a variation on the Malay trope for “man of the woods”) derived from humans who sought refuge from their species in the wilds of the forest. Today, with the last orangutan refuges shrinking drastically, the man of the woods has nowhere else to go. Hundreds of thousands of orangutans once ranged throughout southeast Asia. Now just two orangutan species inhabit just two countries: Indonesia (Kalimantan, the southern part of Borneo and Sumatra Island) and Malaysia (Sabah and Sarawak, in northern Borneo). The Sumatran orangutan is listed as critically endangered; the Bornean, endangered. In Indonesian Borneo and Sumatra, logging operations clear an estimated 5 to 6 million forest acres a year, leaving the apes stranded in isolated stands of trees and the normally fire-resistant rainforest at sudden risk. Another force driving orangutan extinction in Indonesia is the poaching and illegal killing (mothers do not give up their babies without a fight) that secures orangutan babies for the exotic pet trade. In Sabah, Malaysia, the primary threat comes from clearing the forest for agriculture. Conservation efforts depend, among other things, on having reliable data on population size, density, and distribution, but estimates of orangutan numbers in Sabah—which range from 2,000 to 20,000—are outdated. In a new study, Marc Ancrenaz and colleagues report an innovative method of directly estimating orangutan numbers from the number of nests detected during aerial surveys. (Orangutans are tough to spot directly, so researchers count the nests they sleep in at night.) Their survey, which covered the entire orangutan range throughout Sabah, estimates a total population of 11,000 orangutans—a drop of 35% in the past 20 years, based on a 1988 World Wildlife Federation report. Counting orangutans from the ground can be very time-consuming, difficult work, especially when faced with the hip-deep muck and steep slopes of the rainforest floor. Though helicopters obviously cover greater distance and more remote territory than is possible by foot, they're generally used to survey animals in more open landscapes. By using ground survey data to refine their aerial survey results, Ancrenaz and colleagues could directly assess the distribution and size of orangutan populations throughout Sabah. (Sabah covers roughly 72,000 square kilometers.) Over the course of two years, ground surveys—requiring 1,100 hours of field work—and aerial surveys—requiring just 72 hours—were conducted throughout all the major forests of Sabah. Commercial logging occurs in about 76% of all Sabah forests in commercial forest reserves. During the overflights, information was recorded on altitude, forest type, forest disturbance (on a scale from none to active exploitation), and signs of human activity. The authors attribute the 35% decline in Sabah orangutan numbers primarily to habitat loss from agricultural development. The surveys revealed that lowland forests harbored the greatest density of nests and orangutans, with the densest populations found in several highly disturbed, fragmented forests along newly created palm oil plantations. These high orangutan densities could reflect an influx of refugees from recently destroyed forest habitat into areas that are still forested. In logged forests, higher population densities were found in old exploited or sustainably logged forests than in conventionally logged reserves. While the authors acknowledge the density estimates could be more precise—better measures of nest decay and construction rates are needed—their survey reveals crucial information on orangutan numbers and distribution. Most orangutans in Sabah, including those making up one of the largest unfragmented populations in Borneo, live outside protected areas, in commercially exploited forests. These results suggest that orangutans may adapt better to degraded forests than previously thought—provided illegal hunting and agricultural conversion are controlled. More field research will help quantify the impacts of human activity—from logging to stealing babies—on great ape ecology and survival, and determine whether exploited forests can support conservation. It may be, for example, that integrating agricultural fields with forested corridors could sustain orangutan populations over the long term. With time of the essence, these aerial surveys will speed that work, and help sustain orangutans' refuge in the treetops, above their human relatives.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC534815.xml
519005	A Test Case for DNA Barcodes to Identify Species	null	One hundred years before Darwin returned from his voyage on the H. M. S. Beagle “struck with certain facts” that “seemed to throw some light on the origin of species,” Linnaeus published the first systematic taxonomy of life. In Systema Naturae, the Swedish botanist divided organisms into plants, animals, and minerals, eventually assigning scientific names to 7,700 plant and 4,400 animal species, and popularizing the binomial system—as in Homo sapiens —of naming species. In the 1700s and 1800s, naturalists classified organisms based on morphology, devoting their careers to naming newfound plants and animals. Today biologists still use Linnaean taxonomy as the foundation of scientific classification. But with just a fraction of the estimated 5–30 million species on the planet already named and too few specialists to do the job, biologists are looking for high-throughput tools that can rapidly and accurately identify both individuals of a species and entirely new species. That's what some scientists say the DNA barcode will do. The DNA barcode, as the name implies, uses genes to identify species much like supermarket barcodes identify products. The idea is that a short stretch of genetic code from a reference gene is unique enough to one species to distinguish it from every other species, and that comparisons of sequence variations in that stretch of gene can reveal evolutionary relationships among species. Such technology could radically advance biologists' attempts to achieve the long-standing goal of cataloging life on earth, but the approach is controversial, with critics questioning both the method and its applications. (For more on the debate, see “DNA Barcoding: Promise and Pitfalls,” also in this issue.) Paul Hebert and colleagues offer a proof of the utility of the DNA barcoding concept, using a 648-basepair region of a mitochondrial gene (cytochrome c oxidase I, or COI) in a study of 260 North American bird species. Mitochondria—the cell's power generators—contain their own DNA, and mitochondrial DNA (mtDNA) evolves much faster than nuclear DNA. It evolves so quickly, in fact, that mtDNA sequence variation has been found not just between closely related, or sister, species but also within species. Still, the variation is much greater among than within species, which is why mtDNA divergences have become a tool for identifying species. Hebert and colleagues tested the effectiveness of the mtDNA COI barcode by matching bird species flagged by the COI barcode against those already established by taxonomic methods. The litmus test for DNA barcoding is absence of COI sequence overlap between species. Beyond that, differences within species should be significantly fewer than those between species. And that's what the researchers found. All 260 species had unique COI barcodes, with differences between species for the most part much more frequent—on average, 18 times more common—than those within species. In the 130 species represented by two or more individuals, COI sequences were either identical or closest to other sequences within that species. For these 260 bird species (of the 667 bird species that breed in North America), the authors report, the COI barcodes “separate individuals into the categories that taxonomists call species.” The COI barcode, the authors propose, could help resolve problematic classifications based on morphology, as arise when populations of a single species acquire distinct characteristics after geographic barriers prevent their interbreeding. For example, the similar COI sequences found in American and black oystercatchers here support taxonomic studies suggesting that they are actually color morphs of one species. And conversely, highly divergent COI sequences might bolster taxonomic studies indicating that lineages of uncertain status are indeed distinct species. Future studies will have to determine whether these results can be generalized to animals in other climes and ecosystems, but the authors argue that constructing a comprehensive library of barcodes will facilitate such efforts. Hebert and colleagues conclude that the success of DNA barcoding depends not only on such a repository—with sequences pegged to well-characterized species exemplars—but also on the expertise of trained taxonomists. The hope is that large-scale, standardized testing based on a uniform barcode sequence could go a long way toward finishing what Linnaeus started: a full classification of all plant and animal life. To E. O. Wilson, every species is “a masterpiece of evolution, offering a vast source of useful scientific knowledge because it is so thoroughly adapted to the environment in which it lives.” Faced with what Wilson calls the “worst wave of extinction since the dinosaurs died,” the need for a fast and easy way to identify species has never been greater.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC519005.xml
514535	Proteasome inhibitors: Their effects on arachidonic acid release from cells in culture and arachidonic acid metabolism in rat liver cells	Background I have postulated that arachidonic acid release from rat liver cells is associated with cancer chemoprevention. Since it has been reported that inhibition of proteasome activities may prevent cancer, the effects of proteasome inhibitors on arachidonic acid release from cells and on prostaglandin I 2 production in rat liver cells were studied. Results The proteasome inhibitors, epoxomicin, lactacystin and carbobenzoxy-leucyl-leucyl-leucinal, stimulate the release of arachidonic acid from rat glial, human colon carcinoma, human breast carcinoma and the rat liver cells. They also stimulate basal and induced prostacycin production in the rat liver cells. The stimulated arachidonic acid release and basal prostaglandin I 2 production in rat liver cells is inhibited by actinomycin D. Conclusions Stimulation of arachidonic acid release and arachidonic acid metabolism may be associated with some of the biologic effects observed after proteasome inhibition, e.g. prevention of tumor growth, induction of apoptosis, stimulation of bone formation.	Background The proteasome degrades many cellular proteins, several with regulatory functions. It is not surprising that proteasome inhibitors affect many biologic processes [ 1 ] including prevention of cancer [ 2 ]. The effect of proteasome inhibition on cell growth and possible cancer chemoprevention has been reviewed by Adams [ 3 ]. Epoxomicin, an α'-β'-epoxyketone, appears to be the most selective proteasome inhibitor. Based on its anti-tumor activity, this product was originally isolated from an actinomycetes strain C-996-17 [ 4 ]. It inhibits the chymotrypsin-like activity (cleavage after large hydrophobic residues), trypsin-like activity (cleavage after basic residues) and peptidyl-glutamyl peptide hydrolyzing (PGPH) activity (cleavage after acidic residues) of proteasomes. Activities of the Ca ++ -dependent proteases, calpain, papain, chymotrypsin, trypsin and cathepsin are not affected by epoxomicin even at a 50 μM concentration [ 5 ]. The β-lactone, lactacystin, is relatively selective but can inhibit cathepsin A [ 6 ]. Peptide aldehydes, in addition to inhibiting proteasome activity, can also inhibit lysosomal and Ca ++ -activated proteases [ 7 ]. The peptides containing the carboxyvinylsulfone moiety inhibit cysteine proteases [ 8 , 9 ]. I have shown that inhibition of proteolysis by phenylmethylsulphonyl fluoride, the peptide aldehydes carbobenzoxy-leucyl-leucyl-norvalinal and carbobenzoxy-leucyl-leucyl-leucinal (ZLLL) and lactacystin stimulate induced prostaglandin (PGI 2 ) production in rat liver cells [ 10 , 11 ]. Lactacystin stimulates arachidonic acid (AA) release from these cells [ 11 ]. Others have reported that proteasome inhibition up-regulates cyclooxygenase-2 (COX-2) and stimulates PGE 2 production in neuronal cells [ 12 ]. In this report, evidence is presented that proteasome inhibitors stimulate PGI 2 production by rat liver cells as well as the release of AA from rat liver, rat glial, human colon carcinoma and human breast carcinoma cells in culture. The stimulation of AA release from rat liver cells is partially inhibited by preincubation of the cells with actinomycin D. Results and Discussion Of the cells examined (C-9 rat liver, C-6 rat glial, HT-29 human colon carcinoma and BT-20 human breast carcinoma) the prostanoid metabolic profile has been described only for C-9 rat liver cells (95% is PGI 2 and less than 5% is PGE 2 and PGF 2α ) [ 13 ]. At the low cell densities used in this study, only PGI 2 , the main product of COX-mediated synthesis, can be quantitatively estimated. The rat liver cells express COX-2 both constitutively and after induction [ 14 ]. The effect of time on basal and 12-0-tetradecanoylphorbol-13-acetate (TPA) induced PGI 2 synthesis during incubation of cells with epoxomicin is shown in Fig. 1 . Figure 1 Time-course of (A) basal and (B) TPA-induced 6-keto-PGF 1α production during incubation with 1.2 μM epoxomicin. In (B) 16.7 nM TPA was used. Analyses were performed with duplicate and triplicate dishes. Mean values are shown. The stimulation of basal PGI 2 production by epoxomicin and TPA-induced PGI 2 production by epoxomicin and lactacystin as a function of dose is shown in Fig. 2 . As little as 0.3 μM epoxomicin stimulates TPA-induced PGI 2 production significantly (Fig. 2-B ). It is 10 to 15 times more effective than lactacystin (compare Fig. 2-B and 2-C ). Using purified bovine erythrocyte proteasomes, epoxomicin inhibits the chymotrypsin-like activity, about 4 to 5 times more effectively than does clasto -lactacystin β-lactone, the derivative of lactacystin [ 5 ]. They are almost equally effective on inhibiting the trypsin-like and PGPH-like activities [ 5 ]. Assuming that epoxomicin and lactacystin have equal access to the proteasome and that proteasome activity is regulating COX-2 in rat liver cells similarly to neuronal cells [ 12 ] then COX-2 may be degraded in the proteasome by cleavage after large hydrophobic residues. Figure 2 Effect of epoxomicin on (A) basal and (B) TPA-induced 6-ketoPGF 1α production and (C) effect of lactacystin on TPA-induced 6-ketoPGF 1α production. Cells were incubated with the reagents for 6 hours. The analyses were performed with triplicate dishes - statistically significant vs MEM/BSA. - Statistically significant vs TPA. The amplification of PGI 2 production (Figs. 1 and 2 ) after inhibition by epoxomicin could reflect not only stabilization of COX-2 but also an intracellular increase in the concentration of the substrate i.e. the AA that is produced during hydrolysis of the membrane phopholipids by PLase activity [ 15 ]. Extracellular and/or intracellular release of AA will depend, in part, on the localization of the phospholipids in the membrane, e.g. in its inner or outer leaflet [ 16 ]. Release of AA in response to several agonists has been described [ 17 - 20 ]. The effect of a 2, 4 or 6-h incubation on AA release from rat liver and rat glial cells by 1.0 μM epoxomicin was determined. Only after the 6-h incubation were the differences significant statistically. Regulation of PLase activity by the proteasome pathway appears to be a relatively slow process. After a 6-h incubation, epoxomicin, lactacystin and ZLLL stimulate the release of extracellular AA from rat liver cells (Fig. 3 ) and AA release after TPA-induction (3.7% vs 13.5% in the presence of 1.0 μM epoxomicin). Epoxomicin also stimulates the release of AA from rat glial, human colon carcinoma and human breast carcinoma cells (Table 1 ). The stimulation of AA release from the rat liver cells after incubation with epoxomicin is partially inhibited by pre-incubation of the cells for 2-h with actinomycin (Fig. 4 ) suggesting that a fraction of the PLase is induced. As expected, the inhibition of TPA-induced PGI 2 production by actinomycin D is complete (Fig. 5 ). Thus, some mechanisms leading to maximum AA release appear to be genomic. The induced PLase activity, probably PLA 2 , could reflect expression of either a secretory or cytosolic PLA 2 or some combination of both enzymes [ 21 ]. Figure 3 Dose-response of epoxomicin, lactacystin and ZLLL on AA release from rat liver cells. After incubation for 6 hours. The analyses were performed with triplicate dishes. - Statistically significant vs MEM/BSA. Table 1 Effect of Epoxomicin on AA Release from Rat Glial, Human Colon Carcinoma and Human Breast Carcinoma Cells. Cell Type % AA Release MEM/BSA control With epoxomicin Rat glial (C-6) 9.0 ± 0.07 11.4 ± 0.28 Human Colon Carcinoma (HT-29) 7.8 ± 0.05 9.4 ± 0.33 Human Breast Carcinoma (BT-20) 12.1 ± 0.35 14.2 ± 0.31 Cells were incubated in the presence and absence of 4.5 μM epoxomicin for 6-h, 12-h or 9-h (C-6, HT-29, BT-20 respectively). Analyses were performed with quadruplicate (C-6 and HT-29) or quintuplicate (BT-20) dishes. All values are expressed as Mean ± SE (n = 4 or n = 5). All data with epoxomicin are statistically significant vs MEM/BSA. Figure 4 Effect of Actinomycin D on AA release from rat liver cells. Cells were preincubated with and without 1.0 μM actinomycin D for 2-h, then incubated in the presence and absence of epoxomicin and actinomycin D for another 6-h. The analyses were performed with triplicate or quadruplicate dishes. - Statistically significant vs epoxomicin in the presence of actinomycin D. Figure 5 Effects of actinomycin D on TPA-induced 6-Keto-PGF 1α production. Cells were preincubated with 1.0 μM actinomycin D for 2-h and then incubated in the presence and absence of 0.6 μM epoxomicin and/or 16.7 μM TPA for another 6-h. The analyses were performed with triplicate dishes. - Statistically significant vs TPA in the absence of actinomycin D. **- Statistically significant vs epoxomicin plus TPA in the absence of actinomycin D. The release of AA from rat liver cells, most likely resulting from PLase activation, is associated with cancer chemoprevention [ 14 , 17 - 19 ], [ 22 - 24 ]. In addition to its intrinsic biologic activities, AA regulates production of lipoxygenase, cytochrome P-450, and epoxygenase products as well as COX activities. Prostanoid profiles differ with cell type and individual AA metabolites have different pharmacological properties [ 15 ]. COX-2 activity, as measured by PGI 2 production, is stimulated by proteasome inhibition (Fig. 1 and 2 ). Thus, some biologic effects of proteasome inhibition, e.g. stimulation of bone formation [ 25 ], may reflect the metabolism of the intracellular AA. Inhibition of COX-2 activity is one possible mechanism that has been proposed to prevent colon cancer [ 26 ]. However, rather than inhibiting, tamoxifen and raloxifene, statins and epoxomicin stimulate COX-2 activity and AA release from rat liver cells [ 14 , 17 - 19 ]. As shown in Table 1 , epoxomicin stimulates AA release from human colon carcinoma, breast carcinoma and rat glial cells. Tamoxifen and simvastatin also stimulate AA release from the human colon carcinoma and human breast carcinoma cells (unpublished data). These drugs have been reported to prevent cancer [ 27 , 28 ]. At least as measured by the COX activity of rat liver cells, tamoxifen, raloxifene, statins and proteasome inhibitors could be preventing cancer by a COX independent mechanism. AA resulting from proteasome inhibition has many intrinsic biologic properties [reviewed in [ 29 ]]. Some of these activities may trigger PLase activity. The causal relationship of AA to cancer prevention (if any) is unclear. Production of AA by the tumor-suppressive type-II phospholipase A 2 (PLA 2 G 2 A) may be related to the cancer prevention [ 22 - 24 ]. It is not surprising that control of PLase activities present an attractive area for cancer prevention studies [ 30 ]. Methods The rat liver (C-9 cell line) and human breast carcinoma (the BT-20 cell line) were purchased from the American Type Culture Collection (Manassas, VA, USA). The rat liver glial cells (C-6 cell line) was obtained from Dr. Elaine Lai of the Department of Biology, Brandeis University and the human colon carcinoma (the HT-29 cell line) was obtained from Dr. Basil Rigas, American Health Foundation, Valhalla, NY, USA. They were maintained in Eagle's minimum essential medium (MEM) supplemented with 10% fetal bovine serum. [ 3 H] AA (91.8 Ci/mmol) was obtained from NEN Life Science Products, Inc. (Boston, MA, USA). Epoxomicin, lactacystin and ZLLL were purchased from Biomol (Plymouth Meeting, PA, USA). All other reagents were from Sigma Chemical Co. (St. Louis, MO, USA). Rat liver cells incubating with lactacystin (5.4 μM) or ZLLL (1.0 μM) for 6-h have been tested for viability by a tetrazolium-based assay and found not to be toxic [ 31 ]. Two days prior to experiments, the rat liver cells were treated with 0.25% trypsin-EDTA and, after addition of minimal essential media (MEM) containing 10% fetal calf serum, the floating cells were seeded onto 35 mm culture dishes. The plating densities varied from 0.1 to 0.5 × 10 5 cells/35 mm dish. The freshly seeded cultures were incubated for 24-h to allow for cell attachment. After decantation of MEM containing the fetal bovine serum, 1.0 ml fresh MEM containing 10% fetal bovine serum and [ 3 H] AA (0.2 μCi/ml) were added and the cells incubated for another 24-h. The cells were washed 4 times with MEM and incubated for various periods of time with 1.0 ml of MEM containing 1.0 mg BSA/ml (MEM/BSA) and different concentrations of each compound. The culture fluids were then decanted, centrifuged at 2000 × g for 10 min, and 200 μl of the supernate counted for radioactivity. Radioactivity recovered in the washes before the incubation was compared to input radioactivity to calculate the % radioactivity incorporated into the cells [ 31 ]. For PGI 2 production, 1.0 ml of MEM supplemented with 10% fetal bovine serum, void of [ 3 H]AA, was added after the first 24-h incubation. The cells were incubated for another 24-h, washed three times with MEM, then incubated with the compounds in MEM/BSA for various periods of time. The culture fluids were decanted and analyzed for 6-keto-PGF 1α , the stable hydrolytic product of PGI 2 , by radioimmunoassay [ 32 ]. The [ 3 H] AA release is presented as a percentage of the radioactivity incorporated by the cells. Except for the time-course experiments which used duplicate dishes, three to five culture dishes were used for each experimental point. The data are expressed as mean values ± SEM. The data were evaluated statistically by the unpaired Student's t-test . A P value < 0.05 was considered significant.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514535.xml
538752	Disclosure of cancer diagnosis and prognosis: a survey of the general public's attitudes toward doctors and family holding discretionary powers	Background This study aimed to ask a sample of the general population about their preferences regarding doctors holding discretionary powers in relation to disclosing cancer diagnosis and prognosis. Methods The researchers mailed 443 questionnaires to registered voters in a ward of Tokyo which had a socio-demographic profile similar to greater Tokyo's average and received 246 responses (response rate 55.5%). We describe and analysed respondents' attitudes toward doctors and family members holding discretionary powers in relation to cancer diagnoses disclose. Results Amongst respondents who wanted full disclosure about the diagnosis without delay, 117 (69.6 %) respondents agreed to follow the doctor's discretion, whilst 111 (66.1 %) respondents agreed to follow the family member's decision. For respondents who preferred to have the diagnosis and prognosis withheld, 59 (26.5 %) agreed to follow the doctor's decision, and 79 (35.3 %) of respondents agreed with following family member's wishes. Conclusions The greater proportion of respondents wants or permits disclosure of cancer diagnosis and prognosis. In patients who reveal negative attitudes toward being given a cancer disclosure directly, alternative options exist such as telling the family ahead of the patient or having a discussion of the cancer diagnosis with the patient together with the family. It is recommended that health professionals become more aware about the need to provide patients with their cancer diagnosis and prognosis in a variety of ways.	Background Cancer ranks as the third leading cause of death worldwide, accounting for approximately 12 % of all recorded deaths [ 1 ]. As cancer is sometimes fatal and its treatment often involves invasive medical procedures and medication, it has a great impact on patients' lives. The extent to which physicians should inform patients of their diagnosis and prognosis poses a difficult decision in clinical settings. Previous studies show that a patient's cancer diagnosis is not routinely disclosed in many cultures in Africa [ 2 ], Eastern and Southern Europe [ 3 - 6 ], and the Middle East [ 7 ]. Even in the United States, where most doctors follow informed consent guidelines which includes informing patients of their diagnosis as standard clinical practice, problems still exist regarding the accurate provision of prognosis information [ 8 ]. In Japan, historically, physicians have withheld discussing cancer diagnoses directly with patients [ 9 ]. However, since the early 1990s, due to the increased understanding and adoption of informed consent policy and practice, physicians have gradually begun to inform patients of their cancer diagnosis in clinical practice [ 10 , 11 ]. In many cases, however, details regarding prognosis are still concealed from patients, especially if the condition is incurable [ 12 , 13 ]. While some physicians provide full information from the outset, others provide no information at all, even withholding basic diagnosis information [ 14 ]. The National Cancer Centre (The core national institution for developing cancer treatment, research and policy) has compiled a set of guidelines for cancer disclosure. However, each hospital has deferring policy and practice [ 9 ]. No law or regulation stipulates that doctors are required to obtain informed consent from patients. Given this context, there are demonstrated needs to develop concrete guidelines and to promote cancer disclosure based on patients' preferences. In Japan, the patient, family and doctor are the main players in cancer disclosure. According to legal precedents in Japan, doctors are given a wide range of discretionary powers regarding disclosure [ 14 - 16 ]. As a rationale for holding discretionary power, doctors report a number of compelling reasons such as the need to protect patients from psychological distress caused by disclosure of the diagnosis, families' wishes for non-disclosure to patients, and the fact that most patients themselves do not wish to be told the truth [ 9 , 17 , 18 ]. However, several case-control studies report that there is no relationship between cancer disclosure and mental harm [ 19 - 21 ]. As family members are more reluctant than patients to disclose the truth [ 11 , 22 ], patients' needs for information are often unsatisfied in Japan where physicians often discuss the cancer diagnosis with family prior to informing the patient [ 23 , 24 ]. Doctors' discretionary powers and families' powers of attorney need to be reconsidered in the light of patients' preferences. This study's aim was to ask the general population whether they, in the event of developing cancer, preferred doctors' (or family members') discretionary powers regarding disclosure of the cancer diagnosis and prognosis. Methods This study was a cross-sectional, stratified random sampling survey of the general population in their 40s to 50s. As people over 60 years old are epidemiologically more at risk of having cancer, we excluded them not only because it seemed harmful to ask about these experiences, but also because there was a possibility that their responses would be affected by their experiences. Participants were selected from eligible voters in 'A' ward in the Tokyo Metropolitan Area. We chose 'A' ward as a representative area of Tokyo because various social indices such as the proportion of the elderly population, average length of education, and population growth rate were consistent with the Tokyo average [ 25 ]. The researchers mailed 443 questionnaires in October 2002 and received 246 responses (response rate 55.5 %). Amongst the respondents, 26 (10.5%) people had been diagnosed with cancer sometime in the past. As there were no significant differences in the responses of those who had been diagnosed with cancer and those who had not, we included these 26 respondents in the analysis. There were also no significant differences between those who were relatives of a cancer patient or were not and those who had dealt with cancer in their role as medical staff or had not. The sample size was determined by the need to provide adequate numbers to be able to detect differences among disclosure preferences with some degree of statistical certainty. The questionnaire was developed in consultation with 6 medical staff and 19 patients. The questionnaire presented a hypothetical scenario in which "The doctor discovers terminal cancer, but the patient does not know yet." to each respondent, and asked about preferences regarding diagnosis and prognosis disclosure; "How would you want to be told, if you were in such a situation". Answer choices for disclosure preferences regarding diagnosis were: 1."I would not want to be given any information regarding my diagnosis [non-disclosure]", 2. "I would like to obtain information regarding my diagnosis of a general nature but not in detail" 3." "I would like to be given all information regarding my diagnosis [full-disclosure]". Choices for disclosure on the prospects of complete recovery (CR) and expected length of survival (LS) were: 1. "I would not want to be given any information about the prospects of CR and LS [non-disclosure]", 2." I would like to obtain information on the prospects of CR and LS of a general nature but not in detail. [partial-disclosure]", 3. "I would like to be told about my prospects of CR and LS eventually. However, I would like to receive only general information on the prospects of CR (LS) when I am initially informed about the disease [postponed full-disclosure]", and 4. "I would like to be told about my prospects of CR and LS without delay. [immediate full-disclosure]". The reason for providing s answers allowing partial-disclosure was based on research by Akabayashi[ 26 ] which indicated that many Japanese were accustomed to and commonly preferred ambiguous or graded answers rather than polarised ones. Respondents were asked about their attitudes toward doctors and family members holding discretionary powers regarding cancer diagnosis disclosure. In order to compare the attitudes and characteristics of respondents who preferred immediate diagnosis and prognosis and those who did not, analysis was carried out twice. In the first analysis we included respondents who did not choose "full diagnosis and prognosis without delay", and we included the data from the remaining respondents that explained the reason for allowing to receive immediate diagnosis and prognosis. In the second analysis we included respondents who did want to receive diagnosis and prognosis, and we included data from the rest of the respondents about their reasons for preferring the withholding of diagnosis and prognosis. We also asked about preferences regarding the cancer disclosure process, such as whether people would like to obtain information ahead of their family. The questionnaire also included the trait part of the Japanese version of the State-Trait Anxiety Inventory (STAI), which assesses the personality predisposition to anxiety [ 27 - 29 ]. The Japanese version of STAI is a widely used and standardized test. In the present sample, the trait part of STAI for Cronbach's α = 0.90. Firstly, we calculated all respondents' disclosure preferences regarding diagnosis, CR and LS. Secondly, we calculated the attitudes toward doctors and family members holding prognosis discretion of respondents who preferred to be given diagnosis information directly, and those who did not. Wilcoxon's test was used to examine the differences between the attitudes held toward doctors and family members holding discretionary powers between these two groups. Statistical analyses were conducted using SPSS Version 11.5J. Results The socio-demographic characteristics of the respondents are shown in Table 1 . The mean age of the 246 respondents was 49.8 years (± 6.2 years). More than half (N = 143: 58.1 %) were female, 78 (31.7 %) had graduated from college, and 32 (13.0 %) were living alone. Table 1 Characteristics of the respondents. (N = 246) Mean SD Age (yr) 49.8 6.2 STAI (total score) 41.4 9.9 N % Sex (female) 143 58.1 % College graduates 78 31.7 % Living alone 32 13.0 % Married 186 75.6 % Living with adult child 85 26.1 % Living with infant child 98 39.8 % Principal household earner 133 54.1 % Non-religious 185 75.2 % Respondents' preferences regarding diagnosis and prognosis disclosure are shown in Table 2 . Regarding diagnosis, 85.4 % of respondents wanted full-disclosure, 11.3 % wanted partial disclosure and 2.9 % wanted non-disclosure. In the case of the prospect of a complete recovery; 35.7 % of respondents wanted an immediate full-disclosure, 17.2 % wanted a postponed full-disclosure, 39.2 % wanted partial-disclosure, and 2.9 % wanted no disclosure. Regarding the expected length of survival; 32.2 % of respondents wanted an immediate full-disclosure, 11.4 % wanted a postponed full-disclosure, 50.0 % wanted partial-disclosure, and 6.4 % wanted no disclosure. Table 2 Disclosure preferences regarding diagnosis and prognosis Non-disclosure Partial-disclosure Full-disclosure Diagnosis (N = 239) 7 (2.9 %) 27 (11.3 %) 204 (85.4%) Non-disclosure Partial-disclosure Postponed Full-disclosure Immediate Full-disclosure Prospect of Complete recovery (N = 238) 7 (2.9 %) 105 (39.2 %) 41 (17.2 %) 85 (35.7 %) Expected Length of Survival (N = 236) 15 (6.4 %) 118 (50.0 %) 27 (11.4 %) 76 (32.2 %) Regarding the contextual reason for wanting to receive full diagnosis and prognosis information without delay, 117 (69.6 %) respondents agreed to follow the doctor's initiative and 111 (66.1 %) of the respondents agreed to follow the with family member's decision [Figure 1 ]. The Wilcoxon test found no significant difference between these two groups (z = 0.186, p = 0.853). As for the reason for wanting the diagnosis and prognosis information to be withheld, 59 (26.5 %) of the respondents agreed to follow the doctor's initiative, and 79 (35.3 %) of respondents agreed to follow family member's wishes [Figure 2 ]. Wilcoxon test found significant differences between these two groups (z = 6.470, p < 0.001). Figure 1 Preference for who should decide whether to give immediate diagnosis and prognosis. N = 175. Figure 2 Preferences for who should decide whether to withhold diagnosis and prognosis. N = 240. Regarding the cancer disclosure process, more than half the respondents (N = 136; 55.3 %) answered that they would like to obtain diagnosis and prognosis information ahead of their family, a third (N = 82: 32.3 %) answering that would like to receive information with their family together at the same time. Only 26 (32.3%) respondents preferred to obtain this information after the doctor had already informed their family. Discussion Regarding preferences relating to diagnosis and prognosis, only 2.7 % of the respondents wanted no information regarding a cancer diagnosis. In addition to considering to tell or not to tell, the extent to which physicians should inform patient of diagnosis and prognosis poses a difficult decision in clinical settings. However, more than two-thirds (68.7 %) wanted full diagnostic and general prognostic information in a general nature but not in detail or wanted to be told about their prognosis eventually. Less than a third (28.9 %) wanted full information regarding diagnosis and prognosis without delay. These results suggest that a disclosure policy which provides patients with full information on diagnosis and general information on prognosis can satisfy the majority of patients' preferences. The results also suggest that any disclosure policy should also try to acknowledge and meet patients' wishes of being informed together with their families, and of being given information at a later time. Nevertheless, some patients do not want any information regarding their cancer diagnosis. In the clinical setting, medical staff needs to develop policy and procedures that can deal with the needs of patients who do not want any information as well as those patients and who want complete information immediately. The priority in identifying these types of patients over-rides other factors which affect patients preferences regarding diagnosis and prognosis such as patient characteristics and seriousness of cancer (previous research conducted by the authors [ 30 ]). Regarding those respondents who did not want to be given a diagnosis directly, those who preferred to follow a family member's decision were significantly larger than those who would prefer a doctor to decide. If patients reveal negative attitudes toward being given a cancer diagnosis at the time of initial consultation and testing, it may still be effective to tell the patient's family ahead of the patient or to have a discussion of cancer disclosure together with the family. Despite the data that indicates a mix of patients' preferences regarding cancer diagnosis, it may not be necessary for doctors to make choices regarding diagnosis by actually knowing individual patients' preferences. As opposed to those who would prefer no information, a greater proportion of respondents wanted to receive full information, even contrary to their preferences. Two other surveys with the general public show a similar tendency of patients wanting more information regarding cancer diagnosis than they used to. Asahi Newspaper found that regarding one's own cancer diagnosis and prognosis, in 1989, 59% of respondents wanted disclosure, which increased to 76%) in 2000 [ 31 ]. Similarly, Yomiuri Newspaper found that in 1994, 70% of respondents preferred being given information about a cancer diagnosis that increased to 78% in 2001 [ 32 ]. Thus the importance of providing information is widely supported by the majority of the general community. To simulate the fact that cancer results in a variety of disease outcomes for patients, we used scenarios with a range of severities in the outcomes of the cancer. As a result, there is little difference between respondents who had experienced cancer disclosure as a patient and those who did not, and the diagnosis preferences revealed in this study (full-disclosure, 85.4%) are consistent with previous studies (Seo [ 18 ], 85.7%: Miura [ 33 ], 88.1%). These findings suggest that this study's method succeeded in simulating a situation that reflected some degree of reality for respondents who had been given a cancer diagnosis in the past. This study has several limitations. Although the response rate to this study was moderate for a general population survey, we acknowledge that the characteristics of the respondents might not be wholly representative of the general population. Also, because we restricted participants to adult inhabitants in an urban area in Japan, further research is required to test the validity of these findings. It is recommended that health professionals become more aware about the need to provide patients with options to be given their cancer diagnosis and prognosis in a variety of ways. The greater proportion of respondents wants or permits disclosure of cancer diagnosis and prognosis. However, in patients who reveal negative attitudes toward being given a cancer disclosure directly, alternative options should be made available such as telling the family ahead of the patient or having a discussion of the cancer diagnosis with the patient together with the family. Further research with people aged over-sixty is needed to test the applicability of these findings to older age groups. Competing interests Although partial funding for this study was provided by the Education Ministry within the Japanese government, the views and opinions expressed in this report are those of the authors and not those of the funding organisation. Authors' contributions HM planned and conducted the survey, carried out the analysis, and wrote this paper. HT, MT, TS and IK made close supervision and extensive support. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538752.xml
555582	Obituary: Yukio Mano (1943–2004)	Yukio Mano, MD, PhD (1943–2004) Associate Editor, Journal of NeuroEngineering and Rehabilitation	I was terribly shocked to hear of the tragic and sudden passing of Yukio Mano on November 7, 2004, at the age of 61. He had not been well this past year but had been working continuously until just ten days before his death. Yukio Mano (Figure 1 ) was born on August 26, 1943 in Aichi Prefecture, Japan. He studied medicine at Nagoya University School of Medicine, and graduated in 1968. After he completed his basic medical training in Japan, he began his medical career in the USA in 1972. He first worked as a resident at the Institute of Rehabilitation Medicine at New York University for two years, then in 1974 he moved to the Department of Neurology at Baylor College of Medicine working as an assistant instructor and resident for one year. In 1975, he became a research fellow at the University of Maryland School of Medicine, in the Neuromuscular Research Unit. Yukio Mano studied the most advanced techniques in the fields of rehabilitation medicine and neurology during his four-year stay in the USA. Upon returning to Japan in 1976, he resumed his research in rehabilitation medicine at Nagoya University and the National Center of Neurology and Psychiatry, Japan. In 1981, he became an associate professor in the Department of Neurology at Nara Medical University. He was responsible for running the rehabilitation department there as a specialist in rehabilitation medicine. Finally, he was granted a full professorship in Rehabilitation and Physical Medicine at Hokkaido University (Graduate) School of Medicine in 1995. Yukio Mano was committed to helping researchers studying rehabilitation medicine, as well as young medical doctors and therapists in the rehabilitation field. He extensively expanded the Rehabilitation Department of Hokkaido University, and my colleagues and I had expected his leadership to continue into the future. His research interest was rehabilitation medicine, especially brain plasticity. He was the first Japanese developer of an apparatus that could deliver transcranial magnetic stimulation. Using this apparatus, he analyzed changes in the central nervous system resulting from various diseases, including brain plasticity after anastomosis of the musculocutaneous and intercostal nerves following cervical root avulsion, and cortical reorganization in training. The knowledge resulting from his research proved beneficial in the rehabilitation of disabled patients. He also emphasized a multidisciplinary approach to rehabilitation medicine and adopted new techniques from engineering. He received the Best Paper Award from ANNIE (Artificial Neural Networks in Engineering) in 2000 for his work entitled, "Adaptive FES Switching System for Hemiplegics". Yukio Mano served as a council member of the International Society of Electrophysiology and Kinesiology (ISEK) and, in 2000, he organized the XIII Congress of the ISEK in Sapporo, Japan. He was also a member of the editorial board of the Journal of Electromyography and Kinesiology. He served as a council member of many Japanese societies and organized nationwide congresses in Japan even in the year he died. He welcomed the launch of the new Journal of NeuroEngineering and Rehabilitation (JNER) and was honored to be asked to join the Editorial Board as an Associate Editor. He was indeed fully active to his last day. We cannot help praising him for all that he has accomplished in the fields of rehabilitation medicine, neurophysiology and kinesiology. We must also not forget that this remarkable scientist was also a caring family man. He always showed his love for his family as well as for his colleagues and friends. The loss of such an outstanding personality has been met with great sorrow by his family and the international scientific community. We will always remember him with great affection. Figure 1 Yukio Mano, MD, PhD 1943–2004	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555582.xml
554111	Abdominal fat and hip fracture risk in the elderly: The Dubbo Osteoporosis Epidemiology Study	Background Fat mass, which is a major component of body weight, is directly related to bone mineral density and reduced fracture risk. It is not known whether abdominal fat is associated with hip fracture. The present study was designed to examine the association between abdominal fat and hip fracture in women and men aged 60+ years. Methods This was a nested case-control study with one fracture case being matched with two controls of the same age. In women 63 cases were matched with 126 controls, and in men 26 cases were matched with 52 controls. Hip fracture was confirmed by X-ray and personal interview. Other measurements included weight, height, body mass index (BMI), abdominal fat, and femoral neck bone density (FNBMD). Conditional logistic regression model was used to analyse data. Results The odds ratio of hip fracture risk associated with each 10% lower abdominal fat was 1.5 (95% CI, 1.1 to 2.1) in women and 1.2 (95% CI, 0.7 to 2.0) in men. However after adjusting for FNBMD or body weight, the abdominal fat-fracture association was no longer statistically significant. Similarly, body weight and BMI was each significantly associated with hip fracture risk (in women), but after taking with account the effect of FNBMD, the association become statistically non-significant. Conclusion Lower abdominal fat was associated with an increased risk of hip fracture in elderly women, but the association was not independent of FNBMD or weight. The contribution of abdominal fat to hip fracture risk is likely to be modest.	Background Hip fracture is a public health concern, because it is associated with increased mortality, morbidity, reduced quality of life, and incurs significant economic and social costs [ 1 ]. Bone mineral density (BMD), a measure of bone strength, is a strong predictor of hip fracture risk [ 2 ], and is used as a surrogate measure of the severity of osteoporosis [ 3 ], the mechanism of BMD-hip fracture relationship is not well understood. Body weight is strongly related to bone mineral density, such that higher weight is associated with both higher BMD [ 4 - 7 ], and reduced fracture risk [ 8 , 9 ]. Body weight is the sum of lean and fat mass, and the relative importance of each component to hip fracture risk is contentious [ 10 - 14 ]. Lower fat mass was associated with an increase in the risk of hip fracture after adjusting for body weight and age [ 15 ], but it is not clear whether the significant relationship is independent of BMD. Central abdominal fat, which can be derived from dual-energy X-ray absorptiometry (DXA) scans, is highly correlated with, and has been suggested to be a surrogate measure of body fat [ 16 ]. Therefore, it is hypothesized that the BMD-fracture relationship may be partly mediated by fat mass, represented by central abdominal fat. The aim of this study was to test this hypothesis in a sample of elderly men and women of Caucasian background. Methods Setting and subjects The present study was designed as a nested case-control study within the larger Dubbo Osteoporosis Epidemiology Study (DOES), which has been on going since 1989 [ 17 , 18 ]. Briefly, in 1989, all men and women aged 60 or above living in Dubbo, a city of approximately 32,000 people 400 km north west of Sydney (Australia), were invited to participate in the DOES. At that time, the population comprised 1,581 men and 2095 women aged ≥ 60 years, of whom, 98.6 % were Caucasian and 1.4 % were indigenous Aboriginal. Dubbo was selected for the study site because the age and gender distribution of the population closely resembles the Australian population and it is relatively isolated in terms of medical care, so that virtually complete ascertainment of all fractures occurring in the target population is possible. This study has been approved by the St Vincent's Hospital Ethics Committee, and informed written consent was obtained from each participant. By mid 2003, 2560 subjects aged 60 + have participated in the study. Within this population, 89 (63 women and 26 men) hip fracture cases, which had had abdominal fat measured were identified from radiologists' reports from the two centres providing X-ray services as previously described [ 17 ]. Fractures were only included if the report of fracture was definite and, on interview, had occurred with minimum or no trauma, including a fall from standing height or less. Fractures clearly due to major trauma (such as motor vehicle accidents) and due to underlying diseases (such as cancer or bone-related diseases) were excluded from the analysis. For every fracture case, two non-fracture controls of the same age were randomly selected from the database. Age matching tolerance of ± 5 years was applied for women 85 + years and men 81 + years. In total, data from 267 subjects were included in the analysis. Measurements Subjects were interviewed by a nurse co-ordinator who administered a structured questionnaire to collect data including age, life-style factors such as past and present tobacco intake (assessed as pack-years) and alcohol consumption, physical activity. Anthropometric variables (height, weight) were measured and a dietary assessment was performed based on a frequency questionnaire for calcium intake as described elsewhere [ 19 ]. Femoral neck bone mineral density (FNBMD, g/cm 2 ) was measured by DXA using a LUNAR DPX-L densitometer (GE-LUNAR, Madison). The radiation dose with this method is <0.1 μGy. The coefficient of reliability of BMD in our institution in normal subjects is 0.96 and 0.98 at the proximal femur and lumbar spine, respectively [ 20 ]. Abdominal fat of the subjects was directly measured from the spinal DXA scan. Abdominal fat was derived from a standard window extending for 4 cm on either side of the first to fifth lumbar vertebrae. The DXA software expresses the fat mass in this abdominal window as a percentage of the total soft tissue. The coefficient of variation of this measurement as determined for dual scans performed on the same day in 60 people was 1.8% [ 21 ]. Statistical analysis The magnitude of correlation of associations between abdominal fat, body weight and FNBMD were estimated the product moment correlation coefficients and simple linear regression analysis. Differences in these measures between fracture cases and controls expressed as standardized difference (95% confidence interval- CI) were tested by paired t-test or Wilcoxon signed ranks test with significance level of 5%, depending on the distribution of data. The association between abdominal fat and hip fracture risk was assessed by the conditional logistic regression via the PROC PHREG [ 22 ] of the Statistical Analysis System (SAS) [ 23 ]. Results Abdominal fat in both women and men was normally distributed with no significant skewness. In the entire sample, there was no significant difference in percent of abdominal fat between women and men (23.9 ± 9.5 % vs. 24.0 ± 10.1 %, P = 0.993). Abdominal fat significantly decreased with age (r = -0.21, P = 0.003) with 1.4% (SE = 0.47) per 5 year in women. In men the rate of decrease was 0.5% (SE = 0.87) with each 5 year of age; however the decrease was not statistically significant (r = -0.07, P = 0.561). The correlation between abdominal fat and weight (r = 0.7, p < 0.001 for both genders) was higher than that between abdominal fat and FNBMD (r = 0.4, p < 0.001 in women and r = 0.2, p = 0.041 in men), (Figure). Figure 1 Correlations between abdominal fat and weight and femoral neck bone mineral density. (Abdominal fat was expressed as percentage of the total soft tissue). After matching for age and gender, compared to the controls, women with hip fracture had significantly lower weight (-0.7SD, p < 0.001), BMI (-0.6SD, P = 0.001), abdominal fat (-0.5SD, P = 0.014) and FNBMD (-0.9SD, P < 0.001). In contrast, in men there was no significant difference between those with hip fracture and those without fracture with respect to weight, BMI and abdominal fat; however FNBMD in men with a hip fracture was 1.1SD lower than those without a fracture (Table 1 ). Table 1 Baseline characteristics of participants as at 1989 Hip fracture Non fracture P value Standardized difference (95% CI) Women (n = 63) (n = 126) Age (y) 76.3 ± 7.1 76.4 ± 7.2 0.067 b 0.0 (-0.3 to 0.3) Height (cm) 155.8 ± 6.9 158.2 ± 6.1 0.008 b -0.4 (-0.7 to -0.1) Weight (kg) 55.6 ± 11.3 63.7 ± 10.9 <0.001 b -0.7 (-1.0 to -0.4) BMI (kg/m 2 ) 22.8 ± 4.0 25.3 ± 4.7 0.001 b -0.6 (-0.9 to -0.3) Abdominal fat a (%) 21.3 ± 9.2 25.5 ± 8.6 0.014 b -0.5 (-0.8 to -0.2) FNBMD (g/cm 2 ) 0.64 ± 0.10 0.75 ± 0.14 <0.001 b -0.9 (-1.2 to -0.5) Home physical activity (METs) 85.5 ± 34.9 76.6 ± 30.6 0.075 b 0.2 (-0.0 to 0.6) Calcium intake (mg/d) 608 ± 401 580 ± 370 0.455 c 0.1 (-0.2 to 0.5) Duration of smoking (y) 40.8 ± 16.6 33.2 ± 14.4 0.204 c 0.5 (-0.1 to 10.8) Smoking intake (c/d) 13.8 ± 8.0 11.8 ± 7.5 0.384 c 0.3 (-0.3 to 0.8) Men (n = 26) (n = 52) Age (y) 75.2 ± 6.0 75.1 ± 5.9 0.329 b 0.0 (-0.5 to 0.5) Height (cm) 169.8 ± 8.0 172.2 ± 5.5 0.219 b -0.4 (-0.9 to -0.2) Weight (kg) 71.7 ± 15.5 75.2 ± 9.0 0.277 b -0.3 (-0.8 to 0.2) BMI (kg/m 2 ) 24.7 ± 3.9 25.3 ± 2.2 0.442 b -0.2 (-0.7 to 0.3) Abdominal fat a (%) 26.2 ± 10.2 24.8 ± 6.2 0.581 b 0.2 (-0.4 to 0.7) FNBMD (g/cm 2 ) 0.64 ± 0.10 0.75 ± 0.11 0.002 b -1.1 (-1.6 to -0.5) Home physical activity (METs) 78.4 ± 37.4 78.5 ± 25.1 0.990 b 0.0 (-0.5 to 0.5) Calcium intake (mg/d) 546 ± 322 572 ± 242 0.450 c -0.1 (-0.6 to 0.5) Duration of smoking (y) 46.4 ± 13.4 36.2 ± 15.9 0.036 c 0.7 (-0.01 to 1.3) Smoking intake (c/d) 16.3 ± 5.5 16.4 ± 4.8 0.168 c -0.01 (-0.7 to 0.7) Results are expressed as mean ± SD; BMI, Body mass index; FNBMD, femoral neck bone mineral density; METs, metabolic equivalents a Abdominal fat was expressed as percentage of the total soft tissue; b Paired t-test; c Wilcoxon Signed Ranks test The risk of hip fracture was estimated to increase by 1.5-fold (95%CI: 1.1 to 2.1) in women and 1.2-fold (95% CI: 0.7 to 2.0) in men for each 10% lower abdominal fat. However after adjusting for BMD or body weight, the abdominal fat-fracture association was no longer statistically significant. (Table 2 ) Table 2 Odds-ratio (OR) of the risk factors for hip fracture in elderly women and men by conditional logistic regression analysis Unadjusted OR (95% CI) OR (95% CI) adjusted for FNBMD OR (95% CI) adjusted for weight Women Abdominal fat a (-10%) 1.5 (1.1 to 2.1) 1.1 (0.7 to 1.5) 1.1 (0.7 to 1.7) Weight (-10 kg) 2.0 (1.4 to 2.8) 1.3(0.9 to 1.7) - BMI (-4 kg/m 2 ) 1.7 (1.2 to 2.4) 1.2 (0.8 to 1.7) 1.5 (0.7 to 3.0) FNBMD (-0.12 g/cm 2 ) 2.4 (1.7 to 3.9) - 2.1 (1.3 to 3.5) Men Abdominal fat a (-10%) 1.2 (0.7 to 2.0) 1.5 (0.7 to 2.9) 1.8 (0.8 to 4.0) Weight (-10 kg) 1.4 (0.8 to 2.3) 1.5 (0.7 to 3.2) - BMI (-4 kg/m 2 ) 1.3 (0.7 to 2.5) 1.7 (0.6 to 4.5) 1.3 (0.9 to 1.9) FNBMD (-0.12 g/cm 2 ) 2.3 (1.3 to 4.0) - 3.0 (1.3 to 6.5) BMI, Body mass index; FNBMD, femoral neck bone mineral density. a Abdominal fat was expressed as percentage of the total soft tissue. Bold-faced values are statistically significant. Similarly, body weight and BMI was each significantly associated with hip fracture risk (in women), but after taking with account the effect of FNBMD, the association become statistically non-significant. In both women and men, the association between BMD and fracture risk was consistently significant either in unadjusted or in adjusted analysis (Table 2 ). Discussion It has been known for some time that body weight and whole body fat mass are significant predictors of hip fracture risk in women [ 9 , 15 ], however, it is not clear whether this association is independent of BMD. Abdominal fat has been shown to be well correlated with whole body fat mass [ 21 ]. The present study's finding of lower abdominal fat among hip fracture cases compared with the controls is consistent with previous observations [ 9 , 15 ]. However, it further suggests that the association between fat and hip fracture risk is not independent of BMD. Women with lower weight and fat mass may have lower FNBMD because of lower gravitational loading on the bone [ 24 , 25 ], or may have lower level of endogenous estrogens produced in adipose tissue and muscle [ 26 , 27 ]. On the basis of these findings, it may be proposed that BMD is a direct predictor of hip fracture risk, and that central abdominal fat (or fat mass) is a determinant of BMD. Thus, the previously observed relationship between fat and fracture risk is an indirect, rather than a causal association. Interestingly, in men abdominal fat or body weight was not significantly associated with hip fracture risk either before or after adjusting for BMD. Moreover, the magnitude of difference between fracture versus non-fracture cases in body weight or abdominal fat in men was generally more modest compared to that in women. For example, men with hip fracture had 0.2SD lower abdominal fat and 0.4SD lower in weight than non-fracture men. These differences were not significant. In contrast, in women the corresponding differences were significant and were 0.5SD lower for abdominal fat and 0.7SD for weight. This may suggest that the BMD-hip fracture association in men is not mediated via fat mass, despite a similar correlation between fat mass and BMD. However the lack of significance of the association between abdominal fat and hip fracture in men in the present study may be due to the small sample size. In recent years, there has been considerable interest in the relationships between osteoporosis, diabetes and cardiovascular disease [ 28 - 30 ]. A common characteristic of individuals with diabetes and cardiovascular diseases is that the majority have higher body weight and fat mass [ 31 - 38 ], and on this basis together with the well-known relationship between weight and BMD, it is expected these individuals would have higher BMD and lower risk of fracture. However, epidemiological data point out that individuals with cardiovascular diseases have lower BMD, and a higher risk of fracture [ 29 , 39 ]. The present study also found that, without BMD adjustment, men and women with lower body weight had a higher risk of hip fracture. Hypertension has been suggested as a potential contribution to the risk of hip fracture [ 39 ]. A previous study showed that abdominal fat is positively correlated with blood pressure [ 40 ]. However, in this study lower, not higher, abdominal fat was a risk for hip fracture in women. These data suggest that the association between diabetes, cardiovascular diseases and hypertension and fracture risk is also not mediated via fat mass. From a public health point of view, the present study's finding suggests that abdominal fat does not add to the discriminatory value of hip fracture risk that is already provided by BMD. Indeed, in this study, none of the body size measurements (weight, height, BMI and abdominal fat) was a significant predictor of hip fracture risk after adjusting for BMD. This suggests that these measures have limited value in the prediction of hip fracture in a population or an individual. A number of issues should be kept in mind before extrapolating the present finding. First, the participants in this study were Caucasian aged 60 years and above, so it may not be generalizable to younger populations and to different races. Second, neither total body fat, nor waist and hip circumferences (WHC) were measured and these may have had stronger predictive value. However, these measurements may underestimate abdominal adiposity in those with both large waist and hip circumferences, while DXA determination of regional body fat distribution may indeed be more valid than WHC [ 16 ]. Conclusion These data have demostrated that in the elderly, abdominal fat was significantly associated with hip fracture risk in women but the association was not independent of BMD, whereas in men abdominal fat was not a significant predictor of hip fracture risk. Measurement of DXA abdominal fat does not contribute to hip fracture prediction over and above that provided by BMD. List of abbreviations All abbreviations are defined in the text. Competing interests The author(s) declare that they have no competing interest Authours' contributions NDN obtained and analysed the data, and drafted the manuscript. CP had an active role in data analysis and interpretation of results. JRC had an active role in the conduct of the Dubbo Osteoporosis Epidemiology Study and helped with the interpretation of results. JAE established the Dubbo Osteoporosis Epidemiology Study. TVN had an active role in the conception if this project, involved in study design, analysis data and interpretation of results. All authors contributed to the last version of the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC554111.xml
516450	The retroviral RNA dimer linkage: different structures may reflect different roles	Retroviruses are unique among virus families in having dimeric genomes. The RNA sequences and structures that link the two RNA molecules vary, and these differences provide clues as to the role of this feature in the viral lifecycles. This review draws upon examples from different retroviral families. Differences and similarities in both secondary and tertiary structure are discussed. The implication of varying roles for the dimer linkage in related viruses is considered.	Introduction With relatively few genes compared to many other virus families, the retroviridae have evolved over the millenia to maximise the functions of their RNA genome. The genome serves as a versatile template from which various proteins can be translated by the use of splicing and by translational flexibility using scanning, IRES and frameshifting. It is also an RNA molecule capable of interacting with itself, and cellular and viral proteins. By these means, from an RNA around 7 – 12 kilobases long, the retroviridae have evolved to infect a wide range of species and cell types. A unique characteristic of retroviral genomes is the fact that they are dimeric. The reasons for this are as yet unclear, and are discussed below. In brief, it is thought that the diploid genome allows template switching during reverse transcription and may be linked to recombination in some viruses. It may also play a role in translation of proteins and packaging of the RNA. Much of the work on the nature, structure(s), and role of the dimer linkage has been based on Human Immunodeficiency Virus Type 1, and this has been recently reviewed ([ 1 ] and Russell et al this issue [ 2 ]). Whether or not HIV-1 is a representative model for other retroviruses is open to debate. However, there have been important contributions from investigators studying other retroviruses. They have shown both similarities with the HIV-1 motifs, and also, importantly, differences. The fact that distinct RNA structures are used by different retroviruses to perform the same purpose, namely to link their two RNA molecules, tells us something very important. For these viruses, whatever organism or cell they are infecting it has been advantageous to evolve to do so with a double complement of genome in their virion particles. However, diploidy may be used to benefit the virus in a number of ways and for different viruses the priorities may vary. This review will attempt to draw on several examples from viruses other than HIV-1, whilst of necessity drawing comparisons with the latter. The dimeric genome Retroviruses were discovered at the beginning of the 20th century [ 3 , 4 ]. The unique nature of their genome was first discovered in the 1960s [ 5 , 6 ] but the actual dimeric genomes were elucidated, and visualised by electron microscopy, a decade later [ 7 , 8 ]. Bender and colleagues extracted the RNA from several different retroviruses and examined it by electron microscopy under denaturing conditions. The RNA appeared to be joined at a discrete point, termed the dimer linkage site (DLS). Using bromodU to label the RNA at one end, they were able to show that the molecules were joined at their 5' ends [ 9 , 10 ]. Under less stringent conditions the genomes can be demonstrated to interact along their lengths [ 11 ] and it is this that probably contributes to confusing reports on the exact location of the primary DLS in different viruses. RNA dimerisation in the primate lentiviruses, predominantly HIV-1, has subsequently been extensively studied [ 1 ], yet little has been published on this process in the non-primate lentiviruses. Early studies of rapid harvest virions of the prototype lentivirus, Maedi Visna virus (MVV), identified viral RNA with a Svedberg coefficient of 35S immediately post-budding, which increased with time to 70S. It is possible that weakly interacting dimers formed during RNA encapsidation may have been denatured during purification, however these observations are supportive of a progression from monomeric to dimeric RNA associated with viral maturation [ 12 ]. Since 1990 [ 13 ] it has been possible to study in vitro the RNA elements involved in the dimer linkage first observed by EM. It was shown that RNA transcripts comprising sequences from the 5' end of the viral genome would migrate as two species of RNA when subjected to electrophoresis. By this means many subsequent studies were able to focus on isolating the elements and structures involved in dimerisation, and to investigate the role of the viral structural proteins in this process. Multiple functions for the dimeric genome? As yet investigators have not been able to agree on a distinct role for the dimer linkage. The fact that it is conserved amongst the retroviridae does not guarantee that its role will be the same in all retroviral families. The following section of the review will endeavour to explore some of the proposed roles, and examine the evidence from different retroviruses. The dimeric linkage and recombination Several studies have demonstrated that, in HIV-1 and MLV, the dimer linkage serves as a "hotspot" for recombination [ 14 , 15 ]. It is an obvious hypothesis, that in viruses which are known for their hypervariability, there exists the capacity to jump from one RNA molecule to another. Researchers have compared dimerising to non-dimerising controls, and the frequency and distribution of template switching. Templates containing the dimerisation site had a 4-fold higher transfer efficiency than the non-dimerising control [ 14 ]. This result implies that recombination would occur preferentially at the site where the RNA molecules were in close proximity. In the case of HIV-1, whilst it has been shown that template switching is facilitated by template homology [ 16 ], it has also been demonstrated that recombination can occur between viruses of different subtypes which might have different dimer initiation sequences (DIS) [ 17 ]. Bearing in mind the fact that the genome is linked at other sites besides the DIS [ 11 ], it seems probable that other hot spots for recombination exist. Interestingly, it has been suggested that the nucleocapsid protein (NC) promotes or stimulates the strand transfer reaction. As will be discussed below, NC and the precursor Gag protein both bind the RNA close to the DIS in HIV-1. In addition, there is evidence that the presence of a dimer in the virus particle facilitates the first strand-transfer reaction of reverse transcription [ 18 ]. Work in our laboratory has shown that the Maedi Visna Virus DIS is centred on a helix terminating in a GACG tetraloop between positions 281 and 300 in the viral genome; a region which is highly conserved between the ovine and caprine lentiviruses (Monie, personal communication, see Figure 3d ). Intriguingly, this structure shows homology with structural motifs in the Alpha- and Gammaretroviruses , but not with DIS regions identified in the primate lentiviruses. Within the Alpha - and Gammaretroviruses GACG tetraloops are involved in the packaging of viral RNA [ 19 , 20 ] and whilst not a component of the core M-MLV DIS motif [ 21 ], they may contribute to the process of dimerisation and the stability of the resultant dimer [ 22 ]. Importantly, it is possible to form heterodimers between transcripts from these viruses containing the GACG tetraloops and between MVV and M-MLV (personal observations). This raises parallels with recent studies of the dimerisation of murine leukaemia viruses and Harvey Sarcoma virus in which GACG tetraloops were found to regulate inter-species RNA heterodimerisation [ 23 ], whilst other linkage elements were postulated to mediate homodimerisation. Recombination, and the genomic variability it confers cannot be the sole function of the dimeric genome, since retroviruses with highly conserved genomes and little sequence variability such as HTLV-1 [ 24 ] are also dimeric. Translation and packaging? Another possible role is that of the dimer linkage acting as a switch, its presence permitting or restricting the packaging of RNA. In HIV-2 two regions were originally suggested as dimer initiation sites, one analogous to the palindromic sequence identified as the principal DIS in HIV-1 (termed SL1), one close to the PBS [ 25 - 27 ]. Recently, a region upstream of SL1 (also called the DIS, see Figure 1a ) was identified as being critical for packaging [ 28 ]. An extensive deletion analysis of the 5' leader of HIV-2 was carried out, and removal of nucleotides 380–404 (HIV ROD ), termed the DM region, rendered the virus severely packaging deficient. The mutation had been designed based on the mfold [ 29 , 30 ] prediction, that removal of these sequences would disrupt the SL1 structure and hence dimerisation (Figure 1b ). In vitro studies using RNA transcripts comprising the leader region with and without the DM deletion, reveal that it does, indeed, render the viral RNA monomeric (personal observations). Using antisense oligonucleotides, another group have demonstrated that this region may, in fact, play a role in the dimerisation process itself [ 31 ]. By free energy minimisation this region is predicted to be unstructured, so it is not clear how the RNAs would interact with one another. In addition, whilst the SL1/DIS sequence is conserved amongst HIV-2 and SIV sequences in the database, that within the DM region is less so, and the substitutions which exist would affect the auto-complementarity of the sequence. Figure 1 Structure of the HIV-2 leader region. 1a. Secondary structure model of the HIV-2 leader region based on mfold predictions. Indicated are the putative dimer linkage sites (in red). Also highlighted is the DM region defined as being critical for packaging [28], in blue). 1b. The effect of the DM deletion on the SL1/DIS stem loop. The stem is truncated and the internal bulge altered in approximately half the predicted structures. Figure 2 Structure of the key elements involved in HIV-1 RNA dimerisation. 2a. Secondary structure model of the packaging signal of HIV-1 Lai ([64] [65]), containing the principal DLS. 2b Proposed sequence of the RNA dimerisation process in HIV-1 Lai . The initial kissing hairpin interaction (including loop B) followed by formation of the extended duplex ([1]). 2c. Loop B, one of the critical elements in the dimer interaction. The flexibility of this internal loop allows the duplex to form ([44]). Figure 3 Dimer linkages of the retroviridae (excluding the lentiviruses). 3a. Loose and tight dimers ([51]). 3b. Imperfect repeats ([66]). 3c. Palindromes ([38]). 3d. GACG loops ([23]). 3e. CAG tri-loops (see Figure 4). Figure 4 Proposed tertiary structure of the HTLV-1 dimer linkage. 4a Stereoview of 3D molecular modelling of a potential structure of the HTLV-1 DIS from nucleotide A730 to A744 using JUMNA ([61]). 4b. Close up of the terminal loop. Bases are coloured as follows: adenine, grey; cytosine, yellow; guanine, orange; and uracil, cyan. One of the key differences between HIV-1 and HIV-2 replication is their modes of packaging [ 32 ]. Whilst the Gag protein of the former captures the genomic RNA in trans , the latter uses predominantly a cis mechanism. One might postulate therefore, that, if retroviruses must package a dimeric genome, it is critical in the case of HIV-2 that the genome is dimeric before interacting with the Gag polyprotein. Hence, the effect of mutations in the DM region may be to render the RNA monomeric and thus to severely impair packaging. It is attractive to speculate that the reason packaging itself is not affected by DIS mutations to the same degree in HIV-1 [ 33 ] is this difference in protein:RNA interaction. If the RNAs can interact at points other than the principal DIS over time, then perhaps the trans mechanism is less dependent on a high affinity dimer linkage? Particle maturation and viral infectivity A recurring observation amongst investigators is the fact that mutation or deletion of dimer linkage sites causes viral infectivity to be decreased [ 33 ]. One explanation for this might be that a dimeric genome is a prerequisite for maturation of virus particles. Certainly, immature HIV-1 particles are non-infectious, and viruses with their DLS mutated have been demonstrated to form only immature particles [ 34 ]. The DLS of Human T Cell Lymphotropic Virus Type 1 (HTLV-1) was identified as a 14-nucleotide sequence just downstream of the splice donor [ 35 ]. Removal of this region from the leader sequence rendered the RNA monomeric in vitro [ 24 , 36 ]. When this deletion was introduced into the wildtype genome sequence, the only viral replication defect that was observed, following transfection and subsequent infection, was that of impaired infectivity [ 37 ]. Likewise, a similar effect was observed when the DLS of Human Foamy Virus was mutated [ 38 ]. Parent et al showed that if the RNA of Rous Sarcoma Virus (RSV) was engineered so that it was monomeric, the virus was non-infectious [ 39 ]. Interestingly, this group suggested that it might be a difference in localisation of structural proteins and RNA affecting subsequent dimer formation and viral infectivity [ 40 ]. This is an area that has not been explored to any extent. Also working with RSV, Bieth and colleagues found that, in an in vitro system, dimer formation appeared to inhibit synthesis of the Gag polyprotein precursor [ 41 ]. Structure of the dimer linkage Undoubtedly the best defined dimerisation structure is that involved in the dimer linkage of HIV-1. The discovery of the sequences involved, the subsequent description of the RNA:RNA interaction, and the elucidation of the tertiary interaction are described elsewhere [ 1 ]. The initial interaction between the two RNAs appears to be a kissing loop interaction (similar to that seen in the regulation of plasmid replication, [ 42 ]) followed by annealing of the two RNAs into an extended duplex (Figure 2b ). The sequences contained within the palindrome are remarkably conserved. Using an in vitro selection system it has been possible to demonstrate that the DIS has evolved to satisfy both constraints for optimal dimerisation affinity, and the potential to homodimerise [ 43 ]. The dimer linkage is found at the terminal end of Stem Loop 1 (SL1) within the packaging signal region of HIV-1 (Figure 2a ). The tertiary structure of the whole SL1 RNA has been determined [ 44 , 45 ] and the structures have helped to determine exactly how the RNAs interact with one another. A number of elements appear to be critical for the dimer interaction: flanking purines and central nucleotides in the palindromic sequence [ 46 ] and loop B [ 47 - 49 ]. The tertiary structure of the latter has been described (Figure 2c ), and there is some debate as to how flexible this internal loop might be. However, work by Borer et al , examining the interaction of NC with elements of the packaging signal, of which loop B is one, showed that, in fact, both structures might exist, the flexible one allowing NC binding at high affinity [ 50 ]. There are similar linkages in other retroviruses. The Avian Leukosis Viruses also interact firstly in a kissing hairpin manner, and then form an extended duplex (Figure 3a , [ 51 ]). Palindromes remain a theme throughout many of the viruses investigated to-date. As already mentioned, the DIS of HIV-2 is less well defined than that of HIV-1. Whilst there is a palindromic sequence at the top of a stem loop structure that closely resembles the HIV-1 DIS (see Figure 1 ), there are other regions which have also been demonstrated to be important for dimer formation [ 25 , 26 ]. Other viruses with palindromic sequences as their DLS include HFV (Figure 3c ) and MoMLV. In the case of MLV there are other sequences and structures which may play a role in dimer formation, including the GACG tetraloops mentioned previously [ 52 ]. The tertiary structure of this stem loop is the only proposed dimer linkage element yet to be determined in a retrovirus other than HIV-1 ([ 53 ]). RSV and VL30, also have imperfect repeat sequences in their dimer linkages [ 54 , 55 ] (Figure 3b ). Recent work by Monie and colleagues [ 36 ] describes the potential tertiary structure of the HTLV-1 dimer linkage, capped by a novel CAG tri-loop (Figure 3e and Figure 4 ). This tri-loop is formed by an unusual C:synG base pair closing the loop. Other similar loops have been described, in the domain IIId terminal loop of the hepatitis C virus internal ribosomal entry site (IRES) [ 56 ] and in stem loops required for initiation of transcription within the Bromoviridae [ 57 ]. Although sequence heterogeneity between HTLV-1 isolates is rare, distinct mutations identifying individual strains can be identified. Of 101 HTLV-1 sequences identified from the EMBL database, 90 showed sequence homology with HTLV-1 CH , the strain used in the study. The other 11 sequences comprised three different variants. Eight contained a deletion of C736 (see Figure 4 ), two possessed the substitution A737G, and one possessed the substitution C733U. The substitution mutants have minimal impact on regional secondary structure, while the deletion may induce formation of a CAGG tetraloop. Interestingly, the A737G mutation possesses homology with 150 deposited HTLV-II sequences, suggesting a conservation of the DIS between HTLV-I and -II. Conclusions The retroviral RNA genome structure does not stay static during the course of transcription, translation and ultimately packaging. Various investigators have suggested that this constantly changing RNA structure plays an intimate role in the viral replication [ 58 - 61 ]. It seems possible that linkage of the two RNA molecules constituting the genome is integral to the changes in RNA structure. As described in the article above, the dimer also acts as a mechanism for promoting recombination; may be a signal for packaging to occur; may be an inhibitory signal; may direct processes to occur in specific cellular compartments; and lastly, may be capable of interacting with cellular proteins. In vivo data has revealed just how important an intact dimmer linkage may be to a retrovirus. For instance, there are intriguing differences in the effect of dimer mutations on viral infectivity depending on the cell type being infected [ 62 ]. What the significance of this might be in the context of a viral infection is, as yet, unclear. The importance of the dimer linkage is perhaps most clearly exemplified by the observation that a patient infected with a viral isolate having a defective DLS, had a low viral load. The subsequent switch in the predominant virus to that with a competent DLS coincided with a rise in viral load [ 63 ]. One can speculate that, at least in the case of HIV-1, only those viruses with a whole, optimised dimer linkage are capable of efficient infectivity. For the purposes of examining the role of retroviral RNA dimer sequences in the context of animal models, the non-human retroviruses, including the non-primate lentiviruses will be of great importance. To sum up, retroviral dimeric genomes are linked by a variety of RNA structures, including kissing hairpins, GACG tetraloops and unusual CAG-tri loops. The differences in these interactions, and when or where they occur, may reflect different demands upon this unique feature, and highlight the elasticity of the RNA genome. Competing interests None declared.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC516450.xml
543445	Assessment of possible impact of a health promotion program in Korea from health risk trends in a longitudinally observed cohort	Background Longitudinally observed cohort data can be utilized to assess the potential for health promotion and healthcare planning by comparing the estimated risk factor trends of non-intervened with that of intervened. The paper seeks (1) to estimate a natural transition (patterns of movement between states) of health risk state from a Korean cohort data using a Markov model, (2) to derive an effective and necessary health promotion strategy for the population, and (3) to project a possible impact of an intervention program on health status. Methods The observed transition of health risk states in a Korean employee cohort was utilized to estimate the natural flow of aggregated health risk states from eight health risk measures using Markov chain models. In addition, a reinforced transition was simulated, given that a health promotion program was implemented for the cohort, to project a possible impact on improvement of health status. An intervened risk transition was obtained based on age, gender, and baseline risk state, adjusted to match with the Korean cohort, from a simulated random sample of a US employee population, where a health intervention was in place. Results The estimated natural flow (non-intervened), following Markov chain order 2, showed a decrease in low risk state by 3.1 percentage points in the Korean population while the simulated reinforced transition (intervened) projected an increase in low risk state by 7.5 percentage points. Estimated transitions of risk states demonstrated the necessity of not only the risk reduction but also low risk maintenance. Conclusions The frame work of Markov chain efficiently estimated the trend, and captured the tendency in the natural flow. Given only a minimally intense health promotion program, potential risk reduction and low risk maintenance was projected.	Background Evidence was found that health promotion programs affect health risks in the US and in many other countries [ 1 - 4 ]. Also, a consistent association of higher risk individuals with higher medical costs implies a potential impact of risk reduction on cost moderation [ 5 , 6 ]. Musich et al. [ 7 ] showed that participation in health promotion programs can be effective in moderating medical costs. While most health promotion programs in the US focus on the improvement of health rather than direct economic benefits, many economic evaluations claim that there are transfers of benefits between participation, risk reduction and cost savings [ 8 - 12 ]. However, without identified health risks and a systematic evaluation of the needs to provide the quality programs in the Korean society, an implementation of such programs would be unlikely. A previous study with a random sample from a 12-year cohort of civil servants in Korea [ 13 ] provided an insight that lifestyle factors predicted future medical utilization reasonably well. This suggests that risk status, measured by lifestyle and biometric factors at a point in time, could be used as a pivot to estimate future medical utilization as a result of risk progression. Longitudinally observed cohort data can be utilized for health promotion and healthcare planning provided the health risk trend is estimated and it poses a general need for an intervention. This paper attempts to address an issue around the possibility of predicting the impact of a health promotion intervention by applying observed effectiveness data from a population with an intervention to the observed transitions of risk status in a Korean population in the absence of such a program. Methods To help in understanding the overall risk transitions in the Korean population and to implement an effective intervention program, a Markov chain model was utilized, assuming finite risk states at any point in time [ 14 ]. Today's weather affects tomorrow's but yesterday's may be already irrelevant to tomorrow's forecast. Stock price tomorrow may depend on the previous week's stock prices, not just today's. In estimating the chance of getting "sunny" weather tomorrow, most relevant information to it could be today's weather, where today's weather can be described as "rain", 'cloudy", or "sunny", for example. In general, a random process with fixed number of values could be numerically described, with the collection of possible values forming a "state space" (all possible weather, for example) and the possible values being "states" (rain, cloudy, or sunny, for example). Markov chain models the dependent structure of the future state of a random process on previous states as in the weather forecast example. In the current study, the Markov chain model can describe risk transitions over time when future risk transitions depend on the previous risk states. This modelling is used when a decision problem involves risk state change over time, and interest in the event. This also enables one to project the health status of a population [ 15 - 18 ]. This paper utilizes the observed transitions of measured health risks in a cohort of the Korean National Health Insurance Corporation registrants (KNHIC) over 5 years. This longitudinally-followed population trend, without a particular intervention or policy change in place, was used as a basis to estimate a natural history of risk flow. In addition, an intervened transition was simulated, given that a health promotion program was implemented for the KNHIC cohort to project a possible impact on the improvement of health status. Comparing the two transitions also provided directions for a health promotion program that might be implemented in this population. Population The study population consisted of the established KNHIC prospective cohort [ 13 ]. Registrants of KNHIC were invited to complete a health survey prior to each mandatory bi-annual physical examination. The respondents to this preliminary-health risk appraisal (p-HRA) in 1992, were followed bi-annually from 1996 to 2000. Reliability and validation tests were not carried out for the p-HRA. The criteria for inclusion in the study are: (1) actively employed over the period, (2) ages between 30 and 65 in 1996 (N = 180,767). Similarly, a comparison population was selected for the simulation of a program effect from the large longitudinal database of University of Michigan-Health Risk Appraisal (UM-HRA) completers. The UM-HRA was originally a CDC version, which was modified to fit the national trend of cost, and to meet the guidelines over time. Additional conditions applied are: (1) participated in a health promotion program at a minimal intensity, (2) completed the UM-HRA at least three times during 1996–2000, (3) were insured by the same insurance plans during the corresponding years and (4) were actively employed over the period at the same industry, and (5) age under 65 years in 1996 (N = 15,793). Questionnaire The survey questionnaires (p-HRA), on health status, diet, and lifestyles, were sent to work places and homes to encourage a national health screening at designated health care facilities and to measure lifestyle related health behaviors, every two years for the KNHIC registrants. UM-HRA was used to appraise individual health status during the same period (1996–2000) for the US population. The validity of UM-HRA has been addressed elsewhere [ 19 , 20 ]. Health risks and costs Three lifestyle-related health risks were measured by the corresponding questionnaires (p-HRA for the study cohort, and UM-HRA for the comparison population): physical activity, alcohol consumption, and smoking. In addition, pivotal measures of overall health were collected by the questionnaires: perceived health and medical conditions. Three biometric measures were obtained from the appropriate lab tests during physical examination for the KNHIC population and from the self-reported measures for the comparison population: systolic blood pressure (SBP) /diastolic blood pressure (DBP), total cholesterol, and body mass index (BMI) via height and weight measures. The risk criteria for the US population were defined (Table 1 ) following the published guideline by the US–CDC/Carter Center, and some were modified to fit better for prediction of healthcare costs (BMI and physical activity). Table 1 Risk Evaluation Criteria and baseline characteristics. Individuals from US population were classified as Low, Medium, and High risk as in KNHIC population. Within each risk group, random samples were selected each time of sampling from US population stratified with age and gender once they met the similar risk profile of KNHIC. This bootstrap-sampled match would be used as a control (interven ed) population. Baseline characteristics KNHIC population (N = 180,767) Comparative population with intervention (N = 180,767)* Risks Criteria N (%) Criteria N (%) Perceived Health Poor /Fair 24,290(13.4) Poor /Fair 31,814(17.6) Exercise Less than 1/week 100,395(55.5) Less than 1/week 56,399(31.2) Alcohol Drink>7/week 1 17,554(9.7) Drink>14/week 16,630(9.2) Smoking Current smoker 55,052(30.4) Current smoker 29,645(16.4) BMI BMI>25.0 for male, >23.0 for female 2 46,227(25.6) BMI>27.50 69,776(38.6) BP SBP> 120 or DBP>80 3 91,924(63.9) SBP> 139 or DBP>89 48,084(26.6) Cholesterol Cholesterol>220 4 32,118(17.2) Cholesterol>239 7,954(4.4) Medical condition Self-reported disease 9,280(5.1) Self-reported disease 5 35,973(19.9) Baseline Class Average Age = 40.0 Male (61%) Average Age = 40.0 Male (61%) Low Risk (0–2) 62.9% 63.0% Medium Risk (3) 22.2% 22.3% High Risk (4+) 14.9% 14.7% Note 1 a drink of "Soju" = 2 drinks of wine/beer 2 WHO guideline (1999) = 23.5 for Asians 3,4 Korean Medical Association (2000) guideline 5 diabetes, heart problem, cancer, past stroke, bronchitis/emphysema Simulated data after adjusted for age, gender, and baseline risk for KNHIC population Information on eight health risks for the study cohort were systematically evaluated and mapped to the measured risks by UM-HRA (Table 1 ). This was done according to: (1) the published guidelines for Asians, (2) empirical comparison of question by question, and (3) age and gender adjusted association to the respective healthcare costs distribution [ 13 , 21 - 25 ]. In addition, different risk criteria were applied to alcohol consumption and medical condition due to systematic difference in measurement. Corresponding health states per period were defined according to the distribution of the aggregated risk state (sum of individual risks variable states) as low (0–2 risks), medium (3 risks) and high (4+ risks). Inpatient plus outpatient costs per annum were collected from KNHIC database for the medical utilization in association with health risks, and the inflation adjusted average 1996 costs (January 1st, 1996 through December 31, 1996) were used for the T1 costs. Similarly, average T2 costs were calculated from 1998 claims costs. Program Participants of the p-HRA were not given any further information on identified health risks. Neither was it used to gain access to any health intervention programs during the five years (1996–2000). In US, on the other hand, as part of an intervention program, the completers of UM-HRA were given individually tailored health information, followed by encouragement of participation in a health promotion program at no cost during 1996–2000. This nation-wide program included an annual mail-based HRA, personalized follow-up report, identification of top significant risks and referral to health resources. This was defined as minimal level intervention, which differs from KNHIC's p-HRA in providing health information and individual feedback. Trend Provided that the least resources were available for a health intervention of the KNHIC population, a projected health risk transition with the minimal level intervention was utilized for the simulation of possible impact on risk transition. Population health trends were followed over the three time frames (T1, T2, and T3). T1 refers to the baseline year, which is 1996 for both populations. T2 refers to the second time frame, which is 1998 for the KNHIC population, and 1997 or 1998 for the compared US population. T3 refers to the 3 rd time frame, 2000 for the KNHIC population, and 1998 or 2000 for the US population. Population with minimal intervention was matched to the baseline characteristics of KNHIC population, using age, gender, and baseline risk distribution. Trend was defined as the risk state change in each population between the time-points while each transition was annualized. Each change of risk state was estimated and the parameters to trend from the matched intervened population were used to project the possible trend of KNHIC population, following such a program. Analysis An age/gender-cohort model was implemented, following observations on the natural flow of health status over three time frames, and it was compared to the corresponding age/gender-cohort of a US population with the matched risk distributions, where an intervention at minimal level was applied. For a simulation of an intervened transition for KNHIC population, an adjusted estimation of intervened transition was obtained based on random samples drawn from the selected US sample of intervened employees (N = 15, 793). To simulate additional random samples to reduce variances in the estimate of parameters for transition, a Monte Carlo bootstrap [ 26 , 27 ] method was employed to the sub-populations with age, gender, and baseline risk matched to the KNHIC populations (N = 180, 767). For each 10 years apart (30–40, 41–50, 51–60) cohort from the matched boosted random samples of US population, risk transitions following a minimal intervention were estimated while controlling for age, gender, current and previous risk states, and previous year's medical claims costs. Similarly, natural flow of risk transitions was estimated from the observed KNHIC cohort data. The estimated parameters (covariate and baseline effect, corresponding mean, variance and covariance) and fitted model for the intervened transition was applied to the KNHIC population for the transition probabilities for each risk state at T3. Finally, an aggregated weighted probability for transition was calculated based on age-gender-risk state distribution. Computations were carried out using a generalized linear model for the ordered categorical outcomes (low risk<medium risk<high risk), assuming a Markov chain model in which the order of dependency was learnt from the data (Figure 1 ). Figure 1 depicts the transition between risk states at each cycle and the dependency of risk progress on the previous risk state. All risk states are inter-connected and allow feedback cycles at each risk state (staying at the same state). Figure 1 Markov transition with 3 risk states without an exit. Assumptions of Markov chain model, and model fit diagnostics were evaluated for the KNHIC population. Stationarity was tested by likelihood ratio test of the fitted models using a generalized equation estimation with and without time-dependence while controlling for other measured covariates. Order of Markov chain was tested with likelihood ratio test and fit-diagnostics were run. Following the fitted Markov model, we presented the estimated transitions over three time-frames for the natural flow. Cochran-Armitage trend tests controlling for T1 or T2 risk state were performed to test the trend of dependency of future state on T2 or T1 risk state. Average costs in each T1-T2 risk state pair were calculated from the corresponding years' average costs, adjusted for inflation. Healthcare costs were converted to ratios of average costs per T1-T2 risk transition state, compared to low (T1)-low (T2). Adjusting with any appropriate inflation rate in the future, these ratios can be used to project the cost savings over time by comparing the number of projected trends times cost ratios in natural and intervened flows. The projected percentage changes in the two trends (natural vs. intervened) were compared for an estimation of an intervention impact on risk change. Results Feasibility of Markov chain model Individual baseline risk prevalence of the two cohorts is shown in Table 1 . The simulated intervened cohort was adjusted for age, gender, and baseline risk states according to those of KNHIC data. Compared to the US population, lack of physical exercise, and smoking were significantly more prevalent among the Korean cohort at baseline. Biometric risks such as blood pressure and cholesterol were higher in the KNHIC population due to 100% compliance rate of lab test of the study population. Overweight and medical condition was significantly less among the Korean. However, overall baseline risk distributions based on the clustered state (low, medium or high) were about the same. A simulation of an intervention effect follows the adjusted baseline characteristics of the comparative study population and Table 1 shows similarity of the two populations at baseline such as age, gender and risk distribution for comparability. Under the assumption of finite number of health risk states (low, medium and high states), Markovian models were examined for proper order estimation, an association structure, and stationarity. Following a likelihood ratio test, Markov chain (MC) order 2 was preferred for the risk transition of the non-intervened (KNHIC) population (Table 2 ). In other words, future state depends on the current and the most recent past states and when controlled for the dependency on risk state, overall future risk state depends on the individual risk factors far less significantly. This also assured that matching intervened trend is applicable to the KNHIC population without further concern on effect of inherent risk progress to future risk transition. Therefore, for the following estimations of transitions, the dependency specified in Table 2 was used. Table 2 MC order test using cumulative logit model of becoming low risk at T3 Selected Predictors (Significant, P < 0.01) MC order = 1 MC order = 2 Male -0.563 -0.345 Age -0.023 -0.019 Baseline risk (t1) Low 2.73 1.545 Baseline cost (t1) -0.01 -0.011 Risk at t2 Low - 1.938 Model fit log L -283085 -139246 * Log likelihood ratio test of order 2 model, compared to order 1 model (287,678 >> χ 2 (2), pr<0.001) was significantly preferred. Characteristics of the obtained trend When Markov chain order 2 was assumed, higher T1 risk status, controlling for T2 risk state as in Table 3 , was associated with lower percentage of being at low or lower risk state at T3 (T3 state = "0" if at low or lower state than the state at T2. T3 state = "1" if at high or higher state than the state at T2, Pr<0.001). Similarly, higher T2 risk state, controlling for T1 risk, was less likely to be at "0" at T3 (Pr<0.001). Due to a stronger dependency on T2 risk state, medium (T1)-low (T2) is more likely to be at low at T3 than low (T1)-medium (T2). However, additional dependency on the T1 risk state differentiates the trend percentages at "0" state (T3) from medium (T1)-low (T2) and high (T1)-low (T2). Table 3 T1-T2 risk state by T3 risk state. Note that outcome is 0 if T3 state is low or lower than the state at T2, otherwise it is 1. The test for trend controlling for T1 risk state is: a>b>c; d>e>f; g>h>i (Pr>z<0.001). The test for trend controlling for T2 risk state is: a>d>g; b>e>h(Pr>z<0.001)c>f>i (Pr = 0.143) The superscripted numbers in parenthesis represent the order of trend which appeared to be significantly associated with the likelihood to be at "0" at T3. Trend percentage at T3 per T1-T2 risk state T1-T2 risk state (N = 180,767) (0) (1) Comparison of order Trend statistic (Z) Low-Low a 82.4 (1) 17.6 Vs. d -68.2* Low-Medium b 52.5 (3) 47.5 Vs. g -11.4* Low-High c 35.8 (5) 64.2 Vs. e -44.7* Medium-Low d 58.6 (2) 41.4 Vs. b -11.2* Medium-Medium e 34.0 (6) 66.0 Vs. h -18.0* Medium-High f 21.5 (8) 78.5 Vs. i -27.7* High-Low g 43.5 (4) 56.5 Vs. c -26.2* High-Medium h 22.6 (7) 77.4 Vs. f -43.8* High-High i 11.7 (9) 88.3 - - * Significant with α = 0.01 Cochran-Armitage trend tests controlling for T1 risk state or T2 risk state show a strong declining trend of staying at "0" than that at T2 as in the order of appearance in Table 3 . For example, low (T1)-low (T2) is more likely to be at low in T3, compared to low-medium and low-high. This tendency holds for: a>b>c; d>e>f; g>h>i with Pr<0.001, a>d>g; b>e>h with Pr<0.001 except c>f>i (Pr<z = 0.143), where a, b, c; d, e, f; g, h, and i corresponds to low-low, low-medium, low-high; medium-low, medium-medium, medium-high; high-low, high-medium, and high-high risk state, respectively. Also, overall consistency appears significant (one-sided, Pr <z: <0.001). Pairwise comparisons determined the order of likelihood to be at low or lower state at T3 (i.e. outcome = 0) in association with T1-T2 risk state. Low (T1)-low (T2) tops the order, with 82.4 % of the individuals being at low risk at T3, followed by medium (T1)-low (T2) with 58.6%, low (T1)-medium (T2) with 52.5% and so on (a>d>b>g>c>e>h>f>i, Pr<0.001). Application of Markov chain order 2 for estimation of natural and reinforced transition flows We applied the results from the previous sections (regarding the KNHIC population) to the observed risk transitions of the matched sample from a US active employee population, who participated in a minimal level program. Similar assumptions were tested and stationary Markov chain order 2 was also applied. The estimated transition probabilities without an intervention (KNHIC cohort) and those of the same people with a minimal level intervention are shown in Table 4 . The higher probabilities to be at low risk state at T3 were shown in the intervened flow (except medium (T1)-high (T2) and high-low) compared to the natural flow. Table 4 Estimated risk transition probability with and without intervention and medical utilization. Behavioral health risk measured at two times (T1 and T2) T1-T2 Utilization Cost Ratio b KNHIC population without intervention KN HIC population with a simulated intervention at minimal level a Risk state at T3 Risk state at T3 T1 T2 Low Medium High Low Medium High Low Low 1.00 0.81 0.15 0.04 0.97 0.03 0.00 c Medium 1.14 0.53 0.34 0.13 0.69 0.11 0.20 High 1.58 0.38 0.32 0.30 0.65 0.34 0.01 Medium Low 1.16 0.60 0.30 0.10 0.65 0.33 0.02 Medium 1.23 0.35 0.44 0.21 0.39 0.32 0.29 High 1.52 0.23 0.35 0.42 0.05 0.73 0.22 High Low 1.51 0.46 0.31 0.23 0.34 0.65 0.01 Medium 1.52 0.25 0.38 0.37 0.34 0.65 0.01 High 1.79 0.14 0.27 0.59 0.16 0.30 0.54 a Risk transition probability was estimated using Markov model order 2, adjusted for age and gender distribution of KNHIC. b Comparing average cost in each T1-T2 risk state pair (e.g. high-medium) to that of the low-low (T1-T2). c All probabilities were rounded off at the 3 rd decimal place (i.e.0.0014). Overall, the likelihood of maintaining health at low risk (T1-T2-T3) is higher in the intervened transition (0.97 vs. 0.81). Also, the probabilities of being at high (T1-T2-T3) and at medium (T1-T2-T3) are lower in the intervened flow (0.54 vs. 0.59 at high, 0.32 vs. 0.44 at medium). Healthcare cost ratios in relation to the T1 -T2 risk states in the KNHIC population are also shown in the table. In general, cost ratios increased in the order of T1-T2 risk states and cost utilization of high-high group almost doubled (1.79) the cost of low-low. After three years of projection, the transitions, presented in Table 5 show the estimated numbers in each risk state, following the natural and the intervened flows as in Table 4 . Three-year forward projection of populations (non-intervened vs. intervened) was calculated by pre-multiplying the number of people in each T1-T2 risk group to the 3 rd power of the Markov transition matrix of order 2. This is then collapsed by the baseline states as in Table 5 to show the net gain from T1 to T3. The difference was calculated as the projected counts per state from the baseline total counts, and percentage point changes were calculated (Table 6 ). Overall, there was 0.72% point net increase in high risk state following from natural flow and 5.01% points net decrease in the intervened flow. Low risk percentage decreased by 3.07% points following the natural flow but increased by 7.49% points in the intervened flow. Table 5 Projection of population KNHIC following natural vs. intervened flow over 3 waves. Projected 3 – forward years based on MC-order2 Risk state At Baseline At T3 α At T3 β Low 113,605 108,050 127,149 Medium 40,151 44,400 35,675 High 27,011 28,317 17,943 Table 6 Projection of population KNHIC following natural vs. intervened flow over 3 waves. Percentage Point change following Table 4-(a) Risk state At Baseline At T3 α At T3 β Low 113,605 -3.07% +7.49% Medium 40,151 +2.35% -2.47% High 27,011 +0.72% -5.01% α, Following natural risk flow, β Following intervened flow Discussion Markov Chain Model and Transitions of Risk States After controlling for age, gender and other covariates, such as past healthcare costs, variations within groups in predicting the future risk transition were left unexplained. This suggested an additional consideration of dependence on other factors such as past health risk history. This dependent structure of a natural risk flow was best fitted with a Markov model order 2 (due to limitation of the data, no higher order was tested). In other words, current risk states largely depend on the immediate past state and also depend on the one before that. The estimated probability was stable. There was no policy change or environmental effect, which is considered as a period effect, during the study period. In addition, as a result of testing a higher order MC for the available US data, order 2 was preferred to the order 3 (data not shown). Therefore, we concluded that an assumption of MC order 2 was plausible for the health risk transitions regardless of natural flow or reinforced flow at minimal intensity [ 28 , 29 ]. Also, although the time-frames of observations in the two population were different, time of observation turned out to be non-influential to the transition (stationary) and average years of observation was used to calculate an annualized transition for both transitions. Therefore, the difference in observed years would not have implications for the results. Difference in Individual Risk Profile Simulated data for a minimal program participation, adjusted to matched age, gender, and baseline risk distribution of KNHIC population for comparison, showed that sampling disparities were significantly reduced in overall distribution and demographics but still remained in individual risk factors. Exercise and smoking risks are relevant to environmental and cultural adaptation, which showed large differences in prevalence. Typically, overweight is the most prevalent risk factor in the US, and it appeared to be significantly higher even after matching overall risk distribution to the Korean population. Although there exist some differences, these biases in the data were assumed to be smoothed out by collapsing to the aggregated risk levels. There may be inherent differences in risk transitions in the two populations due to two factors: (1) disparities in the individual risk factors, (2) systematic differences due to cultural and environmental factors. Due to the fact that estimation of risk transition was primarily dependent on the aggregated risk state rather than the individual risks (Table 2 ), the potential influences were minimal. Matching on individual risk factors additionally may reduce such a bias in the transition. However, individual risk profiles are inconsistent at times and risk transitions depend more on inter-associations of individual risk profiles. Therefore, matching on individual risk factor and transitions based on such profiles may even create another type of bias in the estimation of transition. On the other hand, by bootstrapping of random samples, matching on age, gender, and risk distribution, variations within each risk state have been reduced and even reduced the potential bias due to disparity in individual risk factors. The remaining disparity in the matched populations would become irrelevant when additional control for those remaining disparity factors is made in the model for the ordered categorical transition. Implementing a program based on found dependent structure of risk transition Without reinforcement, the male population was less likely to be in the lower risk states in future compared to the females from the same past (T1 and T2) risk state, given age (Table 2 ). This suggested an observed potential for improvement of men's health such as an emphasis on the cultural adaptation of health by changing organizations and communities to create supportive environment and re-orienting health services [ 30 ]. Aging has been found to be related to the risk transition because as one gets older, he/she is less likely to be in lower risk state in future without intervention (Table 2 ). There have been concerns about the elderly population, who are under-represented due to unequal resources, and less interest [ 31 , 32 ]. Studies in the US found that current elderly population is healthier than the elderly 10 years ago [ 31 - 33 ]. The observed trend among the elderly may be linked by the changes in health-related individual behaviors in the past. This indicates the potential impact of continuous health promotion on aging adults. Although the immediate past risk states showed the highest importance, when they are the same, risk state at T1 plays a key role in predicting T3 risk state. In other words, individuals currently at a state may have different paths to the current risk state and the weight (by intensity or allocation of resources) of risk reduction program should differentiate this path-variability to maximize the impact (Tables 3 , 4 ). Following the natural flow, the likelihood of those at low at T2 to be at the same low in future (T3) is significantly less than that of those who were on intervention, regardless of their risk state at T1. This suggests a continuing effort on those who once impacted, to maintain their modified health practices until they adapt to their newly improved lifestyle (Table 4 ). Over the period, although there was presumably an aging process in the natural flow, there exists a regression to the previous state. Especially with Markov chain order 2, in controlling for the T1 state, the tendency stays. This implies that even without any external reinforcement for a healthier state, low risk state largely tends to maintain its state, followed by the high risk state. Therefore, an intervention with minimal effort to sustain healthy lifestyle and behavior for those who are at low risk may yield substantial benefit in the long run (Tables 3 , 4 ). Likewise, an intervention program to break high-high cycle to lower the health risks and eventually to optimize the utilization of the health resources is anticipated to be effective and necessary. These, in association with the cost trend following the risk trend (Table 4 ), imply that moderation of healthcare costs is also achievable by a well-targeted health promotion program. Projected effect of a minimal level intervention For this study population, the minimal level intervention appeared to be effective in the low and medium baseline groups (Table 4 ). This is reasoned that the particular program, which was designed to impact people with minimal resources, turned out to be most effective on "low risk maintenance". Thus, efforts to reduce risks were rather undersized. However, people at the high baseline risk except high (T1)-high (T2), were less likely to stay at high risk (at T3) than the non-intervened, showing a potential improvement in risk reduction as well. This was found in similar contexts and quite consistent across studies [ 34 , 35 ]. As a conclusion, these findings suggest that with limited resources, a program that delivers an individual feedback upon completion of an HRA, identifying most significant risks and referring to available resources, could have an impact on the natural health risk flow, especially on low risk maintenance by enhancing awareness and self-efficacy in maintaining good health practice. One of the study goals was to achieve estimate a natural flow of the study population, where a demand and necessity for health promotion program has been increased recently. Natural flow was coined by Edington [ 14 ] to describe a health trend in a population without any intervention for a reinforced effort to improve heath state. By comparing it to various simulated trends with an intervention to the population, one could assess program effectiveness in terms of health change, which was measured as the risk state. This again, with cost-ratios per risk transition as in Table 4 , may be used to calculate the possible savings in association with any program of interest. For example, comparing to low-low group, low-high incurred 158% of the costs of low-low. Trend at low at T3 following the intervention was increased by 16% from low-low, comparing to natural flow. Therefore, the potentially incurred costs due to increased risk state could be saved proportionately by 14% (1.14:1.00, 12% of the low (T1)-low (T2) avoided transition to medium (T3)), 58% (1.58:1.00, 4% avoided transition to high (T3)) of the costs. This concept was used wherever health behavior change and lifestyle modification is applicable in an effort to promote and manage health. Thus, health trends following such a program could be compared to the expected natural flow for a demonstration of "effectiveness". In the current study, the estimated natural flow shows decreases in low risk state by 3.07 % whilst 7.49 % increases by following an intervention. The underlying risk progression may be likely explained elsewhere [ 37 ]. Recommendation It is recommended to set a policy and allocate resources, tailored to the population profile of aggregated risk state, first. Targeting risk factors in the context of risk state or risk clusters could follow next. For example, an overweight population at high risk state requires more resources and more intensive interventions than the overweight population at low risk state. The study provides the base information when planning such a population-based intervention. This approach differentiates from those interventions targeting individual risk profiles, which could be inconsistent at times. Studies suggest that an inter-association over risks is more important in implementing an intervention than individual risks [ 38 ]. Some variables are highly variable such as medical conditions, depending on genetic traits more than the environments. This variation may indicate that the level of a risk factor may not be acceptable in one population while it was so in another. This varying manifestation could have an influence on implementing interventions targeting on particular risk factor, which is beyond the scope of the paper. 5. Conclusions As a reasonable method to project the risk trends, a stationary Markov model of order 2 was fitted for the health risk transitions of the non-intervened population, suggesting a natural risk flow for the population. Utilizing this with the matched intervened population, the reinforced risk transitions of the KNHIC population were estimated. The significant difference in the transitions appeared to be for the low baseline risk population even with a minimal intensity intervention program. Therefore, the difference in the projected numbers by the two transitions (with /without interventions) showed a significant impact on low risk maintenance although higher risk population was also impacted to increase fewer risks and to moderate it to becoming "medium risk". Conventionally, most programs in the past have focused on risk reduction although low risk maintenance has been raised as a practical issue. Since this paper suggests a strong dependence on the previous history of risk status and the instability of high (T1)-low (T2) risk population, compared to low-to-low or medium-to-low population, differentiating efforts for the low risk people (from low to low) and for those moved to low (from high to low), should be sought for maximum impact. The studied intervention with minimal intensity was to diversify the beneficiaries of the program, to increase awareness, to continue educating and motivating individuals to adopt healthier behaviors by individualized feedback, and to provide resources [ 36 ]. The findings demonstrated that even such a minimally intense program could be effective in moderating health risks, preventing relapse and sustaining healthy behaviors over time. Limitation and future research Population characteristics on risk transition were assumed to be similar by adjusting for age, gender, if previous risk states were the same. However, in addition to the measured risks, diet and culturally inherited behavioral differences could make an inherent transitional difference in the two compared populations. Also, some risks such as disease were underestimated in the KNHIC population. Therefore, the presented natural flow estimation in Korean population can be adopted but utilized with caution. It may be conjectured that there are some risk factors are more easily modifiable than others such as exercise while medical condition and overweight may not. The currently shown disparity in the two compared risk profiles may have induced somewhat less significant program impact on risk transition (lowering risk state) due to higher prevalence in medical conditions and overweight in the control population. This study tried to match two populations as close as possible and to model to reduce potential biases but comparing two populations requires more caution in further relevant studies. Despite the fact that individual risk distribution is not consistent often time, not only an overall health status but a multivariate risk state (i.e. a matrix per person) could be utilized to identify an effect of each risk factor to their projected state in future, as in Manton and Stallard [ 37 ]. Based on the estimated transitions, prediction of the effect of different intervention models on the risk transitions and the impact on healthcare costs would be available. Validity of the p-HRA (KNHIC) was not tested although it was matched to and tested by the established UM-HRA. This will be validated and tested against UM-HRA upon availability of cohort data, which were followed longer time. Markov higher order model (3 +) with a longer follow-up time could be explored for exact consistency and stationarity for the cohort. Even the progression rate from no incidence state to symptomatic state utilizing the health care cost database can be added to the cohort to project the proper care of medical conditions. Inclusion of an "exit" state (such as death) and consideration of a reasonable compliance rate would possibly make the model robust over the longer time of an intervention. We will further investigate the possibility of matching individuals in the two compared populations across many covariates including individual risk factors and other confounders such as willingness to improve, to further reduce potential biases. Competing interests The authors declare that they have no competing interests. Authors' Contributions JP carried out the analyses and drafted the manuscript. SHJ collected, cleaned the data and provided initial analyses. DWE participated in the design of the analyses and provided revisions. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC543445.xml
516444	Syndromes with congenital brittle bones	Background There is no clear definition of osteogenesis imperfecta (OI). The most widely used classification of OI divides the disease in four types, although it has been suggested that there may be at least 12 forms of OI. These forms have been named with numbers, eponyms or descriptive names. Some of these syndromes can actually be considered congenital forms of brittle bones resembling OI (SROI). Discussion A review of different syndromes with congenital brittle bones published in the literature is presented. Syndromes are classified in "OI" (those secondary to mutations in the type I pro-collagen genes), and "syndromes resembling OI" (those secondary to mutations other that the type I pro-collagen genes, identified or not). A definition for OI is proposed as a syndrome of congenital brittle bones secondary to mutations in the genes codifying for pro-collagen genes (COL1A1 and COL1A2). Summary A debate about the definition of OI and a possible clinical and prognostic classification are warranted.	Background Besides brittle bones, all other clinical characteristics of osteogenesis imperfecta (OI) are variable, and even different members of the same family may present with a dissimilar degree of severity [ 1 ]. Severity appears to follow a continuum in the OI population, and it is therefore very difficult to establish a clinical prognostic classification in definite categories. This disease has received different names through history, including osteopsathyrosis, Vrolik's disease, fragilitas ossium, mollities ossium, Lobstein's disease, and Van der Hoeve syndrome [ 2 ]. In a first attempt to classify OI, in 1906 Looser divided the disease in two forms [ 3 ]: "congenita" (Vrolik) and "tarda" (Lobstein), depending on the severity of the presentation. In OI congenita, multiple fractures may occur in-utero , whereas in OI tarda, fractures happen at the time of birth or later. OI tarda has also been sub-divided in "gravis" and "levis" [ 3 ]. This classification is no longer tenable because it failed to encompass the obvious clinical variability apparent in this disorder. There has also been an attempt to classify OI according to radiological characteristics [ 4 ]. Some of the features suggested by the authors of this classification are not present until age 5 or 10 years, and children of early ages can not be classified using this scheme. There has also been suggested that OI could be classified OI according to severity [ 5 , 6 ]. The most commonly used classification, initially proposed by Sillence in 1979, divides the patients in four types (I-IV) [ 7 , 8 ]. In fact, it is now widely recognized that there may be more types of OI [ 9 - 11 ]. Some of the forms of congenital brittle bones have been considered OI and have been added as numbers V, VI and VII [ 8 , 12 - 14 ], while others have been designated with eponyms (Cole-Carpenter syndrome [ 15 ], Bruck syndrome [ 16 ]), and others designated with clinical features (OI with denser areas in bones [ 17 ], OI with optic atrophy, retinopathy, and severe psychomotor retardation [ 18 ], OI with microcephaly, and cataracts [ 19 ], osteoporosis-pseudoglioma syndrome). The numeric classification (I-VII) is somewhat confusing, as the characteristics of each type overlap. Furthermore, there is no consensus about basic characteristics of the different types. For example, it is not clear if type I OI includes individuals with short stature, or if bone deformity rules out the diagnosis of Type I OI. Also, it is unclear if an individual with normal height can have type IV OI. Type II OI is defined as lethal, but there are cases of children with clinical characteristics of type II OI who have survived several years. Furthermore, the current classification does not allow for prognostic predictions, as individuals with type I OI may have numerous fractures and chronic pain in the course of their lives. This is part of the lack of a precise definition of OI. Thus, despite the fact that OI has been known for more than 200 years, there is no consensus about the definition for the disease. For some, it includes only those forms of congenital brittle bones secondary to mutations in the genes codifying for type I pro-collagen (COL1A1 and COL1A2). For others, it is a group of conditions with the common feature of congenital brittle bones. The example of the osteogenesis imperfecta-pseudoglioma syndrome illustrates this concept very clearly. This syndrome has been re-named as "osteoporosis-pseudoglioma syndrome" once the causal mutation has been identified to be in the LRP5 gene, elsewhere than the type I pro-collagen genes [ 20 , 21 ]. Following this example, I propose to define osteogenesis imperfecta as syndromes resulting from mutations in either COL1A1 or COL1A2 genes, and to group all other syndromes with congenital brittle bones as "syndromes resembling OI" (SROI), pending the identification of their causal mutation. Here, a review of the different forms of syndromes with congenital brittle bones is presented (Table 1 ). A working-group of international experts (like the Villefrance criteria in Ehlers-Danlos Syndrome) to clarify the issue of definition of congenital brittle bones syndromes and their classification is warranted. Table 1 syndromes with congenital brittle bones Osteogenesis Imperfecta Mild OI with normal stature Moderate OI with short stature Severe OI Lethal OI Congenital brittle bones with dense areas in bones Syndromes resembling OI (SROI) Congenital brittle bones with craniosynostosis and ocular proptosis Congenital brittle bones with congenital joint contractures Osteoporosis-pseudoglioma syndrome Congenital brittle bones with optic atrophy, retinopathy and severe psychomotor retardation Congenital brittle bones with microcephaly and cataracts Congenital brittle bones with redundant callus Congenital brittle bones with mineralization defect Congenital brittle bones with rhizomelia Discussion Here, syndromes with congenital brittle bones are divided in osteogenesis imperfecta (those secondary to mutations in the type I pro-collagen genes) and syndromes resembling OI (those secondary to mutations other than the type I pro-collagen genes, identified or not) Osteogenesis imperfecta (types are ordered based on severity from mild to severe) Type I OI The most common mutation causing OI type I causes a reduction in the production of otherwise normal type I collagen secondary to the effect of a null allele mutation. Patients often have normal stature, and a slightly low stature does not preclude the diagnosis of type I OI. "Type I OI" is not a synonymous of "mild OI". Individuals may have few or no fractures, mostly during the first years of life or even at birth [ 22 ], or numerous fractures throughout their lives. They may have triangular face. They are fully ambulatory, and do not have bowing of the long bones, although vertebral fractures may be present. Most have blue sclera, but it can be white, or blue color may fade as the individual grows older. This condition is transmitted as an autosomal dominant trait. Despite the absence of fractures, bone density can be very low, with no relation with clinical severity, underscoring the relative lack of significance of bone density measurements assessing severity in patients with OI. In many instances bone density is normal during the first months of life, and individuals progressively fail to increase bone mineral density with age. In some cases the diagnosis is an incidental finding after a fracture [ 23 ]. Dentinogenesis imperfecta can be present even in very mild cases. Early hypoacusia [ 24 , 25 ] and cardio-vascular problems, particularly aortic valvular disease [ 22 ] can be present in these subjects. Type IV OI (Moderate OI with short stature) These individuals typically have short stature, bowing of long bones, and vertebral fractures. Scoliosis and joint laxity may be present. Patients with this form of OI are generally able to ambulate, but they may require aids for walking. Based in the presence of DI, moderate OI has been sub-divided in two forms, "a" and "b" [ 26 ]. These patients have white sclera. Precise diagnosis of this type of OI is often difficult, as the clinical characteristics are not clear in the literature, and different centers base the diagnosis on different criteria. Type III OI (Severe OI) These patients have triangular face, product of an enlarged head and under-development of the facial bones. They also have chest deformities, markedly short stature, and severe deformities of the long bones, vertebral fractures, and scoliosis. They are frequently wheelchair-bound, although some are able to walk with aids. Prenatal diagnosis is hard but sometimes possible using ultrasonography [ 27 ]. Long bones are severely bowed, and altered structure of the growth plates lead to a particular structural alteration of the metaphyses and epiphyses described a "popcorn appearance". Severe cases may have respiratory complications that can compromise survival. Type II OI In this form of OI, most newborns do not survive the perinatal period. Causes of death are malformations or hemorrhages of the central nervous system [ 28 ], extreme fragility of the ribs, or pulmonary hypoplasia [ 29 ]. The infants present with multiple intra-uterine fractures, including skull, long bones and vertebrae, beaded ribs, and severe deformity of the long bones [ 30 ]. Prenatal differential diagnosis between severe and lethal OI is usually not possible. Extremely severe cases can be dismembered during delivery [ 31 ]. The vast majority of cases are autosomal dominant new mutations [ 32 - 34 ]. It has been suggested that there may be different clinical forms of lethal OI [ 35 ]. Despite severity, a few patients have survived for several years. Congenital brittle bones with dense areas in bones Described in one infant [ 17 ] who died shortly after birth and presented with an OI phenotype that differed from the usual lethal form. The skeleton had regions of increased bone density, and this girl had dysmorphic facial features, including loss of mandibular angle, low set ears, soft skull, and large anterior and posterior fontanelles. Bilateral upper and lower limb contractures were present with multiple fractures in the long bones and ribs. The metaphysis of the long bones were dense in x-rays. The patient died after a few hours and histopathological studies identified extramedullary hematopoiesis in the liver, little lamellar bone formation, decreased number of osteoclasts, abnormally thickened bony trabeculae with retained cartilage in long bones, and diminished marrow spaces similar to those seen in dense bone diseases such as osteopetrosis and pycnodysostosis. Genetic testing showed that the child was heterozygous for a COL1A14321G → T transversion in exon 52 that changed a conserved aspartic acid to tyrosine (D1441Y). Abnormal proa1(I) chains were slow to assemble into dimers and trimers, and abnormal molecules were retained intracellularly for an extended period [ 17 ]. Because of this mutation, this form of congenital brittle bones should be included in the OI group. Syndromes resembling OI (SROI) Congenital brittle bones with craniosynostosis and ocular proptosis (Cole-Carpenter syndrome) Two boys [ 15 ] and a girl [ 36 ] with this particular form of SROI have been described in the literature. Both boys were normal at birth, but after several months, they developed multiple metaphyseal fractures, associated with low bone density in the entire skeleton and craniosynostosis, hydrocephalus, ocular proptosis, and facial dysmorphism. One of the patients had also hypercalciuria. Neurological development is normal in this form of SROI. Both patients where wheelchair-bound at adult age, with very short stature, severe bone involvement and normal intellectual and neurological development [ 10 ]. No mutation has been identified as causing this syndrome. Congenital brittle bones with congenital joint contractures (Bruck syndrome) First described by Bruck et al in 1897 in an adult patient [ 16 ], in this form of SROI patients are born with brittle bones, leading to multiple fractures and joint contractures and pterygia (arthrogryposis multiplex congenita) due to dislocation of the radial head [ 37 , 38 ]. Wormian bones are present, and inheritance appears to be recessive [ 39 , 40 ]. In three patients that underwent pro-collagen mutation testing, it was not possible to demonstrate any mutations in the COL1A1 and COL1A2 genes [ 38 ]. The basic defect was mapped to locus 17p12 (18 cM interval), where a bone telopeptidyl hydroxylase is located [ 41 ]. The mutation leads to underhydroxilated lysine residues within the telopeptides of collagen type I, and therefore to aberrant crosslinking in bone, but not in cartilage or ligaments. Osteoporosis pseudoglioma syndrome [ 20 , 21 ] This form of SROI was first described in three families in 1972 [ 42 ]. Other report described a South African family of Indian origin with the condition [ 43 ]. Inheritance is autosomal recessive. Individuals with Osteoporosis-pseudoglioma syndrome have mild to moderate OI with blindness due to hyperplasia of the vitreous, corneal opacity and secondary glaucoma. The ocular pathology may be secondary to failed regression of the primary vitreal vasculature during fetal growth [ 44 ]. The genetic defect has been mapped to chromosome region 11q12-13 [ 45 ]. The defect is specifically in the LRP5 gene that encodes for the low-density lipoprotein receptor-related protein 5 [ 44 ]. Treatment with pamidronate has shown promising results in this group of patients [ 46 ]. Other forms of SROI with ocular involvement Two other forms of SROI with ocular involvement have been described: one variant with optic atrophy, retinopathy, and severe psychomotor retardation [ 18 ], another with microcephaly and cataracts [ 19 ]. Congenital brittle bones with redundant callus formation Originally published as "type V OI" [ 12 ]. Patients with this syndrome develop hyperplastic calluses in long bones after a fracture or intramedullary rodding surgery [ 22 ]. These patients present with hard, painful and warm swellings over long bones which initially may suggest inflammation or osteosarcoma. After a rapid growth period, the size and shape of the callus may remain stable for many years [ 47 ]. Microscopically there is increased production of abnormal extracellular matrix, that is poorly organized and incompletely mineralized in the callus area [ 48 ]. Histological studies in bone outside the callus area showed that the bone lamellae are arranged in a mesh-like fashion, as opposed to a parallel arrangement in patients with OI [ 12 ]. A series of case reports of hyperplastic callus formation in patients with clinical characteristics compatible with OI can be found in the literature [ 3 , 47 , 49 - 54 ]. These large calluses may also be present in flat bones [ 55 ]. These patients may also have calcification of the interosseous membrane between the radius and ulna, determining a clinical sign, as affected individuals are unable to pronate and supinate the forearm. The radial head may be dislocated. Patients with this syndrome have white sclera and no DI. Mutations in the pro-collagen genes could not be identified so far. Inheritance appears to be autosomal dominant, with variable penetrance. Congenital brittle bones with mineralization defect Undistinguishable from moderate to severe OI on a clinical basis, this rare form of SROI [ 13 ], has been classified as "Type VI OI". It can only be diagnosed by bone biopsy, where a mineralization defect affecting the bone matrix and sparing growth cartilage is evident. These patients have no DI and no wormian bones. Despite the histological mineralization defect, there are no radiological signs of growth plate involvement. The pattern of inheritance is not clear, but the case of two siblings from healthy consanguineous parents has been described, suggesting gonadal mosaicism or a somatic recessive trait [ 13 ]. No mutations of COL1A1 and COL1A2 genes have been found in these patients, and collagen structure appears to be normal. This form of SROI shares several characteristics with fibrogenesis imperfecta ossium [ 56 , 57 ]. Congenital brittle bones with rhizomelia A particular form of SROI with short humerus and recessive inheritance was described in a First Nations community of Quebec, and published as "type VII OI" [ 14 ]. The individuals affected have short humeri and femora. The phenotype is clinically moderate to severe. Fractures may be present at birth, and the condition progresses to early lower limb deformities, coxa vara and osteopenia. The bone morphology in congenital brittle bones with rhizomelia is not different than that of mild OI by histomorphometry. The genetic defect has been mapped to the short arm of chromosome 3 by linkage studies [ 58 ], where there are no genes that codify for type I pro collagen. Summary A definition for osteogenesis imperfecta is proposed based on the presence of type I pro-collagen mutations. Inclusion of related syndromes with unidentified mutations should be considered as SROI until further defined. Different forms of syndromes with congenital brittle bones are reviewed. A debate about the definition of OI and a possible clinical and prognostic classification are warranted. List of abbreviations OI Osteogenesis Imperfecta SROI syndromes resembling osteogenesis imperfecta DI Dentinogenesis imperfecta Competing interests None declared. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC516444.xml
554105	CNS activity of Pokeweed Anti-viral Protein (PAP) in mice infected with Lymphocytic Choriomeningitis Virus (LCMV)	Background Others and we have previously described the potent in vivo and in vitro activity of the broad-spectrum antiviral agent PAP (Pokeweed antiviral protein) against a wide range of viruses. The purpose of the present study was to further elucidate the anti-viral spectrum of PAP by examining its effects on the survival of mice challenged with lymphocytic choriomeningitis virus (LCMV). Methods We examined the therapeutic effect of PAP in CBA mice inoculated with intracerebral injections of the WE54 strain of LCMV at a 1000 PFU dose level that is lethal to 100% of mice within 7–9 days. Mice were treated either with vehicle or PAP administered intraperitoneally 24 hours prior to, 1 hour prior to and 24 hours, 48 hours 72 hours and 96 hours after virus inoculation. Results PAP exhibits significant in vivo anti- LCMV activity in mice challenged intracerebrally with an otherwise invariably fatal dose of LCMV. At non-toxic dose levels, PAP significantly prolonged survival in the absence of the majority of disease-associated symptoms. The median survival time of PAP-treated mice was >21 days as opposed to 7 days median survival for the control (p = 0.0069). Conclusion Our results presented herein provide unprecedented experimental evidence that PAP exhibits antiviral activity in the CNS of LCMV-infected mice.	Background The broad-spectrum anti-viral agent PAP (Pokeweed antiviral protein) [ 1 ] is a well-characterized 29-kDa plant-derived ribosome-inactivating protein (RIP) isolated from Phytolacca americana [ 2 ]. The anti-viral activity of PAP has been described against a wide range of viruses, including HIV-1, herpes simplex virus, cytomegalovirus, influenza virus and polio virus [ 2 ] (For the most recent review, please see: [ 3 ]). The activity of PAP is attributed to its ability to inhibit protein synthesis by catalytically cleaving a specific adenine base from the highly conserved alpha-sarcin/ricin loop (SRL) of the large ribosomal RNA [ 4 , 5 ] as well as from viral RNA[ 3 ]. The potent anti-HIV activity of PAP at nanomolar ranges taken together with the relative ease of large scale purification has led to the clinical use of PAP [ 2 ]. Since 1985, our group has studied the multifunctional efficacy of this potent agent [ 1 , 2 , 4 - 42 ]. The purpose of the present study was to further elucidate the anti-viral spectrum of PAP by examining its effects on survival of mice challenged with intracerebral inoculations of lymphocytic choriomeningitis virus (LCMV). LCMV is a rodent-borne arena virus that can result in persistent neuronal infection on mice [ 43 , 44 ]. Alpha-Dystroglycan (alpha-DG) was recently identified as a receptor for LCMV as well as for several other arena viruses including Lassa fever virus [ 45 ]. The binding affinity of LCMV to alpha-DG determines viral tropism and the outcome of infection in mice [ 46 ]. LCMV has also been associated with both postnatal and intrauterine human disease. Infection in man is acquired after inhalation, ingestion or direct contact with virus found in the urine, feces and saliva of infected mice, hamsters and guinea pigs. Congenital LCMV infection is a significant, often unrecognized cause of chorioretinitis, hydrocephalus, microcephaly or macrocelphaly as well as mental retardation. Acquired LCMV infection, asymptomatic in approximately one third of individuals, is productive of central nervous system manifestations in one half of the remaining cases. Aseptic meningitis or meningoenceophalitis are the predominant syndromes, although transverse myelitis, a Guillain-Barre-type syndrome, as well as transient and permanent acquired hydrocephalus have also been reported [ 47 ]. In the present study, we describe the significant efficacy of the highly stable, potent and broad-spectrum anti-viral agent PAP in a murine model of LCMV. Our results presented herein provide unprecedented experimental evidence that PAP exhibits antiviral activity in the CNS of LCMV-infected mice. Methods Purification of PAP PAP was purified from spring leaves of the pokeweed plant, P. americanca , in four steps [ 22 , 38 ]. Briefly, spring leaves were homogenized in a neutral pH buffer and centrifuged to sediment the remaining cellular fragments. The supernatant was fractionated between 60–90% saturation of ammonium sulfate and the precipitate dialyzed against a low-ionic strength pH 7.5 buffer. The solution was passed through a DEAE cellulose column and the PAP-containing flow-through fraction was then applied to a cation exchange resin S-Sepharose column. The adsorbed PAP was eluted in a linear KCl gradient. The protein peak which eluted at 0.12 KCl was taken as PAP. This fraction was dialyzed extensively against water and lyophilized for storage at -20°C. The procedure resulted in homogeneous PAP, with a purity of >99% as measured by both SDS -12% PAGE and analytical cation- exchange high-performance liquid chromatography. Purified PAP induced concentration-dependent inhibition of HIV-1 replication in normal human peripheral blood mononuclear cells infected with the HIV-1 strain HTLV IIIB with an IC 50 p24 of 14 ± 2 nM. Animal infection Animal infection was performed in an appropriate Animal BioSafety Level-3 Laboratory (ABL-3) at BRIEM (Research Institute for Epidemiology and Microbiology, MINSK, Belarus) with the technician wearing appropriate facility clothing. The culture was thawed in a water bath at 37°C and then diluted in normal saline to achieve the required concentration. In this study, all mice were challenged with 1000 PFU which is 100-times higher than the LD 50 dose. Each group of animals was placed in a separate cage. LCMV model Three week old CBA mice were intracerebrally infected with 1000 PFU of WE54 strain of LCMV that resulted in lethality of 100% of control (non-treated) animals in 6–8 days after infection. Control animals were given physiologic salt solution as a placebo instead of the compound. In general, for non-treated animals, clinical signs of the disease manifested on the 5 th and 7 th days by presenting: weight loss, immobility, disheveled hair, convulsions, severe decubitus paralysis and death. All subjective measurement of decreased mobility and scruffy fur were done in a blinded fashion as not to influence the results. The protective properties of the experimental anti-viral drugs were assessed by using the following treatment-preventative regimen: Mice were treated either with vehicle (n = 20) or PAP (n = 10) (0.25 mg/kg) administered intraperitoneally 24 hours prior to, 1 hour prior to, and 24 hours, 48 hours, 72 hours, and 96 hours after virus inoculation. Mice were then observed for 21-days post infection. The protective effect of the experimental anti-viral drugs was evaluated according to the rise of the survival rate and prolongation of mean life of the experimental animals as compared with the control animals. Statistical analysis Statistical significance was determined using the Kaplan Meier Log-Rank test. Results In order to evaluate the anti-LCMV activity of PAP, mice were inoculated with intracerebral injections of LCMV at a dose level that is lethal to 100% of mice within 6–8 days. Mice were treated either with vehicle or PAP as described in the methods section. Of the 20 control mice, 3 died on day 1 immediately after intracerebral injection due to accidental brain injury and are excluded from the data analysis. All 17 of the remaining control mice developed clinical signs of LCMV infection between day 4 and 6, including weight loss, disheveled hair, decreased mobility and paralysis (Table 1 ). All control mice developed seizures between day 5 and day7 and died between day 6 and day 8 with a median survival of 7 days (Table 1 , Figure 1 ). Table 1 Anti-LCMV activity of PAP in CBA mice Disease Unset Days after inoculation with LCM Virus Survival (days) Weight loss (grams) Decreased mobility Convulsions Hair disheveled Vehicle Mouse #1* <<1 NA 0 NA NA Mouse #2* <<1 NA 0 NA NA Mouse #3* <<1 NA 0 NA NA Mouse #4 6 1.5 4.5 5.5 4.5 Mouse #5 6 2 4.5 5.5 4.5 Mouse #6 7 2 5 6 5 Mouse #7 7 1.5 5 6 5 Mouse #8 7 2.5 5 6 5 Mouse #9 7 4.5 5 6 5 Mouse #10 7 4.5 5 6 5 Mouse #11 7 4.5 5 6 5 Mouse #12 8 2 6 7 6 Mouse #13 8 2.5 6 7 6 Mouse #14 8 1 6 7 6 Mouse #15 8 1.5 6 7 6 Mouse #16 8 2 6 7 6 Mouse #17 8 2 6 7 6 Mouse #18 7 1.5 5 6 5 Mouse #19 8 2 6 7 6 Mouse #20 8 1.5 6 7 6 PAP 0.25 mg/kg Mouse #1 <<1 NA NA NA NA Mouse #2 <<1 NA NA NA NA Mouse #3 6 1 4.5 5.5 5 Mouse #4 7 1.5 5 6 5 Mouse #5 8 1.5 6 7 5.5 Mouse #6 21 0 10 NO NO Mouse #7 21 0 10 NO NO Mouse #8 21 0 10 NO NO Mouse #9 21 0 10 NO NO Mouse #10 21 0 10 NO NO *Mouse died after intracerebral injection; NA = not applicable, NO = not observed Figure 1 Protective Activity of PAP in CBA Mice Challenged with LCM Virus . CBA mice were inoculated with intracerebral injections of the WE54 strain of LCMV at a 1000 PFU dose level that is lethal to 100% of mice within 7–9 days. Mice were treated either with vehicle or PAP administered intraperitoneally 24 hours prior to, 1 hour prior to and 24 hours, 48 hours 72 hours and 96 hours after virus inoculation. Mice were then observed twice daily for 21 days for morbidity and mortality. ( Top ) Shown are representative survival curves detailing the cumulative proportions (%) of mice surviving after virus inoculation. See Table 1 for more detailed information of the treatment outcome. ( Bottom ) Shown is Kaplan-Meir life-table analysis and statistical comparison using the log-rank test. Of the 10 mice treated with the broad-spectrum antiviral agent pokeweed antiviral protein (PAP), 2 died on day 1 immediately after intracerebral injection due to accidental brain injury and are invaluable. Five of the remaining 8 mice (62.5%) treated with PAP remained alive > 21 days post-LCMV inoculation (median survival >21 days, p = 0.0069) (Table 1 , Figure 1 ). These mice exhibited no LCMV infection-related weight loss or convulsions and showed no signs of scruffy fur. A significant improvement in mobility was also noted. Thus, PAP exhibited significant in vivo anti-LCMV activity in mice challenged intracerebrally with an otherwise invariably fatal dose of LCMV. Discussion In the present study, we describe the significant efficacy of the highly stable, potent and broad-spectrum anti-viral agent PAP in a murine model of LCMV. Our results presented herein provide unprecedented experimental evidence that PAP exhibits antiviral activity in the CNS of LCMV-infected mice. Future studies will examine the therapeutic activity of PAP against LCMV in post-challenge settings. As PAP is well described as an HIV agent, this observation is also relevant as HIV-1 infects the central nervous system (CNS) and it has been feared that the CNS may be a sanctuary site where HIV-1 could hide and continue to replicate despite otherwise effective antiretroviral treatment [ 48 , 49 ]. Antiretroviral therapy of HIV-infected children is associated with a decline in CSF HIV RNA and an improvement in neurological status. The development of genotypic mutations was different in CSF and plasma, suggesting discordant viral evolution. These results suggest that antiretroviral treatment in children should include agents with activity in the CNS [ 50 ]. While these results suggest that PAP crosses the blood brain barrier and may therefore be beneficial in other viral infections affecting CNS, they need to be interpreted with due caution for HIV infections involving the CNS since LCMV affects leptomeninges while HIV affects neurons. In future studies, we must also consider the possibility that PAP penetrates the blood brain barrier only when the latter is impaired by the intracerebral administration of the virus. If this is the case, PAP may not be useful in pre-challenge prophylaxis against LCMV-mediated CNS infection. While these results support the notion that the antiviral activity spectrum of PAP covers LCMV as well, an immunomodulatory effect of PAP may also contribute to the observed prophylactic efficacy of PAP against LCMV. Cases of LCMV infections have been reported in Europe, the Americas, Australia, and Japan. According to the Center for Disease Control, currently there is no specific drug therapy for LCMV. Although the anti-viral agent ribavirin is effective against LCMV in vitro, there is no established evidence to support its use for treatment of LCMV in humans [ 51 ]. The anti-viral activity of PAP has been described against numerous pathogenic viruses, which included poliovirus, HIV-1, herpes simplex virus, cytomegalovirus, influenza virus and now, the negative-strand RNA virus Lymphocytic choriomeningitis virus. The ability of PAP to inhibit viral protein synthesis and depurinate viral RNA and DNA [ 15 , 16 ] as well as capped rRNA and mRNAs [ 52 ] and its ability to inhibit ribosomal frame shifting and retransposition, make it and ideal candidate for anti-viral strategies. Conclusion Treatment of CBA mice with the broad-spectrum anti-viral agent, PAP significantly improved the probability of survival following the LCMV challenge and decreased overall LCMV-related symptoms. The ability of PAP to exhibit anti-viral activity within the central nervous system is also encouraging within the framework of potential HIV-treatment. We have recently described the rational design and engineering of several recombinant PAP mutants with superior anti-HIV activity [ 11 , 17 ]. In our future studies, we plan to describe the potential anti-LCMV activity of these PAP mutants as well as optimizing the prophylactic and post-exposure treatment regimens. In addition, it will be important to determine if PAP or PAP mutants have activity against other viruses associated with lethal viral hemorrhagic fevers and/or encephalomyelitis, such as the Ebola viruses of the Filoviridae family [ 53 - 56 ]. List of abbreviations LCMV: Lymphocytic choriomeningitis virus PAP: Pokeweed anti-viral protein HIV and FIV: Human and Feline immunodeficiency virus, respectively Competing interests FMU is listed as an inventor on a number of PAP patents which are owned by PHI. FMU, AV and HET are salaried employees at PHI. PHI is a non-profit public charity. Authors' contributions FMU designed the research project. LR, AP & LT conducted the Lassa experiments. FMU & AV provided PAP and coordinated the Lassa experiments with LR, AP and LT. FMU and HT wrote the manuscript. All authors read and approved manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC554105.xml
555596	A model of impairment and functional limitation in rheumatoid arthritis	Background We have previously proposed a theoretical model for studying physical disability and other outcomes in rheumatoid arthritis (RA). The purpose of this paper is to test a model of impairment and functional limitation in (RA), using empirical data from a sample of RA patients. We based the model on the disablement process framework. Methods We posited two distinct types of impairment in RA: 1) Joint inflammation, measured by the tender, painful and swollen joint counts; and 2) Joint deformity, measured by the deformed joint count. We hypothesized direct paths from the two impairments to functional limitation, measured by the shirt-button speed, grip strength and walking velocity. We used structural equation modeling to test the hypothetical relationships, using empirical data from a sample of RA patients recruited from six rheumatology clinics. Results The RA sample was comprised of 779 RA patients. In the structural equation model, the joint inflammation impairment displayed a strong significant path toward the measured variables of joint pain, tenderness and swelling (standardized regression coefficients 0.758, 0.872 and 0.512, P ≤ 0.001 for each). The joint deformity impairment likewise displayed significant paths toward the measured upper limb, lower limb, and other deformed joint counts (standardized regression coefficients 0.849, 0.785, 0.308, P ≤ 0.001 for each). Both the joint inflammation and joint deformity impairments displayed strong direct paths toward functional limitation (standardized regression coefficients of -0.576 and -0.564, respectively, P ≤ 0.001 for each), and explained 65% of its variance. Model fit to data was fair to good, as evidenced by a comparative fit index of 0.975, and the root mean square error of approximation = 0.058. Conclusion This evidence supports the occurrence of two distinct impairments in RA, joint inflammation and joint deformity, that together, contribute strongly to functional limitations in this disease. These findings may have implications for investigators aiming to measure outcome in RA.	Background Physical disability is an important outcome of rheumatoid arthritis (RA) [ 1 , 2 ]. The American College of Rheumatology preliminary definition of improvement in RA, used primarily to assess the short term response to medical interventions, includes disability as one of its seven key outcomes [ 3 ]. An understanding of disability in RA requires an appreciation of the interrelationship between the biology of the disease, the person and his or her psychology, and the social environment [ 4 - 7 ]. Inquiries into physical disability in RA, needing to weigh the influence of numerous variables interacting over time in complex ways, benefit from a conceptual framework, or model [ 8 ]. A model informs research by clarifying the relationships between variables, and facilitates communication of ideas related to the research in question [ 9 ]. In studying the development of disability in RA, we proposed a theoretical framework [ 10 ], which we based on the disablement process that occurs with aging [ 8 ]. Initially based on purely theoretical grounds, our model proposed strategies to quantify the four sequential stages of the main disease-disability pathway in RA: pathology → impairment → functional limitation → physical disability [ 10 ]. A useful device to facilitate the understanding of these stages of disablement, is to think of them in terms of the level at which they occur, and can be quantified. Thus, pathology occurs at the level of molecules, cells, or tissues, and is measured using tests such as the erythrocyte sedimentation rate, the C-reactive protein concentration, cytokine expression patterns, or images of the joints obtained with X-ray or MRI. Impairments are dysfunctions or structural abnormalities that occur at the level of organs or organ systems. They include signs and symptoms of disease such as pain, morning stiffness, joint tenderness, swelling and deformity. Functional limitations are restrictions in basic physical or mental actions, and they involve the whole person. Although they can be measured in a number of different ways, we have chosen to use performance-based functional tests such as the grip strength, walking velocity and the timed shirt-button test to measure functional limitations [ 11 ]. Disability involves difficulty with a physical or mental activity, within a social context. The measurement level therefore should include the person and the societal environment. We have used self report measures of physical disability such as the Health Assessment Questionnaire, or the physical function scale of the SF-36, to measure physical disability [ 12 ]. As noted above, we based our initial model and its proposed measurement strategies on purely theoretical grounds [ 10 ]. Since its publication, however, we have provided initial empirical evidence to support two of the model's main disease-disability pathway stages, and our approach to measuring them, using data from a clinical sample of RA patients. Those two stages are functional limitation and physical disability [ 11 - 14 ]. In the present report, we show additional data to support our definition of impairment in RA. Methods Patients From 1996 to 2000, we enrolled consecutive patients meeting the 1987 RA criteria [ 15 ], into a study of the disablement process in RA We have described our sample in previous publications [ 13 , 14 ]. Here, we will show cross-sectional results obtained during the recruitment evaluation of each participant. Settings We recruited patients from six outpatient rheumatology clinics in San Antonio, Texas: 1) An Army Medical Center, 2) An Air Force Medical Center, 3) A private, university-based clinic, 4) A community-based, seven-rheumatologist private practice, 5) A county-funded clinic and, 6) A Veterans Administration clinic. All evaluations were done on location in these facilities. Data collection procedures Our study was approved by the Institutional Review Board of each of the clinical facilities were we went to recruit patients, and all patients gave written, informed consent. A physician or a research nurse, assisted by a trained research associate, conducted evaluations at the clinic where the patient was recruited. The evaluation lasted approximately 90 minutes, and consisted of a comprehensive interview, physical examination, review of medical records, and laboratory and X-ray tests. Interviews were conducted in either English or Spanish, as preferred by patients. Data elements Impairments We measured impairments using self-report and physical examination. We used a validated, one-page joint mannequin for patients to mark the joints that were painful or swollen [ 16 ]. This variable is expressed as the painful joint count. A physician or research nurse, trained in joint examination techniques, assessed 48 joints in each patient for the presence or absence of tenderness or pain on motion, swelling or deformity. Each of these variables is expressed as a count for the number of affected joints [ 17 ]. Functional limitations We measured functional limitations using the following performance-based rheumatology function tests: 1. Grip strength. We measured grip strength with a hand held JAMAR ® Dynamometer (Sammons Preston, Bolingbrook, Illinois). In a sitting position, with the elbow held at 90 degrees, and the forearm supported on a flat horizontal surface, patients were asked to squeeze the handle with as much as strength as possible. Three repetitions from each hand were recorded, in kilograms. The mean value of all repetitions for both hands is shown; 2. Walking velocity. Starting from a standing position, patients were asked to walk at their usual pace, for a distance of 50 feet, or 25 feet if they had difficulty covering the full distance. No effort was made to conceal the stopwatch used to time the patients. Results are expressed in feet per second. Patients unable to walk were assigned a velocity of 0 feet per minute; 3. Timed button test. Patients were asked to don and fasten the front buttons of a standard eight-button, men or women's extra-large, denim shirt (Wal-Mart, San Antonio, Texas). A stop watch was activated when the patient took the shirt as it was offered by the examiner, and stopped when the last button was fastened. This test quantifies the performance and large and small upper extremity joints. Results are expressed as buttons per minute. Patient unable to don the shirt were assigned a value of 0 buttons per minute. Analysis We began the analysis by inspecting summary statistics and histograms of all study variables. Skewed variables were square root transformed toward normality. We specified a structural equation model (SEM) of impairment and functional limitation in RA [ 18 ]. We hypothesized that impairment in RA is represented by two distinct constructs. The first of these is characterized by joint inflammation & pain, the second by joint deformity. Each of these two constructs is represented by a latent variable in the model. The inflammation-pain construct is measured by a physical examination joint count for tenderness and swelling [ 17 ], and a self-reported joint count of painful joints [ 16 ]. The joint deformity construct is measured by the deformed joint count [ 17 ]. Because a latent variable is more reliably measured by two or more measures, we disaggregated the deformed joint count into upper and lower extremities counts, and a count for other joints. The latter included the temporo-mandibular, acromio-clavicular and sterno-clavicular joints. We assessed the influence of joint impairment on functional limitation, the next stage in the RA disablement process, by positing a direct path from the former to the later in the SEM. As we have shown previously [ 11 ], we defined functional limitation as a latent variable measured by three rheumatology function tests: grip strength, shirt-button speed and walking velocity over 50 feet [ 11 ]. We used a maximum likelihood procedure to estimate the model parameters. Once these were estimated, we examined modification indices seeking parameters not estimated in the initial model, that may increase model fit if added to the model. Here, it is important to keep in mind that structural equation modeling is not meant to be a data-driven technique, rather, it should be a theory-driven one [ 18 ]. We therefore only considered parameters that would not substantially change the basic structure of the model, i.e. one with two impairment latent variables and one functional limitation latent variable. Also after specifying the initial model, we evaluated a the effect of adding a direct path from the joint inflammation to the joint deformity latent variables. We quantified the degree of fit of the hypothetical model to our empirical data using the comparative fit index (CFI), the normed fit index (NFI), and the root mean square error of approximation (RMSEA). Interpretation of these fit indices is subjective, and there are no universally accepted guidelines. Generally, fit index values ≥ 0.95 are considered to indicate acceptable fit of a model to data [ 18 ]. RMSEA values ≤ 0.05 indicate close fit, values ≤ 0.08 indicate reasonable error of approximation [ 19 ]. We used the parameter estimates for the latent variables to compute their values and plot frequency distributions for each one. We used the Amos 5.0 statistical pathway package to specify and test the SEM. (SmallWaters Corporation, Chicago, Illinois). Results We recruited 779 patients from 1996 to 2000. We have described the clinical characteristics of the study sample in earlier publications [ 13 , 14 ]. Briefly, from 1996 to 2000, we recruited consecutive patients who met the 1987 criteria for the classification of RA [ 15 ], from six rheumatology clinics in San Antonio, Texas. In addition to having RA, patients had to be 18 years of age or older. No other inclusion or exclusion criteria were applied. The median age of the patients was 57 years (min to max 19 to 90), 70% were women and 56% were Hispanic. The median number of years of formal education was 12 (min to max 0 to > 16), 21% were working full- or part-time, 27% were disabled from work. The median disease duration was 8 years (min to max 0 to 52). Mean joint counts were 15 for tender, 7 for swollen and 10 for deformed. Subcutaneous nodules were present in 30%, and rheumatoid factor in 89%. The joint counts displayed skewed distributions. Square root transformation reduced skewness from tender, swollen and painful joint counts, but not from the deformed joint counts. We used the transformed values for the former three variables, but used the unstransformed deformed joint count in the SEM. A graphical display of the proposed model is shown in Figure 1 . We hypothesize that two distinct impairments take place in RA, each represented by a latent variable in the SEM. The two impairments, joint inflammation and joint deformity, are shown as ovals on the left side of Figure 1 . Joint inflammation is measured by the extent of joint tenderness, joint swelling and joint pain, shown in boxes in Figure 1 . Table 1 lists the path coefficients from joint inflammation to the measured variables. All three were large, and statistically significant. The standardized coefficient was > 0.5 for each measured variable, suggesting the a standard deviation change in the latent variable of joint inflammation, is associated with a change in the measured variables of at least one half standard deviation (Table 1 ). Figure 1 Identification diagram of a structural equation model of the relationship between the stages of impairment and functional limitation in rheumatoid arthritis. Two types of impairment, joint inflammation and joint deformity are shown as ovals on the left. Measurements for these latent variables include joint tenderness (JT), joint swelling (JS) and joint pain (JP), for joint inflammation; and joint deformities. We disaggregated joint deformities into upper limb (DUL), lower limb (DLL) joints, and other joints (DOJ). Several of the parameters were constrained to enable estimation. Circles represent residuals or disturbance terms, for each variable. See Table 1 for parameter estimates. Table 1 Parameter estimates from a structural equation model of joint impairment on functional limitations in RA. Parameter estimates Direct Paths Unstandardized SE P-value Standardized Joint inflammation → Functional limitation -.377 0.029 ≤ 0.001 -0.576 Joint deformity → Functional limitation -0.088 0.007 ≤ 0.001 -0.564 Joint tenderness → Joint inflammation 1.000 0.872 Joint pain → Joint inflammation 0.837 0.046 ≤ 0.001 0.758 Joint swelling → Joint inflammation 0.445 0.034 ≤ 0.001 0.513 Upper limb joint deformity → Joint deformity 1.000 0.849 Lower limb joint deformity → Joint deformity 0.478 0.029 ≤ 0.001 0.785 Other joint deformity → Joint deformity 0.011 0.001 ≤ 0.001 0.308 Functional limitation → Grip strength 1.000 0.773 Functional limitation → Walking velocity 0.958 0.051 ≤ 0.001 0.740 Functional limitation → Button speed 0.046 0.002 ≤ 0.001 0.757 Correlations* Upper extremity deformity ←→ Other joint deformity -0.22 --- 0.004 --- Walking velocity ←→ Shirt button speed 0.274 --- ≤ 0.001 --- Parameters estimated using maximum likelihood with Amos 4.0. See Figure 1 for a diagram of the specification model. * Correlation shown are between the residuals of the measured variables. The latent variable of joint deformity is measured by the deformed joint count, which we disaggregated into counts for the joints in the upper and lower extremities, and other joints. The path coefficients for upper and lower extremity deformities were stronger than that of the other joints (i.e. temporo-mandibular, acromio-clavicular and sterno-clavicular joints), but all three remained significant Table 1 ). To obtain a better understanding of the properties of the latent variables, we used the path coefficients from the two impairment latent variables to compute their estimated values, generating a new variable for each one. Figures 2 and 3 show the distribution of the two impairment latent variables, after rescaling them variables to vary from 0 to 100. Joint inflammation displayed a characteristic Gaussian distribution (Figure 2 ). This was not the case for joint deformity, however, which remained skewed by the a substantial number of patients who lacked deformities on physical examination (Figure 3 ). Figure 2 Frequency distributions of the joint inflammation (JI) impairment latent variable. This was computed from JI = JT 1/2 + JS 1/2 × 0.445 + JP 1/2 × 0.837, where JT = joint tenderness, JS = joint swelling, and JP = joint pain. Weights were estimated using maximum likelihood with Amos, after constraining the coefficients for JT and DUL to 1. The latent variable was then rescaled to vary from 0 to 100. Figure 3 Frequency distributions of the joint deformity (JD) impairment latent variable. This was computed from JD = DUL + DLL × 0.478 + DOJ × 0.011, where DUL = deformity upper limb, DLL = deformity lower limb, and DOJ = deformity other joints. The weights for the equation were estimated using maximum likelihood with Amos, after constraining the coefficients for JT and DUL to 1. The latent variable was then rescaled to vary from 0 to 100. In the structural equation model, both of the impairment latent variables displayed strong direct paths toward functional limitations. The standardized coefficients were -0.5 or less, representing a change of more than half a standard deviation in the functional limitations for every standard deviation change in the impairments (Table 1 ). The squared multiple correlation of functional limitation was 0.65, suggesting that impairments explain 65% of functional limitation's variance. The initial CFI and NFI of the model were both 0.95, suggesting a close fit of the model to the data. The initial RMSEA was 0.07, suggesting reasonable fit [ 19 ]. Modification indices suggested a number of potential parameters that could increase model fit if added to the estimation model. Most of these did not make clinical sense, or ran counter to the a priori model we were testing, and we therefore did not specify them. However, we noted two covariances that would increase model fit without altering the overall structure of the model. The first of these was between the residuals of the measured variables for deformities in the upper extremity and deformities in other joints; the second, between residuals for walking velocity and the timed button test. After specifying these two covariances, CFI increased to 0.975, NFI to 0.966, and RMSEA decreased to 0.058. Also post hoc , we tested a direct path between joint inflammation and joint deformity, but the resulting coefficient was small and did not reach statistical significance. We therefore omitted an inflammation → deformity path from the final model. Discussion Several models have been proposed to study disability in the general population [ 8 , 9 , 20 ], and can be applied to study RA and other types or arthritis [ 10 , 21 , 22 ]. The World Health Organization (WHO) has developed the International Classification of Functioning, Disability and Health (ICF), with a corresponding set of core measures for a number of chronic conditions [ 21 ]. One set of ICF core measures has been proposed for RA [ 21 ]. It includes a comprehensive range of body structures and functions, activities and participation that can be assessed in studies of disability in RA [ 21 ]. The ICF classifies its components into categories that are analogous to the disablement process stages: body structures and functions in the ICF are analogous to the stages of pathology and impairment, while activities and participation correspond to functional limitations and disability in the disablement process [ 22 ]. We chose the disablement process over other disability models, among other reasons, because it considers the stages in disablement as an explicit sequence of linked events, one leading to the next. Rather than suggesting specific functions or activities to measure, it offers broad definitions of each stage, leaving it up to investigators to find ways to test them. Several of the variables we include in the present analysis are represented in the ICF classification, including joint pain and deformity, and walking. Variables not represented in the ICF, but which we did obtain, include joint tenderness and swelling, and grip strength. Our measurement of the stage of disability, described elsewhere, has considerable homology with that of the ICF [ 12 ]. The most important difference between our RA disablement model and the ICF classification, is that the latter does not contemplate the stage of pathology with sufficient detail for our aim, to map pathways from disease to disability in RA. We used the disablement process framework to build a model of impairment and functional limitation in RA [ 8 ]. Impairments occur when pathology at the level of the molecule, cell or tissue, crosses the clinical horizon causing symptoms or signs of disease. They represent derangement of structure or function at the organ level. Consistent with this framework, we used the articular manifestations of RA, joint pain, tenderness, swelling and deformity, to measure impairment. It should be noted that we consider impairment to be theoretical construct that cannot be directly quantified. We studied it as a latent variable, the articular signs and symptoms listed above serving as the tools we used to tap into the impairment construct. The concept of impairment as a stage in the disablement process is not intended to oversimplify the anatomical or physiological derangements that occur within that stage. In fact, the derangements can be quite complex, depending on the nature of the initial pathology, and the organ system under study. In the case of rheumatoid joints, the initial pathology can be broadly classified into two discrete, but related groups: inflammation and damage [ 10 ]. It should be noted that we did not include measures of these two pathological processes here. However, because impairments are tied to their underlying pathology, we posited two types of impairments, one for each type of joint pathology. The first type is related to inflammation in the joints, and the other to damage. The former, we measured using tender, swollen and painful joint counts, the latter, using the deformed joint count. Both impairment latent variables displayed strong, statistically significant path coefficients toward the measured variables, providing evidence that the measures we chose adequately tap into the proposed impairments. In the diagram of our model, these paths are shown as arrows from the latent to the measured variables, to indicate that it is the joint inflammation or damage (both of them latent variables), that are "causing" the joint signs and symptoms that we are able to measure. Not included in our final model because it did not reach statistical significance, was a path from joint inflammation to deformity. This is likely because we restricted the present analysis to the stages of impairment and functional limitation. Although there is considerable evidence that inflammation leads to damage in RA joints, the link between the two processes occurs during the stage of pathology, not impairment. We expect to find a strong link between inflammation and damage when we extend our analyses to include pathology measures such as the erythrocyte sedimentation rate, C-reactive protein, joint erosions and joint space narrowing. According to disablement theory, impairments lead to functional limitations [ 8 , 10 ]. We expected that this should translate into a link between variables representing these two stages. We thus posited direct paths from each type of impairment to a latent variable representing functional limitations. We have shown previously that functional limitations can also be represented as a latent variable [ 12 ], and that it can be measured satisfactorily using the performance-based rheumatology function tests, grip strength, walking velocity and the shirt-button test [ 12 ]. We found strong path coefficients from both impairment latent variables to the functional limitation latent variable (Figure 1 ). Moreover, the impairments accounted for 65% of the variance in functional limitations. Both these findings provide additional support for our definition of rheumatoid impairments and functional limitations. The disablement process was proposed as a framework to aid investigators in their efforts to understand the development of disability in aging, and in specific disease states [ 8 ]. The framework's acknowledgement of the sequential nature of a disease's manifestations make it especially informative to inquiries into disability in chronic diseases. It is worthy of attention that the disablement model can also be applied advantageously to outcomes other than disability. One of our goals with this and earlier efforts to model RA's stages [ 10 - 12 ], is to improve the current interpretation of the disease's outcome. Current systems used to assess RA's outcome mix stages of the disease process, without regard to their sequential nature, or omit some stages altogether [ 3 , 23 ]. For example, the improvement criteria of American College of Rheumatology (ACR), define response on the basis of the ACR's core set of RA disease activity measures [ 24 ]. These measures, although empirically tested in clinical trials, were adopted without reference to an explicit model of the disease. The improvement criteria include measures of pathology (i.e. the erythrocyte sedimentation rate or the C-reactive protein), impairment (i.e. the tender and swollen joint counts, global assessment of disease activity, pain scale), and disability (i.e. the Health Assessment Questionnaire). They do not include measures of functional limitation [ 3 ]. The success with which the ACR and similar response measurement systems have been used in clinical trials, does not preclude the possibility that they could be improved [ 25 ]. Mixing or omitting disease stages may dilute a response measurement system's ability to detect the effect of treatments targeted at the early stages of the disease process. Although treatments that are primarily anti-inflammatory may indeed affect late disease stages such as physical disability, their effect is indirect, mediated through their primary effect on the inflammatory process. We have proposed what we believe would be a more rational approach to outcome assessment, using the stages of the disablement model to inform the selection of outcome measures [ 10 ]. Thus, measures of pathology and impairment would best capture response to anti-inflammatory therapies, while measures of functional limitation would best capture response to joint surgery or other rehabilitation interventions [ 10 ]. Empirical data to test our proposal would be of great interest. Certain constraints apply to the interpretation of our findings. The maximum likelihood estimator that we used for SEM assumes multivariate normality, a requirement that is not strictly met by some of the variables we used in this analysis. Non-normality may affect standard errors, and thus significance testing, about the parameter estimates, albeit not the value of the parameters themselves. The overall structure of the model we propose, i.e. two distinct impairments linked to functional limitations, is thus likely to be unaffected by this deviation from assumptions. The patient sample we studied is sufficiently large that the potential deleterious effect of non-normality on the significance of the path coefficients may be offset. It should also be noted that data we used here to test the impairment → functional limitation relationship are cross-sectional. The sequential link between the two stages we propose has face validity in that joint tenderness, swelling and deformity are causes, not consequences, of diminished grip strength, walking velocity and shirt button speed. Nevertheless, confirmation of our findings in a longitudinal dataset would strengthen the evidence for the model. Conclusion We conclude that two distinct impairments occur in RA, one characterized by signs and symptoms of joint inflammation, the other by joint deformity. Both of these contribute substantially to the functional limitations that occur in this disease. List of abbreviations RA = Rheumatoid arthritis; ÓRALE = Outcome of rheumatoid arthritis longitudinal evaluation; CFI = Comparative fit index; NFI = Normed fit index; RMSEA: Root mean square error of approximation; SEM: Structural equation model; ACR = American College of Rheumatology Competing interests The author(s) declare that they have no competing interests. Authors' contributions AE designed and obtained funding for the study, directed the statistical analysis, and drafted initial and final versions of the manuscript; RWH performed the statistical analysis and edited the manuscript; IDR designed and obtained funding for the study, supervised its implementation, and edited the manuscript. Table 2 Squared multiple correlations of impairments and functional limitations VARIABLE R 2 Joint Deformity‡ 0.000 Joint Inflammation‡ 0.000 Functional Limitation‡ 0.650 Upper limb joint deformities 0.617 Lower Limb Joint Deformities 0.722 Other joint Deformities 0.095 Shirt-button time 0.574 Walking velocity 0.547 Grip strength 1/2 0.598 Tender joint count 1/2 0.761 Swollen Joint Count 1/2 0.263 Painful Joint Count 1/2 0.574 Posited relationships between variables are shown graphically in Figure 1. ‡ Identifies latent variables Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555596.xml
434154	Identifying Genes Involved in Innate Immunity through RNAi	null	An organism's ability to sense and respond to potentially harmful pathogens is key to its survival. To fight off disease and infection, organisms must detect pathogens, activate immune cell signaling pathways, and produce molecules able to thwart a pathogenic attack. So fundamental is this need that molecules and protein domains related to innate immunity are evident in organisms as diverse as plants, flies, and humans, highlighting the ancient origins of defense mechanisms. Once detected, the pathogen's presence triggers a cascade of signaling events that generate a rapid response tailored to specific classes of pathogens. When pathogens attack the Drosophila fruitfly, they elicit a range of defensive reactions in the fly, including the production of antimicrobial proteins, phagocytes (which engulf the pathogen), and toxic metabolites. A microbe sets off one of these responses by interacting with a receptor, triggering a pathway that activates a special class of transcription factors, which in turn activate genes needed to make antimicrobial peptides, say, or toxins. Drosophila attacked by fungi and a certain class of bacteria activate these transcription factors through a pathway (the Toll pathway) that also operates in mammals. A second pathway is activated when a different class of bacteria attack. While the general steps of this pathway, called the Immune deficiency (Imd) pathway, are known—Dredd-mediated activation of the Relish transcription factor, for example, is central to this antibacterial response—the details and mechanisms remain unclear. Signaling pathways are notoriously complex and the Imd pathway is no different. In the current model, bacterial pathogens stimulate a transmembrane receptor, which activates the Imd protein, which then transmits the signal through intermediary proteins, which ultimately activate Dredd, sending Relish into the nucleus to activate genes required for an immune response. In this issue of PLoS Biology , Edan Foley and Patrick O'Farrell use a genome-wide approach to characterize the Imd pathway, with an eye toward understanding what regulates the Dredd-Relish interaction. To identify pertinent genes and their roles, Foley and O'Farrell took advantage of a technique, called RNA interference (RNAi), that can selectively target and “silence,” or inhibit, nearly any gene. After silencing a gene, researchers can then track how a cell responds and infer the gene's function. The technique uses double-stranded RNA (dsRNA) molecules with nucleic acid sequences that match a gene of interest (RNA can bind to DNA through complementary base pairing). The dsRNAs precipitate a series of steps that ultimately degrade the messenger RNA associated with the gene, preventing messenger RNA translation, thereby silencing the gene. The authors produced over 7,000 dsRNAs corresponding to most of the Drosophila genes with counterparts in mammals or the worm C. elegans (that is, conserved genes), then developed a cell culture model to analyze the components of the Imd pathway. Since the cells in these cultures respond to the presence of bacterial proteins by generating antimicrobial peptides, they provide a good testbed for identifying genes involved in the pathway. Foley and O'Farrell's RNAi screen identified many molecules involved in signaling, including both signal inhibitors and activators. Among these signaling components, they discovered two new genes: one, which they named sickie , is required to activate Relish; the second, called defense repressor 1 (dnr1) , appears to inhibit Dredd activity and thus inhibit the pathway. Based on molecular analysis of these genes—which involved exposing cells to bacterial proteins—the authors propose a model of Imd signaling in which Dredd and the protein produced by dnr1 , called Dnr1, operate through a negative feedback loop: Dredd activity appears to promote the accumulation of its own inhibitor, Dnr1. Since suppression of Dnr1 through RNAi can trigger an immune response, the authors explain, it appears that interruption of this feedback loop activates the signaling pathway. Without the inhibitory action of Dnr1, Dredd can activate Relish, which dissociates from Dredd, enters the nucleus, and activates the transcription of antimicrobial genes. It's likely that the two genes described here play a key role in activating Relish, the authors conclude, and that others identified in this screen will also prove significant. Since many of these components are conserved in mammals, the pathway may likewise operate in humans. Future experiments will be the test of these assumptions, but until then, Foley and O'Farrell have demonstrated the soundness of using RNAi screens on a large scale to dissect the elements of a complex signaling pathway.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC434154.xml
539302	Penile metastasis from primary transitional cell carcinoma of the renal pelvis: first manifestation of systemic spread	Background Almost one-third of all penile metastases are detected at the same time as a primary tumor, whereas the remaining two-thirds are detected a mean of 18 months after the discovery of the primary tumor. Cutaneous metastasis of transitional cell carcinoma (TCC) is extremely rare and generally accepted as the late manifestation of a systemic spread. Case presentation We report the first case of simultaneous penile and lung metastases from a primary TCC of the renal pelvis in a 76-year-old man, that occurred 8 years after a left nephroureterectomy. Conclusions This case report underscores the importance of physical examinations of the skin of patients who undergo surgical procedures for TCC from bladder as well as from the upper urinary tract, including those seemingly without metastatic disease, because of the possibility of skin and penile metastatic spread.	Background Approximately 300 cases of penile metastases are reported worldwide [ 1 ]. Seventy percent of these metastases have their origin in the genitourinary tract, and the primary tumor is most frequently located in the bladder [ 2 ]. We present the first case of penile metastasis from primary transitional cell carcinoma (TCC) of the renal pelvis, and discuss the striking feature of discovering lung metastasis almost simultaneously. Case presentation A 76-year-old man with a painful, ulcerative swelling of the glans that had appeared less than 2 months previously was brought to our attention. Physical examination revealed a 3.5-cm fungating lesion involving the upper half of the glans [Fig. 1 ]. The lesion was malodorous with varying amounts of seropurulent discharge, partially ulcerated, erythematous, and slightly itchy. The superficial inguinal lymph nodes were not palpable. No similar lesions were evident on the head, neck, trunk, or arms. The patient had been treated in another hospital 8 years previously for TCC of the left renal pelvis by nephroureterectomy (pT2 Nx M0 G2). Subsequent postoperative follow-ups (every 6 months for the first 3 years after surgery and then once every year) consisting of a clinical examination, total body CT scan, and regular endoscopic examination showed no evidence of recurrence and the patient did well until he noticed the painful penile nodule. Pathological examination of incisional biopsies confirmed the lesion to be urothelial carcinoma [Fig. 2 ]. A subsequent CT workup revealed a 4-cm lesion localized in the upper-right pulmonary lobe and enlarged pelvic lymph nodes suggestive of multiple metastases. The patient began combination chemotherapy (gemcitabine at 1250 mg/m2 on days 1, 8, and 15, and then every 28 days for six courses) and external beam radiotherapy to the mass, which promptly relieved the penile pain. At an 8-month follow-up the patient was still alive with no remarkable changes to the pulmonary and lymph node metastases, or the penile swelling. Conclusion Cutaneous metastases from primary TCC of the urinary system are extremely rare and are generally accepted as late manifestations of systemic spread [ 1 ]. Spector et al. (1987) suggested that skin metastases in TCC were the result of increased longevity in successfully treated patients [ 3 ]. Various mechanisms by which lesions may metastasize to the penis have been reported. Recently Bordeau and Lynch and Berger et al. reported three cases of penile metastases from TCC of the bladder in which they assumed that metastatic spread from primary bladder cancer to the penis occurred mainly via the retrograde venous route [ 2 , 4 ]. There are also a few reports of TCC seeding outside the urinary tract after iatrogenic procedures (i.e., partial cystectomy, suprapubic cystostomy, pyelotomy, and laparoscopy) as the cause of cutaneous metastasis [ 1 , 5 , 6 ]. Between 5% and 20% of patients with superficial bladder cancer have vascular or lymphatic spread [ 5 ], and the reported rate of skin metastasis from TCC of the bladder is between 0.2% and 2% [ 7 ]. The rate of skin metastasis from TCC of the renal pelvis is currently unknown, and to the best of our knowledge ours is the first reported case of penile metastasis from TCC of the upper urinary tract. Almost one-third of all penile metastases are generally detected at the same time as a primary tumor, whereas the remaining two-thirds are detected a mean of 18 months after the discovery of the primary tumor [ 8 ]. Because the disease-free survival of this patient was 8 years following the diagnosis of the primary tumor, this case report represents a description of the biology of an unusual tumor. Moreover, the penile metastasis in this case was not the late manifestation of systemic spread, as is usually reported in the literature [ 1 ]; in fact, the patient was accurately evaluated with a standard follow-up scheme. Therefore, it is plausible that the lung metastasis appears almost simultaneously with the penile metastasis. The optimal treatment of penile metastasis requires a multidisciplinary approach that is correlated with the disease extent. The average survival in patients with penile metastasis is 3.9 months from diagnosis and, with extensive surgery and chemotherapy, a survival of 9.2 months has been reported [ 4 ]. This patient was not eligible for platinum-based regimens due to impaired renal function (prior nephrectomy) and advanced age. As a result, he was treated with gemcitabine as a single agent. This treatment has demonstrated objective response rates of 25–29% with minimal toxicity as compared to standard regimens [ 9 - 11 ]. In conclusion, this case report suggests that particular attention should be paid to physical examination of the skin of patients who undergo surgical procedures for TCC from bladder as well as from the upper urinary tract, including those seemingly without metastatic disease, because of the possibility of skin and penile metastatic spread. Competing interests The author(s) declare that they have no competing interests. Authors' contributions GP drafted the manuscript and coordinated the co-authors. IP, MS, PC participated in the sequences alignment. GM carried out the histological features of the lesion. FF participated in the sequences alignment and coordinated the co-authors. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC539302.xml
524353	True good	null	"The open society, the unrestricted access to knowledge, the unplanned and uninhibited association of men for its furtherance – these are what may make a vast, complex, ever growing, ever changing, ever more specialized and expert technological world, nevertheless a world of human community [ 1 ]." –J. Robert Oppenheimer (1904–1967) Biomedical scientific communities increasingly articulate concern about the unsustainable status quo of traditional publishing paradigms [ 2 ], feeding an ongoing debate and consideration of alternative publishing models [ 3 ]. Seizing an opportune moment to contribute to the ongoing assessment and refinement of scholarly publishing in biomedical education, research, and patient care, an energetic group of librarians and faculty researchers endeavour to launch Biomedical Digital Libraries . While traditional biomedical library concern revolves around the role of collection stewardship and delivering published knowledge to students, scholars, researchers, and clinicians, information professionals in library or information center settings are also concerned with the cultivation of their own knowledge domains and evidence-based library practice [ 4 , 5 ]. Perhaps more than any other knowledge profession, librarians understand that their own opportunities for collaboration and barrier-free exchange of ideas and research results has dramatically altered and ultimately improved the provision of services and resource management in their professional environment. Open access scientific literature arrived in several pre-print guises over the previous two decades and in a recent contemporary commercial model with BioMed Central [ 6 ] (BMC). Separating itself from the pre-print unfiltered era, BMC provides biomedical researchers with peer review, retention of copyright, permanent redundant digital archiving in repositories such as PubMed Central (PMC) [ 7 ], and rapid global distribution of their ideas. Harnessing the innovations of the web, BMC also provides online submission, article history, and support for multiple languages. To sustain and expand an open access business model and maintain timely equitable global access, BMC publishing income derives from a combination of author fees, institutional memberships, and advertising. BMC now has 425 institutional members in 40 countries [ 8 ]. Researchers from member institutions have the right to publish an unlimited number of research articles in journals published by BMC without paying any article processing charges. Because of the BMC institutional membership, biomedical librarians and information professionals have unexpectedly been thrust into a position of advocacy for an alternative publishing model. More often than not, librarians are more aware than their faculty or researchers that open access does not mean the absence of peer review. In a web-based information age of unfiltered content and one-stop search engine shopping, both biomedical scientists and information professionals are justifiably concerned that the open access movement without peer review simply adds to the morass of unfiltered, unproven hyperbole. Attention to peer review provides credibility, and BMC even offers their journals the opportunity to publish review reports and preliminary drafts for each article as "publication history." The decision to found a BMC journal for the world of organized biomedical information was both spontaneous and practical. At the core of the founding of Biomedical Digital Libraries is the conviction that open access will push biomedical librarianship forward in new and improved ways. Biomedical Digital Libraries provides a legitimate alternative to traditional specialty journals in the field, which have subscription fees and assumption of copyright by the publisher. This journal will stress peer-reviewed open access to research and practice in digital collection and services settings and will permit rapid and unimpeded dissemination of knowledge, only weeks after manuscript submission. Over 30 information professionals and biomedical scholars will be engaged as editorial staff to conduct blinded peer review reports for original research, as well as integrate commentaries, resource reviews, and debates into a forum for the discussion of unique considerations of biomedical information needs. These include both opportunities and constraints presented by health care settings, including collaborative initiatives with information technology and informatics partners. We hope the advocates, philosophers, caretakers, and architects of biomedical library digital content take immediate advantage of rapid peer-review and publication, extensive BMC content promotion, permanent URL, redundant public archiving, and retention of copyright when they submit to Biomedical Digital Libraries . Beyond our immediate narrow spheres of digital library practice and service, the community of open knowledge has the immediate and timely potential to inspire, inform, and create value on a global scale through permanent, uninhibited access. Some seek good in authority, others in scientific research, others in pleasure. Others, who are in fact nearer the truth, have considered it necessary that the universal good, which all men desire, should not consist in any of the particular things which can only be possessed by one man, and which, when shared, afflict their possessor more by the want of the part he has not, than they please him by the possession of what he has. They have learned that the true good should be such as all can possess at once, without diminution and without envy, and which no one can lose against his will [ 9 ]. – Blaise Pascal(1623–1662)	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524353.xml
555569	Inclusion of Scar/WAVE3 in a similar complex to Scar/WAVE1 and 2	Background The Scar/WAVE family of proteins mediates signals to actin assembly by direct activation of the Arp2/3 complex. These proteins have been characterised as major regulators of lamellipodia formation downstream of Rac activation and as members of large protein complexes. Results We have investigated the interactions of the three human Scar/WAVE isoforms with several previously described binding partners for Scar/WAVE 1 or 2. We find that all three Scar/WAVE isoforms behave similarly and are likely to participate in the same kinds of protein complexes that regulate actin assembly. Conclusion Differences between Scar/WAVE proteins are therefore likely to be at the level of tissue distribution or subtle differences in the affinity for specific binding partners.	Background Rearrangements of the cortical actin cytoskeleton are essential for numerous cell processes such as cell migration, phagocytosis [ 1 , 2 ], and adhesion [ 3 ], in which formation of dendritic networks of polymerised actin plays a key part. The Arp2/3 complex is required in the formation of dendritic networks to provide sites for de novo nucleation of actin filaments and to form branch points from existing filaments [ 4 - 6 ]. The mammalian Arp2/3 complex can be activated downstream of Rho family small GTPases by several known interacting proteins, in particular members of the Wiskott Aldrich Syndrome Protein Family (haematopoietic WASP, ubiquitous N -WASP, and Scar/WAVE 1, 2, and 3 [Suppressor of Cyclic AMP Receptor mutation/Wiskott Aldrich VErprolin homologous protein]) [ 7 , 8 ]. Homology between these proteins consists of a core proline rich region, and C-terminal WH2 (Wiskott Homology 2) and Acidic (A) domains, of which the WH2 and A domains together are sufficient to activate the Arp2/3 complex [ 9 - 11 ]. N -WASP and the Scar/WAVE proteins differ the most at the amino-terminus, designated the Wiskott-Homology 1 (WH1) domain and the Scar/WAVE homology domain (SHD) [ 9 ]. Regulation of WASP-family proteins involves many interactions and is still the subject of intensive research. WASP and N-WASP are found in an inhibitory complex with WIP/CR16 proteins and can be activated by the small GTPase Cdc42 and a co-activating protein TOCA-1, recently described by Ho et al. [ 12 ]. Phosphorylation on tyrosine residues in the N-terminal half of WASP/N-WASP can also enhance the activation and perhaps serve to prolong its duration ([ 13 - 15 ]). Original reports had suggested that WASP/N-WASP were autoinhibited in their pure form, but whether this form is ever present in cells is unclear, so the physiological relevance of autoinhibition is unclear as well ([ 16 ]). Recombinant Scar/WAVE proteins were constitutively active toward Arp2/3 complex in vitro , but have been postulated to be regulated by the small GTPase Rac in vivo [ 17 , 18 ]. In 2002, Eden et al purified Scar/WAVE1 from bovine brain extracts in association with a complex of four other proteins; p140-SRA1, p125 Nck-Associated Protein (NckAP1), HSPC300, and Abi2 [ 19 ]. They proposed a model whereby active Rac could bind to p140-SRA1 and cause it plus NckAP1 and Abi2 to dissociate, releasing active Scar/WAVE in a complex with HSPC300. Some of the members of this so-called "Scar/WAVE complex" had been previously studied by other groups and implicated in signalling. SRA1 (also called Cytoplasmic FMRP Interacting protein [ 20 ]) is a Rac associated protein which also binds to the SH3 domain containing adapter protein Nck, and the Fragile-X Mental Retardation Protein (FMRP) [ 20 - 23 ]. NckAP1 also interacts with active Rac and Nck, although the interaction appears to be indirect in both cases [ 22 , 23 ]. HSPC300 is a small, ubiquitously expressed, uncharacterised Haematopoeitc Stem/Progenitor Cell protein, bearing considerable homology to the Maize actin cytoskeletal associated protein Brk1 [ 24 - 26 ]. Abi2 (Abl interactor 2) is a neuronally expressed SH3 domain containing protein associated with Abl tyrosine kinase [ 27 ]. It has a close relative, Abi1, also referred to as e3B1 (eps8 binding protein 1) in early literature [ 28 , 29 ]. Since Eden et al., a similar complex has also been isolated in association with Scar/WAVE2 [ 30 , 31 ]. This complex contains the more ubiquitously expressed Abi1 (as opposed to Abi2) consistent with the more ubiquitous expression of Scar/WAVE 2 compared to the other Scar/WAVE isoforms [ 31 - 33 ]. In these studies, the complex does not dissociate upon Rac activation and is not inhibitory during in vitro actin polymerisation assays [ 31 ]. Recombinant Abi1 is rather found to increase the Arp2/3 complex activating ability of recombinant Scar/WAVE 2 [ 31 ]. Following these studies, the mechanism for regulation of the different Scar/WAVE proteins has been called into question. It has been proposed that differences between Eden et al. and the other two studies may be due to experimental conditions or to genuine differences between Scar/WAVE isoforms [ 19 , 30 , 31 , 34 , 35 ]. Knockdown of Abi1, SRA1 and NckAP1 in mammalian tissue culture cells causes severe defects in formation of lamellipodia [ 31 , 36 ], similar to loss of Scar/WAVE 2 activity [ 37 , 38 ]. Mutations or RNAi knockdown of the only Drosophila Scar/WAVE protein severely affects the ability of cultured cells to produce lamellipodia, ruffles and filopodia [ 39 - 41 ]. Mutational and RNAi studies used to produce Drosophila cells deficient in the NckAP1 homologue, Kette, reveal a lack of actin based protrusions in the absence of functional protein [ 42 , 43 ]. RNAi knockdown of Drosophila Abi1 and SRA1 also prevents the formation lamellipodia in tissue culture cells [ 39 , 41 , 43 ]. Dictyostelium knock out of Scar protein exhibits severe defects in chemotaxis and motility, but cells can still extend pseudopods and migrate directionally [ 44 , 45 ], whereas knockout of the PirA gene encoding an homologue of SRA1 causes an excessive lamellipodial protrusion phenotype, thought to be due to unregulated Scar protein activity [ 46 ]. While the Scar/WAVE complex is currently thought to be the most likely regulator of Scar/WAVE activity in cells, several other Scar/WAVE binding proteins have been identified and proposed to regulate its activity. IRSp53, for example, binds to Scar/WAVE2 and also to Rac and was proposed to be a Scar/WAVE2 regulator prior to Eden et al. ([ 47 - 50 ]). It was unclear whether IRSp53 binding was specific to Scar/WAVE 2, or ubiquitous among the Scar/WAVEs and where/when Scar/WAVE was associated with IRSp53 as opposed to the other binding partners. IRSp53 is also a scaffold protein, with several partners, raising the possibility of at least two different types of large protein complexes in association with Scar/WAVE proteins [ 49 , 51 - 55 ]. Scar/WAVE3 is the most tissue-specific of the three mammalian Scar/WAVE isoforms, being found in haematopoietic cells and brain tissue, but not yet being characterised in cells [ 9 , 56 ]. It has not previously been shown whether Scar/WAVE3 interacts with either Abi1/2 or HSPC300. This report presents data indicating that the interaction of Scar/WAVE proteins with Abi1, HSPC300 and IRSp53 is conserved among all three Scar/WAVE proteins, suggesting that multiple different protein complexes likely exist in cells that contain different Scar/WAVE isoforms. This observation suggests that the different Scar/WAVE isoforms are not regulated by exclusive participation in specific complexes, but rather that the regulation is likely to be more subtle, at the level of affinity or modifications/binding partners not yet discovered. Alternatively, Scar/WAVEs may have largely overlapping functions when present in the same cell type, although there is already some evidence against this idea [ 57 ]. Results Association of the Scar/WAVE complex with Scar/WAVE 3 Protein sequence conservation between the three Scar/WAVE proteins in the N-terminal Scar Homology Domain (SHD) indicates a potential for association of binding partners with all three members of the family. Scar/WAVE1 and 2 have been shown to associate with the key adapter protein Abi1 through the SHD, so it is possible that Abi1 would also associate with Scar/WAVE3 [ 19 , 30 ]. To test for an interaction between Scar/WAVE3 and Abi1, immunoprecipitations were performed with antibodies recognising Abi1 (Fig. 1A ) and antibodies recognising Scar/WAVE3 (Fig. 1B ). Anti-Abi1 immunoprecipitates co-precipitate Scar/WAVE1 and 3 with endogenous Abi1 from mouse brain extracts, while Abi1, Scar/WAVE1 and Scar/WAVE3 are not present in immunoprecipitates with an unrelated control antibody or without antibody (Fig. 1A ). Abi1 also co-immunoprecipitates with Scar3, but is not detected in the absence of antibody or in unrelated control immunoprecipitates (Fig. 1B ). Since Abi1 is thought to bind directly to Scar/WAVE proteins and to connect them with the rest of the Scar/WAVE complex, interaction with Abi1 is a strong indicator Scar/WAVE3 associates in an analogous complex to Scar/WAVE1 and 2 isoforms [ 35 ]. Figure 1 Co-immunoprecipitation of Scar3 and Abi1 from mouse brain extract. Protein G beads were used to precipitate Abi1 from mouse brain extracts in the presence or absence of (A) anti-Abi1 or an unrelated control antibody or (B) anti-Scar3 or an unrelated control antibody. (A) The beads fractions were probed with anti-Abi1, anti-Scar/WAVE1 and anti-Scar/WAVE3. Anti-Abi1 immunoblotting confirmed precipitation of Abi1 only with specific antibody and immunoblotting with anti-Scar/WAVE1 shows association with a known binding partner. Immunoblotting with anti-Scar/WAVE3 revealed Scar/WAVE3 to also be present in Abi1 immunoprecipitates. Anti-Abi1 pulls down both Scar1 and Scar3 together with Abi1. (B) Beads fractions from anti-Scar/WAVE3 immunoprecipitates were probed with both anti-Scar/WAVE3 and anti-Abi1. Both Scar3 and Abi1 are detected in anti-Scar/WAVE3 immunoprecipitates, but not with controls. Conservation of Scar complex association between all three human Scar/WAVE isoforms Our data and previously published studies indicate that Abi1 exhibits the potential to bind to all three mammalian Scar/WAVE isoforms. To test whether the association of Scar/WAVE1, 2, and 3 with Abi1 is mediated by the Scar Homology Domain, Myc-Scar/WAVE SHD of each of the three family members were transiently expressed in Cos cells with HA-Abi1 or HA-HSPC300 and immunoprecipitated using a 9E10 anti-myc monoclonal antibody. HA-Abi1 precipitated with each of the three SHDs, but not in the absence of antibody or Myc-SHD (Fig. 2A ). HA-HSPC300 also co-immunoprecipitated with all three SHDs, but not the negative controls (Fig 2B ). Pull-down experiments performed from HA-Abi1 or HA-HSPC300 transfected Cos cell lysates using GST-fusion proteins of Scar/WAVE1 and 2 (Fig. 3A ) and Scar/WAVE3 (Fig. 3B ) also show conservation of interaction of all human Scar/WAVE isoforms with Abi1 and HSPC300. Supporting data showing the specific interaction of Abi1 with Scar/WAVE1 SHD is shown in Table 1 and Additional file 1B . Data supporting interaction of Scar/WAVE1 SHD with HSPC300 is shown in Table 1 and in Additional file 1C . Figure 2 Co-immunoprecipitation of Abi1 and HSPC300 with Scar Homology Domain. Cos 7 fibroblasts transiently transfected as indicated. Protein G beads were used to immunoprecipitate Myc-Scar1 SHD, Myc-Scar2 SHD or Myc-Scar3 SHD from lysates in the presence (+) or absence (-) of anti-Myc (9E10) monoclonal antibody. Empty pRK5-Myc vector was used as an additional negative control. (A) HA-Abi1 was detected in the beads plus antibody fractions for Myc-Scar1 SHD, Myc-Scar2 SHD, and Myc-Scar3 SHD, but not in the negative controls. (B) HA-HSPC300 was detected in the beads plus antibody fraction of immunoprecipitations from cells co-transfected only with Myc-Scar1 SHD, Myc-Scar2 SHD, and Myc-Scar3 SHD, but not empty vector controls or in the absence of 9E10 antibody. Figure 3 Recombinant SHD pulls down Abi1 and HSPC300. Lysate of Cos7 fibroblasts transfected with HA-Abi1 or HSPC300 was incubated as indicated with GST alone, GST-Scar1 SHD ( A ), GST-Scar2 SHD ( A ), or GST-Scar3 SHD ( B ) on glutathione-s-agarose beads. Beads bound fractions were analysed by SDS-PAGE and immunoblotting with a monoclonal anti-Myc (9E10) antibody. HA-Abi1 and HA-HSPC300 were found in pull-downs with all three Scar Homology Domains, but not with GST alone. Table 1 Yeast two-hybrid analysis of interactions between Abi1, HSPC300, and various domains of Scar1. (+) Indicates a positive interaction backed up by a beta-Gal reporter gene assay. (-) Indicates negative for an interaction between the constructs. pAct-Scar1 PYTH9 HSPC300 Abi1 Full Length (FL) + + SB + + SP + + BPWA - +/- PWA - - Conserved interaction of Scar/WAVE isoforms with IRSp53 Conservation of Scar Homology Domain binding partners raises the question of whether other binding partners of Scar/WAVE are also capable of interacting with the entire Scar/WAVE family. IRSp53 has been implicated in regulation of actin dynamics through binding to Scar/WAVE proteins [ 49 , 50 , 54 ]. GST-IRSp53 was used for pull down experiments to test for binding to Scar/WAVE1, 2, and 3 (Fig. 4 ). None of the Scar/WAVE isoforms were detected in the beads fraction of a negative control (constitutively active GST-L61 Cdc42), but all were detected in the beads fraction of GST-IRSp53. An ability to bind to IRSp53 is conserved between all three human members of the Scar/WAVE family, although the binding appears to be strongest for Scar/WAVE2. Conservation of this interaction may indicate that all three Scar/WAVE proteins are regulated by common mechanisms, but the increased binding of Scar/WAVE2 to IRSp53 highlights the possibility that differing affinities for binding partners between the three Scar/WAVE isoforms may affect their roles in cells. Figure 4 IRSp53 interacts with Scar/WAVE1, 2 and 3. Equivalent amounts of GST-IRSp53 or constitutively active GST-L61 Rac were used for pull-down assays from lysates of Cos cells transfected with Myc-Scar1, 2 or 3. Bead fractions and whole cell lysates, indicating loading with each Scar/WAVE isoform, were analysed by SDS-PAGE and immunoblotting with monoclonal (9E10) anti-Myc antibody. None of the Scar/WAVE isoforms bound to active Cdc42, but all were detected in GST-IRSp53 bead fractions. Cellular localisation of Scar/WAVE3 As an activator of the Arp2/3 complex, Scar/WAVE3 would be expected to localize to the same areas of the cell as polymerized actin, subunits of the Arp2/3 complex, and members of the Scar/WAVE complex. C2C12 cells were used for immunocytochemistry with anti-Scar3 polyclonal antibody (Fig. 5Ai, Bi, Ci ), phalloidin to detect filamentous actin (Fig. 5Aii ), anti-Arp2/3 monoclonal antibody (Fig. 2Bii ), and anti-Abi1 antibody (Fig. 5Cii ). C2C12 cells were used because they were the only standard tissue culture cell line that we found to contain a detectable amount of Scar/WAVE3 protein (see Methods). Scar/WAVE3 mainly exists in a peri-nuclear pool with enrichment to areas of polymerised actin (Fig. 5Aiii ). Scar/WAVE3 co-localises with the Arp2/3 complex subunit, Arp3 (Fig. 5Biii ), and with Abi1 (Fig. 5iii ) in areas of lamellipodial protrusion. Our results show Scar/WAVE3 to localise to areas of polymerised actin near the cell periphery and to co-localise with a known component of the Scar/WAVE complex, indicating that Scar/WAVE3 may have a similar role and method of regulation to Scar/WAVE1 and 2. Figure 5 Cellular localisation of Scar3. C2C12 cells were stained with anti-Scar/WAVE3, anti-Arp3, anti-Abi1 and fluorescently labelled phalloidin to investigate the cellular localisation of endogenous proteins. (A) C2C12 cells stained with (i) anti-Scar/WAVE3 (green), (ii) phalloidin (red) (iii) reveal co-localisation of Scar/WAVE3 with cortical polymerised actin in membrane ruffles. Scar/WAVE3 also colocalises with other areas enriched in polymerised actin, but not stress fibres. (B) Co-staining with (i) anti-Scar/WAVE3 and (ii) anti-Arp3 reveals co-localisation of Scar/WAVE3 (green) with the Arp2/3 complex (red) at protruding areas of the cell membrane. (C) Cells stained with (i) anti-Scar/WAVE3 and (ii) anti-Abi1. (iii) Scar/WAVE3 (green) co-localises with Abi1 (red) in some areas of membrane protrusion. A (i), B (i), and C (i) all show a cytoplasmic and perinuclear and cytoplasmic pool of Scar/WAVE3, with enrichment at areas of lamellipodial protrusion. B (ii) and C (ii) show similar patterns of staining for Arp3 and Abi1. Scale bars are equal to 20 μm in all pictures. Cellular localisation of Scar/WAVE SHD, Abi1, and HSPC300 If the SHD of Scar/WAVE protein mediates the interaction with a regulatory complex it can be expected that the SHD will co-localise with binding partners in cells. Ectopic expression of full length Scar/WAVE isoforms in cells reveals a peri-nuclear and cytoplasmic pool, causing an abundance of polymerised actin in the cytoplasm, but with enrichment of Scar/WAVE protein in lamellipodia [ 50 , 58 ]. A sizeable amount of Myc-Scar 1 SHD (Fig 6Aiii ), Myc-Scar 2 SHD (Fig 6Aiv ), and Myc-Scar 3 SHD (Fig. 6Av ) localises to the cell periphery in lamellipodia, as well as in a cytoplasmic pool similar to full length Scar/WAVE proteins (see Figure 5Ai for comparison). HA-Abi1 is present throughout the cytoplasm in spots (Fig 6Aii ), previously described as characteristic of the reticulovesicular system [ 59 ]. HA-HSPC300 exhibits a similar cellular localisation to ectopically expressed Scar/WAVE1, 2 and 3 being abundant in and around the nucleus, and diffusely around the cytoplasm with enrichment at lamellipodia (Fig 6Ai ). Figure 6 Cellular localization of Abi1, HSPC300, and the Scar homology domain. Transfected Cos cells were stained with a monoclonal anti-Myc antibody and polyclonal anti-HA antibodies to detect over-expressed proteins. (A) Cos7 cells were co-transfected with (i) HA-HSPC300, (ii) HA-Abi1, (iii) Myc-Scar1 SHD, (iv) Myc-Scar2 SHD, or (v) Myc-Scar3 SHD to examine localization of these proteins. HSPC300, and all three SHDs localise in a diffuse cytoplasmic pool with some enrichment at protrusive edges of cells. Abi1 is not detected at the edges of cells, but appears in vesicle-like spots throughout the cytoplasm. (B) Cos7 cells were co-transfected with Myc-Scar1 SHD (i) and HA-HSPC300 (ii) shown as red and green respectively in a merged image (iii). Scar1-SHD and HSPC300 exhibit the same diffuse staining throughout the cytoplasm but also co-localize at protruding edges of cells. (C) Cos7 cells were co-transfected with Myc-Scar1 SHD (i) and HA-Abi1 (ii). (iii) Shows a merge image with myc-Scar1 SHD in red and HA-Abi1 in green. Scar1 SHD and Abi1 both colocalize to punctate spots similar to those seen for Abi1 alone. (D) Cos7 cells co-transfected with Myc-Scar2 SHD (i) and HA-HSPC300 (ii) shown as red and green respectively in a merge image (iii). Both Scar2 SHD and HSPC300 exhibit peri-nuclear staining and localization to protrusive edges of cells. (E) Cos7 cells co-transfected with Myc-Scar2 SHD (i) and HA-Abi1 (ii). (iii) Scar2 SHD (red) and Abi1 (green) show co-localization to cytoplasmic spots and the edges of cells. (F) Cos7 cells co-transfected with (i) Myc-Scar3 SHD (red) and (ii) HA-HSPC300 (green), in a merged image (iii). HSPC300 and Scar3 SHD show diffuse cytoplasmic staining with enrichment at protrusive edges of cells. (G) Cos7 cells co-transfected with (i) Myc-Scar3 SHD and (ii) HA-Abi1. (iii) In a merge image Abi1 (green) and Scar3 SHD (red) co-localize to punctate cytoplasmic spots and to the edges of lamellipodia. Scale bars in all panels are 20 μm. When expressed with Myc-Scar 1 SHD (Fig. 6Bi ), HA-HSPC300 (Fig. 6Bii ) co-localises (Shown as yellow in merge pictures [6Biii]) with enrichment in ruffles. Myc-Scar/WAVE 2 (Fig. 6Di ), and 3 (Fig 6Fi ) SHD also co-localises with HA-HSPC300 (Fig. 6Dii, 6Fii ) in the nucleus, cytoplasm, and particularly strongly at areas of ruffles, shown as yellow in merge images (Fig. 6Diii, 6Fiii ). All three human isoforms behave the same in the presence of HSPC300 consistent with a conservation of the interaction between Scar/WAVEs and HSPC300. HA-Abi1 (Fig. 6Cii ) strongly co localises with Myc-Scar 1 SHD (yellow, Fig 6Ciii ), in unidentified punctate spots (Fig 6Ci ). Myc-Scar 2 SHD (Fig. 6Ei ) shows a similar co localisation with Abi1 (Fig. 6Eii and 6iii ), whilst Myc-Scar 3 SHD (Fig. 6Gi ) shows partial co-localisation to Abi1 (Fig. 6Gii and 6iii ). Although Myc-Scar3 SHD (Fig. 6Gi ) co-localises to the HA-Abi1 spots (Fig. 6Giii ) to a lesser extent than Scar/WAVE1 and 2 SHD, leaving some diffuse staining in the cytoplasm and nucleus. Abi1 expression affects the localisation of all three human Scar/WAVE isoforms in Cos7 cells, providing further evidence for a role of the Scar/WAVE complex in regulation of the activity and localisation of Scar/WAVE proteins. The staining pattern seen for overexpressed HA-Abi1 (Fig. 6Aii, Cii, Eii and 6Gii ) does not strongly correlate with the cellular localisation of endogenous Abi1 (Fig. 5Cii ). Although the punctuate staining pattern seen in cells over-expressing Abi1 has been described as being characteristic of the reticulovesicular system (Ziemnicka-Kotula, 1998), it is also possible that these spots are protein aggregates induced by overexpression of Abi1. The relocalisation of Scar/WAVE1, 2 and 3 SHDs to these vesicles or aggregates may or may not be biologically important, but it does demonstrate the possibility that Abi1 can affect Scar/WAVE protein localisation in cells. Discussion We show that Abi1 and HSPC300 interact with three human Scar/WAVE isoforms, via the Scar Homology Domain. Conservation of Abi1 binding between all three Scar/WAVEs is indicative of the association of the "Scar/WAVE complex" with all three isoforms, as Abi1 seems to be the link between Scar/WAVE and NckAP1 and SRA-1/PIR121 in previous reports [ 30 , 31 ]. Association of the complex with all three Scar/WAVE proteins is a possibility, as all the complex members are widely expressed throughout the body and seem to be found in all tissues where a Scar/WAVE protein is expressed [ 26 , 60 ]. Little is known about the role of Scar/WAVE3 in cells. Expression of the mammalian Scar/WAVE3 gene is largely limited to the brain and lung with relatively little expression in other adult tissues [ 9 , 32 , 56 ]. The major suggestion has been a role in lamellipodial and filopodial protrusion indicated by localisation in neuronal growth cones and interaction with the Arp2/3 complex [ 9 , 57 ]. Association with a Abi1-NckAP1-SRA1 does indicate a further potential role for Scar/WAVE3 in lamellipodial protrusion given the requirement for these proteins in the ruffling process [ 31 , 36 ]. Scar/WAVE3 is also implicated as a potential tumour suppressor protein in some ganglioneuroblastomas when the gene is down-regulated [ 56 ]. The role of interaction of HSPC300 with the Scar/WAVE3 complex is unknown. Loss of function of the brk1 gene product, the Maize homologue of HSPC300, causes aberrations in cellular morphology and filamentous actin distribution of leaf epidermal cells [ 24 , 25 ]. This could indicate a role in regulating the localisation of actin polymerisation machinery, but this idea awaits further testing. HSPC300 is also implicated in the frequency of occurrence of renal cell carcinoma [ 26 ]. Association of all three Scar/WAVE isoforms with an Abi-NckAP1-SRA1 complex may show that there is some functional redundancy between Scar/WAVE family members but data from mice deficient in a Scar/WAVE isoform indicates that Scar/WAVE1 and 2 do not have completely overlapping functions [ 37 , 38 , 57 , 61 ]. The identification of specific binding partners for Scar/WAVE1, such as WRP, BAD or the RII subunit of PKA suggests a potential for differing roles between Scar/WAVE proteins [ 62 - 64 ]. The conservation of the Abi1-NckAP1-SRA1 complex between three Scar/WAVE isoforms suggests that regulation of individual Scar/WAVE proteins is likely to involve components that we have not yet studied [ 37 , 57 , 58 , 61 ]. Other proteins that associate with the complex may perform this regulation, or tissue specificity may occur between the complexes formed where differentially expressed related proteins, such as Abi1 or 2, bind to alternative Scar/WAVE isoforms with different affinities. Further study of the Scar/WAVE family and associated proteins will be required to fully understand the true roles and mechanisms of function of these complexes in cells. Conclusion We conclude that Scar/WAVE3 is likely to participate in similar signalling complexes to Scar/WAVE1 and 2 and that the differences between these Scar/WAVE proteins is likely to be at the level of tissue expression, differences in affinity for certain binding partners and possibly interaction with yet undiscovered binding partners. Methods Reagents and chemicals All chemicals were purchased from Sigma-Aldrich, UK unless otherwise specified. Antibodies were from the following sources: 9E10 anti-Myc monoclonal (Cancer Research UK), 12CA5 anti-HA monoclonal (Cancer Reaearch UK), anti-Myc polyclonal 1:500 anti-HA polyclonal (Santa Cruz), 1:500 anti-Scar/WAVE3 (Upstate), 1:100 anti-Abi1, or 1:100 anti-Scar/WAVE1 [ 65 ], anti-Abi1 (gift from Giorgio Scita, Instituto Europeo di Oncologia, Milan, Italy), HRP-conjugated secondary antibody anti-mouse or anti-rabbit (Jackson labs), anti-Arp3 (Sigma Israel), Goat anti-mouse or rabbit, FITC, TRITC, Alexa-488 or Alexa-546 Conjugates (Molecular Probes). The specificity of the anti-Scar/WAVE3 antibody for Scar/WAVE3 has been demonstrated by Oda and colleagues [ 47 ]. Vectors and cloning HSPC300 and Abi1 expression vectors were created using the Gateway Cloning system (Invitrogen). Open reading frames for HSPC300, and Abi1, derived from I.M.A.G.E. clones 4519512 (HSPC300) and 4158413 (Abi1) (UK-Human Genome Resource Centre, Babraham), were amplified by PCR and ligated into the pENTR-Topo entry vector (Invitrogen). N-terminally HA- and Myc- tagged HSPC300, and Abi1, were created in pRK5 DEST-Myc or pRK5 DEST-HA Gateway destination vectors by recombination from entry vectors. pRK5 DEST-Myc or pRK5 DEST-HA were created by ligation of the Gateway Vector Conversion System (Invitrogen) reading frame A cassette into pRK5-myc or pRK5-HA cut with SmaI. Scar/WAVE1, Scar/WAVE2, and Scar/WAVE3 full length and deletion constructs were created by PCR and cloned into pRK5-myc or pGEX4T-2. Scar/WAVE-1 constructs were derived from KIAA00429[ 7 ]; Scar/WAVE2 constructs were derived from pDSRed Scar2, a kind gift from Giorgio Scita (Innocenti et al., 2004); Scar/WAVE3 constructs were derived from pcDNA Scar3, a kind gift from John Scott, Portland, U.S.A. L61 Cdc42-pGex2T and IRSp53-pRK5-Myc were a gift from Alan Hall (LMCB, London) [ 54 ]. IRSp53 was subcloned into pGex4T-2 using BamH1 and EcoR1 restriction sites. For Yeast Two-Hybrid analyses, Abi1 and HSPC300 were fused in-frame to the C-terminus of the GAL4 DNA-binding domain in pYTH9. Myc-tagged Scar/WAVE1 deletion constructs have been previously described[ 7 ]. Scar1 deletion mutants (described in Additional file 1 ) were fused to the C-terminus of the GAL4 activation domain in pACT-II [ 66 ]. Yeast 2-hybrid analyses were performed as previously described [ 66 ]. Cell culture, transfection, and lysis Cos 7 and C2C12 cells were grown in DMEM plus 10% Foetal Calf Serum and antibiotics. Transfections were performed using Genejuice transfection reagent (Novagen) following the manufacturers instructions. Cell lysis was performed from confluent cells in a 1% Triton X-100 lysis buffer (1% Triton X-100, 50 mM Tris-HCl pH 7.5, 150 mM NaCl,, 1 mM EDTA, 10% Glycerol, 1 mM PMSF, and 1 μg/ml each of chymostatin, aprotinin, leupeptin, and pepstatin). Lysates were cleared by centrifugation and equalised to 1 mg/ml total protein concentration. Preparation of mouse brain extract Mouse brains were homogenised on ice in 2 μl of 1% Triton X-100 lysis buffer per mg of tissue using a fine gauge needle. Homogenates were centrifuged once for 10 minutes at 13,000 rpm in a microcentrifuge, then for 30 minutes at 100,000 g in a Beckmann TLA-100.2 rotor. Supernatants were equalised to 1 mg/ml total protein concentration with 1% Triton X-100 lysis buffer. Immunoprecipitations Lysates or brain extracts (500 μg of total protein) were incubated with 30 μl bed volume of pre-washed protein G beads (Cancer Research UK) for 1 hour, designated beads minus antibody controls. Beads were precipitated and supernatants removed. 5 μg of antibody monoclonal 9E10 anti-myc, 12CA5 anti-HA, Rabbit anti-Scar/WAVE3, or Mouse anti-Abi1 or an irrelevant control antibody (rabbit or mouse anti-myc) was added to the supernatant and incubated for 1 hour, before the addition of 30 μl bed volume of pre-washed protein G beads, designated beads plus antibody, for 1 hour. Beads were precipitated, and the supernatant removed. Beads were washed 3 times in 30-fold bed volume of lysis buffer. Beads were resuspended in an equal volume of 2× SDS-PAGE loading buffer [ 67 ]. 10 μl of beads sample and supernatant was analysed by SDS-PAGE and western blotting. GST-production and pull downs GST-fusion proteins were expressed in BL21 E.coli induced for 3 hours with 0.2 mM IPTG, or 15 hours at 25°C for GST-IRSp53. Cells were spun down and sonicated in 1 ml of 1% Triton PBS (1% Triton X-100, PBS pH 7.5, 1 mM PMSF, and 1 μg/ml each of chymostatin, leupeptin, aprotinin, and pepstatin) per 100 ml bacterial culture. GST fusion proteins were batch purified on 100 μl Glutathione-s-agarose (Sigma) per 100 ml culture and washed 5 times with 10-fold bed volume of sonication buffer. Beads were resuspended in an equal volume of 1% Triton lysis buffer. SDS-PAGE analysis and coomasie staining were used to qualify GST fusion proteins. 500 μg total protein of transfected Cos7 cell lysate was incubated with GST-fusion protein or GST alone for 1 hour. Beads were precipitated, lysates removed and the beads were washed 3 times in 1% Triton lysis buffer. Beads were resuspended in and equal volume of 2× SDS-PAGE loading buffer. Samples were analysed by SDS-PAGE and western blotting. SDS-PAGE and Western blotting Proteins were separated by SDS-PAGE and transferred onto nitrocellulose membrane by western blotting. Blots were saturated with 5% dried milk in PBS 0.2% Tween-20, probed with 1:500 9E10 anti-Myc monoclonal, 1:500 12CA5 anti-HA monoclonal, 1:500 anti-Myc polyclonal 1:500 anti-HA polyclonal, 1:500 anti-Scar/WAVE3, 1:100 anti-Abi1, or 1:100 anti-Scar/WAVE1 primary antibody in 2.5% BSA 0.2% Tween-20 PBS followed by 1:10'000 dilution of HRP-conjugated secondary antibody diluted in 0.2% Tween-20 PBS. Bound secondary antibody was visualised using SuperSignal (Pierce) according to the manufacturers instructions. Immunofluorescence microscopy Cells cultured on glass coverslips were fixed in 4% paraformaldehyde and permeabilized with 0.1% Triton PBS. Cells were stained with 1:200 Rb anti-HA polyclonal, 1:200 12CA5 anti-HA monoclonal, 1:200 9E10 anti-myc monoclonal, 1:100 anti-Scar\WAVE3, 1:200 monoclonal anti-Arp3, or 1:25 anti-Abi1 primary antibodies diluted in 1% BSA PBS. Secondary antibodies used were Goat anti-mouse or rabbit, FITC, TRITC, Alexa-488 or Alexa-546 Conjugates at 1:200 dilution in 1% BSA PBS. Filamentous actin was visualised using Alexa-546 conjugated phalloidin (Molecular Probes) diluted 1:500 in 1% BSA PBS. Slides were viewed using a BioRad MRC100 confocal laser scanning microscope. For endogenous Scar/WAVE3 labelling, it was important to test which cell lines expressed Scar/WAVE3. Among the following cell lines: NIH 3T3, N1E 115 neuroblastoma, N19 neuroblastoma, PC6 neuronal precursor, J774.A1 bone marrow macrophage, RAW 264.7 alveolar macrophage and C2C12 myoblast, we found that only C2C12 expressed endogenous Scar/WAVE3 by western blotting of whole cell lysates (M. Vartiainen, unpublished results). We thus used C2C12 for localisation of endogenous Scar/WAVE3 protein. Abbreviations GST glutathione-S-transferase, WH2 WASP-homology 2, A acidic sequence motif, SHD Scar homology domain, NckAP1 Nck-associated protein 1, HSPC300 haematopoeitc stem/progenitor cell protein, Abi 2 abelson tyrosine kinase interactor 2 Authors' contributions C.F.S. performed most of the experiments and contributed intellectually to the design and interpretations. C.F.S. also drafted the manuscript. T.H.M. performed the experiments shown in Figure 4 and contributed intellectually to the project as a whole. L.M.M. Contributed to the design and analysis of the experiments and data and edited the manuscript. Supplementary Material Additional File 1 Characterization of HSPC300 and Abi1 binding domain of Scar/WAVE1. ( A ) Schematic representation of the domain architecture of Scar/WAVE family proteins and the deletion constructs used in these experiments. Numbers indicate start or end point of Scar/WAVE1 deletion mutants in amino acid residue position. Start/End points for Scar/WAVE2 deletion mutants are FL = 498, delta A = 456, SPW = 455, SP = 435, SB = 245, SHD = 170, BPWA = 171, PWA = 246. Start/End points for Scar/WAVE3 deletion mutants are FL = 499, delta A = 488, SPW = 457, SP = 439, SB = 238, SHD = 171, BPWA = 172, PWA = 239. FL; Full length. SHD; Scar Homology Domain. B; Basic Rich Region. PRR; Proline rich region. W; Wiskott Homology 2 domain. C; Central or connecting region. A; Acidic Rich Region. ( B ) Co-immunoprecipitation of HA-Abi1 with Myc-Scar/WAVE deletion constructs. ( C ) Co-immunoprecipitation of HA-Abi1 or HA-HSPC300 respectively with Myc-Scar/WAVE deletion constructs. Protein G beads were used to immunoprecipitate protein from transiently transfected Cos 7 fibroblasts before (-Ab) and after addition (+Ab) of an anti-Myc (9E10) monoclonal antibody. Bead bound fractions and supernatants (Sup) were analysed by SDS-PAGE followed by immunoblotting. Blots were probed with an anti-HA polyclonal antibody for the presence of HA-Abi1 or HA-HSPC300 in immunoprecipitated complexes. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555569.xml
548289	In memoriam: Celso-Ramon Garcia, M.D. (1922–2004), reproductive medicine visionary	This article traces the career of Celso-Ramon Garcia (1922–2004), noted physician, educator, and internationally renowned pioneer in the field of reproductive endocrinology. His work helped to formulate oral contraceptives used by millions of women throughout the world. Garcia's research collaborators included Gregory Pincus and John Rock, who together finalized the landmark clinical data needed to secure initial FDA approval for "the pill" in 1960. In addition to Garcia's monumental work in contraceptive endocrinology, his scholarly interests encompassed physiology of the menopause, minimally invasive reproductive surgery, as well as psychological aspects of infertility. Closely identified with the University of Pennsylvania, Garcia was instrumental in establishing the first formal clinical program in reproductive biology and influenced countless young scientists whose training he supervised and mentored. His distinguished career was emblematic of the best of the medical profession, characterized by compassion, intellect, and a sincere desire to help others. Our manuscript outlines Garcia's wide range of interests, acknowledges his superior fund of knowledge, and honors his humanitarian spirit – all of which contributed to an impressive legacy of medical discoveries. The impact of Prof. Garcia's work will continue to be felt for many years.	Introduction Celso-Ramon Garcia (Figure 1 ), who departed this world in 2004 at the age of 82, was a remarkable human being who did much to shape the specialty of reproductive medicine as it is known today. His research yielded important discoveries that continue to influence the daily lives of millions of women and men around the globe. Reviewing his career and his accomplishments is like looking through a kaleidoscope – elements of greatness and the purest embodiment of the medical arts are brought together and with a small twist transformed into a new vision. Hard work and determination brought him from a relatively modest but academically-oriented Spanish immigrant family to medical training in community hospitals in Brooklyn, and then into the elite circles of East Coast academic medicine. Figure 1 Celso-Ramon Garcia, M.D. (1922–2004) Background Garcia was raised speaking Spanish, but his English diction and elocution were so honed by his teachers in grammar and high school that hardly a trace of accent was left to be detected by the Boston Brahmans he would later encounter. Along the way his interests took seemingly improbable turns, yet there was a clear rationale that was beyond the imagination of the average physician. Indeed, his work presaged several important developments in the field of reproductive medicine. Garcia began his career with an interest in obstetrics, but ended up conducting clinical research on contraception that on more than one occasion was nominated for the Nobel Prize in Medicine or Physiology. He was revered as a meticulous abdominal infertility surgeon, yet he embraced the fledgling field of endoscopy and endoscopic surgery. While being celebrated for his surgical technique, he promoted research into the psychological aspects of infertility. Towards the end of his career he became interested in sexuality in the menopause, bringing his career through the full cycle from obstetrics, to family planning, to infertility to reproductive aging. His accomplishments in each of these spheres were well recognized by his peers who chose him President of the American Association of Planned Parenthood Physicians, Chairman of the Board of Directors of Planned Parenthood of America, Founding President of the Society of Reproductive Surgeons, and President of the American Society of Reproductive Medicine and bestowed upon him the Scientific Leadership Award from the Global Alliance for Women's Health, a non-governmental organization of the United Nations. Early years Garcia's course to greatness was based on a diverse set of skills that he used creatively throughout his career. He majored in chemistry at Queens College where he revealed a remarkable aptitude for analysis of organic compounds. During his undergraduate studies, Garcia served as a teaching fellow in chemistry at New York University and authored two papers on qualitative and quantitative analysis. These are the earliest public manifestations of his passion for rigor and precision, which subsequently guided his clinical practice and his scholarship. After receiving his baccalaureate degree he entered medical school at the Long Island College of Medicine (now SUNY-Downstate) and graduated in 1945. He completed a rotating internship at the Norwegian Hospital in Brooklyn and then served two years in the U.S. Army Medical Corps in Fairbanks Alaska and then the Phoenixville General Army Hospital outside of Philadelphia. Garcia started a residency in pathology at Cumberland Hospital in Brooklyn where he was introduced to Sam Lubin, who became his friend and mentor. Lubin's work focused on uterine pathology and contractility and this sparked Garcia's interest in obstetrics and gynecology. With Lubin, Garcia would later publish papers on the effects of meperidine on labor. Further studies and research After a year of pathology residency and a year as a research fellow, Garcia transferred to the obstetrics and gynecology program at Cumberland Hospital. With a pathologist's keen eye and an appreciation of tissue biology/healing, Garcia was well-equipped to excel in the art of reproductive surgery. In the operating rooms at Cumberland Hospital Garcia met Shirley Stoddard, the newly appointed Director of Surgical Nursing; she was to become his wife and lifelong companion. After completing residency, Garcia explored openings in academia but few opportunities were available. One exception was an advertisement for assistant professor of obstetrics and gynecology at the newly established medical school at University of Puerto Rico, for which Garcia applied and was accepted in 1953. Garcia's research during residency, mentored by Lubin, had focused on obstetrical issues – something that he could have easily developed as a junior faculty member at the University of Puerto Rico. However, Garcia was about to embark on the first of several sharp turns that would lead him into seemingly improbable arenas. His arrival in Puerto Rico fortuitously coincided with the time for the first large scale clinical trials of the oral contraceptive pill. Within a year of Garcia's arrival, Gregory Pincus (director of the Worcester Foundation for Experimental Biology in Shrewsbury, Massachusetts) was in Puerto Rico looking for sites to conduct these trials. It was there that Pincus was first introduced to Garcia. Pincus had an uncanny ability to sense talent; he always surrounded himself with the brightest people. Pincus, undoubtedly impressed by Garcia's fund of knowledge in organic chemistry, training in pathology, and of course, the fact that he was bilingual in addition to his training in obstetrics and gynecology, promptly recruited Garcia to oversee his important research program in Puerto Rico. Pincus would subsequently introduce Garcia to John Rock, who arranged for Garcia to be named Sydney Graves Fellow in infertility and gynecology at the Free Hospital for Women in Brookline, Massachusetts. There, Garcia continued to work on oral contraceptives but also pursued clinical work in infertility. Although the Graves fellowship no longer exists, contemporary reproductive medicine was developed by its incumbents. Rock had an interesting manner of selecting these fellows. He usually interviewed candidates over breakfast in his hotel room at the annual meeting of the American College of Obstetricians and Gynecologists. He could identify the smart ones within a minute, and had little patience for individuals who were not well-spoken. The diction and elocution drilling Garcia had endured and his insistence on precision brought him through with flying colors. Garcia quickly became indispensable to Rock. Garcia moved to Boston, where he held assistant and instructor faculty posts at Harvard Medical School. During this time, Garcia also commuted to Puerto Rico to manage the oral contraceptive trials. Rock had a thriving infertility practice and Garcia did much of the surgery. Garcia's contributions to Rock's clinical practice were, however, underappreciated and when it came time for Garcia to be proposed for a full-time position at the Free Hospital for Women, the job did not materialize. This scenario is not unfamiliar to practitioners in our field, and it elicited the obvious reaction: Rock and Garcia pulled up stakes and took their hugely lucrative practice to the Faulkner Hospital in Boston. But after a short interval, the Rock/Garcia partnership was persuaded to return to the Free Hospital where an appropriate appointment was arranged for Garcia. Mrs. Stanley McCormick, a benefactor of remarkable vision who had provided initial funds to Pincus and colleagues at the Worcester Foundation to begin studies on oral contraceptive steroids, helped establish the Rock Reproduction Study Center at the Free Hospital campus, greatly facilitating research and training. The Worcester Program and Massachusetts General Hospital years By 1960, Pincus had recruited Garcia to the Worcester Foundation as a senior scientist and director of the training program in Physiology of Reproduction. Sponsored by grants from the Ford Foundation and National Institutes of Health, this was the first training program of its kind. The discipline of reproductive biology as we know it today had its roots in the Worcester program. Although Garcia was appointed chief of the infertility clinic at Massachusetts General Hospital in 1962, he continued to visit Puerto Rico to perform follow-up examinations on women who had participated in the oral contraceptive trials in gratitude for their role and with genuine concern for potential untoward long-term effects. That was a personal mission, not a mandated component of the study protocol. This post-trial surveillance helped identify rare side effects that ultimately led to product reformulation and lowering of the steroid content of oral contraceptives. Garcia subsequently carried out groundbreaking studies on patient acceptance of oral contraceptives, recognizing that future developments in family planning must address the social and demographic aspects, including cultural differences in preferences for methods. The technologies that gave men and women the ability to control their own reproduction through contraception and infertility treatments are arguably among the most important medical advances of the 20 th century. This is in good part Garcia's legacy [ 1 - 3 ]. While his career embraced the full spectrum of reproductive medicine and surgery, undoubtedly Garcia's most significant role was as co-developer of the first oral contraceptive approved by the U.S. Food and Drug Administration. Important medical advances like "the pill" or IVF do not come without controversy, societal baggage, and rhetoric. The pill's creators fully understood these implications, and knew such peripheral distractions would take away from the significance of their work. Consequently, there was no phone call from Sweden despite several nominations to the Nobel committee. But it is best to look back on the history of the development of the oral contraceptive with awe and wonder. Indeed, the commercial availability of oral contraceptives was also important because it established a formal approval and regulatory framework for pharmacologic agents of an entirely new class: drugs designed to be used by healthy people for long periods of time. It provided a clear understanding of how post-marketing surveillance can lead to product reformulations that reduce side effects without reducing efficacy. It yielded an understanding of how sociology and demography must be incorporated into the development of pharmaceuticals and the practice of medicine. Garcia the peerless surgeon Garcia maintained his connections with Rock and Rock's family of protégés, one of whom (L.M.) later lured him and other Rock/Pincus trainees, Edward Wallach and John V. Kelly, to the Department of Obstetrics and Gynecology at the University of Pennsylvania. Garcia spent the rest of his career at Penn developing the specialty of reproductive surgery, particularly microsurgical approaches. Here again came the totally un-expected. Garcia was at his best when confronted with the most challenging open abdominal procedures, which always ended with the pelvis in a pristine state, not a drop of blood, all surfaces meticulously re-peritonealized, no risk of inflammation or adhesion. Garcia's novel regimen of intraperitoneal corticoids and antihistamines (an early anti-adhesion therapy) was merely icing on the cake. The two-stage tuboplasty procedure was celebrated both locally and internationally. Yet this master of the open abdomen quickly recognized the potential of minimally invasive surgery, establishing one of the first endoscopic surgery programs in USA after a brief sabbatical with his colleague, Professor Kurt Semm in Kiel, Germany. During his career, Garcia authored numerous textbook chapters on surgery of the ovaries, tubes and uterus that guided a generation of gynecologists in reconstructive surgery. He was also an early advocate of laser and electrosurgical methods, as well as microsurgical techniques designed to minimize tissue trauma. When Garcia's residents and fellows under-performed, he could not mask his displeasure, yet when they excelled he could not hide his joy. His concern for the outcomes of a rapidly evolving medical specialty led him to develop one of the first electronic infertility data registries [ 4 ]. In 1968, Garcia made yet another change in tack, developing an interest in psychological aspects of infertility. While this may have puzzled colleagues who knew Garcia only by reputation as a surgeon, those of us who worked with him knew well his holistic commitment to medicine and his devotion – unprecedented even in the current era – to his patients and their families. He had a profound understanding of the human frailties that emerge during the rigors of infertility treatment. One of us (J.F.S.) assisted Garcia in the care of an attorney's wife who underwent a two-stage tuboplasty. The husband demanded to be present during all medical examinations of his wife, he tape recorded all sessions, and refused to pay for Garcia's services, despite the fact that she conceived within three months after the hoods were removed from her reconstructed tubes. Garcia did not press the attorney for payment – he knew that the treatment course was exceptionally arduous with considerable post-surgery discomfort. When this same lawyer called eight months later asking if Garcia would serve as an expert witness in a plaintiff's medical malpractice case, Garcia thanked him for his call and politely said his schedule would not permit it, never mentioning the unpaid bill. He forgave, but he did not forget. Final years and retirement Later in his career, Garcia and his colleagues undertook studies on the menopause that presaged the current interest in female sexual dysfunction. In retrospect, this was a logical extension of his long-standing interest in the social and psychological aspects first of contraception and family planning, and later of infertility. Garcia published on perimenopausal sexuality, and reported the first studies the relationship between sex steroid levels and sexual behavior. Garcia also recognized the need for multidisciplinary collaboration in the care of women, particularly in the post-reproductive years, and co-authored an early self-help manual for menopausal women [ 5 ]. The Women's Wellness Program he established at the Hospital of the University of Pennsylvania opened years ahead of similar facilities elsewhere. Garcia gracefully retired from practice to devote time to his wife and family. This life transition, seamlessly executed, surprised many who were familiar with Garcia's intensity. But there should have been no surprise. Garcia demanded the highest technical standards from residents and fellows in the operating room, the same standards that he held for himself. He would not carry on if he could not perform to those standards, and he would not neglect his wife who needed increasing medical attention. Epilogue Throughout his professional life, Celso-Ramon Garcia was a major force driving innovation in clinical care, and ultimately the evolution of the field of reproductive medicine. Let all who follow in the arena of reproductive endocrinology honor his contributions, appreciate the unique attributes that made him a great physician and educator, and gain insight from a truly remarkable career. Competing interests The author(s) declare that they have no competing interests. Authors' contributions JLS and LM contributed equally to this work and its revisions.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC548289.xml
517941	Cytotoxic effects of Gemcitabine-loaded liposomes in human anaplastic thyroid carcinoma cells	Background Identification of effective systemic antineoplastic drugs against anaplastic thyroid carcinomas has particularly important implications. In fact, the efficacy of the chemotherapeutic agents presently used in these tumours, is strongly limited by their low therapeutic index. Methods In this study gemcitabine was entrapped within a pegylated liposomal delivery system to improve the drug antitumoral activity, thus exploiting the possibility to reduce doses to be administered in cancer therapy. The cytotoxic effects of free or liposome-entrapped gemcitabine was evaluated against a human thyroid tumour cell line. ARO cells, derived from a thyroid anaplastic carcinoma, were exposed to different concentrations of the drug. Liposomes formulations were made up of 1,2-dipalmitoyl-sn-glycero-3-phosphocholine/cholesterol/1,2-distearoyl-sn-glycero-3-phosphoethanolamine-MPEG (8:3:1 molar ratio). Cell viability was assessed by both trypan bleu dye exclusion assay and fluorimetric analysis of cell DNA content. Results A cytotoxic effect of free gemcitabine was present only after 72 h incubation (ARO cell mortality increased of approximately 4 fold over control at 1 μM, 7 fold at 100 μM). When gemcitabine was encapsulated in liposomes, a significant effect was observed by using lower concentrations of the drug (increased cell mortality of 2.4 fold vs. control at 0.3 μM) and earlier exposure time (24 h). Conclusion These findings show that, in vitro against human thyroid cancer cells, the gemcitabine incorporation within liposomes enhances the drug cytotoxic effect with respect to free gemcitabine, thus suggesting a more effective drug uptake inside the cells. This may allow the use of new formulations with lower dosages (side effect free) for the treatment of anaplastic human thyroid tumours.	Background At present, the prognosis of anaplastic thyroid carcinomas is very poor [ 1 , 2 ]. Since they are usually unable to concentrate radioiodine, the therapeutical approach is based on combination of aggressive surgery, external beam radiations and chemotherapy. Only rarely, however, this treatment results effective especially for treating metastatic disease [ 1 , 2 ]. A major limit of the chemotherapeutic agents presently used in these tumours, including doxorubicin, paclitaxel and various drug combinations [ 1 , 2 ], is represented by their low therapeutic index. Thus, the identification of systemic antineoplastic drugs effective against these carcinomas has particularly relevant implications. Gemcitabine is a new fluorinated nucleoside analogue provided with a potent anti-tumour activity, tested in vitro and in vivo against a variety of solid malignancies [ 3 - 6 ]. In addition, gemcitabine is a potent radiosensitizing agent [ 7 ]. In a preclinical study, gemcitabine showed a marked cytotoxic activity against poorly differentiated human thyroid carcinoma cell lines [ 8 ]. In contrast, no appreciable response was observed in four patients with aggressive thyroid anaplastic carcinoma treated with combination of gemcitabine and vinorelbine [ 9 ]. Since hematological and other toxicities have been reported when effective anti-tumour doses of gemcitabine were tested [ 10 ], the aim of this study was to verify the possibility of maintaining the drug cytotoxicity at lower doses, by using a liposomal drug delivery system. Our data demonstrate that incorporation of gemcitabine in liposomes enhances the cytotoxic effect of the drug against a well-established human thyroid cancer cell line, as compared to the free drug. Methods Chemicals Gemcitabine (2,2 1 -difluorodeoxycytidine) hydrochloride (HPLC purity >99%) was a kind gift of Eli-Lilly Italia S.p.A. (Sesto Fiorentino, Firenze, Italy) and was used without further purification. Cholesterol was obtained from Sigma Chemicals Co. (St. Louis, USA). 1,2-dipalmitoyl-sn-glycero-3-phosphocholine monohydrate and N-(carbonyl-methoxypolyethyleneglycol-2000)-1,2-distearoyl-sn-glycero-3-phosphoethanolamine sodium salt (MPEG-2000-DSPE) were purchased from Genzyme products (Suffolk, United Kingdom). Double-distilled pyrogen-free water from Sifra S.p.A. (Verona, Italy) was used. Sterile saline was a product of Frekenius Kabi Potenza S.r.l. (Verona, Italy). All other materials and solvents were of analytical grade (Carlo Erba, Milan, Italy). Liposome preparation Liposome formulations were made up of DPPC/Chol/MPEG-2000-DSPE (8:3:1 molar ratio). To maximize the entrapment efficiency of a hydrophilic molecule such as gemcitabine, various liposome preparation procedures [ 11 , 12 ] were applied together. Namely, liposome colloidal suspensions were prepared by dissolving the lipid mixture (40 mg) in chloroform-methanol (3:1 v/v). Organic solvent was removed by a rotavapor thus allowing the formation of a thin lipid film on the inner surface of a pyrex glass vial. To remove any trace of organic solvent, lipid films were stored overnight in a Büchi T-50 under high vacuum at 30°C in the dark. Lipid films were hydrated with a solution (2 ml) of 250 mM ammonium sulphate by vortexing at 45°C for 15 min. The resulting colloidal suspension of multilamellar vesicles was submitted to ten cycles of freeze in liquid nitrogen and thaw in warm (35°C) water in order to allow the homogenous distribution of the ionic species. The entrapped ammonium sulphate solution was removed by centrifugation at 14000 rpm (IEC, International Equipment Company, mod MP4R, equipped with a mod. 851 rotor) for 1 h at a temperature of 4°C. The pellet was re-suspended in 400 μl of a 1 mM gemcitabine aqueous solution and stored at room temperature for 3 h. Double-distilled pyrogen-free water (1.6 ml) was added to the liposome suspension and then a freeze-drying procedure was carried out (Edwards freeze-dryer Modulayo). Then, the non-incorporated gemcitabine was removed and the freeze-dried liposomes were re-suspended just before the experiments with 2 ml of the culture medium. Scalar dilutions were prepared in the same medium to obtain the final concentrations used in the experiments. Liposome loading capacity The loading capacity of liposomes was evaluated after removing the free gemcitabine (centrifugation at 14000 × g for 1 h at 4°C) from the vesicular colloidal suspension coming from the re-suspension of freeze-dried liposomes. The amount of gemcitabine in the supernatant was spectrophotometrically determined at 268.8 nm (Shimadzu UV-1601). UV calibration straight-line ( y = 6.958 × 10 -3 + 0.3971 x , where y is the UV adsorbance and x is the drug concentration) presented a r 2 value of 0.9993. The amount of gemcitabine entrapped in liposomes was determined as a difference between the amount of drug added during liposome preparation and the amount of the entrapped drug present in the supernatant. The encapsulation efficiency was expressed as percentage of the total amount of gemcitabine that became entrapped. Cell cultures and cell viability The ARO cells, from a human anaplastic thyroid carcinoma, were grown as previously described [ 13 ]. All experiments were performed in 12-well culture dishes, when cells reached approximately 50% confluence. Cell viability was evaluated by trypan bleu dye exclusion assay. Briefly, after incubation with gemcitabine or liposomes containing the drug at different doses, cells were trypsinized and the pellet re-suspended in a 0.4% trypan bleu buffer and counted in the hemocytometric chamber. Cell mortality was calculated as the percentage of stained cells over the total and expressed as the ratio between treated and untreated (control) cells. DNA cell content was assayed by using a fluorimetric DNA assay kit (Bio-rad laboratories, Segrate, Milano, Italy). Results We first analysed the effects of different concentrations of gemcitabine on the viability of ARO thyroid cancer cells. As shown in Figure 1 the cytotoxic effect was observed only after 48 or 72 h incubation of the drug. After 72 h, the cell mortality increased 4.6 and 7.9 fold over control using 1 and 100 μM of gemcitabine. In accordance with previous reports [ 8 ], no significant effect was visible after 24 h exposure (Fig. 1 ). In order to improve the drug entry into the cells, gemcitabine entrapment in a liposome capsule was next performed, using a combination of various liposome preparation procedures. In particular, a pH gradient method was used [ 14 ]. Namely, the presence of ammonium sulphate in the internal compartments of liposomes provide an acidic environment that elicit the protonation of gemcitabine in order to drastically reduce the drug back-diffusion (leakage) from liposomes (Figure 2 ). The combination of various liposome preparation procedures together with the application of a pH gradient provided an encapsulation efficiency of ~90%. The gemcitabine-loaded liposomes were then tested for their cytotoxic activity against ARO cell, and the results compared with those of gemcitabine alone. As shown in Figure 3 , a cytotoxic effect appeared after 24 h incubation, with an initial increased cell mortality at 0.3 μM (2.4 fold over control), and a maximum at 10 μM (death of almost all the cells) (Figure 3 ). Similar results were obtained by analysing the DNA content of ARO cells treated with free or liposome-encapsulated gemcitabine (data not shown). Discussion Despite an intense application of aggressive multimodal therapeutic regimens, no standardized protocol has shown success in the treatment of anaplastic thyroid carcinoma [ 2 ]. A critical need for new approaches is therefore vital for the systemic therapy of metastatic anaplastic thyroid carcinoma. At present one promising option is provided by the gene therapy. Studies are currently in progress attempting to: a. reintroducing iodine uptake by viral-driven Iodide/sodium symporter gene expression [ 15 , 16 ]; b, inducing tumour cell death by expressing p53 or by suicide gene prodrug systems [ 17 , 18 ]. An alternative approach is represented by increasing the effectiveness of chemotherapy, using novel agents, combination therapy or enhancing drug delivery inside the tumour cell. The latter option has been taken advantages in the progress related to liposome research. Traditional dosage forms of anticancer drugs have the limitation of many potential obstacles (barriers) before they reach their target site, one of these is the large volume of distribution after i.v. administration. This situation elicits a low therapeutic index and a high toxicity in healthy tissues [ 19 ]. After encapsulation into liposomes similar to those used in our investigation the volume of distribution of a drug is reduced, and its concentration in the tumour tissue is increased [ 20 ]. In fact, liposomes can protect drugs from metabolic inactivation, and, due to size limitation, they are not able to transport encapsulated molecules across healthy endothelium, thus avoiding anticancer drug accumulation in a large extent in healthy tissues and hence a decrease of side-effects [ 21 ]. On the other hand, the increased permeability of the tumour endothelium allows the extravasation of liposomes, with a consequent accumulation of anticancer drugs in the site of action [ 22 ]. Furthermore, it should be considered that liposomes are non-toxic being mostly constituted by naturally occurring lipids. In the last years, gemcitabine has emerged as a very potent anti-tumour drug and it is currently used, alone or in combination, in the treatment of patients with different malignancies, including ovarian, pancreatic, non-small cell lung and other cancers [ 3 - 6 ]. Despite its promising effectiveness derived from a preclinical study [ 8 ], no appreciable response was observed in a clinical study including four patients with aggressive thyroid anaplastic carcinoma treated with combination of gemcitabine and vinorelbine [ 9 ]. Although low, compared to other anti-neoplastic drugs, hematological and other toxicities appeared when effective anti-tumour doses of gemcitabine were tested. In this study, we prepared a liposome formulation with a high entrapment efficiency of gemcitabine and tested its efficacy against a well established human poorly differentiated thyroid carcinoma cell line. Our data demonstrate that, in ARO cells, incorporation of gemcitabine in liposomes enhances the cytotoxic effect of the drug as compared to free gemcitabine, suggesting a more effective uptake of the drug inside the cells. This activity was demonstrated at concentrations even lower than serum levels obtained in clinical trials. In our knowledge, no data are available in the literature concerning the relationship between gemcitabine uptake and its cytotoxicity, since neither liposomes or other carrier system have been tested against tumour cells. Our preliminary experiments (unpublished observations) show that a similar effect (improved uptake and cytotoxicity at lower doses) may be obtained also in colon cancer cells. It is note worthing that the various component used for the preparation of the liposomal delivery device were already approved by regulatory offices to be used in the preparation of liposome-based formulations, i.e. Doxil ® . Conclusion The clinical predictive value of the in vitro cell line preclinical cancer model is well established [ 23 ] and the ARO cells are one of the cell model of anaplastic thyroid cancer more exploited both for investigating the mechanism of tumour progression and for testing new molecules with antiproliferative effect. However, these promising observations need to be confirmed in studies using in vivo xenograft cancer models, prior to propose the use of such preparations at lower (presumably side effect free) doses in a clinical trial of human thyroid tumours. Competing interests (medicine) None declared. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517941.xml
545949	A novel method for prokaryotic promoter prediction based on DNA stability	Background In the post-genomic era, correct gene prediction has become one of the biggest challenges in genome annotation. Improved promoter prediction methods can be one step towards developing more reliable ab initio gene prediction methods. This work presents a novel prokaryotic promoter prediction method based on DNA stability. Results The promoter region is less stable and hence more prone to melting as compared to other genomic regions. Our analysis shows that a method of promoter prediction based on the differences in the stability of DNA sequences in the promoter and non-promoter region works much better compared to existing prokaryotic promoter prediction programs, which are based on sequence motif searches. At present the method works optimally for genomes such as that of Escherichia coli , which have near 50 % G+C composition and also performs satisfactorily in case of other prokaryotic promoters. Conclusions Our analysis clearly shows that the change in stability of DNA seems to provide a much better clue than usual sequence motifs, such as Pribnow box and -35 sequence, for differentiating promoter region from non-promoter regions. To a certain extent, it is more general and is likely to be applicable across organisms. Hence incorporation of such features in addition to the signature motifs can greatly improve the presently available promoter prediction programs.	Background Accumulation of a huge amount of genome sequence data in recent years and the task of extracting useful information from it, has given rise to many new challenges. One of the biggest challenges is the task of gene prediction and to fulfil this need, several gene prediction programs have been developed (For reviews see [ 1 - 5 ]). Most of these prediction programs require training based on prior knowledge of sequence features such as codon bias, which in turn are organism specific. In such cases, lack of large enough samples of known genes, as typically seen in a newly sequenced genome, can lead to sub optimal predictions. On the other hand, some gene prediction methods are based on the homology between two or more genomes but these methods are not of much help for gene prediction in case of genomes with no homologues. In addition, most of the gene prediction programs concentrate on the protein-coding regions and RNA genes, that can make up to 5 % of total protein coding genes, are neglected. Hence it is important to design ab initio gene prediction programs. One of the important steps towards ab initio gene prediction is to develop better promoter and TSS (transcription start site) prediction methods. Although reasonable progress has been achieved in the prediction of coding region, the promoter prediction methods are still far from being accurate [ 6 - 9 ] and there are some very obvious reasons for these inaccuracies. One of the major difficulties is that the regulatory sequence elements in promoters are short and not fully conserved in the sequence; hence there is a high probability of finding similar sequence elements elsewhere in genomes, outside the promoter regions. This is the reason why most of the promoter prediction algorithms, which are based on finding these regulatory sequence elements, end up predicting a lot of false positives. Thus it is likely that incorporation of additional characteristics, which are unique to the promoter region, will help in improving the currently available promoter prediction methods. In our earlier analysis, we observed that in case of bacteria as well as in eukaryotes, various properties of the region immediately upstream of TSS differ from that of downstream region [ 10 ]. There are differences in sequence composition as well as in different sequence dependent properties such as stability, bendability and curvature. The upstream region is less stable, more rigid and more curved than downstream region. Some of these observations are supported by other studies carried out independently on genomic sequences [ 9 , 11 - 17 ]. Among all types of promoters, the most prominent feature is the difference in DNA duplex stabilities of the upstream and downstream regions. Here, we propose a prokaryotic promoter prediction method, which is based on the stability differences between promoter and non-promoter regions. Results and discussion Lower stability of promoter regions in bacterial sequences It is well known that the stability of a DNA fragment is a sequence dependent property and depends primarily on the sum of the interactions between the constituent dinucleotides. The overall stability for an oligonucleotide can thus be predicted from its sequence, if one knows the relative contribution of each nearest neighbour interaction in the DNA [ 18 ]. The average stability profiles for three sets of bacterial promoter sequences calculated (using 15 nt moving window) based on this principle is shown in Figure 1 . It is interesting that the promoters from diverse bacteria, which have quite different genome composition (A+T composition: E. coli 0.49, B. subtilis 0.56 and C. glutamicum 0.46), show strikingly similar features. Promoters from all the three bacteria show low stability peak around the -10 region. The second prominent feature in the free energy profiles of all the three bacteria is the difference in stabilities of the upstream and downstream regions. In all the three groups of promoter sequences, the average stability of upstream region is lower than the average stability of downstream region. But the three sets of promoter sequences differ in their basal energy level, which seems to be dependent on the nucleotide composition of the bacteria. Detailed analysis of E. coli promoter sequences In order to get a better insight into the stability feature, we carried out a detailed analysis of E. coli promoter sequences. Our statistical analysis using "Wilcoxon signed test for equality of medians" (see METHODS) shows that the free energy distribution corresponding to a fragment extending from position -148 to 51 in the E. coli sequences is appreciably different from the energy distribution calculated in randomly selected windows, at a significance level as high as 0.0001. A comparison of free energy distribution at position -20 (corresponding to the promoter region) with distributions at positions -200 (corresponding to the region upstream of promoter region) and +200 (corresponding to the coding region) is shown in Figure 2 . It is clearly seen that the region immediately upstream of TSS is much less stable than the other two regions. The average free energy at -20 position is -17.48 kcal/mol while average free energies at the -200 and +200 positions are -19.42 kcal and -20.19 kcal/mol respectively. The Kolmogorov-Smirnov test also confirms that the free energy distribution at position -20 significantly differs from that at -200 and +200 positions at a very high significance level (alpha = 10 -10 ). Details of methodology This difference in free energy and the stability of promoter regions as compared to that of coding and other non-coding regions can be used to search for the promoters. Based on this consideration, a new scoring function D(n) is defined, which will look for differences in free energy of the neighbouring regions of position n: D(n) = E1(n) - E2(n) where, Thus, E1(n) and E2(n) represent the free energy (see METHODS) average in the 50 nt region starting from nucleotide n and neighbouring 100 nt region starting from nucleotide n+99, respectively. The E1 value represents the basal energy level, which is characteristic of the given bacterial genome (e.g. in this case E. coli ) and the D value represents the free energy difference in the two neighbouring regions. A stretch of DNA is assigned as promoter only if the average free energy of that 50 nt region (E1) and difference in free energy as compared to its neighbouring region (D) is greater than the chosen cut-offs. The protocol followed to calculate the true and false positives and hence sensitivity and precision is presented in the form of a flowchart in Figure 3 . Identical sensitivity values can be achieved using different combinations of D and E1 cut-off values, which is obvious from the contour plot shown in Figure 4A . Similarly, different combinations of D and E1 cut-offs can lead to similar precisions (Figure 4B ). But we observe that the use of different D and E1 cut-offs, corresponding to a given sensitivity level, results in a wide range of precisions (Figure 5 ). Hence, in order to attain a desired level of sensitivity the D and E1 cut-off values are chosen such that the number of false positives is minimum and the precision is maximum. Initially, we divided the E. coli sequence data into two sets. The E1 and D cut-off values corresponding to different sensitivity levels were obtained for 100 randomly selected sequences (1 st set). These cut-off values were then applied to a second set consisting of remaining 127 sequences. The sensitivity and precision values calculated for the first and second set match very well. We also found that very similar results can be obtained when we use the whole dataset (Figure 6 ). Hence, we present the results for the whole dataset rather than separately for two sets. The D and E1 cut-offs and the number of false positives corresponding to different levels of sensitivity are given in Table 1 . To confirm the validity of our choice, we used another set of 1000 nt long sequences extracted from the centre of the ORFs, which were more than 2000 nt long. The results corresponding to this set of control fragments are also given in Table 1 and show very few false positives. In principle, D can also be calculated using equal sized windows, i.e. 50 nt, for both E1 and E2 instead of a 50 nt window for E1 and a 100 nt window for E2. However, our calculations show that use of equal sized windows, for E1 as well as E2 calculations, results in a slightly lesser precision than when 100 nt window is used for E2 calculations (Figure 7 ). Hence, in our promoter predictions, we chose a 100 nt window for E2 calculations. Comparison with other promoter prediction programs A large number of promoter prediction programs have been developed for eukaryotic sequences and are easily accessible, while NNPP [ 19 , 20 ] is the only available prokaryotic promoter prediction program. It is a neural network based method where prediction for each sequence element constituting promoter sequence is combined in time-delay neural networks for a complete promoter site prediction. Some other prokaryotic promoter prediction methods are based on weight matrix pattern searches [ 21 - 24 ]. One of the representative weight matrix method, proposed by Staden [ 21 ], uses three weight matrices corresponding to the -35 sequence, the -10 sequence and the transcription start site. It also takes into account the spacing between the -35 and -10 motifs, as well as the distance between the -10 motif and the transcription start site. A brief comparison of the results obtained by our method and the other two methods (Staden method and NNPP program) is given in Table 2 . It can be clearly seen from Table 2 that for similar sensitivity, our program gives much better accuracy than the other two programs. It is pertinent to mention here that our method differs from the other two methods in one major respect, namely our method tries to find a promoter region while the other two programs try to pinpoint the transcription start site. It may be argued that the lesser number of false positives in our prediction method, as compared to the other two algorithms, may be due to this difference. But even after taking this difference into consideration, the number of false positives predicted by our protocol turns out to be smaller than those predicted by the other two methods. For example, Figure 8 represents the case of argI and argF genes, where the NNPP program predicts a few extra TSS as compared to our method which correctly picks up a region in the vicinity of TSS. A combination of both the methods can therefore help in reducing the false predictions in the upstream and downstream regions. In principle, by restricting the pattern recognition using NNPP and Staden's methods only to the promoter region located initially with the help of our method, one can reduce the number of false positives. This composite approach will also help in pinpointing the TSS, which is not possible by use of our method alone. But at the same time it should be noted that both types of predictions fail to identify some of the promoters (Figure 8 ), e.g. for csiE gene, our program could correctly predict the promoter region but the NNPP program could not locate it. On the other hand, our program failed to find the promoter region for gyrA gene while NNPP could correctly position it. And in case of ilvA gene both the programs did not succeed in identifying the promoter region. Very recently a study on improvement of NNPP prediction (TLS-NNPP), by combining this method with additional information such as distance between TSS and translation start site (TLS), has been published [ 25 ]. With the use of additional information regarding TLS, Burden et al. could significantly increase the precision of NNPP. The TLS-NNPP method was tested on 510 E. coli sequences of length 500 bp. For comparable sensitivity levels, the precision achieved by TLS-NNPP was 0.188 (sensitivity = 0.452) as compared to 0.109 precision (sensitivity = 0.443) achieved by NNPP. It can be seen that, for similar sensitivity levels, the precision achieved by our method (~0.7) is higher as compared to both TLS-NNPP and NNPP (Figure- 9 ). Presence of high densities of promoter like signals in the upstream region of TSS may be one of the reasons why pattern matching programs result in low level of precision. This has been shown recently by a systematic analysis of sigma70 promoters from E. coli [ 24 ]. In this study a number of weight matrices were generated by analysis of 599 experimentally verified promoters and these were tested on the 250 bp region upstream of gene start site. It was found that each 250 bp region on an average has 38 promoter-like signals. The study also presented a more rigorous patter searching method for locating promoters. With the use of this function the authors reach a sensitivity values of 0.86 but the corresponding precision achieved is only ~0.2. In case of our method, for a sensitivity of 0.9 we obtained a precision of 0.35 (as shown in Figure - 9 ). Recently Bockhorst et al. [ 26 ] proposed a very accurate method for predicting operons, promoters and terminators in E. coli . This method is based on sequence as well as expression data, but requires prior knowledge of coordinates of every ORF in the genome. We would like to emphasize here that our method is different from other methods in that it is independent of any such prior knowledge about the test gene or the organism and hence holds promise as being useful for promoter prediction in a newly sequenced genome. The eukaryotic promoter prediction method proposed by Ohler et al. [ 27 ] is also worth mentioning here. Ohler et al. showed that a 30 % reduction of false positives can be achieved by use of physical properties, such as DNA bendability, in addition to other sequence properties of promoters. Interestingly, our method which also uses a physical property gives much smaller number of false positives as compared to Ohler et al. 's method. (For similar sensitivity, number of false predictions in case of Ohler et al. 's method are 1/4740 nt while in case of our method these are 1/8407 nt). Another vertebrate promoter prediction program, 'Promfind' [ 28 ] identifies differences in hexanucleotide frequencies of promoter and coding region and is algorithmically quite similar to our method. But Promfind differs from our method in two important aspects. First, the Promfind program is developed mainly for vertebrate promoters and second, it assumes that in a given sequence, a promoter is always present and merely predicts its location. This need not necessarily be the case, as some of the sequences may not have any promoter at all. Our program differs from Promfind in that a promoter is predicted only when the sequence satisfies certain criteria and hence is much more appropriate for carrying out genome scale analysis. Promoter predictions in case of RNA genes In addition to protein coding genes there are genes present for the non-coding RNAs (ncRNAs), which play structural, regulatory and catalytic roles. It is a difficult task to find out ncRNA genes in a genome because unlike protein coding regions they lack open reading frames and also they are generally smaller in size. In addition, it is also difficult to do a homology sequence search as only the structure of ncRNA is conserved and not the sequence. There are around 156 E. coli RNA genes reported on the NCBI site [ 29 ] and in addition many more small RNA genes are known to exist. Argaman et al. [ 30 ] recently identified 14 novel sRNA genes by applying a heuristic approach to search for transcriptional signals. We have checked the performance of our algorithm with respect to the 42 RNA transcription units (TUs) reported in Ecocyc database. Our method could pick up around 57 % RNA TUs, at a cut-off corresponding to 60 % sensitivity. The program works much better in case of rRNA operons than tRNA transcription units. We could correctly pick up promoter regions in 6 out of 7 rRNA transcription units, 17 out of 33 tRNA TUs and 1 out of the 2 remaining RNA types. Promoter prediction in Bacillus subtilis and Corynebacterium glutamicum Finally, it is very important to see whether the method works equally well for other organisms which have genome compositions substantially different from that of Escherichia coli . Hence, we also tested our method using the promoter sequences from 1) the A+T-rich bacteria, Bacillus subtilis and 2) a G+C rich bacteria such as Corynebacterium glutamicum . Figure 9 gives a summary of the predictions in case of bacillus and corynebacterium promoters, along with those of Escherichia coli . It can be clearly seen that, at present our method performs optimally for the Escherichia coli promoters and also performs quite well in case of Bacillus subtilis . The prediction accuracy in case of Corynebacterium glutamicum promoters is not as good as that for the other two classes of promoters. However, it should be noted that the number of experimentally determined Corynebacterium promoters is much smaller as compared to other two bacteria and a larger dataset is required to arrive at any firm conclusion. Conclusions It has often been suggested that use of certain properties of promoters, other than just the sequence motifs, which can distinguish promoters from other genomic regions, could significantly improve the gene prediction methods. Although the lower stability of promoter regions as compared to non-promoter regions has been reported previously, this observation was not incorporated into a promoter prediction program. We have been able to successfully use the differential stability of promoter sequences to predict promoter regions. Our method performs better as compared to currently available prokaryotic prediction methods and is also moderately successful in predicting RNA and bacillus promoter regions. The method certainly needs to be further improved to reduce the number of predicted false positives. This can be achieved by combining the approach presented here, with the earlier reported sequence analysis methods. Such a composite method will also help in pinpointing the TSS within the promoter region identified by our method. Methods Promoter sequence sets All the promoter sequences used in this study are 1000 nt long, starting 500 nt upstream (position -500) and extending up to 500 nt downstream (position +500) of the TSS. In order to avoid having multiple TSS in a given 1000 nt sequence, we have excluded all the transcription start sites which are less than 500 nt apart. Our promoter set has 227 E. coli promoters, 89 B. subtilis promoters and 28 C. glutamicum promoters. a) Escherichia coli promoter sequences We tested our algorithm using the Escherichia coli promoter sequences, which were taken from the PromEC dataset [ 31 ]. The PromEC dataset provides a compilation of 471 experimentally identified transcriptional start sites. As mentioned above, after excluding all the transcription start sites which are less than 500 nt apart, the dataset contains 227 promoters. With the help of TSS information, promoter sequences were extracted from Escherichia coli genome sequence (NCBI accession no: NC_000913). b) Bacillus subtilis promoter sequences The transcription start sites for Bacillus subtilis promoters were obtained from the DBTBS database [ 32 ]. The required length sequences around transcription start sites were extracted from the Bacillus genome sequence (NCBI accession no: NC_000964). c) Corynebacterium glutamicum promoter sequences Analysis of Corynebacterium glutamicum promoters is carried out on a set of promoters compiled by Pàtek et al. [ 33 ] based on experimentally determined transcription sites. d) RNA promoter sequences The transcription start positions of RNA transcription units are obtained from the ecocyc dataset. In this set, both computer predicted as well as experimentally determined transcription start sites, are included. In total, we have 7 rRNA TUs, 33 tRNA TUs and 2 TUs of other RNAs. Free energy calculation The stability of DNA molecule can be expressed in terms of free energy. The standard free energy change (ΔG o 37 ) corresponding to the melting transition of an 'n' nucleotides (or 'n-1' dinucleotides) long DNA molecule, from double strand to single strand is calculated as follows: where, ΔG o ini is the initiation free energy for dinucleotide of type ij. ΔG o sym equals +0.43 kcal/mol and is applicable if the duplex is self-complementary. ΔG o i,j is the standard free energy change for the dinucleotide of type ij. Since our analysis involves long continuous stretches of DNA molecules, in our calculation we did not consider the two terms, ΔG o ini and ΔG o sym , which are more relevant for oligonucleotides. In the present calculation, each promoter sequence is divided into overlapping windows of 15 base pairs (or 14 dinucleotide steps). For each window, the free energy is calculated as given in the above equation and the energy value is assigned to the first base pair in the window. The energy values corresponding to the 10 unique dinucleotide sequences are taken from the unified parameters proposed recently [ 34 , 35 ]. Statistical tests a) Wilcoxon signed test for equality of medians The free energy distribution at a given position, in the 1000 nt E. coli sequences ranging from -500 to +500, was compared to the distribution in a randomly selected set. For this comparison, we followed a similar procedure as adopted by Margalit et al. [ 11 ]. The random set was chosen such that an energy value per sequence was selected arbitrarily, independent of its position in the sequence. The comparison between the energy distributions was carried out using Wilcoxon signed test for equality of medians. This is a nonparametric test, which is used to test whether the two samples have equal medians or not. b) Two-sample Kolmogorov-Smirnov test We compared the free energy distribution at position -20 (with respect to TSS) with the distributions at the positions -200 and +200 using Kolmogorov-Smirnov two sample test [ 36 ]. All the calculations related to the statistical tests were carried out using MATLAB 6.0 ® . Implementation and scoring of NNPP and Staden's method The promoter predictions were also carried out using two other methods viz. NNPP and Staden's method. NNPP program is available at [ 20 ]. All the NNPP predictions were carried out at a score cut-off 0.80. The implementation of Staden's method was carried out as described in [ 21 , 37 ]. The weight matrix search was carried out with the help of PATSER program [ 38 ]. In case of NNPP as well as Staden's method, the true and false positives were scored as in case of our method (Figure 3 ), with a prediction in -150 to 50 region being considered as a true prediction. Sensitivity and precision The sensitivity and precision for the predictions are calculated using the following formulae: Authors' contributions AK performed the analysis, evaluated the results, and drafted the manuscript. MB suggested the problem, helped with evaluation of the results and the manuscript, also provided mentorship. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545949.xml
545785	Consensus clustering and functional interpretation of gene-expression data	Consensus clustering, a new method for analyzing microarray data that takes a consensus set of clusters from various algorithms, is shown to perform better than individual methods alone.	Background There are many practical applications that involve the grouping of a set of objects into a number of mutually exclusive subsets. Methods to achieve the partitioning of objects related by correlation or distance metrics are collectively known as clustering algorithms. Any algorithm that applies a global search for optimal clusters in a given dataset will run in exponential time to the size of problem space, and therefore heuristics are normally required to cope with most real-world clustering problems. This is especially true in microarray analysis, where gene-expression data can contain many thousands of variables. The ability to divide data into groups of genes sharing patterns of coexpression allows more detailed biological insights into global regulation of gene expression and cellular function. Many different heuristic algorithms are available for clustering. Representative statistical methods include k-means, hierarchical clustering (HC) and partitioning around medoids (PAM) [ 1 - 3 ]. Most algorithms make use of a starting allocation of variables based, for example, on random points in the data space or on the most correlated variables, and which therefore contain an inherent bias in their search space. These methods are also prone to becoming stuck in local maxima during the search. Nevertheless, they have been used for partitioning gene-expression data with notable success [ 4 , 5 ]. Artificial Intelligence (AI) techniques such as genetic algorithms, neural networks and simulated annealing (SA) [ 6 ] have also been used to solve the grouping problem, resulting in more general partitioning methods that can be applied to clustering [ 7 , 8 ]. In addition, other clustering methods developed within the bioinformatics community, such as the cluster affinity search technique (CAST), have been applied to gene-expression data analysis [ 9 ]. Importantly, all of these methods aim to overcome the biases and local maxima involved during a search but to do this requires fine-tuning of parameters. Recently, a number of studies have attempted to compare and validate cluster method consistency. Cluster validation can be split into two main procedures: internal validation, involving the use of information contained within the given dataset to assess the validity of the clusters; or external validation, based on assessing cluster results relative to another data source, for example, gene function annotation. Internal validation methods include comparing a number of clustering algorithms based upon a figure of merit (FOM) metric, which rates the predictive power of a clustering arrangement using a leave-one-out technique [ 10 ]. This and other metrics for assessing agreement between two data partitions [ 11 , 12 ] readily show the different levels of cluster method disagreement. In addition, when the FOM metric was used with an external cluster validity measure, similar inconsistencies are observed [ 13 ]. These method-based differences in cluster partitions have led to a number of studies that produce statistical measures of cluster reliability either for the gene dimension [ 14 , 15 ] or the sample dimension of a gene-expression matrix. For example, the confidence in hierarchical clusters can be calculated by perturbing the data with Gaussian noise and subsequent reclustering of the noisy data [ 16 ]. Resampling methods (bagging) have been used to improve the confidence of a single clustering method, namely PAM in [ 17 ]. A simple method for comparison between two data partitions, the weighted-kappa metric [ 18 ], can also be used to assess gene-expression cluster consistency. This metric rates agreement between the classification decisions made by two or more observers. In this case the two observers are the clustering methods. The weighted-kappa compares clusters to generate the score within the range -1 (no concordance) to +1 (complete concordance) (Table 1 ). A high weighted-kappa indicates that the two arrangements are similar, while a low value indicates that they are dissimilar. In essence, the weighted-kappa metric is analogous to the adjusted Rand index used by others to compare cluster similarity [ 16 , 19 ]. Despite the formal assessment of clustering methods, there remains a practical need to extract reliably clustered genes from a given gene-expression matrix. This could be achieved by capturing the relative merits of the different clustering algorithms and by providing a usable statistical framework for analyzing such clusters. Recently, methods for gene-function prediction using similarities in gene-expression profiles between annotated and uncharacterized genes have been described [ 20 ]. To circumvent the problems of clustering algorithm discordance, Wu et al. used five different clustering algorithms and a variety of parameter settings on a single gene-expression matrix to construct a database of different gene-expression clusters. From these clusters, statistically significant functions were assigned using existing biological knowledge. In this paper, we confirm previous work showing gene-expression clustering algorithm discordance using a direct measurement of similarity: the weighted-kappa metric. Because of the observed variation between clustering methods, we have developed techniques for combining the results of different clustering algorithms to produce more reliable clusters. A method for clustering gene-expression data using resampling techniques on a single clustering method has been proposed for microarray analysis [ 19 ]. In addition, Wu et al. showed that clusters that are statistically significant with respect to gene function could be identified within a database of clusters produced from different algorithms [ 20 ]. Here we describe a fusion of these two approaches using a 'consensus' strategy to produce both robust and consensus clustering (CC) of gene-expression data and assign statistical significance to these clusters from known gene functions. Our method is different from the approach of Monti et al. , in that different clustering algorithms are used rather than perturbing the gene-expression data for a single algorithm [ 19 ]. Our method is also distinct from the cluster database approach of Wu et al [ 20 ]. There, clusters from different algorithms were in effect fused if the consensus view of all algorithms indicated that the gene-expression profiles clustered independently of the method. In the absence of a defined rule base for selecting clustering algorithms, we have implemented clustering methods from the statistical, AI and data-mining communities to prevent 'cluster-method type' biases. When consensus clustering was used with probabilistic measures of cluster membership derived from external validation with gene function annotations, it was possible to accurately and rapidly identify specific transcriptionally co-regulated genes from microarray data of distinct B-cell lymphoma types [ 21 ]. Results Cluster method comparison Initially we assessed cluster method consistency for HC, PAM, SA and CAST using the weighted-kappa metric and a synthetic dataset of 2,217 gene-expression profiles over 100 time points that partitioned into 40 known clusters. The weighted-kappa values derived from the metric indicate the strength of agreement between two observers (Table 1 ). To interpret two weighted-kappa scores, for example, from two cluster arrangements, the broad categories from Table 1 are used, together with an assessment of relative score differences. If the two scores in question were 0.2 and 0.4, one could say that the former is poor (worse) and the latter is fair (better), but not that one is twice as good as the other. To allow defined clusters to be extracted from the tree structure of HC we used the R statistical package [ 22 ] implementation of HC. This implementation uses the CUTTREE method to convert the tree structure into a specified number of clusters. With the synthetic dataset, all clustering algorithms had a 'high' weighted-kappa agreement (data not shown) [ 18 ]. It is possible that the highly stylized nature of synthetic data resulted in higher than expected cluster-method agreement compared to experimentally derived data. This effect has been observed previously [ 10 , 12 ]. Therefore, we used a repeated microarray control element Amersham Score Card (ASC) dataset as a semi-synthetic validation standard. We also used an experimentally derived microarray dataset for cross-cluster-method comparison. To facilitate cross-method comparison, we used fixed parameters where appropriate (see Materials and methods). Consistent with other studies, we observed that clustering-method consistency varied between methods and datasets (Figure 1 ). As expected, the repeated gene/probe measurements present in the ASC dataset resulted in higher levels of cluster agreement between methods than the single gene probe B-cell data. With the ASC data there was in general a 'good' level of agreement between all different clustering algorithms, with only CAST compared to HC scoring as 'moderate'. This shows that most clustering methods are able to group highly correlated data accurately, and that repeated measurements of gene-expression values can aid cluster partitioning [ 12 ]. Nevertheless, even with such high gene-expression correlation not all cluster assignments were consistent. This effect is magnified with the single probe per gene B-cell lymphoma data, where the degree of agreement for cluster partitioning was less, with no comparison scoring above 'fair'. This observation emphasizes the need for the current desired practice in microarray analysis of using many different clustering algorithms to explore gene-expression data, thereby not over-interpreting clusters on the basis of a single method [ 23 ]. Algorithms The partial agreement of the different clustering algorithms must reflect the clustering of highly similar gene-expression vectors regardless of the clustering methods used. Where algorithm-based inconsistency problems occur in other aspects of computational biology, such as protein secondary structure prediction, consensus algorithms are often used [ 24 ]. These can either report a full or a majority agreement. This consensus strategy has also been applied to explore the effect of perturbing the gene-expression data for a single clustering algorithm [ 19 ]. We have therefore designed a similar strategy to identify the consistently clustered gene-expression profiles in microarray datasets by producing a consensus over different clustering methods for a given parameter set (see Materials and methods). Extracting such consistently clustered robust data from a large gene-expression matrix is extremely useful, increasing overall analysis confidence. Robust clustering We initially developed an algorithm called robust clustering (RC) for compiling the results of different clustering methods reporting only the co-clustered genes grouped together by all the different algorithms - that is, with maximum agreement across clustering methods. For two genes i and j , all clustering methods must have allocated them to the same cluster in order for them to be assigned to a robust cluster. This gives a higher level of confidence to the correct assignment of genes appearing within the same cluster. Robust clustering works by first producing an upper triangular n × n agreement matrix with each matrix cell containing the number of agreements among methods for clustering together the two variables, represented by the row and column indices (Figure 2 ). This matrix is then used to group variables on the basis of their cluster agreement (present in the matrix). Robust clustering uses the agreement matrix to generate a list, List , which contains all the pairs where the appropriate cell in the agreement matrix contains a value equal to the number of clustering methods being combined (that is, full agreement). Starting with an empty set of robust clusters RC, where RC i is the i th robust cluster, the first cluster is created containing the elements of the first pair in List . Then the pairs in List are iterated through and checked to see if one of the members of the current pair is within any of the existing clusters, RC i . If one element of the current pair is found and the other element of the pair is not in the same cluster, then the other element is added to that cluster. If neither element of the pair is found in an existing RC i in RC, then a new cluster is added to RC containing each element of the pair. When the end of the list is reached, the set of robust clusters, RC, is the output. The robust clustering algorithm is as follows: Input:Agreement Matrix ( n × n ), A (1) Set List = all pairs ( x , y ) in the matrix, with agreement = the number of methods (2) Set RC to be an empty list of clusters (3) Create a cluster and insert the two elements ( x , y ) of the first pair in List into it (4) For i = 2 to size of List -1 (5) For j = 1 to number of Clusters in RC (6) If x or y of List i is found within RC j (7) If the other member of the pair List i is not found in RC j (8) Add the other member to RC j (9) End If (10) Else If the other member of the pair List i is not found in RC j (11) Add a new cluster to RC containing x and y (12) End If (13) End For (14) End For Output:Set of Robust Clusters RC Application of robust clustering Robust clustering was applied to both the ASC and B-cell lymphoma datasets and the partitioning of the gene-expression profiles observed. As expected, the robust clusters do not contain all variables because of the underlying lack of consistent clustering by all methods. As a result, the weighted-kappa cannot be calculated. This metric requires both clustering arrangements being compared to be drawn from the same set of items. This is not the case with robust clustering because many items will not be assigned to a cluster. However, approximately 80% of the total ASC data variables and 25% of the B-cell lymphoma variables are assigned to a robust cluster. Robust clustering further subdivides the datasets into smaller clusters, with 24 rather than 13 clusters being defined for ASC, and 154 rather than 40 being defined for the B-cell lymphoma data (Table 2 ). Robust clusters are therefore valuable for allowing a rapid 'drilling down' in a gene-expression dataset to groups of genes whose coexpression pattern is identified in a manner independent of cluster method. The robust clustering algorithm is, by definition, subject to discarding gene-expression vectors if only one clustering method performs badly in the co-clustering. This effect of single method under performance on a given dataset has been previously observed for single linkage hierarchical clustering [ 10 , 13 ]. Therefore, to generate clusters with high agreement across methods but not so restrictive as to discard majority consistent variables, we adapted the algorithm to generate consensus clusters, making use of the same agreement matrix. Consensus clustering Consensus clustering relaxes the full agreement requirement by taking a parameter, 'minimum agreement', which allows different agreement thresholds to be explored. Rather than grouping variables on the basis of full agreement only, consensus clustering maximizes a metric, which rewards variables in the same cluster if they have high cluster method agreement and penalizes variables in the same cluster if they have low agreement. Consensus clustering maximizes agreement using the function f ( G i ) in Equation (1) to score each cluster of size s i where A is the agreement matrix, G ij is the j th element of cluster i ( G i ) and β is a user-defined parameter (the agreement threshold), which determines whether the score for the cluster is increased or decreased. The score for a clustering arrangement is the sum of the scores of each cluster, which consensus clustering attempts to maximize. If β is equal to Min , the minimum value in A , then the function is maximized when all variables are placed into the same cluster (that is, a single large cluster). Alternatively, when β is equal to Max , the maximum value in A , the function is maximized when each variable is placed into its own cluster. Essentially all clusters produced by Consensus Clustering are scored by f ( G i ), rewarding and preserving clusters with high agreement between members, while penalizing and discarding clusters containing low agreement between members. A value for β should lie between the minimum and the maximum agreement so as not to skew the scoring function. A suitable value for β is ( Max + Min )/2, where Max is the maximum value in A and Min is the minimum. For a uniformly distributed agreement matrix, ( Max + Min )/2 is the mean value; therefore we penalize values below the mean agreement and reward above it. For both the ASC and B-cell lymphoma data β was 2, as Max = 4 (four clustering algorithms giving complete agreement) and Min = 0 (no agreement). In order to maximize the scoring function for consensus clustering, a search over possible cluster membership is needed. There are many methods for performing a search and it was decided that SA was best because it is an efficient search/optimization procedure that does not suffer from becoming stuck in local maxima. The consensus algorithm is as follows: Input: Agreement Matrix ( n × n ), A; Maximum Number of Clusters sought, m; Number of Iterations, Iter; Agreement Threshold, β; Initial Temperature, θ 0 ; Cooling Rate, c (1) Generate a random number of empty clusters (<m) (2) Randomly distribute the variables (genes) 1.. n between the clusters (3) Score each cluster according to Equation (1) (4) For i = 1 to Iter do (5) Either Split a cluster, Merge two clusters or Move a variable (gene) from one random cluster to another (6) Set Δ f to difference in score according to Equation (1) (7) If Δ f < 0 Then (8) Calculate probability, p , according to Equation (2) (9) If p > random(0,1) then undo operator (10) End If (11) θ i = cθ i -1 (12) End For Output:Set of Consensus Clusters Note that random (0,1) (line 9) returns a random uniformly distributed real number between 0 and 1. The 'split', 'merge' and 'move' operators (line 5) are as follows and used with equal probability: Split a cluster: Input: Cluster g of size n (1) Randomly shuffle the cluster (2) Set i to be a random whole number between 1 and n -1 (3) Create two empty clusters g 1 and g 2 (4) Add elements 1.. i of g to g 1 (5) Add elements i+1.. n of g to g 2 Output:Two new clusters g 1 and g 2 Here the old cluster is deleted and the two new clusters are then added to the set of clusters. Merge two clusters: Input: Two Clusters g 1 and g 2 (1) A new cluster g is created by forming the union of g 1 and g 2 Output: A new cluster g Here the old clusters are deleted and new cluster is then added to the set of clusters. Move a gene: Input:A set of clusters G (1) Two random clusters g 1 and g 2 are chosen where the size of g 1 is greater than one (2) A random element of g 1 is moved into g 2 Output:The updated set of clusters G The probability ( p ) (line 8) is calculated by: In the following experiments we found θ 0 = 100, c = 0.99994 and iter = 1,000,000 as the most efficient parameters for SA. These parameter settings for SA are effectively determined by the iter setting. We denote the change in fitness during the SA algorithm as Δ f and the starting temperature as θ 0 which is always positive. From equation 2 it can be clearly seen that if Δ f = θ 0 then the (worse) solution will be accepted with probability 0.368 ( e -1 ). As the temperature cools, this probability will reduce. Here we set θ 0 to be the average of Δ f over 1,000 trial evaluations, so that at the beginning of the algorithm, the average worse solution (Δ f = θ 0 ) will be accepted with the probability stated above. It can be seen from the consensus algorithm that during the i th stage of the SA algorithm θ 0 = θ 0 c i . The SA algorithm works by assuming that the temperature reduces to zero over an infinite number of iterations. As it is not practical to run the SA algorithm to infinity the method is usually terminated after a fixed number of iterations, ( iter ). At this time the temperature will not be zero, but very small and positive, say ε . Therefore, Hence if some small positive value for ε is chosen, and the algorithm is to run for a defined number of iterations ( iter ), then the decay constant c is calculated as above. Application of consensus clustering As consensus clustering relaxes the 'complete agreement' criteria we would expect the majority but not necessarily all robust cluster members to be assigned to the same consensus clusters. This was indeed true for the B-cell data where consensus clustering of the datasets showed that 98.5% of the B-cell robust clusters were assigned correctly to their respective consensus clusters. With the more consistent ASC data 100% of the robust clusters were assigned to the correct consensus clusters. The advantage of consensus clustering over all single-cluster methods was evident when comparing consensus clustering to the mean weighted-kappa score for each pairwise combination of individual clustering algorithms (derived from Figure 1 ). Comparisons for the ASC dataset (Figure 3a ) and B-cell lymphoma data (Figure 3b ) show that consensus clustering improves on all single methods regardless of dataset, except in the case of CAST compared to SA for the ASC dataset (Figure 3a ). It is interesting to observe that consensus clustering has higher agreement with SA compared to SA agreement with all other methods in the B-cell data (Figure 3b ). The reasons for this are unclear, but suggest that with datasets similar to the B-cell data, SA captures a reliably partitioned subset of the data. To determine if consensus clustering was consistently superior to the use of single clustering methods, particularly the stochastic methods CAST and SA, we performed 10 independent runs of CAST, SA and consensus clustering. From the resulting clusters we determined the mean weighted-kappa scores for 45 possible comparisons (that is, the number of unique pairs formed from 10 objects = 10 × 9/2) (Table 3 ). Consensus clustering provided an extremely high degree of consistency over all 10 runs, with a mean weighted-kappa score of 0.96. Importantly, there was little variation between each of the 10 runs with a standard deviation of the mean weighted-kappa of 0.0015. SA had a similar low standard deviation, but produced lower inter-run consistency (mean weighted-kappa of 0.816). CAST was the least consistent of the methods (mean weighted-kappa of 0.646). The differences in the consensus clustering mean compared to SA and CAST are significant at greater than the 99.9% confidence level, thereby showing consensus clustering identifies a reliable data partition, which is significantly better than multiple runs of single clustering methods. We wished to confirm that the benefit of consensus clustering was not simply due to the parameter settings chosen for the dataset used. This could be confirmed by extensively varying each algorithm's parameter settings and comparing cluster partitioning using the same dataset; however, the large number of combinations of possible parameter settings between all methods makes this unrealistic. An alternative approach is to compare all methods on additional datasets. We therefore tested consensus clustering on two different simulated datasets containing 60 defined clusters of genes. The first synthetic dataset was generated from a vector autoregressive process (VAR) and the second using a multivariate normal distribution (MVN). The number of genes in each cluster varied from 1 to 60, with the number of conditions (arrays) set to 20. The datasets therefore contained 1,830 genes over 20 conditions. As the structure of each dataset is known, the results of each clustering method can be evaluated for accuracy using the weighted-kappa metric. Cluster accuracy using the single methods ranged between a weighted-kappa of 0.505 to 0.7 (mean weighted-kappa of 0.614) (Table 4 ). It is interesting to note that the single clustering methods performed differently on the two synthetic datasets, with HC, SA and CAST performing better on the MVN synthetic data and PAM better on the VAR synthetic data. Consensus clustering was superior to all single clustering algorithms with weighted-kappa scores of 0.725 and 0.729 for VAR and MVN respectively, demonstrating that consensus clustering is accurate regardless of subtleties in the data structure (Table 4 ). Interpretation of consensus clustering Consensus clustering greatly improves the accuracy of identifying cluster group membership based solely on the gene-expression vector, but as with other clustering algorithms still produces essentially unannotated clusters which require further external validation by gene function analysis. To address this problem, we derived a probability score to test the significance of observing multiple genes with known function in a given cluster against the null hypothesis of this happening by chance. This identifies clusters of high functional group significance, aiding assignment of functions to unclassified genes in the cluster using the 'guilt by association' hypothesis. The probability score is based on the hypothesis that, if a given cluster, i , of size s i , contains x genes from a defined functional group of size k j , then the chance of this occurring randomly follows a binomial distribution and is defined by: where n is the number of genes in the dataset. As k j and x may potentially be very large, Pr from the above equation would be difficult to evaluate. Therefore the normal approximation to the binomial distribution can be used as defined by: Large positive values of z mean that the probability of observing x elements from functional group j in cluster i by chance is very small, (for example z > 2.326 corresponds to a probability less than 1%). Note that we perform a one tailed test as we are only interested in the case where a significantly high number of co-clustered genes belong to the same functional group. This cluster function probability score was used to identify statistically significant (at the 1% level) B-cell consensus clusters containing defined genes known to be associated with 10 functional groups [ 21 ]. To determine if consensus clustering was better able to identify important functional group clusters we determined the functional group probability scores produced by individual clustering algorithms analogous to the strategy of Wu et al . [ 20 ]. For each functional group, the mean lowest probability scores (using Equation (4)) were calculated for the signal clustering methods and compared to consensus clustering (Figure 4a ). Consensus clustering always produced equivalent or lower probabilities for each functional group, indicating that it produced more informative clusters. One potential confounding factor in this analysis is that consensus clustering achieves a lower probability score by finding smaller clusters. This would decrease the ability to associated new genes with a given functional group. In the worst case the number of genes defining a functional group (FG) would equal the cluster size ( s i ) (FG/ s i = 1). Alternatively, single clustering methods may produce lower probability scores by increasing the cluster size, thereby pulling many genes into the cluster resulting in a FG/ s i ratio tending towards zero. This would also reduce the usefulness of the clusters. We determined the cluster size and functional group size for two representative functional groups where the consensus clustering probability was similar to the single method probability score, namely the endoplasmic reticulum (ER) stress response (also known as the unfolded protein response) (ER/UPR) functional group, or the markedly better nuclear factor-κB (NFκB) functional group (Figure 4b ). Apart from SA, all single clustering methods tended to produce larger clusters, thereby decreasing the FG/ s i ratio. In the most extreme case of the ER/UPR functional group, the HC cluster size was 310 compared to the consensus clustering size of 40. SA tended to produce similar cluster sizes as consensus clustering but with higher overall probabilities. Therefore, consensus clustering identifies significant functional clusters while achieving a workable balance between large or small cluster sizes. We further investigated the two groups NFκB and ER/UPR to assess what additional insights consensus clustering allowed. These two functional groups represent important B-cell functions at different stages of the B-cell development pathway. The consensus cluster associated with NFκB also contained genes either not previously associated with or only tentatively associated with NFκB activity in subsets of B-cell lymphomas. The gene-expression profiles from this consensus cluster were visualized by average linkage HC using the programmes Cluster and Treeview [ 5 ] (Figure 5 ) and clustered gene functions were investigated further using the annotation resources DAVID [ 25 ] and GeneCards [ 26 ]. From GeneCards each gene was identified in the complete human genome sequence using Ensembl [ 27 ] and 1,000 base pairs (bp) upstream of the predicted transcriptional start site extracted for promoter analysis using the program TESS from the Baylor College sequence analysis software BCM [ 28 ] (Figure 6 ). This consensus cluster is predominantly overexpressed in the cell lines Raji, Pel-B, EHEB, Bonna-12 and L-428. These cell lines have in common the induction of the NFκB pathway, either through expression of Epstein-Barr virus LMP-1 protein (Raji, Pel-B, EHEB and Bonna-12) or the loss of function of the inhibitor of NFκB, namely IκB (L-428). This implies that many of these genes could be NFκB responsive. Twenty-four putative promoter regions were analyzed and NFκB-binding sites were identified in 12 of these. As expected, NFκB-binding sites were found in the CD40L receptor gene, Bfl-1 , BIRC3 , EBV-induced gene 3 ( EBI3 ), and the genes for class I MHC-C and lymphotoxin α , as these have been previously characterized as NFκB responsive and were present in the initial NFκB-defined gene set. Interestingly, NFκB-binding sites were also found in six additional promoters for which accurate mapping of promoter transcription factor binding is not available (Figure 6a ). All but four NFκB-binding sites conform precisely to the canonical consensus binding site (Figure 6b ) [ 29 , 30 ] and of the variants with T at position 1, two genes, lymphotoxin α and BIRC3 are known to be NFκB responsive. Overall, this indicates that the six additional genes identified are likely to be NFκB responsive. The consensus cluster associated with the ER/UPR functional group contained genes not previously associated with ER stress-induced upregulation. The gene-expression profiles were visualized and annotated as described for the NFκB functional group (Figure 7a ). Annotation showed that of the 32 genes within the ER/UPR consensus cluster (23 defining the original functional group), 16% (5) were involved in calcium-ion binding within the ER and 13% (4) were involved with N -glycan biosynthesis. This functional group was overexpressed in cell lines of plasmablast or plasma-cell tumors, where physiological upregulation of the ER is required for cellular function. This process is controlled by two transcription factors, ATF6 and XBP1 [ 31 ]. The ATF6 transcript was present as a defining signature gene in the ER/UPR functional group. This suggests that ATF6 and XBP1 may be responsible for upregulation of the calcium-ion binding and N -glycan biosynthetic genes. Two responsive elements have been defined for ATF6 and XBP1 respectively, the ER stress-response element (ESRE), comprising the binding site CCAATN 9 CCACG and the unfolded protein response element (UPRE), comprising the binding site TGACGTG(G) [ 32 ]. ATF6 and XBP1 can bind to the CCACG region of ESRE in conjunction with the general transcription factor NF-Y/CB. XBP1 can bind to the UPRE, but ATF6 can only bind to the UPRE when expressed to high (possibly non-physiological) levels [ 33 ]. ESRE sites were identified in two of the five calcium-ion binding proteins, namely, calnexin and the tumor rejection antigen (gp96) 1(TRA1) (Figure 7b ). Interestingly, XBP1 (UPRE) binding sites were identified in two of the N -glycan biosynthetic genes but no ESRE sites were found. This suggests that these two groups of genes are regulated through two distinct mechanisms by transcription factors ATF6 and XBP1 as a result of ER stress. Discussion Grouping data into sets based on a consistent property is a common occurrence in biological analysis. This has recently increased in importance with the production of large microarray datasets. Implicit in the experimental rationale is the fact that patterns of coexpressed genes should be identifiable in a gene expression matrix and these can be linked to shared biological processes. However, different clustering algorithms are known to partition data into different groups [ 10 - 13 ]. We also observe a similar lack of cluster-method concordance using a weighted-kappa metric. This metric effectively scores how well different cluster method pairs assign the same genes to the same clusters. The weighted-kappa metric readily shows that, even for highly correlated gene-expression profiles present in the ACS dataset, no two clustering algorithms have complete agreement, although the global search methods such as SA seem to produce the most consistent results. Overall this emphasizes that no single analysis method will identify all patterns in the gene-expression data; therefore multiple analyses should be performed and compared [ 23 ]. We and others recently described the use of consensus clustering to improve confidence [ 34 ] or as a re-sampling method for microarray analysis [ 19 ]. It was suggested that a natural extension of Monti et al . was to use a meta-consensus across different clustering algorithms rather than to re-sample over the same algorithm. Our results represent this extension and confirm the validity of consensus clustering. We have developed both robust and consensus clustering, with these methods offering specific advantages over the use of individual clustering algorithms for microarray analysis. The robust clustering algorithm is useful for creating clusters of genes with high confidence and is extremely effective for reducing the dimensionality of large gene-expression datasets. However, robust clustering can be restrictive in discarding genes that do not have full agreement. Consensus clustering overcomes this problem, requiring a minimum-agreement parameter to generate clusters based on the combined results of a number of existing clustering methods. This strategy enables the effective identification of cluster groups that are of high reliability and cluster method independent. The choice of clustering algorithms and parameter settings is a major stumbling block for all gene array cluster analysis. The effect of varying the parameters depends on the cluster method used. The performance of cluster methods has been extensively investigated [ 12 ]. The authors show that model-based methods and certain partitional methods, when used with optimal distance matrices, perform well on synthetic and real-world data. From our study, SA, an optimization method, also performs well as a clustering method. Therefore, the individual algorithms used as input to consensus clustering should ideally consist of representative algorithms from optimization (for example, SA), graph theoretical (for example, CAST), model-based (for example, MCLUST [ 12 ]) and partitional (for example, HC). Some methods (for example, CAST, SA and MCLUST) can determine the number of clusters directly from the input data. However, some other methods require the number of clusters to be specified as a parameter (for example, PAM and HC). In principle, methods such as CAST, SA and MCLUST can be used to determine this parameter for methods such as PAM and HC. Consensus clusters are likely to contain gene subsets that are co-regulated by common transcriptional control networks or are coexpressed to participate in cellular processes that together manifest a global phenotype of the cell or tissue. In either case, these clusters are of high biological value. To facilitate further analysis it is useful to know which clusters are involved in a given biological process. By supplying a list of genes from a given biological process or network, the use of the normal approximation to the binomial distribution of these genes over all consensus clusters, allows the identification of clusters of high functional significance. Similar statistical assignment of gene function based on cluster analysis was performed by Wu et al. using a database of clusters [ 20 ]. To assign significance Wu et al. used the hypergeometric distribution. This distribution can be formally shown to asymptotically become the binomial distribution when the population size increases. Therefore, when used on large gene-expression datasets our methods are directly analogous to Wu et al . However, consensus clustering has the advantage over a database of clusters by producing low-probability clusters containing a significant percentage of known elements from functional groups. Two functionally significant clusters, the NFκB-responsive cluster and the ER/UPR cluster, were investigated further here. Within the NFκB-responsive cluster 50% of the putative promoters of genes investigated had canonical NFκB-binding sites within 1,000 base pairs of the transcriptional start site, suggesting that they are NFκB responsive [ 29 , 30 ]. The majority of these genes had NFκB-binding sites within 500 bp of the transcriptional start sites, consistent with the location in other NFκB-responsive genes [ 35 ]. Of the remaining two genes with NFκB-binding sites greater than 800 bp from the transcription start site, one, Bfl-2 , has been experimentally verified [ 36 ]. Analysis of the ER/UPR consensus cluster also provided information on gene regulatory elements, but more interestingly provides insights into the control and effect of the ERSR/UPR. In cells, the presence of unfolded proteins in the ER is associated with induction of the ER/UPR. However, during the maturation of B-cells to antibody-secreting plasma cells, expansion of the ER to accommodate increased secretion of immunoglobulins is thought to be coupled to the final stages of plasma-cell maturation. The induction of the ER/UPR occurs via the coordinated activation of the transcription factors ATF6 and XBP1 [ 31 , 33 ]. ATF6 is normally maintained as an inactive, ER-resident, transmembrane protein that is cleaved, after translocating to the Golgi upon ER stress, by the site proteases S1P and S2P [ 37 , 38 ]. The cleaved transcriptionally active ATF6 is then free to translocate to the nucleus, where it can activate target genes such as XBP1 and the ER chaperon protein GRP78/BiP [ 39 ]. XBP1 mRNA is cleaved by the ER stress activated protein IRE1 to yield the transcriptionally active form of XBP1, inducing further genes of the UPR [ 32 ]. The activation of both the ATF6 and IRE1/XBP1 pathways results in enhanced transcription of ESRE-responsive genes; however, only XBP1 appears able to transactivate the UPRE. The identification of ESRE binding sites in the promoter regions of genes for calcium-ion binding protein and UPRE binding sites in the promoter regions of N -glycan biosynthesis genes suggests that these genes are differentially regulated by ATF6/XBP1 and XBP1 respectively. The only known UPRE target gene is ER degradation-enhancing α -mannoside-like protein (EDEM), whose induction depends solely on IRE1/XBP1 activity [ 33 ]. Induction of the two UPRE-containing genes, UDP-GlcNAc:dolichol phosphate N -acetylglucosamine-1 phosphate transferase ( DPAGT1 ) and UDP-GlcNAc: α -6-D-mannoside β -1-2- N -acetylglucosaminyltransferase II ( MGAT2 ), which catalyze essential steps in the biosynthetic pathway of complex N -linked glycans, supporting a clear link between the dolichol pathway and the UPR [ 40 ]. In addition, the ER/UPR functional group suggests that DPAGT1 and MGAT2 expression is regulated solely by the IRE1/XBP1 pathway. Altogether, these results show that consensus clustering and gene functional group analysis provide a highly accurate way of mining gene-expression data for novel insights into different genes within the cluster. Robust and consensus clustering provide a platform for more efficient microarray analysis pipelines. There is effectively no limit to the number of different clustering algorithms that can be used to feed into the consensus clusters, and each clustering algorithm could be run under different parameter sets to fully explore a microarray dataset [ 19 ]. In addition, different distance matrices could be used as input into the range of clustering algorithms. In each case the consensus clustering algorithm effectively acts as the collation and interpretation point for the different individual analysis methods. This environment is ideal for use in parallel processing computer farms and the GRID [ 41 ]. In such an environment, each node of the farm could perform a range of analyses with a subset of clustering algorithms, with the master node compiling the consensus results. This would greatly increase computational speed and allow a thorough, single data entry point, access to an extensive range of clustering methods. All areas of functional genomics that produce high-dimensional datasets with inherent patterns will require data partitioning to allow interpretation. Consensus clustering in the context of statistically defined functional groups could allow a consistent analysis platform for such diverse data types. Materials and methods Clustering methods We implemented and compared a representative sample of methods from the statistical, AI and data-mining communities. The methods used were average linkage HC, PAM, SA and CAST. As all the clustering techniques use correlation between variables, we used the Pearson's correlation coefficient, r , to measure the linear relationship between two variables, x 1 and x 2 , where the variable can be either discrete or continuous [ 42 ]. HC and PAM methods were implemented using the statistical package R [ 22 ], while CAST and SA were implemented locally in C++. HC is an agglomerative method that produces a hierarchical (binary) tree or dendrogram representing a nested set of data partitions. It has been applied successfully to many gene-expression datasets [ 43 ]. Sectioning a hierarchically clustered tree at a particular level leads to a partition with a number of disjointed groups, thereby yielding different clustering of the data. The tree was sectioned using the CUTTREE method, to yield 13 clusters for the ASC dataset and 40 for the B-cell dataset. The method PAM works by first selecting m out of n total objects that are the closest (according to a distance matrix) to the remaining ( n - m ) objects. The fitness of these medoids is calculated by placing the remaining ( n - m ) objects in a group according to the nearest medoid and summing the distances of the group members from this medoid. These m selected objects are the initial medoids. A Swapping procedure is then applied to reassort the objects until there is no improvement in the fitness of the medoids [ 3 ]. As with HC, PAM is set to search for 13 and 40 clusters. The choice of 13 clusters for the ASC data was determined by the number of repeated genes, whereas 40 clusters for the B-cell data was based on previous exploratory data analysis [ 21 ]. SA [ 6 ] is an iterative improvement search technique that starts with a random solution to a given problem, and then tries to increase its worth by a series of small changes in cluster membership. If such a small change is better than the previous solution, then further changes are made from this new point. However, if the new solution is worse than the old one, it is not discarded, but accepted with a certain probability. The measured worth of the SA clustering arrangement is based here on the EVM metric [ 44 ]. SA has recently been applied to the clustering of gene-expression data [ 45 ]. The performance related parameters for SA were set as follows: θ 0 = 100, c = 0.99994 and number of iterations = 1,000,000. CAST [ 9 ] is a heuristic algorithm that uses an affinity measure to determine whether variables are assigned to clusters. It requires a threshold parameter, which determines whether variables are assigned or moved to new clusters. Once CAST is complete, a clean-up operation is applied to ensure that the affinity of every variable to its cluster is greater than a user defined threshold. The only parameter CAST needed was the affinity level, which was set to 0.5 as recommended [ 9 ]. Methods such as CAST and SA require the differences/relationships between a pair of observations, x 1 and x 2 , to be expressed as binary ( b ). As Pearson's correlation coefficient is bounded, it provides a good basis for defining a binary relationship between two variables as defined by: where 0 < α ≤ 0 is a constant and is a floor function that returns the largest integer less than or equal to the real number y . Datasets for evaluation Two datasets were used for evaluating the cluster methods. The first is a set of multiply repeated control element spots relating to the Amersham Score Card (ASC) probe set on the Human Genome Mapping Project Human Gen1 clone set array [ 46 ]. The ASC probes are present as a single row of 32 elements in each of the 24 array sub-grids. Of these elements, 13 gene probes consistently give signals above background in both the Cy5 and Cy3 channels. Therefore, each array has effectively 24 repeat measurements of 13 spots. After filtering for low signal-to-noise ratio (SNR) probes, a dataset of 30 arrays was examined by treating each positional repeat probe element across the 30 array set as an individual gene, which together with the remaining 23 same-gene probes per array, represents a highly correlated gene-expression profile. Therefore, we assume the repeated probes should cluster together; thus, this dataset becomes 308 genes/probe elements, which would cluster into 13 known groups, with each group having between 6 and 24 members after SNR filtering. In essence, the ASC data represent a semi-synthetic dataset for internal cluster method validation. The second dataset consists of a series of 26 arrays (1,987 filtered genes) measuring gene-expression difference across a set of human B-cell lymphomas and leukemias [ 21 ]. The dataset is available via the URL indicated in Jenner et al. [ 21 ]. Each probe on the array detected a single gene transcript. This dataset contains a number of genes that correspond to known cellular functions, for example cell proliferation and NFκB response. The four clustering techniques described above were applied to both the datasets, with each method being set to find 13 clusters for the ASC and 40 clusters for the B-cell data. Synthetic datasets Two synthetic datasets were generated using a vector autoregressive process (VAR) or a multivariate normal distribution (MVN). The two datasets contained 1,830 genes and 20 conditions (arrays). The VAR process of order p is a linear multivariate time series defined by where x ( t ) is the n -dimensional vector of observations at time t , A i is the n × n autoregressive coefficient matrix at time lag i , and ε ( t ) is the zero mean n -dimensional noise vector at time t (drawn from a normal distribution). Therefore x ( t ) is a linear combination of the previous observations plus some random noise. For the synthetic dataset, each cluster was generated by a vector autoregressive model of order p = 1 and size n equal to the number of genes in the cluster. For the MVN dataset, a vector of random variables x has a MVN distribution if every linear combination of that vector is also normal. Under such conditions we use the notation x ~ N ( μ , Σ) to denote that x follows the MVN distribution, where μ is the mean vector and Σ is a positive definite matrix of covariance. The probability density function of x is given by where \|Σ\| = det(Σ). For the synthetic dataset, each cluster was drawn from an MVN distribution with varying mean μ and covariance Σ. Weighted-kappa metric To compare the resultant clusters for each method, a statistic known as weighted-kappa was used [ 18 ]. This metric rates agreement between the classification decisions made by two or more observers. In this case the two observers are the clustering methods. The classification from each observer for each unique pairing of variables (within the clusters) is divided into a 2 × 2 contingency table. Rows and columns within this table are indexed according to whether the two variables are in the same group or in different groups. The total number of comparisons, N , is defined in the following equation, where Count ij is the number of elements in the matrix cell indexed by ( i , j ), and n is the number of variables (genes) in the clusters as this represents the number of unique variable pairings. The weighted-kappa metric is calculated from the contigency table by: where, w ij is the weights for each category comparison; p o ( w ) and p e ( w ) represent the observed weighted proportional agreement and the expected weighted proportional agreement; Count ij is the i th, j th element of the 2 × 2 contingency table; N is the sum of the elements within this table; Row ( i ) and Col ( i ) are the row and column totals for this table respectively and K w is the weighted-kappa value. The interpretation of weighted-kappa values indicates the strength of agreement between two observers (Table 1 ) is used to compare cluster method agreement in both datasets.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545785.xml
535538	Vaccination with EphA2-derived T cell-epitopes promotes immunity against both EphA2-expressing and EphA2-negative tumors	Background A novel tyrosine kinase receptor EphA2 is expressed at high levels in advanced and metastatic cancers. We examined whether vaccinations with synthetic mouse EphA2 (mEphA2)-derived peptides that serve as T cell epitopes could induce protective and therapeutic anti-tumor immunity. Methods C57BL/6 mice received subcutaneous (s.c.) vaccinations with bone marrow-derived dendritic cells (DCs) pulsed with synthetic peptides recognized by CD8+ (mEphA2 671–679 , mEphA2 682–689 ) and CD4+ (mEphA2 30–44 ) T cells. Splenocytes (SPCs) were harvested from primed mice to assess the induction of cytotoxic T lymphocyte (CTL) responses against syngeneic glioma, sarcoma and melanoma cell lines. The ability of these vaccines to prevent or treat tumor (s.c. injected MCA205 sarcoma or B16 melanoma; i.v. injected B16-BL6) establishment/progression was then assessed. Results Immunization of C57BL/6 mice with mEphA2-derived peptides induced specific CTL responses in SPCs. Vaccination with mEPhA2 peptides, but not control ovalbumin (OVA) peptides, prevented the establishment or prevented the growth of EphA2+ or EphA2-negative syngeneic tumors in both s.c. and lung metastasis models. Conclusions These data indicate that mEphA2 can serve as an attractive target against which to direct anti-tumor immunity. The ability of mEphA2 vaccines to impact EphA2-negative tumors such as the B16 melanoma may suggest that such beneficial immunity may be directed against alternative EphA2+ target cells, such as the tumor-associated vascular endothelial cells.	Background EphA2 is a member of Eph family of receptor tyrosine kinases comprised of two major classes (EphA and EphB), which are distinguished by their specificities for ligand (ephrin-A and ephrin-B, respectively [ 1 - 3 ]). Recent reports suggest that EphA2 is frequently overexpressed and often functionally dysregulated in advanced cancers, where it contributes to multiple aspects of malignant character. These changes in EphA2 have been observed in a wide array of solid tumors, including melanoma [ 4 , 5 ] and prostate [ 6 ], breast [ 7 ] and lung [ 8 ] carcinomas. Indeed, the highest degree of EphA2 expression among tumors is most commonly observed in metastatic lesions [ 6 , 9 ]. These data suggest that EphA2 may serve as an attractive target for cancer vaccines. In this regard, we have identified five human leukocyte antigen (HLA)-A2 binding and three HLA-DR4-binding peptides derived from EphA2 that are capable of inducing specific, tumor-reactive CD8 + or CD4 + T-cell responses, respectively [ 10 ]. A more recent report has identified two additional HLA-A2 restricted T-cell epitopes encoded by EphA2 [ 11 ]. These observations and findings support our rationale for near future implementation of EphA2-targeted vaccine clinical trials. For pre-clinical evaluation of EphA2-targeted vaccines, however, there is little information on immune responses against EphA2 in mouse models. Therefore, we hypothesized that identification of mouse T-cell epitopes in mEphA2 would allow us to determine the effect of EphA2-targeted vaccinations in mice bearing tumors. In this study, we examined whether novel T-cell epitope peptides identified in the mEphA2 protein sequence could elicit protective and therapeutic anti-tumor immune responses in murine models. Our results indicate that DC-based vaccines incorporating these peptides elicit effective CTL responses that can inhibit the growth both EphA2+ and EphA2-deficient tumors. Materials and methods Animals Female 6–8-week-old C57BL/6 mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Animals were handled under aseptic conditions in microisolator cages within the Central Animal Facility at the University of Pittsburgh per an Institutional Animal Care and Use Committee-approved protocol, and in accordance with recommendations for the proper care and use of laboratory animals. Cell Lines and Culture Glioma cell lines KR129, KR130, KR233 and KR158D were derived from spontaneously arising gliomas in F1 between B6 × CBA (KR129), B6 × SJL (KR130), and C57BL/6 (KR233 and KR158D)-background NPcis mice that express mutations in two tumor-suppressor genes, Nf1 and Trp53 [ 12 ]. B16 melanoma, NPcis-derived glioma cells, GL261 glioma and MCA205 sarcoma (H-2 b ) cell lines were cultured in complete medium (CM) [RPMI 1640 supplemented with 10% heat-inactivated fetal bovine serum, 100 units/ml penicillin, 100 μg/ml streptomycin, and 10 mM L-glutamine (all reagents from Life Technologies, Inc., Grand Island, NY)] in a humidified incubator in 5% CO 2 at 37°C. Generation of DCs in Vitro from Bone Marrow The procedure used to generate DCs has been previously described [ 13 ]. Briefly, C57BL/6 bone marrow cells were cultured in CM supplemented with 1000 units/ml recombinant mouse granulocyte/macrophage colony-stimulating factor and recombinant mouse interleukin-4 (Schering-Plough, Kenilworth, NJ) at 37°C in a humidified, 5% CO 2 incubator for 7 days. DCs were then isolated at the interface of 14.5% (w/v) metrizamide (Sigma, St. Louis, MO) in CM discontinuous gradients by centrifugation. DCs typically represented >90% of the harvested population of cells based on morphology and expression of the CD11b, CD11c, CD40, CD54, CD80, CD86, and class I and class II MHC antigens as assessed using flow cytometry (data not shown). Peptides and immunization The protein sequences of mEphA2 was obtained from GenBank and analyzed for H-2K b -, H-2D b -, and I-A b -binding binding motifs using BIMAS , and a proteosomal cleavage site prediction system [ 14 ], respectively. Peptide sequences that were given high binding scores and predicted proteosomal cleavage sites at the ends of the sequences were chosen [ 15 ]. The H-2D b -binding mEphA2 671–679 (FSHHNIIRL), H-2K b -binding mEphA2 682–689 (VVSKYKPM), and I-A b -binding mEphA2 30–44 (LLDFAAMKGELGWLT) epitopes were synthesized using an automated solid-phase peptide synthesizer (Applied Biosystems, Foster City, CA) by the protein synthesis facility at the University of Pittsburgh Cancer Institute, purified (to greater than 95%) by reverse phase HPLC, and characterized for amino acid sequences by mass spectrometry. Day 7 DCs were pulsed with 10 μM each of the indicated peptides for 4 hours at 37°C, as previously described [ 13 ]. Cells were then washed twice with Hank's balanced salt solution (HBSS), with animals receiving injections of the indicated numbers of peptide-pulsed DCs in 0.1 ml HBSS s.c. Tumor challenge In the s.c. model, animals were injected with the indicated numbers of tumor cells in the right flank. Anti-tumor responses were assessed based on comparative longitudinal measurements of tumor area. In the lung metastasis model, animals were injected i.v. with 2 × 10 5 B16-BL6 tumor cells on day 0, and they subsequently received s.c. vaccinations of peptide-loaded DCs on days 3, 10 and 17. Animals were sacrificed on day 28 post-tumor injection and analyzed for the assessment of pulmonary metastases by enumerating the number of surface tumor-nodules. Western Blotting Protein lysates isolated from normal mouse brain, spleen, liver, lung, heart, skeletal muscle, and mouse tumor lines were separated by SDS-PAGE, blotted onto nitrocellulose and analyzed for expression of mEphA2 using EphA2 monoclonal antibody (C-20 Ab; Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Blots were imaged on Kodak X-Omat Blue XB-1 film (NEN Life Science Products, Boston, MA) after using horseradish peroxidase (HRP)-conjugated goat anti-rabbit Ig (Biorad, Hercules, CA) and the Western Lighting™ chemiluminescence detection kit (Perkin Elmer, Boston, MA). CTL activity assay Single cell suspensions of SPCs were cultured at 2 × 10 6 cells/ml with 2 μg/ml mEphA2 671–679 or mEphA2 682–689 in presence of 10 U/ml human IL-2 (Chiron, Emeryville, CA), 50 μM 2-mercaptoethanol (Sigma), and 50 μM N G mono-methyl-L-arginine (Cyclopss, Salt Lake City, UT) in 24 wells plates (Corning, Corning, NY) for 5 days. Specific CTL activity was determined in 4 h 51 Cr release assays against the indicated target cells, as previously described [ 16 ]. Matrigel plug assay Eight-week-old mice were injected s.c. twice (days -14 and -7) with DCs loaded with either EphA2-derived peptides (EphA2 671–679/30–44 ) or OVA-derived peptides (OVA 257–264/265–280 ). Each animal then received 400 μl of Matrigel (BD Biosciences) supplemented with 400 ng/ml vascular endothelial growth factor (VEGF) in the dorsal area [ 17 ]. The animals were sacrificed 10 days later, then the plugs were removed and photographed. Statistical analysis Comparative numbers of lung metastasis, growth of s.c. tumors and T cell responses were compared by Student's t test for two samples with unequal variances. Statistical significance was determined at p value <0.05. Results Expression of mEphA2 in murine tumor cells We evaluated the expression of mEphA2 in H-2 b -syngeneic murine tumor cell lines and normal organs from C57BL/6 mice by Western Blotting (Figure 1 ). MCA205 sarcoma, the NPcis mice-derived spontaneously developed glioma cell lines KR129, KR130, KR233 and KR158D (Figure 1A ), as well as, the GL261 glioma (Figure 1B ) expressed detectable but variable levels of mEphA2 protein. In marked contrast, none of melanoma lines tested, including B16 melanoma and its more metastatic sub-clones B16V1 and B16BL6, expressed detectable levels of mEphA2 (Figure 1A ). The control β-actin staining on the same blots demonstrated that equal amount of protein was loaded in each lane. With regard to the normal organs examined, moderate levels of mEphA2 expression were identified in the spleen, liver and lung, whereas no detectable expression was observed in brain, heart and skeletal muscle (Figure 1B ). Figure 1 Expression of EphA2 in murine tumor cell lines and tumor/normal tissues. (A) Aliquots of protein lysates (10 μg/lane) were analyzed by Western Blotting for expression of mEphA2 protein using a specific monoclonal antibody (C-20 Ab; Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Lysate samples were obtained from mouse tumor lines including the B16, B16V1 and B16BL6 melanomas and the KR129, KR130, KR233 and KR158D glioma cell lines derived from spontaneously developed glioma tissues in NPcis mice. Control β-actin labeling demonstrated equal amount of protein loading in each lane. (B) Protein lysates isolated from normal mouse brain, spleen, liver, lung, heart, skeletal muscle, and the G261 glioma and MCA 205 sarcoma cell lines were separated by SDS-PAGE, blotted onto nitrocellulose and analyzed for expression of mEphA2 using C-20 anti-EphA2 monoclonal antibody. Vaccination of mice with synthetic mEphA2 T cell epitopes promotes specific CTL responses We have previously characterized the immunogenicity of the H-2D b -binding mEphA2 671–679 , H-2K b -binding mEphA2 682–689 , and I-A b -binding mEphA2 30–44 epitopes based on delayed-type hypersensitivity responses in B6 mice (Storkus et al ., unpublished results). We immunized syngeneic C57BL/6 mice twice s.c. with syngenic DCs loaded with the K b -binding and the I-A b -binding epitopes, or DCs loaded with the D b -binding and the I-A b -binding epitopes on a weekly basis. Control animals received DCs loaded with the cancer-irrelevant K b -binding OVA 257–264 and the I-A b -binding OVA 265–280 epitopes [ 18 ]. One week after the secondary immunization, the animals were sacrificed, and SPCs were harvested. The SPCs were then stimulated in vitro with the MHC class I peptide that was used for in vivo immunization, in the presence of 10 U/ml hIL-2, for 7 days. SPCs from control animals immunized with OVA peptides were stimulated in vitro with the mEphA2 671–79 to provide an index for a control, "primary" level of specific CTL induction. CTL activity against mEphA2-expressing, and H-2 b -syngeneic MCA205 sarcoma, GL261 glioma and KR158D glioma cells were assessed by standard 51 Cr-release assays (Figure 2 ). All three mEphA2+ cell lines demonstrated susceptibility to CTLs generated by the immunizations with K b - or D b -binding mEphA2 derived epitopes. With 20 h – incubation time, the percent lysis by EphA2-induced CTLs increased remarkably in comparison to 4 h-incubation regimen, whereas the lysis by OVA-induced CTLs do not, suggesting that the specific lysis can be more appreciated with the prolonged incubation time (up to 20 h) for the CTLs raised against EphA2-epitopes. B16 melanoma and YAC cells, which were used as negative control target cells, displayed constantly less than 10% of specific lyses in all groups (data not shown). These data support the conclusion that these mEphA2-derived peptides may serve as effective in vivo immunogens for anti-tumor T cell activation. Figure 2 Immunizations with mEphA2-derived peptides elicit anti-tumor CTLs. C57BL/6 mice received two cycles of s.c. immunization with 5 × 10 5 DCs loaded with mEphA2-derived peptides or control OVA-derived peptides. One week after the second immunization, the animals were sacrificed, and SPCs were re-stimulated in vitro with the indicated peptides in the presence of 20 IU/ml hIL-2 for 7 days. Standard 4-hr (upper panel) and 20-hr (lower panel) 51 Cr-release assays were used to assess cytotoxicity of the responder SPCs against the EphA2+ MCA205, GL261 and KR158D tumor cell lines. Each value represents the average of triplicate determinations for each group. YAC cells were evaluated as non-specific target cells, while EphA2-negative B16 melanoma cells were examined as negative control target cells in all groups, with the lysis of each of these targets constantly less than 10% (data not shown). mEphA2 peptide-pulsed DC-based vaccines effectively protect/treat s.c MCA205 sarcoma and B16 melanoma We evaluated the efficacy of protective and therapeutic immunizations with mEphA2-derived peptides in s.c. tumor models. These included the mEphA2-positive MCA205 sarcoma and the mEphA2-negative B16 melanoma models. First, animals bearing established s.c. MCA205 tumors received therapeutic vaccinations with mEphA2-derived CTL (H-2K b or H-D b presented) + Th epitope peptides on days 3 and 10 following tumor inoculation (5 animals/group). Control animals received DCs loaded with the OVA-derived CTL + Th epitopes. As displayed in Figure 3A , vaccination with either combination of mEphA2-derived CTL (H-2K b or H-D b presented) + Th epitope peptides significantly inhibited the growth of tumors in comparison to control vaccinations with irrelevant (OVA)-derived T cell epitopes. Figure 3 Immunizations with DCs loaded with Eph-A2 derived peptides inhibit the growth of both EphA2-positive and EphA2-negative tumors. (A) C57BL/6 mice received 1 × 10 5 MCA205 cells s.c. on day 0. These animals were then immunized with 1 × 10 6 DCs loaded with the indicated peptides on days 3, 10. Tumor growth was monitored for 28 days following the tumor inoculation (N = 5/group). P < 0.05 for both mEphA2 671–679/30–44 and mEphA2 682–689/30–44 treatments in comparison to OVA control for the tumor based on a two-tailed Student-t test (). (B) C57BL/6 mice received s.c. pre-immunizations with 1 × 10 6 DCs loaded with either mEphA2 671–679 , mEphA2 682–689 , mEphA2 30–44 or control OVA 257–264 on days -14 and -7 (N = 5 /group), followed by s.c. challenge with 5 × 10 4 B16 melanoma cells on day 0. On day 14, the size of tumors and the number of animals that had measurable tumors were assessed. P < 0.01 for EphA2 671–679 , EphA2 682–689 and EphA2 30–44 treatments in comparison to OVA 257–264 treated group (). We next examined whether vaccination with mEphA2-derived peptides could induce anti-tumor effects against mEphA2-negative tumors, expecting that this would not be the case. Since s.c. B16 melanomas grow aggressively, we employed a pre-immunization model to assess the anti-tumor response. Syngeneic C57BL/6 mice were pre-immunized with either mEphA2 671–679 , mEphA2 682–679 , or mEphA2 30–44 peptides on days -14 and -7 before s.c. injection with 5 × 10 4 B16 cells on day 0 (5 animals/group). The control group received irrelevant OVA 257–264 instead of mEphA2-derived peptides. Tumor size and the number of animals that had measurable tumors are assessed on day 21 (Figure 3B ). All control animals receiving OVA-immunization developed large tumors, whereas, surprisingly, 3 of 5 animals in each of mEphA2-derived peptide-treated groups rejected the tumor growth. Indeed, the average tumor size in each of mEphA2-immunized groups was significantly smaller than the control groups (P < 0.01). As B16 melanoma cells did not express mEphA2 protein in vitro (Figure 1A ), these data suggest that vaccination with mEphA2-derived peptides may induce in vivo CTL activity not only directly against mEphA2 expressed in tumors, but also against mEphA2 expressed by other critical components of tumor-structure, such as tumor vascular endothelial cells. With regard to EphA2 expression in vivo , we attempted immunohistochemistry on s.c. B16 tumors to directly validate EphA2 expression. Although deposition of endogenous melanin in B16 tissues posed interference for the substrate-development, we did not observe EphA2-specific staining in s.c. B16 tumor tissues (data not shown). EphA2 peptide-pulsed DC-based vaccines protect against/treat B16-BL6 lung metastasis We next evaluated a lung metastasis model using B16-BL6 melanoma cell line. Syngeneic C57BL/6 mice received 2 × 10 5 B16-BL6 melanoma cells via tail vein injection. The animals subsequently received s.c. immunizations with 1 × 10 6 DCs loaded with mEphA2 CTL (H-2K b or H-D b presented) + Th on days 3, 10 and 17 (5 animals/group). Control animals received DCs loaded with the OVA CTL + Th epitopes. On day 28, the animals were sacrificed, and the lungs harvested from each animal. The number of surface pulmonary metastases and lung weight were measured and compared between the groups. As demonstrated in Figure 4A and 4B , both of mEphA2-immunization regimens resulted in remarkable suppression of B16-BL6 lung metastases in comparison to the OVA-immunized control group in this early-stage treatment model. Figure 4 Inhibition of B16-BL6 lung metastasis and VEGF-induced angiogenesis by s.c. immunizations with DCs loaded with EphA2-derived peptides . (A and B), C57BL/6 mice received i.v. injections of 2 × 10 5 B16-BL6 cells followed by s.c. immunizations with 1 × 10 6 DCs loaded with the indicated peptides on days 3, 10 and 17 (N = 4/group). The animals were sacrificed on day 28, with representative macroscopic pictures of the harvested lungs from each group (A) and the numbers of pulmonary surface metastases counted (B) P < 0.001 for both mEphA2 671–679/30–44 () and mEphA2 682–689/30–44 (*) treatments in comparison to OVA control for the number of metastasis based on two-tailed Student-t test. (C), mEphA2-targeted immunizations inhibited the vascular formation in the Matrigel plug assay. Syngeneic mice pre-immunized with control OVA- (left) or EphA2- (right) peptides were injected s.c. with Matrigel containing VEGF. After 10 days, the plugs were removed and photographed. The picture shows representative VEGF-containing plugs excised from total 5 mice/group. VEGF has been implicated to be a major mediator of tumor-angiogenesis [ 19 , 20 ], and EphA2-blockade has been shown to inhibit VEGF-induced angiogenesis [ 21 ]. To examine whether immunizations with mEphA2-derived peptides inhibit VEGF-induced angiogenesis, Matrigel plugs containing VEGF were implanted in pre-immunized mice and recovered 10 days later. Macroscopic analysis revealed that the plugs from mice pre-immunized with mEphA2-derived peptides were much paler than those from control mice (Figure 4C ). These results suggest that vaccinations with mEphA2-derived T cell epitopes may induce potent anti-tumor responses in mEphA2-deficient tumor models through inhibition of tumor angiogenesis. Discussion In this report, we have demonstrated that mEphA2-derived peptide epitopes can elicit specific CTL responses and anti-tumor response in vivo . Of major note, the anti-tumor response was not restricted to mEphA2-postive models, but was also observed in the case of both s.c. and metastatic B16 melanoma models. While the melanoma cell lines were negative for EphA2 expression in vitro as assessed by Western Blotting, it was conceivable that they had become EphA2+ after in vivo injection, which could more easily explain the efficacy of the mEphA2 peptide-based vaccinations. However, we were unable to detect mEphA2 expression even in the in vivo grown B16 tumors (data not shown), suggesting that the T-cell immunity against mEphA2-derived epitopes suppress the in vivo tumor-growth not only through the direct anti-tumor cell mechanisms but also through indirect mechanisms, such as inhibition of tumor-angiogenesis. The Eph kinases and ephrins play a crucial role in vascular development during embryogenesis and tumor formation [ 1 ]. In two independent tumor types, the RIP-Tag transgenic model of angiogenesis-dependent pancreatic islet cell carcinoma and the 4T1 metastatic mammary adenocarcinoma, Ephrin-A1 ligand is expressed in both tumor and endothelial cells, and EphA2 receptor was localized primarily in tumor-associated vascular endothelial cells [ 3 ]. Soluble EphA2-Fc or EphA3-Fc receptors inhibit angiogenesis and tumor formation in multiple models [ 3 , 22 ]. Further molecular analyses revealed that the EphA-ephrinA interaction is necessary for VEGF-induced angiogenesis [ 21 , 22 ]. These published observations suggest that immunological targeting of EphA2 may also inhibit tumor growth by suppression of tumor-neoangiogenesis. Immunization against angiogenic-associated antigens selectively expressed on tumor vasculature may provide for a novel strategy to block tumor growth. The feasibility of this approach has been borne out in reports demonstrating that the protein or DNA encoding angiogenic molecules, such as VEGF-R2, can be used as a vaccine to generate effective CTL and antibody responses against tumor vessels, thereby limiting tumor growth and metastasis [ 19 , 20 ]. In this regard, our results with Matrigel assays suggest that immunization of C57BL/6 mice with mEphA2-derived peptides may inhibit VEGF-induced angiogenesis. We are currently confirming these observations with quantitative analyses of vessel formations in the Matrigel plugs. As expected, based on our Western Blotting, several normal organs examined expressed EphA2 including the spleen, liver and lung. This could provide concern for the potential clinical translation of EphA2-based vaccines, given concerns for autoimmune pathology that might occur in these anatomic sites. However, despite the effective induction of antigen-specific CTLs that are capable of mediating DTH-type reactivity (unpublished data) and anti-tumor responses in situ , our careful observation of EphA2-vaccinated animals at autopsy did not reveal any evidence for tissue infiltration by inflammatory leukocytes or the destruction of EphA2+ tissues. In our assessment of CTL and anti-tumor response, in most cases, we employed both MHC class I and II associated peptides in our treatment regimens based on our previous study demonstrating that the combination of MHC class I and II-restricted OVA-immuno-epitopes induced higher levels of anti-tumor efficacy than MHC class I-restricted OVA-epitopes solely [ 18 ]. With regard to the immunogenicity of each mEphA2-derived peptide, our data with preimmunizations against s.c. B16 challenge indicated that each of mEphA2 671–679 , mEphA2 682–689 and mEphA2 30–44 could induce anti-tumor response. It was noteworthy that mEphA2 30–44 , which was found to bind I-A b and induce DTH and specific CD4+ T cell responses, also suppressed the growth of s.c. (MHC class II-deficient) B16 melanomas as a single agent. This may suggest that neovessels in the B16 microenvironment are MHC class II+ either constitutively or under inflammatory conditions under which IFN-γ is elaborated, that B16 melanomas become class II+ after in vivo transfer or upon provision of IFN-γ in situ , or that the mEphA2 30–44 Th peptide also contains an embedded CTL epitope. We are currently assessing each of these possibilities. It is also intriguing to determine whether there is a correlation between expression levels of mEphA2 in the target cells and their susceptibility to CTLs raised against EphA2-derived peptides. Our preliminary data indicated EphA2 expression levels did not simply correlate with the CTL susceptibility, suggesting other factors, such as MHC class I expression and intracellular anti-apoptotic proteins, may contribute in the CTL-mediated cytotoxicity (data not shown). To address this issue, our ongoing studies with human glioma cells employ small inhibitory RNA for EphA2 to determine whether specific inhibition of EphA2-expression affects the CTL response. Conclusions Although the precise mechanisms that underlie the anti-tumor efficacy of EphA2-based vaccines remain to be elucidated, our results demonstrate that the mEphA2-derived epitopes may represent important vaccine candidates for the development of clinical trials for the treatment of (both EphA2+ and EphA2-negative) cancers. List of abbreviations s.c., subcutaneous; i.v., intravenous; CTL, cytotoxic T lymphocyte; Th, T helper; DCs, dendritic cells; APCs, antigen-presenting cells; OVA, ovalbumin; CM, complete medium; HBSS, Hank's balanced salt solution; PBS, phosphate buffered saline; SPCs, splenocytes; HLA, human leukocyte antigen; ELISPOT, enzyme-linked immuno-spot; DTH, delayed-type hypersensitivity; VEGF, vascular endothelial cell growth factor Disclosure of competing interests The author(s) declare that they have no competing interests. Authors' contributions NK carried out the Western Blotting and immuno-assays. NK, JED, TT, MH and FH carried out assessment of in vivo anti-tumor response and preparation of peptide-pulsed DCs. WJS participated in peptide identification, the design of the study and critical review of data and the manuscript. HO conceived of the study, and participated in its design and coordination. All authors have read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535538.xml
545961	Protocol for the evaluation of a decision aid for women with a breech-presenting baby [ISRCTN14570598]	Background There is now good evidence about the management options for pregnant women with a breech presentation (buttocks or feet rather than head-first) at term; external cephalic version (ECV) – the turning of a breech baby to a head-down position and/or planned caesarean section (CS). Each of these options has benefits and risks and the relative importance of these vary for each woman, subject to her personal values and preferences, a situation where a decision aid may be helpful. Decision aids are designed to assist patients and their doctors in making informed decisions using information that is unbiased and based on high quality research evidence. Decision aids are non-directive in the sense that they do not aim to steer the user towards any one option, but rather to support decision making which is informed and consistent with personal values. The ECV decision aid was developed using the Ottawa Decision Support Framework, including a systematic review of the evidence about the benefits and risks of the options for breech pregnancy. It comprises an audiotape with a supplementary booklet and worksheet, a format that can be taken home and discussed with a partner. This project aims to evaluate the ECV decision aid for women with a breech presenting baby in late pregnancy. Study design We aim to evaluate the effectiveness of the decision aid compared with usual care in a randomised controlled trial in maternity hospitals that offer ECV. The study group will receive the decision aid in addition to usual care and the control group will receive standard information on management options for breech presentation from their usual pregnancy care provider. Approximately 184 women with a single breech-presenting baby at greater than 34 weeks gestation and who are clinically eligible for ECV will be recruited for the trial. The primary outcomes of the study are knowledge, decisional conflict, anxiety and satisfaction with decision-making that will be assessed using self-administered questionnaires. The decision aid is not intended to influence either the uptake of either ECV or planned CS, however we will monitor health service utilisation rates and maternal and perinatal outcomes.	Background Breech presentation Breech presentation occurs when a baby presents with the buttocks or feet rather than head first (cephalic presentation). As breech presentation is related to both fetal size and gestational age, the incidence decreases as pregnancy progresses to 3–4% by full-term[ 1 , 2 ]. Decades of controversy over the safe management of breech birth at term has recently been resolved by an international multicentre randomised controlled trial (the Term Breech Trial, TBT) of planned vaginal breech birth versus planned caesarean section (CS)[ 3 ]. This trial was stopped prematurely because of overwhelming benefit favouring planned CS, with a relative risk of 0.33 (95%CI 0.19–0.56) for perinatal/neonatal mortality or serious neonatal morbidity[ 3 ]. The TBT results were subsequently added to two small trials in a Cochrane Systematic Review[ 4 ]. The reduction in perinatal morbidity and mortality was even more pronounced when the analyses were limited to births in low perinatal mortality countries, such as Australia (RR 0.13; 95%CI 0.05–0.31)[ 4 ]. However, planned caesarean section was associated with increased maternal morbidity (RR 1.29; 95% CI 1.03–1.61)[ 4 ]. These results have dramatically altered a woman's options if she has a breech presentation at term, as CS is now offered as the safest and in many institutions the only, management option. This change has occurred rapidly: The TBT was published in October 2000 and the rate of vaginal breech birth in NSW declined from 17% in 1999 to 14% in 2000 and 4.5% in 2001[ 5 ]. However, while safer than vaginal breech birth, planned CS is not without risk[ 6 , 7 ]. Complications include increased risk of pulmonary embolism, infection [ 8 - 10 ], bleeding[ 9 , 11 ], damage to bladder and bowel[ 12 ], slower recovery from the birth[ 12 , 13 ], longer hospitalisation[ 11 ], respiratory difficulties for the baby [ 14 - 16 ], delayed bonding and breastfeeding[ 17 , 18 ], and compromise of future obstetric performance[ 17 , 19 - 21 ]. Therefore, the best way to avoid the increased risks associated with term breech presentation is to avoid it altogether, and this is possible via external cephalic version. External cephalic version (ECV) of the breech-presenting baby External cephalic version (ECV) is the turning of a breech baby to a cephalic presentation. Systematic review of six well designed randomised controlled trials demonstrates that among women with breech presentation in late pregnancy, ECV reduces both breech presentations in labour (RR = 0.42, 95%CI 0.35–0.50) and caesarean sections (RR = 0.52. 95%CI 0.39–0.71)[ 22 ]. Despite clear evidence of effectiveness and potential benefit, many women decline ECV for a variety of reasons. Both breech presentation and ECV success rates are strongly influenced by parity, with success rates reported as low as 25% for women having their first baby[ 23 ]. For other women, the inconvenience of extra clinic visits and the need for an IV line for tocolysis may be deterrents[ 24 , 25 ]. Approximately 35% of women undergoing ECV report mild or moderate discomfort during the procedure[ 26 ]. Other complications of ECV are either uncommon (e.g. transient fetal bradycardia [12%] or dizziness and palpitations from tocolysis [4%]) or rare (<1%) (e.g. profound fetal bradycardia, preterm labour, premature rupture of membranes and bleeding)[ 26 ]. The remote possibility of emergency CS (e.g. because of placental abruption following the procedure) is also recognised. For these and other reasons, women may have a preference for planned CS. An Australian study of decision making for CS conducted in 1996 included 62 women with a breech presentation[ 27 ]. Of these, 39 women were offered ECV and 12 (31%) "decided against it". Further, 37 women were offered vaginal breech birth but 14 (38%) women chose CS[ 27 ]. Women's views and information needs To obtain data on Australian women's views and information needs about ECV we undertook a cross-sectional study of women's knowledge, attitudes and decision-making preferences for the management of breech presentation[ 28 ]. Of 174 pregnant women respondents (97% response rate), almost 90% preferred vaginal delivery but only 66% had heard of ECV. After a brief written explanation of ECV 39% would choose ECV, 22% were unsure and 39% would not choose ECV. The reasons for not choosing ECV included concerns about safety for the baby (13%), that ECV doesn't guarantee vaginal delivery (12%) and preference for a caesarean section anyway (8%). Importantly, 95% of pregnant women wanted involvement in decision-making about breech presentation. Patient participation in clinical decision making It is now recognised that many consumers want to participate in clinical decisions about their health [ 29 - 31 ]. NHMRC states that good medical decision making should take account of patients' preferences and values[ 29 ], thus challenging health professionals to find ways of involving consumers/patients in decisions about their health. Yet little is currently known about how this can be effectively achieved. One method is to provide information to consumers about treatment options and likely outcomes. To assist informed decision making, such information must be unbiased and based on current, high quality, quantitative research evidence. However, patient information materials are often outdated, inaccurate, omit relevant data, fail to give a balanced view and ignore uncertainties and scientific controversies[ 31 , 32 ]. To help patients take a more active role in important clinical decisions, decision aids based on latest research evidence are being developed by several centres (for example the Ottawa Health Decision Center in Canada and the Foundation for Informed Medical Decision Making in the USA). Decision aids are defined by the Cochrane Collaboration[ 33 ] as "interventions designed to help people make specific and deliberative choices among options by providing (at minimum) information on the options and outcomes relevant to the person's health status". Additional strategies may include providing: information on the disease/condition; the probabilities of outcomes tailored to a person's health risk factors; an explicit values clarification exercise; examples of others' decisions; and guidance and coaching in the steps of decision making[ 33 ]. Decision aids are non-directive in the sense that they do not aim to steer the user towards any one option, but rather to support decision making which is informed, consistent with personal values and acted upon[ 34 ]. Decision aids have been found to improve patient knowledge and create more realistic expectations, to reduce decisional conflict (uncertainty about the course of action) and to stimulate patients to be more active in decision making without increasing anxiety[ 35 ]. Currently only 38 decision aids worldwide have been developed and carefully evaluated in randomised controlled trials[ 33 , 35 ]. Examples include hormone replacement therapy for postmenopausal women, anticoagulants for atrial fibrillation, PSA testing for prostate cancer and prenatal genetic screening. Until the publication of a series of evidence-based leaflets in the United Kingdom in 2002[ 36 , 37 ], no decision aids have been developed and evaluated in the context of obstetric care, although this is an area in which consumers are known to want to participate actively in decision making[ 38 ]. An Australian survey of 790 postpartum women found not having an active say in decisions about pregnancy care was associated with a sixfold increase in dissatisfaction among primiparas and a fifteen fold increase among multiparas[ 38 ]. Similarly in the UK, postpartum women rated an explanation of procedures and involvement in decision making as most important to satisfaction with care[ 39 ]. Further, neither obstetricians nor midwives appreciated the importance to women of "being told the major risks for each procedure"[ 39 ]. Decision making and breech presentation The management of breech presentation is a clinical decision that fulfils Eddy's criteria for a decision in which patients' values and preferences should be included[ 40 ]. The outcomes for the breech management options (ECV and planned CS), and women's preferences for the relative value of benefits compared to risks are variable and could result in decisional conflict. For such a clinical decision, a decision aid would be expected to improve patient knowledge and create more realistic expectations, to reduce decisional conflict and to stimulate patients to be more active in decision making without increasing anxiety[ 35 ]. Development and pilot-testing of the decision aid In 2002 we developed an evidence-based decision aid for women with a breech presentation in late pregnancy. In developing the decision aid we utilised the NHMRC guideline "How to prepare and present information for consumers of health services"[ 41 ] (developed in 1999 by a team led by Dr Barratt), and the Ottawa framework established and rigorously tested by the Ottawa Health Decision Center[ 34 ]. The decision aid includes a Workbook, Audiotape/CD and Worksheet. The workbook highlights key points (similar to a slide presentation) and the audiotape/CD connects these points in a narrative format, providing more detail than the workbook. The worksheet is a one-page sheet to be completed by the woman to record her decision making steps, to list any questions she needs answered before deciding, and to indicate her preferred role in this decision (she should decide, her health care provider should decide, they should decide together). Most importantly, the DA was designed to be non-directive in that it did not aim to steer the user towards any one option or increase or decrease intervention rates but rather act as an adjunct to care The decision aid was designed for women to use at home or in the clinical setting, and takes about 30 minutes to complete. The aural component is available on both audiotapes and CDs so participants can choose which they prefer to use. After working through the decision aid, the woman brings her completed worksheet to her next antenatal appointment to discuss her provisional decision with her health care provider before arriving at her final decision. The worksheet is also useful for the practitioner, who can see rapidly from it what evidence the patient has considered, what her values and preferences on this topic are and which way she is leaning in her decision. The decision aid was developed, pilot tested and revised with extensive consumer involvement, as outlined in the NHMRC guideline on preparing information for consumers[ 41 ]. Content was largely driven by consumers' questions and information needs as determined from the cross-sectional study[ 28 ] and from the process of drafting, pilot testing and re-drafting. A number of draft decision aids (including workbook, audiotape/CD script, and worksheet), were developed and each subjected to pilot testing and revision as we obtained feedback. The process of testing and revising started with the project group. The next phase included a review by a group of national and international content experts, including decision aid experts, obstetricians, midwives, perinatal epidemiologists and psychologists. Once we were convinced that the content was accurate the decision aid was pilot-tested amongst consumers. There were several rounds of consumer review and refinement. We pilot-tested with members of consumer organisation (Maternity Alliance) and in a convenience sample of pregnant and recently pregnant women. The next draft was pilot-tested amongst pregnant women attending the antenatal clinic, who may or may not have had a breech presentation. And finally we formally pilot tested the decision aid with women who had a breech presentation in late pregnancy and were at the point of decision making. Pilot-testing results included: 95% of participants found the decision aid clear and easy to understand and 80% thought there was enough information for them to make a decision. Over 90% found it very helpful and nearly all women would recommend it to others. After reviewing the decision aid, women experienced a significant increase in their knowledge scores, less anxiety, had no difficulty making decisions and were satisfied with their decision. This study aims to evaluate the ECV Decision Aid for women with a breech-presenting baby in late pregnancy. The decision aid is based on the most recently available evidence and will be evaluated to assess the impact on women's satisfaction with decision-making, knowledge, anxiety and pregnancy outcomes. If successful, the results could be applied to a improve consumer information and participation in clinical decisions across a wide spectrum of pregnancy care issues. Methods Specific aim To compare the relative effectiveness of the decision aid with standard care in relation to women's knowledge, expectations, satisfaction with and participation in decision making, anxiety and decisional conflict. Secondary outcomes will include service utilisation and perinatal outcomes. Hypotheses The primary hypotheses of the study are: Use of the ECV decision aid by women with a singleton breech-presenting infant in late pregnancy 1. increases knowledge about breech presentation and the management options 2. reduces decisional conflict (uncertainty about the course of action) 3. increases satisfaction with decision making 4. reduces anxiety The secondary hypotheses of the study are: Use of the ECV decision aid by women with a singleton breech-presenting infant in late pregnancy does not influence 1. uptake of ECV 2. the proportion of women having a planned caesarean section for breech presentation at term 3. maternal and infant outcomes Study design We will use a randomised trial with the following study groups to assess the impact of the decision aid: Group 1: Usual care (usual antenatal care provider counselling on the management of breech presentation) Group 2: Usual care + Decision aid with review by a research midwife As randomisation will be done at the individual level, there is a risk of contamination of the usual care group if the usual care provider also reviews the decision aid with women in the study group. Therefore the decision aid will be reviewed with a research midwife and the usual antenatal care providers will be blinded to the exact content and format of the decision aid. Setting Australian obstetric hospitals that offer external cephalic version. Participants/eligibility criteria Women with a single breech-presenting baby at ≥ 34 weeks gestation and who are clinically eligible for ECV will be invited to participate in the trial. Exclusion criteria are therefore those for ECV and include women presenting with a breech in labour, multiple pregnancies, previous CS, severe fetal anomaly, ruptured membranes and indications for CS anyway. The decision aid will be produced in English and will be designed to be simple and accessible for women with low levels of literacy. The use of audiotapes and graphics will further aid comprehension and ensure that, as far as possible, women with low English literacy need not be excluded. Procedures, recruitment, randomisation and collection of baseline data The study procedure will draw upon the usual schedule of antenatal visits becoming weekly in late pregnancy and the usual management of a breech-presentation in the late pregnancy (Figure 1 ). Women diagnosed with a breech-presenting baby at ≥ 34 weeks gestation will be asked to participate. The research midwife will explain the trial and obtain informed consent, collect baseline data and randomly allocate (using telephone randomisation) women to study or control groups. This is only a minor deviation from current practice. As women of child-bearing age are known to be very mobile, participants will be asked to provide alternate contact details (eg friend or relative) to enhance subsequent follow-up. Private obstetricians will be asked to offer their patients participation in the study. Those interested will be requested to come to the antenatal clinic for randomisation and recruitment. The private obstetrician will provide usual care. Figure 1 Schema of ECV decision aid trial Intervention The study group will receive the decision aid (workbook, tape or CD, and worksheet) and the control group will receive standard information on management options for breech presentation from their usual carer. The study group will be given the opportunity to work through the decision aid while in the antenatal clinic and/or to take home, which ever is most convenient. Many women will also want to discuss the decision with their partner. This pragmatic approach aims to assess the decision aid under the conditions most likely to be applied in real practice. At the next antenatal visit, women in the study group will review their decision aid worksheet and any questions with the research nurse. Follow-up 1 All women will be given a follow-up questionnaire to complete prior to their next antenatal clinic visit (see Outcome Measures below for more detail). All women choosing ECV will be given the opportunity to discuss the procedure with the obstetrician providing the service. This discussion is likely to include some of the probabilistic information in the decision aid but will occur after the follow-up questionnaire and will not influence the outcome measures. Follow-up 2 At 12–16 weeks post-partum all participants will be mailed a second follow-up questionnaire. This will assess women's satisfaction with their decision and the decision making process when the events are past and the outcomes known. (See Outcome Measures below for more detail). Questionnaires will be mailed with reply paid envelopes, with up to two reminder telephone prompts to non-responders. Blinding and contamination As in many obstetric interventions double blinding is virtually impossible. The main outcomes of this study are self-reported and the women are clearly not blinded to their treatment allocation. However, we will institute a number of measures aimed at keeping antenatal staff blind to the treatment allocation and preventing contamination of the control group: • Women will review the decision aid with the research midwife and complete the first questionnaire (primary outcome measures) prior to their next antenatal consultation • Usual antenatal care providers will be blinded to the content and format of the decision aid • Regular in-service (educational training) for the antenatal care providers to explain the trial protocol and to make clear the potential effect of unmasking or contamination. • Monitoring decision aid distribution and keeping them locked up and only accessible by the research midwife • Asking participants not to reveal their treatment allocation, or share their decision aid material with antenatal staff or other women. If participants do not want to keep their decision aid they will be asked to return it. • Monitoring the "usual care" (control) arm by conducting a run-in period in which women found to have a breech presentation will be asked to complete the 1 st follow-up questionnaire. Thus we will have a baseline record of knowledge about ECV, anxiety and decisional conflict about the decision and satisfaction with the decision before the DA is in use. Comparison of the data obtained from this run-in period and the control arm will allow us to judge whether, and to what extent, contamination has occurred. Outcome measures Baseline data collection Brief baseline data will be collected to assess comparability of the study groups. The baseline assessment will include age, parity, brief socio-demographic data, highest level of education achieved, knowledge and anxiety as assessed by the state component of the short Spielberger anxiety scale[ 42 ]. Primary outcomes The effectiveness of risk communication to aid patient decision making is best assessed by a combination of cognitive, affective and behavioural outcomes[ 43 ]. Thus the primary outcomes of the this study will be • cognitive: change in knowledge and realistic expectations of the management options and possible benefits and risks of each option • affective: anxiety, satisfaction with the decision, participation in decision-making and the amount of decisional conflict (uncertainty about which course of action to choose) experienced • behavioural: actual decision taken and acted upon (see secondary outcomes) Measures of knowledge and realistic expectations about options for the management of breech presentation and the benefits and risks of ECV will be specific to this project. Thus we will need to develop, and test these measures as part of the project. Anxiety will be measured by the state component of the short Spielberger anxiety scale which has been extensively used and validated[ 42 ]. We do not anticipate the decision aid will increase women's anxiety but it is nevertheless important to document any increase or decrease in anxiety attributable to the decision aid. Satisfaction with the decision will be assessed using the Satisfaction with Decision Scale. Satisfaction with Decision Scale (a very brief six item scale with high reliability) was developed specifically to assess satisfaction with health care decisions[ 44 ]. Participation in decision-making will be ascertained using the five-item Degner Control Preferences Scale[ 45 ]. This allows respondents to specify the degree of control in decision-making they wish to assume with their doctor. Decisional conflict will be assessed by the Decisional Conflict Scale which has established reliability, good psychometric properties and is short (16 items)[46]. It has been used to evaluate a range of decision aids[ 35 ]. Because the decision about ECV must be made within a short timeframe, the outcomes will be measured as soon as practical after the consultation in which the ECV decision was made – prior to the next antenatal visit. For the primary outcomes this will be within one week of the decision being made (Figure 1 , 1 st follow-up). Satisfaction with the decision and anxiety will be measured again at 12–16 weeks postpartum as the last weeks of pregnancy and the week after birth are associated with a reduction in state anxiety[ 47 ]. We are interested to explore whether women's views of the decision making process, and the decision they ultimately made, may change with time to reflect on the experience (Figure 1 , 2 nd follow-up). Secondary outcomes The aim of the decision aid is to assist patient decision making, and not to influence the direction of the decision taken. Nevertheless, we think it is important to collect service utilisation and pregnancy outcome data so we will record and compare the numbers of ECVs undergone and ECV success rate in both arms of the study, as well as recording and comparing rates of pregnancy complications and perinatal outcomes. Data on ECVs are already prospectively collected for quality assurance, these include fetal lie, parity, success rates and complications. Other perinatal outcomes will be obtained (with informed consent) from the existing computerised obstetric database. These outcomes include mode of delivery (vaginal, emergency or planned CS), enrolment to delivery interval, gestational age, birthweight, Apgar scores, perinatal deaths, Neonatal Intensive Care Unit admission, maternal haemorrhage (antepartum or postpartum) and length of stay. Statistical issues Sample size Sample size calculations for the trial (significance 0.05, power 0.8) were determined using the mean difference we would like to detect in women's decisional conflict scores and knowledge of options and outcomes. Compared with usual care, decisional conflict was shown in the most current systematic review to be significantly reduced by decision aids; the meta-analysed mean difference was -5.75, 95%CI -8.63, -2.87 (on a scale ranging from 1 lowest to 5 highest decisional conflict; median standard deviation 13.25)[ 48 ]. Assuming a mean difference of -5.75 and standard deviation 13.25, we would need approximately 84 women in each arm to demonstrate changes in decisional conflict. The meta-analysis also showed that for nine trials comparing decision aids and usual care, decision aids improved average knowledge scores by 18.75 points (out of 100) (95%CI 13.1 to 24.4, median standard deviation 20)[ 48 ]. To show such a difference, assuming mean difference of 18.75 and standard deviation of 20, would require only 18 women in each arm. Because we would like to be able to show differences in decisional conflict if they exist, we have used the larger sample size estimate. Although follow-up will be relatively short term, there will inevitably be some loss to follow-up. To allow for 10% loss to follow-up the sample size calculated above (84) is inflated by 10% to give the effective sample size of 92 women per arm and a total sample of 184 for the trial. This sample size is different from our original application for funding. Originally, we estimated a sample size of 310 women (155 in each arm). This was based on results from a 1999 systematic review of only 2 trials of decision aids versus usual care that had assessed decisional conflict[ 35 ]. Subsequent to the submission (January 2001) and funding of this protocol (March 2002) the systematic review was updated (2002) incorporating 6 trials that assessed decisional conflict[ 48 ]. At that time we revised the sample size estimate to incorporate the most current research evidence available. Ethics approval was obtained for the protocol amendment. Data analysis Analyses will be by intention to treat, including withdrawals and losses to follow-up. Study groups will be compared in terms of baseline characteristics. As this is a randomised trial, we would anticipate minimal differences in baseline characteristics. If however, important differences are found, these potential confounders will be adjusted for in the analysis of outcomes. For the primary outcomes, the mean score for each measure for each group will be compared using t-tests. If adjustment for confounders is needed a multiple linear regression model will be used. The secondary outcomes will be compared using chi-square tests of significance for categorical data and t-tests for continuous data. If adjustment for confounding is necessary logistic regression and multiple linear regression will be used respectively. Interim analysis An interim analysis will be conducted part way through the study and the results will be reviewed by an independent Data Monitoring Committee. Specifically the incidence of anxiety and decisional conflict in the two randomised groups will be determined after the first 150 women have been enrolled and data have been collected. If there is a significant increase in either of these outcomes at p < 0.01 (1-tailed) with the decision aid, the trial will be stopped. The trial will also be stopped of it is evident that no clear outcome will be obtained. Ethical considerations We expect the project to provide ethical benefit. It is possible that some women may experience heightened anxiety as a result of receiving the decision aid during its evaluation by randomised trial. However, a systematic review of decision aids found they improved knowledge without increasing anxiety[ 33 ]. Nevertheless we will measure anxiety levels at baseline and follow-up to document any adverse effects. A trained research midwife will interview all women and obtain written consent for the trial. Women will be encouraged to discuss any concerns or anxiety about the project with the research midwife and/or with their usual antenatal care provider. Women will be reassured that they are able to drop out of the study at any time with no adverse effects on the management of their pregnancy. Participation will require women to complete brief self-report questionnaires during and after pregnancy. Working through the decision aid will take ~ 30 minutes and review of the decision and any outstanding questions will be at a routine antenatal visit. The study has been approved by the Central Sydney Area Health Service Ethics Review Committee (Protocol no. X01-0067) and the University of Sydney Human Ethics Committee (Ref No. 3806). Confidentiality and data security Participants in the trial will be identified by a study number only, with a master code sheet linking names with numbers being held securely and separately from the study data. To ensure that all information is secure, data records will be kept in a secure location at the University of Sydney and accessible only to research staff. As soon as all follow-up is completed the data records will be de-identified. De-identified data will be used for the statistical analysis and all publications will include only aggregated data. The electronic version of the data will be maintained on a computer protected by password. All hard copy patient identifiable data and electronic backup files will be kept in locked cabinets, which are held in a locked room accessed only by security code and limited staff. Data files will be stored for 7 (seven) years after completion of the project as recommended by the NHMRC. Disposal of identifiable information will be done through the use of designated bags and/or a shredding machine. Outcomes and significance This project will make an important contribution to a largely neglected aspect of pregnancy care, assisting informed participation by women in clinical decisions that affect their pregnancy. Involvement in decision making is a strong predictor of satisfaction with care in pregnancy and childbirth, yet there are only a few published decision aids for maternity care. A decision aid for the management of breech presentation is both timely and practical as there is new evidence supporting planned CS, dramatically altering the management options. Further, the randomised trial will provide high quality evidence about the effectiveness of the decision aid in supporting shared clinical decision making during pregnancy. If successful, the results of this project could be applied to improve consumer information and participation in clinical decisions across a wide spectrum of pregnancy care. Finally, if the decision aid increases the utilisation of ECV, in addition to reducing breech presentation and CS for breech presentation (and the associated increased hospitalisation and potential morbidities), some women may have more choice of where they give birth as breech presentation precludes birth in birth centres and small rural hospitals. Competing interests The author(s) declare that they have no competing interests. Authors' contributions CR, AB, CRG, BP, DHS were involved in the conception and design of the study. CR, NN and CRG were responsible for the drafting of the protocol and NN and CR were involved in the development and implementation of the study. All authors have read and given final approval of the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545961.xml
539300	Management of acute renal colic in the UK: a questionnaire survey	Background There is great variation in the Accident and Emergency workload and location of Urology services in UK hospitals. This study investigated the relationship of the initial management of acute renal colic with the department workload plus local facilities including location of X-ray and urology services in UK Accident and Emergency (A&E) departments. Methods A&E departments in each of the 11 UK Deanery regions were stratified based on departmental workload, namely <30,000 ( small ); 30,000 to 50,000 ( medium ); 50,000 to 80,000 ( large ) and >80,000 ( very large ) patients per year. One third of departments were selected in each group leading to a sample size of 106. A questionnaire was administered. Associations between categorical variables were investigated using the chi-squared test and when not valid, Fisher's Exact test was employed. Differences between groups in ordinal variables were investigated using the Mann-Whitney test. Results All questionnaires were returned. Twenty-nine units (27.4%) did not perform any radiological investigation on renal colic patients. The number of radiological investigations that were available to departments was associated with workload ( P = 0.003); with 57.1% of the small departments performing none and at least 82.8% of units in the other categories performing at least one. Of those departments with X-ray facilities in or adjacent to the department, 63% performed an intravenous urography (IVU) compared to 25% of those departments without ( P = 0.026). Of those departments with on-site urology services, 86% performed at least one radiological investigation compared to 52% of units without such services ( P = 0.001). Department workload was associated with the first choice analgesia (NSAIDs or parenteral opiates) ( P = 0.011). Of the small departments, 64.3% used NSAIDs, 21.4% used parenteral opiates and 14.3% used neither. In comparison, NSAIDS were used by at least 87%, and opiates by at most 12.5% of units in each of the other three categories of department workload. Conclusions Over a quarter of UK A&E departments did not perform any radiological investigations and some departments do not even offer renal colic patients any analgesia. Patient management was associated with departmental workload, location of X-ray and Urology services. National guidelines are needed to ensure optimum care for all patients.	Background Upon presentation to the A&E department, suspected acute renal colic patients must have a clinical examination and radiological investigations to confirm the diagnosis. Without radiological investigations, life-threatening conditions such as abdominal aortic aneurysm and ectopic pregnancy may be misdiagnosed as renal colic. However, a delay in the diagnosis is possible as the facilities needed for the diagnosis are sometimes not based in the same hospital as the A&E department. In the UK, great variation exists between Accident and Emergency services in their workload as measured by the number of new patients seen per year[ 1 ]. Larger A&E departments tend to be located in those teaching hospitals that have most specialist services on-site, therefore facilitating adequate investigation of suspected renal colic. Every A&E department in the UK should be able to perform initial assessment and investigation of suspected renal colic, provide pain relief and refer appropriately, irrespective of the location of urology services. If acute renal colic presented to primary care then the patient would be rapidly referred to secondary care, namely an A&E department. The aims of this study were to investigate the initial management of acute renal colic in UK A&E departments and if practice was related to department workload, plus the location of X-ray and urological services in relation to the A&E department. Methods The handbook of the British Association for Accident and Emergency lists a total of 311 A&E departments in the UK[ 1 ]. The UK is divided into 11 so-called Deanary regions that represent geographical areas. These units were categorised according to their workload (number of new patients seen per year) as follows: Small - less than 30,000, Medium - 30,000 to 50,000, Large -50,000 to 80,000 and Very Large -more than 80,000. For each of the four-workload categories in each of the 11 Deanery regions, every third unit was selected resulting in a total of 106 (34.1%) departments. A questionnaire was administered by post to the 106 departments (see Appendix 1) requesting details about the location of X-ray, location of Urology services plus current practice in the investigation and management of pain in acute renal colic patients. A covering letter was included indicating that the purpose of the survey was to collect information about practice when patients present to the A&E department and not their subsequent management. The most senior medical member of each department was invited to complete the questionnaire. Over a period of nine months, each of the 106 departments returned a completed questionnaire. A total of 35 departments did not respond initially and they were sent a second questionnaire by post as a reminder. Ten departments did not respond to the second questionnaire and these were followed up with a telephone call. Consultants completed the questionnaire in 74.5% (n = 79) of units. Middle grade doctors who had been in post for at least six months completed the remaining 27 (25.5%) questionnaires. Statistical methods The Chi-squared test (test statistic denoted by χ 2 ) was used to investigate the following associations: a) the location of X-ray and intravenous urography; b) location of Urology services and total number of investigations performed and c) categorised departmental workload with the investigations performed and also the analgesics used. When the Chi-squared test was invalid, Fisher's Exact test (test statistic denoted by FI) was employed. The Chi-Squared was considered to be invalid if more than 20% of the cells had an expected value less than five or if one of the cells had an expected value less than one [ 2 ]. The categorised departmental workload groups were compared in the number of films used during an IVU procedure using the Kruskal-Wallis test (test statistic approximated to the Chi-Squared distribution and denoted by χ 2 ) [ 3 ]. Degrees of freedom were abbreviated to df. The critical significance level was 0.05. All statistical analyses were performed using SPSS for Windows (version 11). Results On-site services Of the 106 departments, a total of 94 (88.7%) had X-ray facilities located in the department. A greater proportion of those departments that have X-ray facilities within their premises used the Intra-Venous Urogram (IVU) option compared to those departments without these facilities [n = 59 (62.5%) versus n = 3 (25%); FI = 6.03, df = 1, P = 0.026]. Urology was located within the hospital for 64 (60.4%) departments. The total number of radiological investigations [IVU, Ultrasound Scan (USS) or Computed Tomogram (CT)] that were available to units was categorised as none, one and two or more. Those departments that had urology on-site had more radiological options available than those without ( P = 0.001) (see Table 1 ). At least one radiological option was used by 85.9% (n = 55) of those units with on-site urology services compared to 52.3% (n = 22) of units without. Table 1 Association between total number of radiological investigations performed and location of urology services. Total Number of Investigations Location of Urology services None One Two or three Total Within Hospital 9 (14.1%) 36 (56.3%) 19 (29.7%) 64 (60.4%) Outside Hospital 20 (47.6%) 13 (31.0%) 9 (21.4%) 42 (39.6%) Percentages in brackets are those within the category of the location of urology services; those in the 'total' column are those for the whole sample (n = 106). There was a significant difference between hospitals as regards their location of services in the number of investigations performed (χ 2 = 14.6, df = 2, P = 0.0007). None of the departments in our study routinely used nuclear medicine to investigate renal colic. Radiological investigations Intra-Venous Urogram (IVU) A significant relationship existed between department workload and if an IVU option was available ( P = 0.001) (see Table 2 ). An IVU option was available to 28.6% of the small departments, compared to at least 62.5% of those units in the larger categories, namely the medium , large and very large departments. Table 2 Tabulation of department workload by radiological investigations performed plus total number of investigations, and number of films used in IVU investigations. Number of new patients per year < 30,000 30,000 to 50,000 50,000 to 80,000 >80,000 All departments Test statistics IVU performed No 20(71.4%) 8 (22.9%) 13 (37.1%) 3 (37.5%) 44 (41.5%) FI = 15.54, df = 3, P = 0.001 Yes 8 (28.6%) 27 (77.1%) 22 (62.9%) 5 (62.5%) 62 (58.5%) USS performed No 21 (75.0%) 25 (71.4%) 17 (48.6%) 4 (50.0%) 67 (63.2%) χ 2 = 6.52, df = 3, P = 0.089 Yes 7 (25.0%) 10 (28.6%) 18 (51.4%) 4 (50.0%) 39 (36.8%) CT performed No 27 (96.4%) 33 (94.3%) 32 (91.4%) 5 (62.5%) 97 (91.5%) FI = 6.87, df = 3, P = 0.056 Yes 1 (3.6%) 2 (5.7%) 3 (8.6%) 3 (37.5%) 9 (8.5%) Total number of investigations None 16 (57.1%) 6 (17.1%) 6 (17.1%) 1 (12.5%) 29 (27.4%) FI = 18.85, df = 6, P = 0.003 One 9 (32.2%) 21 (60.0%) 16 (45.7%) 3 (37.5%) 49 (46.2%) Two or three 3 (10.7%) 8 (22.9%) 13 (37.2%) 4 (50.0%) 28 (26.4%) If IVU, total Number of films n 8 27 22 5 62 χ 2 = 6.68, df = 3, P = 0.083 mean 3.0 2.9 2.6 1.6 2.7 standard deviation 1.14 0.97 1.05 0.89 1.10 median 3 3 3 1 3 lower quartile 2.0 3.0 2.0 1.0 2 upper quartile 4.5 3.0 3.0 2.5 3 Percentages in brackets are those of the grouped department workload; those in the "All departments" column are of the 106 units. The test statistics comparing the four groups of department size are displayed. The relationship between department workload and the average number of films used when an IVU was performed is shown in Table 2 . Of the 106 departments, 43 (40.6%) did not undertake an IVU leaving a total of 63 (59.4%) units for analysis. All of these 63 departments used between one and five films per IVU investigation except for the very large category where the greatest number of films used by a department was three. The very large departments used a median number of a single film whilst the other three categories of department size used a median number of three films. Although there was a tendency for fewer films to be used as departmental size increased, this just failed to reach statistical significance at the 5% level ( P = 0.083). Ultrasound Scan (USS) There was no statistically significant relationship between department workload and if an USS option was available ( P = 0.089). However, at least half of the large and the very large units used USS compared to less than 30% of the departments in the small and medium sized categories (see Table 2 ). Overall, 36.8% (n = 39) of departments were able to perform an USS. Computerised Tomogram (CT) Scan-Helical CT The relationship between department workload and if a CT scan was available just missed statistical significance ( P = 0.056) (see Table 2 ). Of the very large departments, 37.5% (n = 3) could perform a CT scan compared to less than 10% of the units in each of the small , medium and large categories. Total number of radiological investigations A total of 29 units (27.4%) did not perform any radiological investigations. The relationship between the total number of investigations available and department workload was statistically significant ( P = 0.003) (see Table 2 ). No radiological investigations were carried out by 16 (57.1%) of the small departments whilst at least 83% of the units in each of the other three departmental workload categories were able to perform at least one radiological investigation. Exactly half (n = 4) of the very large departments had at least two options available. Choice of analgesia There was a statistically significant relationship between department workload and the first choice analgesia: either NSAIDs (Diclofenac or Ketorolac) or parenteral opiates ( P = 0.011) (see Table 3 ). Parenteral opiates were used by 21.4% (n = 6) of the small departments compared to at most 12.5% of units in the other workload categories. Neither NSAIDs nor parenteral opiate was used by four (14.3%) of the small departments and one large department; one of these small units plus the large department reported using codydramol (a combination of paracetamol with dihydrocodeine). Of the 106 departments, 91 (85.8%) used NSAIDs including 86 (81.1%) – diclofenac and five (4.7%)- ketorolac as the first choice analgesia. Of the 86 departments that used diclofenac, 68 (79.1%) routinely used the intra-muscular route, 17 (19.7%) the rectal route and one (1.2%) administered it orally. Table 3 First choice analgesic (either NSAIDs, Parenteral opiates or neither) by department workload (n = 106). First choice analgesia (NSAIDs or Parenteral opiates) Number of new patients per year All departments < 30,000 30,000 to 50,000 50,000 to 80,000 > 80,000 None used 4 (14.3%) 0 1 (2.9%) 0 5 (4.7%) NSAIDs 18 (64.3%) 34 (97.1%) 32 (91.4%) 7 (87.5%) 91 (85.8%) Parenteral opiates 6 (21.4%) 1 (2.9%) 2 (5.7%) 1 (12.5%) 10 (9.4%) Percentages in brackets are those of the grouped departmental workload; those in the "All departments" column are of the 106 units. There was significant difference between department workloads in first choice analgesia (FI = 13.49, df = 6, P = 0.011). Discussion This study reports the initial management of renal colic irrespective of which specialty team carried out the management. Traditionally, renal colic was confirmed by IVU alone [ 4 ] although the use of USS and helical CT scans has increased in current practice [ 5 ]. A study of a single department reported that up to 37% of patients with suspected renal colic were investigated with ultrasound, although this included mainly patients with allergy to the contrast used in IVU and those in early pregnancy when irradiation needs to be avoided [ 6 ]. There may to be an upward trend in the use of USS in A&E departments due to the current drive for USS by non-radiologists [ 7 - 9 ]. Our study found that only a quarter of UK units used USS although these included at least half of each of the large and very large departments. Radiological investigations confirm or refute a diagnosis of renal colic. If the diagnosis is refuted, then the clinician is prompted to consider other diagnoses. We found that over a quarter of departments (27.4%; n = 29) did not perform any radiological investigation (see Table 2 ). This is of concern since it has been reported that renal colic is one of the most common misdiagnoses in catastrophic abdominal conditions including ectopic pregnancy and abdominal aortic aneurysm [ 10 ]. The concern is greatest for those departments in the small category; nearly 60% of them did not routinely perform any radiological investigations and they may be located in remote areas lacking specialist surgical facilities such as on-site vascular surgery. When departments are isolated with minimal specialist back up, an early diagnosis is crucial, so that patients with other abdominal conditions requiring urgent specialist management can be appropriately referred. An IVU can be easily done in the X-ray department; a negative IVU should prompt the clinician to consider an alternative diagnosis to renal colic and this, in our opinion should not be beyond the reach of any A&E department in the UK. Intra-venous urography was performed by a significantly greater proportion of those departments with X-ray facilities within the unit compared to those with X-ray facilities located elsewhere. This would suggest that if all A&E departments had X-ray facilities located within the unit, the potential for misdiagnosis would be minimised since all units would then be more likely to perform at least an IVU. Those hospitals that had on-site urology services performed more investigations than those sites without (see Table 1 ). In particular nearly half of those hospitals with urology services located outside the hospital did not perform any investigations at all compared to 14% of those hospitals with on-site services. This potentially means that patients with conditions other than renal colic are sent to a urology clinic with the consequence that their management is delayed. We found that when an IVU was performed, the larger units used fewer films although this relationship just missed statistical significance (see Table 2 ). This finding suggests that adequate information to diagnose renal colic might be obtained by using only one film, in keeping with previous findings [ 11 ]. However, these suggestions need to be verified by further research. We found that less than 10% of UK A&E departments use a CT scan in the assessment of renal colic and the association with departmental workload just missed statistical significance at the 5% level (see Table 2 ). A CT scan was available to 37.5% of the very large departments compared to less than 10% in the other sized categories. The main difficulty with performing a CT scan in an A&E setting is that interpretation of the images requires urologists or radiologists who are not always available [ 5 ]. When appropriate personnel are available, CT should be the favoured investigation as it has been shown not only to diagnose urinary tract calculi accurately but also provide other diagnoses. The choice of investigation in some of the units that reported using more than one type of radiological investigation may have been influenced by availability of the required personnel, as both USS and CT require some expertise. However, this study did not investigate this aspect of practice. Previous research has shown the efficacy of NSAIDS in renal colic [ 12 - 17 ]. We found that 85.8% of UK A&E departments use NSAIDS. Intra-venous ketorolac reported to have the fastest onset of action and equal analgesic properties to other NSAIDS, was used by only 4.7% of units although its use may have been precluded by difficulty with venous access. Intra-muscular diclofenac was routinely used in 64% of departments despite the problems associated with this route including discomfort at the injection site and the potential for sterile abscess formation [ 16 ]. Whilst the rectal route is favoured over the intra-muscular route since it is equally effective and avoids possible injection site problems, only 16% of all departments in this study reported using it. In spite of the proven efficacy of NSAIDS, we found that nearly 10% of all A&E departments used parenteral opiates as the analgesic of first choice. Given that opiate administration requires checking and crosschecking by at least two nurses, there will inevitably be a delay in relieving the patients' pain. Parenteral opiates would be better as second-choice analgesic in our opinion. We found five departments, including four in the small and one in the large workload categories that did not use either NSAIDs or parenteral opiates in suspected renal colic. The large and one of the small departments prescribed Codydramol. The other three small departments referred renal colic patients directly to a urology team off-site without even offering analgesia. Why these departments adopted this approach was unclear. Nonetheless, these findings were of concern since an A&E department would be expected to at least consider offering analgesia to patients that present to them in pain irrespective of the patients' final destination. There is no reason to suspect that the departments in this study were not representative of A&E units in the UK, since the sample was derived from each of the workload categories in the 11 UK Deanery regions. However, as with any questionnaire study it is difficult to assess the reliability of the answers provided. The most senior individual in the department was invited to complete the questionnaire. However, it is not possible to quantify the bias, if any, that may be introduced by the variation in grade of the respondents. Since nearly three quarters of the questionnaires were completed by consultants one might expect that that the results were reliable. However, it is possible that consultants may not be fully aware of routine practice and therefore the information provided could be inaccurate. Furthermore, as soon as a unit was asked about its current practice through the questionnaire, it may subsequently have adjusted it, particularly if it was sub-optimal. Therefore, the information provided on the questionnaire may reflect the altered, rather than original practice. Conclusions The management of acute renal colic differs between A&E departments in the UK. Local factors may contribute to these differences. The total number of radiological procedures that were available to a unit was positively associated with departmental workload. Of great concern was that a significant proportion of departments overall (27.4%) did not perform any radiological investigation. The concern is greatest for those departments in the small category with nearly 60% performing no radiological investigation. Location of X-ray facilities within the A&E premises is associated with whether an IVU is ever performed. Departments with on-site urology have a greater range of radiological investigations to choose from. Furthermore, the very large units tended to routinely use fewer films per IVU, with a median number of one compared to three in all other smaller units. The first choice analgesic used by most units is NSAIDS in keeping with the literature; more departments, however, need to adopt the use of the rectal route for diclofenac in order to avoid the potential complication of the intra-muscular route. The low percentage of departments using parenteral opiates as first-choice analgesic is encouraging as parenteral opiates are better used as second choice in view of the unavoidable delay that occurs before their administration. The practice in over a quarter of A&E departments in the UK is below standard. There is significant association with departmental workload and location of services such as radiology and Urology relative to A&E. We suggest that national guidelines be developed for the management of acute renal colic in A&E departments to ensure optimum care for all patients. Subsequent to the implementation of any guidelines, we suggest that UK practice is regularly reviewed. Competing interests The author(s) declare that they have no competing interests. Authors' contributions TAL undertook the literature search, designed the questionnaire, participated in the collection and analysis of data, and wrote the paper. PMS undertook the statistical analysis and contributed to the writing of the paper. NN initiated the research, participated in data collection and contributed to the paper. CS performed the initial stratification and selection of units for the study. NP participated in data collection. Appendix 1 Questionnaire on renal colic 1. Name of hospital: ____________________________________________ 2. Where is X-ray located? Within or adjacent to A&E □ Distant □ 3. Where are urology services located? Same Site as A&E □ Separate Site from A&E □ 4. Are the following investigations performed on suspected cases of renal colic? a) Urinalysis No □ Yes □ b) IVU No □ Yes □ If IVU is performed, please indicate how many films are used □ c) CT No □ Yes □ d) USS No □ Yes □ e) Nuclear Medicine No □ Yes □ 5. Which of the following analgesics are given on presentation? a) Codydramol Other Oral No □ Yes □ b) NSAIDS No □ Yes □ If NSAIDS used; which one: (indicate route below) i. Intra-muscular □ ii. Oral □ iii. Rectal □ iv. Intra-venous □ c) Parenteral opiate No □ Yes □ If parenteral opiates are used then please indicate if first or second choice: i. First choice □ ii. Second choice □ Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC539300.xml
524379	Supervised Treatment Interruptions Fail to Control HIV-1 Viremia	null	Highly active antiretroviral therapy (HAART) for the treatment of individuals infected by HIV-1 is limited by high costs, drug resistance, and drug-related toxicities. This has led researchers to investigate new treatment options, including ways to boost immune responses to better control HIV. One such approach has been termed supervised treatment interruption (STI)—in which HAART is intermittently stopped once viral load has been reduced to a low level, in order to boost natural immunity by brief exposure to virus. The goal is to allow for the eventual discontinuation of drug treatment. ELISPOT assays detect HIV-specific cytotoxic T lymphocyte responses Preliminary evidence, published by Bruce Walker and his colleagues from Harvard Medical School in Nature in 2000, suggested that this approach worked in persons treated in the earliest stages of acute HIV infection. HIV-1 viral loads in newly infected patients remained suppressed for a median of six months after therapy had been stopped. However, a follow up paper, published this month in PLoS Medicine by the same research group, shows that the viral load rebounded in eight of the 14 patients by one year. “The findings are very straightforward and very important,” comments Danny Douek from the Vaccine Research Center, National Institutes of Health, United States, who was not involved in the study. “In almost every case, virus rebounded and no clinical benefit from the interruption could be determined.” Walker's team first considered the possibility of STI in 1997 after they demonstrated that HAART given to patients recently infected with HIV could protect T helper cells, which are normally destroyed in the earliest stages of infection. They hypothesized that early treatment of acute HIV-1 infection with HAART might boost the immune response, allowing it to control the HIV-1 infection without the need for continuous therapy. “We did not know at that time whether the T helper cells would be functional,” explains Walker. “The only way to tell this was to stop medications and see if the immune response could control the virus.” To test this hypothesis the researchers did an open-label trial of STIs; they published data from six months follow-up in the Nature paper. “The key finding was that we were able to get at least transient control of virus in all eight persons studied, and in five of eight the viral load was less than 500 copies (very low!) at the time of publication,” explains Walker. However, at that point they did not know how long the protective effects would last. The first evidence that protection was not complete came two years later when Walker's team reported a case of superinfection; one of the patients in the original experiment was infected with a second strain of HIV, even though the first virus was still well controlled. “This paper was important because it indicated that the amount of immunity might be enough for the person's own virus, but might not protect against closely related viruses circulating in the population,” says Walker. The PLoS Medicine study adds more concern since it shows that although most persons can indeed transiently control their own virus, they do so for only a limited amount of time. “We expanded the study to 14 persons, and now have about five years of follow-up on some of the patients,” says Walker. “Although we were able to use early treatment and structured treatment interruption to boost immunity and have 11 of 14 patients control their virus, most of the persons ultimately ‘broke through,’ meaning that they had a recurrence of viremia.” At the present time the researchers do not know what causes the loss of viral control. Walker and colleagues conclude that treatment interruptions should probably be avoided outside the setting of controlled clinical trials, whereas Douek goes a step further: “The study shows that even early short-term treatment and structured treatment interruptions, using current strategies, impart only transient benefit and are unlikely to serve as a reasonable therapeutic option in the future.”	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524379.xml
549195	Home injuries and built form – methodological issues and developments in database linkage	Background The aim of this body of research is to determine whether injuries in the home are more common in particular types of housing. Previous home injuries research has tended to focus on behaviours or the provision of safety equipment to families with young children. There has been little consideration of the physical environment. This study reports methodological developments in database linkage and analysis to improve researchers abilities to utilise large administrative and clinical databases to carry out health and health services research. Methods The study involved linking a database of home injuries obtained from an emergency department surveillance system with an external survey of all homes in an area and population denominators for home types derived from a health service administrative database. Analysis of injury incidence by housing type was adjusted for potential biases due to deprivation and distance to hospital. For non-injured individuals data confidentiality considerations required the deprivation and distance measures be imputed. The process of randomly imputing these variables and the testing of the validity of this approach is detailed. Results There were 14,081 first injuries in 112,248 residents living in 54,081 homes over a two-year period. The imputation method worked well with imputed and observed measures in the injured group being very similar. Re-randomisation and a repeated analysis gave identical results to the first analysis. One particular housing type had a substantially elevated odds ratio for injury occurrence, OR = 2.07 (95% CI: 1.87 to 2.30). Conclusions The method of data linkage, imputation and statistical analysis used provides a basis for improved analysis of database linkage studies.	Background Home injuries are frequent and result in greater mortality and morbidity than road traffic injuries[ 1 ]. It is evident that the risks of some types of injuries, such as falls or injuries resulting from fire, could well be related to the built form of the home. Nevertheless the availability of robust evidence based on large numbers of cases is quite limited[ 2 ]. Previous home injuries research has tended to focus on behaviours or the provision of safety equipment to families with young children. There has been little consideration of the physical environment. Linkage of existing large datasets has the potential to yield substantial evidence. The present study was designed to utilise three such datasets relating to all residential properties in a defined geographical area in the United Kingdom, the resident population, and their attendances at local hospital emergency departments (EDs). These datasets could not be linked comprehensively at individual level on account of constraints on identifiability of individuals. This article describes the novel challenges that result, and the methodology used to obviate them. Methods This study was carried out as part of the wider Housing and Neighbourhood and Health (HANAH) project[ 3 ]. This is a long-term partnership between academia and local authorities to elucidate the relationship between the social and built environment and health and to develop interventions to improve health. Data from an injury surveillance system on injury events in residents of the Neath-Port Talbot County Borough Council area was linked to a register of property types and denominator data. Ethical permission for the study was granted by the Morgannwg Local Research Ethics Committee. All properties in the study area were viewed externally. Domestic properties were classified into categories based on floor area (four groups), five period groups, and five build types, viz. detached, semi-detached, flat conversions, purpose-built flats, and terraced housing. Ninety-four of the 100 combinations of the three housing type variables were found in the study area. Analysis was carried out at individual property level: analysis at postcode or zip code level was not possible because very few postcodes (13%) comprised a single property type. Postcodes contained an average of 14 properties. Data on injuries treated at EDs were obtained for the period 1999 – 2000 from the All Wales Injury Surveillance System (AWISS) which routinely obtains individual level data from EDs surrounding the study area and is described in detail elsewhere[ 4 ]. Briefly, the data comprises the patient's address, age, sex, date of occurrence, type and anatomical site of injury (up to three diagnoses and three sites can be coded), and includes a code indicating whether the injury occurred at home or not. No information on precipitating factors such as falls, fires or drug abuse is included. To obtain denominator data for each property type we used the National Health Service Administrative Register (NHSAR), a list of all people registered with the free-to-use primary care health service in Wales. Data on this system are highly confidential and denominator population profiles were obtained by providing a list of all properties in the ninety four different groups to the NHSAR staff who then matched these with their system and obtained the number of people in each of the property types, subdivided by age and sex. This system has previously been used to obtain small area population data[ 5 ]. In analysing data on injury attendance at hospital EDs it is important to take into account the potential confounders of deprivation and access, which are known to be strongly related to injury occurrence and ED attendance respectively[ 6 ]. For each of the injured individuals it was possible to assign an exact value for the Townsend Index of Material Deprivation and distance to hospital by road as the individual addresses were available[ 7 ]. The Townsend Index is a small area based deprivation index, commonly used in epidemiological studies in the UK, and derived from four census variables: home ownership, overcrowding, access to a car, and unemployment. It has been shown to be strongly related to the incidence of specific types of injuries[ 5 , 6 ]. For non-injured individuals this was not possible due to confidentiality constraints described above – only data aggregated at groups of address level was available. Linkage had to be performed in an indirect manner because a small number of properties in a single electoral division meant that data on age/sex compilation could be considered potentially identifiable and so could not be released. During 1999 and 2000, 14,171 out of 112,248 residents made one or more emergency department visits for a home injury. We sought to combine three files comprising individual-level injuries data from an emergency department surveillance system; an external assessment of the built form in all 54,801 homes in the area; and denominator populations for each of ninety-four property types delivered from a health service registration system. Application of logistic regression to model injury risk on built form, property size and age, subject age, sex, deprivation and distance from emergency department jointly necessitates construction of a single linked data file. The objective was to construct an appropriately linked database and hence determine whether injuries occur more commonly in different types of home. The study population of 112,248 individuals made 16,358 ED attendances for home injuries during the study period. The vast majority (99.5%, n = 16,277) of these attendance records included adequate data on age, gender, proximity and Townsend score. These 16,277 attendances involved 14,171 residents of the study area. Thus the average number of visits during the study period among those who ever visited the ED was 1.15, in other words 15% of the visits were repeat visits. The main analyses were constructed to compare the 14,171 first injury records identified as the subject's first attendance at ED for a home injury during the study period with those of the remaining 98,077 subjects. This has the effect of identifying all subjects who had one or more ED attendance for home injury during the study period. The other datasets to be linked were the population distribution by housing group, age and sex (112,248 subjects, no missing data) and a file listing 54,913 properties, of which we restrict attention to the 54,801 with complete data. Table 1 shows how the three sets of data include different subsets of the variables. It is not possible to construct a comprehensively linked dataset at individual level enabling comparison of 14,171 injured and 98,077 uninjured subjects, because individuals in the latter group lack deprivation and proximity data. Table 1 Variables included in the 3 datasets to be linked. Variable Included in dataset for Injuries Properties Population Housing type (build type, period, floor area) Y Y Y Sex Y N Y Age Y N Y Townsend score Y Y N Proximity to hospital Y Y N Type of injury (3 variables) Y N N Anatomical site of injury (3 variables) Y N N Following discussion of preliminary results it was decided to use actual deprivation and proximity scores for the injured, and to impute values randomly according to property type for the uninjured. This is appropriate because risk scores for the remaining variables, age and sex taken together as a forty two category categorical variable, were uncorrelated with risk scores for all other variables (see later), hence imputing according to property type alone and disregarding age and sex is a reasonable strategy. The process of randomly imputing property records, and hence Townsend and proximity scores, to the 112,248 population according to housing type is not trivial. Using ordinary stratified sampling does not work, simply because the number of residents is larger than the number of properties. The overall occupancy ratio was 112,248 residents in 54,801 properties, i.e. 2.05 individuals per property, but this figure varied widely between the 94 property types. For example, for property type 1 there were just under 2 people per property, 114 properties and 226 population. We then choose randomly 114 of the 226 population to match one-to-one to the 114 properties in a random order, leaving the remaining 112 to be matched to a random sample of the 114 properties also in a random order. This simply achieves a maximal degree of representativity with an appropriate degree of randomness. A multi-stage randomisation and linkage process (further details available from the authors) successfully linked the vast majority (14,081/14,171) of first attendances with the merged population-properties file. This has the effect of producing a merged data file in which the deprivation and distance scores randomly imputed to the uninjured both incorporate the correct means to produce an appropriate degree of adjustment for confounding, and also the correct amount of variation to produce appropriate logistic regression coefficients to perform the adjustment. The resulting linked file, comprising reconstructed data for all 112,248 residents, containing elements originating in population, properties and events files, has complete data for demographics, property type and randomly imputed Townsend and proximity measures. Among the 112,248 subjects, 14,081 are identified as having had one or more attendances for injuries following home accidents. Actual Townsend and proximity measures and the code for injury type 1 are available for each of the 14,081 injured individuals. The actual and imputed Townsend and proximity values were summarised for the injured and uninjured and compared. This was done by t-test, paired or unpaired as appropriate, rather than by non-parametric methods which were associated with a serious loss of power due to the gross discreteness of the small area based, randomly imputed Townsend score. Associations between random and actual Townsend and proximity scores were characterised by Spearman rank correlations. Examination of these results, together with the low correlations of a risk score for age and sex with those for property type variables, Townsend and proximity, suggested it was appropriate to include in the logistic regression model composite Townsend and proximity scores, defined as the actual value in the injured and the randomly imputed value in the uninjured. Property size, age and build type were entered as categorical variables. Preliminary analyses indicated that Townsend score should be included as a continuous variable, but distance from hospital discretised into five categories. On account of the known marked sex-age interaction, sex and subject's age were entered together as a 42-group categorical variable, age being discretised into groups under 1, 1–4, 5–9, ..., 90–94 and 95 and over. The main analysis proceeded as above, using the 14,081 who were ever injured during the study period as the group with the outcome of interest. Both univariate analyses for each explanatory variable in turn and a multivariate analysis were produced. Also, the main analysis was repeated after re-running the randomisation parts of the linkage process. This step is more radical than might appear. In particular, due to the small amount of missing data, the randomised matching program does not pick up exactly the same set of 14,081 events records on both occasions. Essentially, the process draws 14,081 out of 14,114 potentially matchable events. Results Table 2 shows summary statistics for actual and randomly imputed Townsend score and proximity. The summary statistics for the randomly imputed Townsend and proximity scores based on all 112,248 subjects are very similar but not identical to those for the 54,801 properties, from which they have been drawn. The mean randomly imputed Townsend score for the injured is very similar to the mean actual score in the 14,081 injured subjects, 0.85 (p = 0.78). Conversely, the randomly imputed distance measures are similar in injured and uninjured (p = 0.12), but the randomly imputed values are significantly greater than the actual ones in the 14,081 injured (p < 0.001). For the Townsend score, the randomly imputed scores are highly significantly higher (i.e. more deprived) for the injured (mean 0.86) than the uninjured (mean 0.70, p < 0.001). For Townsend score and distance alike, the difference between the actual mean in the injured and the mean of randomly imputed values in those not injured is approximately correct to adjust for in the subsequent multivariate analysis, and the process incorporates the appropriate degree of variation at individual level. Table 2 Summary statistics (mean and SD) for actual and randomly imputed Townsend score and proximity to hospital. Basis Series n Townsend score Distance (km) All properties 54,801 +0.74 (2.75) 8.33 (5.06) Randomly imputed All 112,248 +0.72 (2.77) 8.36 (5.05) Not injured 98,167 +0.70 (2.77) 8.37 (5.06) Injured 14,081 +0.86 (2.76) 8.30 (5.03) Actual Injured 14,081 +0.85 (2.77) 7.73 (4.87) Table 3 gives Spearman rank correlations between random and actual Townsend and proximity measures. While all of these are statistically significant (p < 0.001), most are quite small. The correlation of nearly 0.2 between random and actual Townsend scores reflects the unsurprising, substantial variation in Townsend score between property types. Table 3 Spearman rank correlations between random and actual Townsend and proximity measures. Within groups between variables N Spearman rank correlation 95% confidence interval Townsend v. distance Randomly imputed 112,248 0.107 0.101 to 0.113 Actual 14,081 0.051 0.034 to 0.067 Randomly imputed v. actual Townsend 14,081 0.193 0.177 to 0.209 Distance to ED 14,081 0.055 0.039 to 0.072 The effect of the random imputation process was explored in the main logistic regression model. In the final model the effect of age and sex jointly was dominant, followed by build type, then distance, property age, Townsend score (all p < 0.001) and floor area (p = 0.007). Table 4 shows the univariate and adjusted results with regard to build type, the housing variable with the clearest association with the proportion of subjects who ever attended for a home injury. Adjustment for the confounding effects of the other variables made some difference to the odds ratios, but the doubled risk of injury in build type D remained essentially unaltered. Table 4 Main logistic regression model results. All first injuries (14,081 subjects out of 112,248). Odds ratios and X 2 tests for effect of build type on proportion of subjects ever injured (a) unadjusted; (b) adjusted for other factors after random imputation of deprivation and distance scores to the uninjured; and (c) adjusted for other factors after re-randomisation. Build type Number of residents Univariate model Adjusted for other factors, original random imputation Adjusted for other factors, re-randomisation Odds ratio Odds ratio (95% CI) Odds ratio (95% CI) A 15,877 0.790 (0.746–0.837) 0.890 (0.831–0.954) 0.892 (0.832–0.955) B 35,791 1.049 (1.009–1.091) 1.108 (1.055–1.165) 1.109 (1.055–1.166) C 280 1.003 (0.703–1.431) 1.106 (0.768–1.592) 1.113 (0.774–1.600) D 2,695 2.046 (1.863–2.247) 2.074 (1.870–2.301) 2.074 (1.869–2.301) E 57,605 1.000 1.000 1.000 X 2 (4 df) 327.5 254.5 254.1 p-value <0.001 <0.001 <0.001 Distance to ED (km) 4.26 and below 1.431 (1.352–1.514) 1.469 (1.384–1.560) 1.467 (1.381–1.557) 4.27 – 5.57 1.288 (1.211–1.369) 1.413 (1.317–1.516) 1.409 (1.314–1.512) 5.58 – 8.69 1.385 (1.308–1.467) 1.356 (1.276–1.440) 1.352 (1.273–1.437) 8.70 – 13.25 1.168 (1.100–1.240) 1.243 (1.166–1.325) 1.248 (1.170–1.330) 13.26 and above 1.000 1.000 1.000 X 2 (4 df) 194.8 (p < 0.001) 179.9 (p < 0.001) 176.9 (p < 0.001) Townsend score 1.020 (1.014–1.027) 1.016 (1.008–1.024) 1.016 (1.008–1.024) X 2 (1 df) 36.8 (p < 0.001) 15.4 (p < .0.01) 15.9 (p < 0.01) The results of the main multiple logistic regression model in table 4 were used to construct risk scores for each individual representing age and sex, the three property variables, and the composite distance measure. Table 5 shows parametric correlations of the risk score for sex and age with those for the three property variables and the composite distance measure, and the composite Townsend score. (This is, of course, equivalent to using a risk score based on it as it is entered as a linear factor in the model). Even though all but one of these correlations is highly significant, all of them are sufficiently small that we can regard the random imputation process as reasonable. Table 5 Parametric correlations of risk scores for age and sex with those for other factors. Parametric correlation of risk score for age and sex with: 95% confidence interval Risk score for floor area +0.027 +0.021 to +0.033 Risk score for age of property -0.020 -0.026 to -0.014 Risk score for type of property +0.028 +0.022 to +0.034 Townsend score working +0.026 +0.020 to +0.031 Risk score for distance -0.003 -0.009 to +0.003 Discussion The main methodological finding of the study is that the random imputation process developed here is a reasonable one. This approach enabled us to base analyses on a very large dataset notwithstanding confidentiality issues precluding comprehensive linkage directly at individual level. It is feasible to incorporate randomisation into the linkage process, even when the target group is larger than the source. The reasonableness of the imputation process can be judged by comparison of the actual and imputed variables for the injured population. For the Townsend score the actual and imputed scores are essentially identical in distribution, thus showing that the methodology does not produce a biased result. For the distance variable the mean imputed value is 7.4% higher than the actual distance. Whilst this difference is statistically significant the magnitude of the effect on residual confounding cannot be large, given the odds ratios for attendance rates by distance in Table 4 , which indicate that a 1% change in distance produces around a 1% change in the odds ratio for attendance. The low correlation between randomly imputed and actual values, for distance from hospital and Townsend score, could result in attenuated regression coefficients and hence underadjustment for the confounding effect of these variables. For distance, which is by far the more influential of the two variables, a logistic regression in which the entire population of 112,248 is assumed to have the same distribution into the 5 distance groups as applies to the 54,801 properties gives odds ratios 1.469, 1.318, 1.445, 1.200 for the first 4 distance categories relative to the 5 th (most distant) one. These figures are similar to those obtained in the univariate logistic regression based on the composite distance measure, and suggest that the latter regression coefficients, and hence also those in the multiple regression, may be attenuated by around 10% only. Of primary importance to the validity of the methodology set out here, the results obtained after a second randomisation were almost identical. The unadjusted analyses for age-sex and housing type were unaltered, as these variables do not come from the random imputation. The analyses for deprivation and proximity and the results of multivariate analyses for build type and other variables were altered, but only to a very minor degree. These results provide considerable reassurance that the random element that was necessary in order to achieve the linkage process introduced very little additional uncertainty into the final analyses. It appears that injured people tend to live in property types more associated with deprivation than the uninjured. Their actual Townsend scores are in line with what we would expect from their property types. On the composite data, i.e. when we replace random by actual Townsend scores for the injured only, there is a substantial difference in mean Townsend score, 0.85 v. 0.70, and all the 0.15 points difference is attributable to a real effect of deprivation on risk. Conversely, the injured and uninjured tend to live in property types equally distanced from hospital. The actual distance is less for the injured than the randomly imputed distance, which is in line with the known tendency for hospital attendance for less serious types of injury to be related to proximity (6) . On the composite data (with means 7.73 v. 8.37 km), nearly the whole of the difference (0.57 out of 0.64 km) is attributable to this self-selection effect. Conclusions This process is an important methodological development to increase the power of linkage studies when all individual data elements are not available for all individuals. As a result the analysis was based on 112,248 subjects and not on ninety-four groups. Thus, the power to detect important differences is substantially enhanced. Further work is continuing in the relationship between specific features of built type and injury occurrence, using the methodology described in this paper. Competing interests The author(s) declare that they have no competing interests. Authors' contributions All authors contributed to the design of the study. RGN devised the conceptual design of the data imputation and carried out the statistical analysis. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC549195.xml
517943	Antibodies from malaria-exposed pregnant women recognize trypsin resistant epitopes on the surface of Plasmodium falciparum-infected erythrocytes selected for adhesion to chondroitin sulphate A	Background The ability of Plasmodium falciparum -infected erythrocytes to adhere to the microvasculature endothelium is thought to play a causal role in malaria pathogenesis. Cytoadhesion to endothelial receptors is generally found to be highly sensitive to trypsinization of the infected erythrocyte surface. However, several studies have found that parasite adhesion to placental receptors can be markedly less sensitive to trypsin. This study investigates whether chondroitin sulphate A (CSA) binding parasites express trypsin-resistant variant surface antigens (VSA) that bind female-specific antibodies induced as a result of pregnancy associated malaria (PAM). Methods Fluorescence activated cell sorting (FACS) was used to measure the levels of adult Scottish and Ghanaian male, and Ghanaian pregnant female plasma immunoglobulin G (IgG) that bind to the surface of infected erythrocytes. P. falciparum clone FCR3 cultures were used to assay surface IgG binding before and after selection of the parasite for adhesion to CSA. The effect of proteolytic digestion of parasite erythrocyte surface antigens on surface IgG binding and adhesion to CSA and hyaluronic acid (HA) was also studied. Results P. falciparum infected erythrocytes selected for adhesion to CSA were found to express trypsin-resistant VSA that are the target of naturally acquired antibodies from pregnant women living in a malaria endemic region of Ghana. However in vitro adhesion to CSA and HA was relatively trypsin sensitive. An improved labelling technique for the detection of VSA expressed by CSA binding isolates has also been described. Conclusion The VSA expressed by CSA binding P. falciparum isolates are currently considered potential targets for a vaccine against PAM. This study identifies discordance between the trypsin sensitivity of CSA binding and surface recognition of CSA selected parasites by serum IgG from malaria exposed pregnant women. Thus, the complete molecular definition of an antigenic P. falciparum erythrocyte surface protein that can be used as a malaria in pregnancy vaccine has not yet been achieved.	Background Rapid clearance of parasitaemia following transfusion of IgG from malaria immune adults to clinically ill recipients illustrates that naturally acquired antibodies have a parasite clearing role in human malaria infection [ 1 - 3 ]. Neither the nature of the protective immune response nor the target antigens and epitopes recognized by infection clearing antibodies are fully understood. Evidence is accumulating to suggest that the acquisition of antibodies binding the VSA on infected erythrocytes plays a major role in the development of age and exposure dependent immunity [ 4 - 8 ]. The evidence for protective anti-VSA responses is particularly strong for the PAM syndrome [ 9 , 10 ]. PAM is characterized by the sequestration of Plasmodium falciparum infected erythrocytes in the intervillous spaces of the placenta. Infected erythrocytes adhere to low-sulphated forms of CSA present on the extracellular proteoglycan matrix of syncytiotrophoblasts [ 11 ]. In vitro selection of infected erythrocytes for adhesion to CSA concomitantly selects for expression of VSA that share characteristics with postnatal placental isolates. Thus plasma antibodies from malaria exposed pregnant, or multi-gravid women, recognize the VSA of CSA binding parasites (here referred to as VSA PAM ). These sera can also block adhesion of CSA-selected infected erythrocytes to CSA in vitro [ 12 ]. Interestingly, antibodies that bind CSA-selected parasites and block adhesion are not acquired by malaria-exposed males. There is a striking female-specific antibody response recognizing both in vitro CSA-selected parasites [ 12 , 13 ] and P. falciparum isolates taken from infected placentae at delivery [ 14 - 16 ]. Furthermore, the levels of CSA-adhesion blocking plasma IgG have been shown to increase with adult female parity. Recent immuno-epidemiological studies also show a strong positive correlation between the levels of antibodies that recognize the infected erythrocyte surface[ 15 ], the level of CSA-adhesion blocking antibody [ 17 ] and positive birth outcomes as measured by birth weight. PAM is, thus, the clearest example in malaria pathology research of a strong association between infected erythrocyte sequestration and a particular disease syndrome. The VSA recognized by female-specific, parity-dependent antibodies are, therefore, rational and exceptionally interesting candidates for inclusion in an experimental vaccine to protect women against PAM, a major cause of stillbirth, maternal anaemia and low birthweight. To date, the best characterized VSA is P. falciparum erythrocyte membrane protein 1 (PfEMPl), a polymorphic, high molecular weight membrane protein (200–450 kDa) encoded by the var multi-gene family [ 18 - 20 ]. Members of the PfEMP-1 family function as adhesion molecules binding to various host endothelial receptors. They are situated in the knob-like protrusions associated with the parasitized erythrocyte surface. Since var genes encode large extracellular domains rich in lysine and arginine residues, it is not surprising that PfEMP-1 molecules and adhesion to endothelial receptors have been reported to be highly sensitive to trypsin treatment [ 18 , 21 - 24 ]. Less expected was the finding that parasite adhesion to the placental receptor CSA, when immobilized [ 25 - 27 ] or when cell surface associated [ 28 , 29 ], can be relatively trypsin resistant. This study investigates the protease-sensitivity profile of the VSA PAM expressed by CSA-selected parasite clone FCR3 with regard to recognition by antibodies acquired during PAM and adhesion to placental receptors. Methods Parasite isolates Parasites were maintained in group O erythrocytes under standard conditions [ 30 ], using RPMI 1640 medium containing 25 mM HEPES, supplemented with 20 mM glucose, 2 mM glutamine, 25 μg/ml gentamycin and 10% pooled normal human serum. The pH was adjusted to between 7.2 and 7.4 with 1 M NaOH. Culture flasks at 5% haematocrit were gassed with 96% nitrogen, 3% carbon dioxide and 1% oxygen. The laboratory clone FCR3 originates from peripheral blood collected in the Gambia. FCR3CSA was obtained from the Malaria Research and Reference Reagent Resource Centre (ATCC) [ 31 ], and was confirmed, using genetic markers to be identical to the laboratory clone FCR3 kept in the original W.H.O. strain registry collection in Edinburgh (D. Walliker, pers. comm.). CSA binding was maintained by panning late stage infected erythrocytes fortnightly on bovine tracheal CSA (10 μg/ml) (Sigma) immobilized on polystyrene Petri dishes (Falcon), as previously described [ 26 ]. Prior to protease treatment and analysis by flow cytometry, cultures were synchronized by sorbitol treatment to obtain cultures enriched for late stage parasites. Plasma donors Serum samples from 20 men living in a malaria endemic region of Ghana were pooled to produce the male serum pool. Serum samples collected at the time of birth from the placentas of 15 women living in a malaria endemic region of Ghana were pooled to produce the pregnant female serum pool. This pool included five primigravidae, nine secundigravidae and one multigravid woman. Serum samples from six Scottish malaria naïve individuals were pooled and used as a control. Protease treatment Protease treatment of infected erythrocytes was carried out as previously described [ 26 ]. Briefly, samples containing 3 × 10 6 cells from sorbitol treated late stage cultures of 8–10 % parasitaemia were washed twice with phosphate-buffered saline (PBS) and then incubated with the appropriate concentration of trypsin-TPCK (Worthington Biochemicals) or pronase (Boehringer-Mannheim) in a final volume of 1.0 ml in PBS, for 10 minutes at 37°C. The reaction was terminated either by adding soybean trypsin inhibitor (Worthington Biochemicals) to a final concentration of 1 mg/ml or by adding 10% human serum. Cells were washed twice with PBS before further use. Analysis of VSA specific antibodies by flow cytometry Flow cytometry was used to measure the levels of plasma IgG binding to the VSA of late stage parasites essentially following the method previously described by Staalsoe et al [ 13 , 32 ]. 3 × 10 6 cells form late stage P. falciparum cultures of 8–10 % parasitaemia were washed twice with PBS. Cells were incubated sequentially with plasma antibodies diluted 1:20 in PBS, goat anti-human IgG diluted 1:200 in PBS (Dako) and fluorescein isothiocyante (FITC)-conjugated rabbit anti-goat (Dako) diluted 1:25 in PBS. All incubations were in a total volume of 100 μl for 30 minutes at room temperature and were followed by two washes with 1 ml of PBS. Samples were analysed immediately on a FACSCAN apparatus (Becton-Dickinson). FITC fluorescence due to cell surface antibody recognition was determined for 5000–10000 ethidium bromide gated infected erythrocytes. Modified labeling procedure for FACS analysis In order to circumvent the non-specific labeling of the VSA by the tertiary antibody, new reagents have been introduced. The procedure follows the method detailed above with the following modifications. A biotinylated rabbit anti-human IgG antibody (DAKO) was used diluted 1:25 to replace the secondary antibody. In the place of a tertiary antibody, FITC-conjugated streptavidin (DAKO) was used at a 1:2000 dilution. In these experiments the control sera was a pool of malaria naïve Danish volunteer serum. Binding assays Human umbilical cord hyaluronic acid (Sigma) and bovine trachea CSA (Sigma) were used at a concentration of 10 μg/ml in PBS (pH 7.2). 20 μl of each receptor was spotted in triplicate onto 5 cm diameter petri dishes (Falcon). Receptors were adsorbed onto the plastic petri dishes overnight at 4°C. 10 μg/ml BSA in PBS was similarly adsorbed as a negative control. Plates were then blocked by removing the receptor solution and adding 20 μl of 2% BSA in PBS. Following the removal of this blocking solution late stage parasites, suspended in 2 ml of complete RPMI-HEPES medium (8–10% parasitaemia, 5% haematocrit), were added to the petri dish. Parasites were incubated with the immobilized receptor for 60 minutes at 37°C with occasional agitation. Unbound cells were removed by four gentle washes with incomplete RPMI-HEPES medium; bound cells were fixed with 0.5% (v/v) glutaraldehyde in PBS for 10 minutes and Giesma Stained. Bound cells were counted by light microscopy. Protease treatment of intact cells was carried out as described above. Statistical analysis Statistical analyses were performed using Analyses of Variance in Minitab 13.30 (Minitab Inc.), using protease, protease concentration and serum pool as explanatory variables. Statistical models were tested for homogeneity of variance and normality of error distributions. Where possible, maximal models with interactions between these variables were fitted first, after which models were minimized by removing nonsignificant (p > 0.05) terms. Results Concomitant selection of a trypsin-resistant VSA following parasite selection for CSA adhesion It was first established that selection of clone FCR3 for adhesion to CSA resulted in the concomitant selection for VSA specifically recognized by plasma IgG from malaria exposed Ghanaian pregnant women (IgG preg ) (figure 1 ). However there was no increase in the binding of IgG from a pool of plasma from malaria exposed Ghanaian men (IgG male ). The unselected FCR3 clone expressed VSA that were equally well recognized by antibodies in the IgG male and IgG preg serum pools (figure 1 ). These interactions between serum antibody binding and selection for CSA adhesion were highly significant (F 2,24 9.5, P = 0.001). Figure 1 IgG recognition profiles of parasite clone FCR3 before and after selection for adhesion to CSA. Following selection of parasite clone FCR3 for adhesion to CSA the expression of variant surface antigens was investigated using FACS. 5000–10000 late stage parasites were gated using ethidium bromide and FITC fluorescence due to serum IgG binding was measured. Serum samples from six Scottish malaria naïve individuals were pooled and used as a control (IgG control ). Sera from 20 Ghanaian men were pooled to produce the malaria exposed male serum pool (IgG male ). Sera collected at the time of birth from the placentas of 15 Ghanaian women were pooled to produce the malaria exposed pregnant female serum pool (IgG preg ). The bar chart shows mean and standard error of the means for five independent experiments. The trypsin sensitivity of this VSA/IgG binding interaction and of parasite adhesion to CSA was then measured. Parasitized erythrocyte surface trypsinization at a concentration of 0.1 mg/ml showed that the IgG preg binding of FCR3CSA was significantly more trypsin-resistant than was binding of the same serum to the unselected clone (figure 2A & 2B ; F 1,4 16.4, p = 0.015). Although the mean surface fluorescence due to the IgG preg binding of FCR3CSA was slightly reduced by 0.1 mg/ml trypsin this reduction was not significant (figure 2A ; F 1,2 11.3, p = 0.078). The effect of 0.1 mg/ml trypsin on VSA recognition by IgG male and IgG control was comparable before and after CSA selection of the parasite (figure 2 ). Figure 2 Serum IgG from malaria exposed pregnant women recognises trypsin-resistant surface epitopes. Intact infected erythrocytes were treated with 0.1 mg/ml trypsin prior to FACS analysis. Panels A and B show serum IgG binding to the surface of FCR3CSA and FCR3 infected erythrocytes respectively. Serum pools are the same as those described in Figure 1. The bar chart shows mean and standard error of the means for two independent experiments. The effect of a 10-fold higher trypsin concentration and the effect of the non-specific protease, pronase, on IgG recognition of FCR3CSA was also determined. Trypsinization with 1 mg/ml did not significantly reduce the mean surface fluorescence due to IgG preg binding to FCR3CSA (figure 3 ; F 1,4 0.35, p = 0.587). However, treatment of the intact infected erythrocyte with 0.1 mg/ml pronase did significantly reduce IgG preg recognition of FCR3CSA (figure 3 ). Pronase treatment also significantly reduced binding of the IgG male and IgG control serum pools (figure 3 ; F 4,18 3.1, p = 0.041). Figure 3 FCR3CSA expresses surface antigens exhibiting differential protease sensitivity. Intact infected erythrocytes were treated with 1.0 mg/ml trypsin or 0.1 mg/ml pronase prior to FACS analysis. Serum pools are the same as those described in Figure 1. The bar chart shows mean and standard error of the means for three independent experiments. Surprisingly, IgG control binding to the infected erythrocyte surface increased following CSA selection of the parasite (figure 1 ); however, this non-immune recognition was found to be significantly more trypsin sensitive than IgGpreg recognition (figure 3 ; F 4,18 3.11, p = 0.041). This indicates that the epitopes recognized by the IgG control serum pool and the epitopes recognized by the IgG preg serum pool are distinct entities. An increase in apparent non-immune immunoglobulin binding to the infected erythrocyte surface has been observed for a number of parasite clones after selection for adhesion to CSA (data not shown). The source of this background labelling of FCR3CSA by naïve sera was found to be due to non-specific binding by the FITC-labelled tertiary rabbit anti-goat antibody. By using the modified antibody labelling procedure, which employs a biotin-labelled secondary antibody and FITC-labelled streptavidin, binding of malaria naive IgG to FCR3CSA (mean fluorescence index = 16) was comparable to the unselected parasite (mean fluorescence index = 17). Thus the recognition of VSA PAM by malaria naive IgG was abolished (figure 4 ). Figure 4 A modified antibody labelling procedure for FACS analysis of CSA selected parasites. In order to circumvent the non-specific labelling of FCR3CSA VSA seen when using the FITC rabbit anti-goat tertiary antibody, a biotinylated rabbit anti-human antibody in combination with FITC-conjugated streptavidin was used. Panels A and B show FCR3CSA and FCR3 infected erythrocytes respectively. In these experiments the control serum was a pool of sera from malaria naïve Danish volunteers, here shown as a solid grey histogram. The IgG male serum pool is shown as a lightweight line and the IgG preg serum pool as a heavyweight line. Discordance between the protease sensitivity of the CSA adhesion interaction and IgG binding Following the identification of trypsin-resistant epitopes that appear to be concomitantly selected with CSA adhesion, the trypsin sensitivity of CSA adhesion itself was determined. FCR3CSA binding to immobilised CSA was markedly more sensitive to trypsin than IgG preg recognition of the infected erythrocyte surface (figure 5 ). Parasite adhesion was reduced by 81% and 91% following treatment with 0.1 mg/ml trypsin and 1 mg/ml trypsin respectively (figure 5 ). A trypsin concentration of 1 mg/ml reduced binding as efficiently as 0.1 mg/ml pronase, and although 0.1 mg/ml pronase significantly reduced cell surface fluorescence due to IgG preg antibody binding, 1 mg/ml trypsin had no significant effect on IgG preg antibody binding. There is, thus, significant discordance between the high trypsin sensitivity of CSA adhesion and the relatively trypsin-insensitive binding of IgG preg serum antibodies to the infected erythrocyte surface (F 1,8 14.4, p = 0.005). Figure 5 The effect of increasing concentrations of trypsin on parasite adhesion to immobilised CSA and HA. Parasite adhesion to 10 μg/ml human umbilical cord HA and bovine trachea CSA, adsorbed onto the plastic petri dishes, was determined following protease treatment of the intact infected erythrocyte. Bound cells were Giemsa stained and counted by light microscopy. Panels A and B show receptor binding for FCR3 and FCR3CSA infected erythrocytes respectively. The bar chart shows mean and standard error of the means for three independent experiments. Human umbilical cord hyaluronic acid (HA) was also included in these assays to investigate the binding capacity of the CSA selected clone with respect to this receptor. FCR3CSA was found to bind both HA and CSA, although binding to HA was significantly lower (figure 5B ; F 3,19 20.44, p < 0.001), at 71% that observed for CSA. Interestingly, as has previously been shown for other P. falciparum isolates [ 27 ], the trypsin-sensitivity of parasite adhesion to HA and CSA differed at low trypsin concentrations (0.01 mg/ml) (figure 5B ; F 1,8 7.7, p = 0.024). Parasite adhesion to hyaluronic acid was found to be more sensitive to trypsinization than adhesion to CSA. Discussion The acquisition of antibodies to the surface of placental isolates correlates with protection from malaria in pregnancy and the targets of these antibodies are potential vaccine candidates [ 13 , 15 ]. Two variants of the well characterized VSA, PfEMPl, have been shown to have distinct CSA-binding domains [ 29 , 33 ] and antibodies raised against these domains have been reported to recognize the infected erythrocyte surface [ 34 ] and in some cases block parasite adhesion [ 35 , 36 ]. However, in a recent study of var gene transcription in CSA-selected clones, a third potential CSA-binding PfEMPl (var2csa) was identified. Var2csa is predicted to possess distinctly different DBL domains and appears to be the major var expressed by CSA-selected parasites that are recognized by parity-dependent antibodies [ 14 ]. Proteomic analysis of CSA-selected parasites has also identified four additional potential CSA binding PfEMPl molecules [ 37 ]. The molecular identity of the surface antigens expressed at the infected erythrocyte surface remains unclear [ 38 ]. However, the differential protease sensitivity of the epitopes described here would allow treatment of the infected erythrocyte surface with trypsin thereby simplifying the surface complexity, thus, potentially making proteomic approaches more straightforward. Although PfEMPl-mediated CSA adhesion appears to play a role in placental malaria the molecular interactions triggering this syndrome are more complex than initially thought. Several studies implicate additional receptors and binding phenotypes of placental parasites, such as non-immune IgM [ 39 ], hyaluronic acid [ 25 , 27 , 40 ] and non-immune IgG [ 41 ]. CSA-binding laboratory clones and placental CSA binding isolates also appear to express some parasite encoded surface antigens other than PfEMPl, such as ring surface proteins 1 and 2 (RSP 1 and 2) [ 42 ]. Interestingly, a gene 'knock-out' of the CSA binding var (FCR3 var CSA) in parasite clone FCR3 abolishes CSA binding, but the 'knock-out' parasites still bind the syncytio-trophoblast of ex vivo placental cryosections [ 43 ]. Monoclonal antibodies raised against the CSA binding DBLγ domain also show this domain to be sensitive to surface proteolysis using relatively low trypsin concentrations (100 μg/ml) [ 34 ]. It is certainly possible that the trypsin-resistant VSA described here are not of the PfEMPl/CSA binding type. Surface epitopes of the FCR3CSA parasite are both highly resistant to trypsin and are recognized by antibodies from malaria-exposed pregnant women. This agrees with a number of studies that have found parasite adhesion to placental receptors to be resistant to surprisingly high trypsin concentrations. However, binding assays with the parasite clone used in this study showed CSA and HA adhesion to be relatively trypsin-sensitive. This is also compatible with the results of Beeson and his colleagues who demonstrated trypsin-resistant CSA adhesion to be a clone dependent phenomenon [ 27 ]. Another recent study by the same group showed sera that is strongly reactive to the surface of CSA selected parasites is not always capable of inhibiting CSA adhesion [ 44 ]. Thus this study supports the view that erythrocyte surface epitopes distinct from those involved in CSA adhesion may be targets of the antibodies acquired during PAM and suggests that these two epitopes could be on different molecules. One further implication for vaccine development is that a candidate vaccine raising only CSA adhesion blocking antibodies may not mimic protective surface reactive gender-specific immune responses. Conclusion This study supports the view that major differences exist between VSA PAM and previously characterized VSA. Apart from being recognized only by female sera in a parity-dependent manner, VSA PAM show other distinct characteristics such as; i) VSA PAM rarely form infected erythrocyte rosettes when compared to CD36 binding VSA [ 27 , 45 ], ii) with the exception of rosetting isolates, non-immune IgM binding is a phenomenon only seen with CSA-binding clones [ 39 ], iii) VSA PAM do not generally mediate adhesion to CD36 [ 27 , 46 ], and iv) VSA PAM mediated adhesion to the placenta and CSA can be resistant to concentrations of trypsin known to remove most PfEMPl molecules from the infected cell surface. In combination with the findings of this study, these distinct properties of VSA PAM suggest the involvement of either an unusually protease-resistant PfEMPl structure, such has been shown to exist in the A4tres PfEMPl molecule [ 47 ] or an alternative class of VSA in placental adhesion. The differential protease sensitivity exhibited by VSA PAM can be exploited in comparative proteomic analysis to aid in the identification of the molecules whose phenotype is described here. List of abbreviations TPCK – L-(tosylamido-2-phenyl) ethyl chloromethyl ketone, CSA – chondroitin sulphate A, PfEMPl – P. falciparum erythrocyte protein 1, PAM – Pregnancy associated malaria, VSA – variant surface antigens, VSA PAM – variant surface antigens expressed by placental or CSA binding parasites, IgG – immunoglobulin G, DBL-γ-Duffy like binding domain-gamma, FITC – fluorescein isothiocyanate. Authors' contributions LS conceived of the study, maintained P. falciparum culture, performed FACS analysis and binding assays, AE performed the modified labelling FACS experiments and participated in manuscript preparation, MS participated in the design of the study, TS helped develop some methodologies used in this study. DA helped conceive and fund the study and write the manuscript. All authors read and approved the final manuscript. Declaration None declared.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517943.xml
549197	Role of Leishmania (Leishmania) chagasi amastigote cysteine protease in intracellular parasite survival: studies by gene disruption and antisense mRNA inhibition	Background The parasitic protozoa belonging to Leishmania (L.) donovani complex possess abundant, developmentally regulated cathepsin L-like cysteine proteases. Previously, we have reported the isolation of cysteine protease gene, Ldccys2 from Leishmania (L.) chagasi . Here, we have further characterized this cysteine protease gene and demonstrated its role during infection and survival of Leishmania (L.) chagasi within the U937 macrophage cells. Results The amastigote specific Ldccys2 genes of L. (L.) chagasi and L. (L.) donovani have identical gene organization, as determined by southern blots. In vivo expression analyses by Northern blots showed that Ldccys2 is amastigote specific. Western blot using anti-Ldccys2 antibody confirmed the amastigote specific protein expression. Recombinant expression of Ldccys2, a 30 kDA protein, was functionally active in a gelatin assay. Results from Ldccys2 heterozygous knockout mutants showed its role during macrophage infection and in intra-macrophage survival of the parasites. Since attempts to generate null mutants failed, we used antisense RNA inhibition to regulate Ldcccys2 gene expression. Not surprisingly, the results from antisense studies further confirmed the results from heterozygous knockout mutants, reiterating the importance of amastigote specific cysteine proteases in Leishmania infection and pathogenesis. Conclusions The study shows that Ldccys2 is a developmentally regulated gene and that Ldccys2 is expressed only in infectious amastigote stages of the parasite. The collective results from both the heterozygous knockout mutants and antisense mRNA inhibition studies shows that Ldccys2 helps in infection and survival of L. (L.) chagasi amastigotes within the macrophage cells. Finally, antisense RNA technique can be used as an alternate approach to gene knockout, for silencing gene expression in L. (L.) chagasi , especially in cases such as this, where a null mutant cannot be achieved by homologous recombination.	Background Leishmania are the etiological agents of a variety of disease manifestations, collectively termed as leishmaniasis. Visceral leishmaniasis caused by Leishmania (L.) donovani and Leishmania (L.) chagasi is a serious health problem in many tropical and subtropical countries [ 1 - 3 ]. During the digenetic life cycles of Leishmania , it alternates between gut of sand fly vector as an extra cellular promastigote and in the acidic phagolysosome of macrophage as an intracellular amastigote. However, the most intriguing question is the capacity of Leishmania to withstand the hydrolytic conditions of the macrophages, the mechanism of which is still unclear. Thus, identifying the genes expressed specifically in the amastigote stage of the parasites and elucidating their biological function is very important as it would provide new insights into the role of these gene products in the intracellular life cycle of the parasites. Further, this will also help in designing specific drugs and identifying vaccine candidates. Cysteine proteases play an important role in the infection, replication, development and metabolism of protozoan parasites [ 4 , 5 ]. They have been implicated in the invasion of human erythrocyte by Plasmodium falciparum [ 6 ] and considered as virulence factors in the pathogenesis of Entamoeba histolytica [ 7 ]. Cysteine protease activity is necessary for the survival of Leishmania (L.) mexicana [ 8 , 9 ] and related protozoan, Trypanosoma cruzi , within the macrophages, in vitro [ 10 ]. Knockout studies in L. (L.) mexicana have shown that cysteine proteases not only are virulence factors but also act as modulators of host immune responses [ 11 , 12 ]. Thus, cysteine proteases have become a potential target for chemotherapy and a candidate for vaccine development. Initial studies have confirmed the efficacy of cysteine protease inhibitors in treatment of T. cruzi , P. falciparum and L. (L.) major [ 13 - 15 ]. Immunization with the hybrid protein vaccine, consisting of L. (L.) major cysteine proteases CPB and CPA, partially protected against leishmaniasis [ 16 ]. So far, functionally well characterized cysteine proteases are from the New World species of Lesihamania causing the cutaneous forms of leishmaniasis. The members of L .(L.) donovani complex also possess multiple classes of cysteine proteases, which are developmentally regulated [ 17 , 18 ] and are not functionally well characterized. Therefore, there is a need to study the function of these proteases and their role in visceral leishmaniasis. Studies aimed at defining the function of protozoan parasite components have often used gene disruption approach by homologous recombination. Recently an alternative antisense RNA method was followed that may easily and rapidly answer the complex biological questions. Anti sense RNA approach has been used to study the functions of certain gene products in Entamoeba and Leishmania , to elucidate the functions of cysteine protease [ 19 , 20 ], A2 protein [ 21 ] and gp63 [ 22 ]. Previously, we have isolated and characterized two distinct cysteine protease cDNA clones Ldccys1 and Ldccys2 from promastigote and amastigote specific cDNA libararies of L. (L.) chagasi [ 17 ]. Ldccys1 , a member of multi gene family was characterized both in L. (L.) chagasi ( Ldccys1 ) and L. (L.) donovani ( Lddcys1 ) parasites [ 18 ]. In the present study, we have characterized the functional role of amastigote specific cysteine protease gene ( Ldccys2 ) of L. (L.) chagasi . We have generated Ldccys2 heterozygous knockout mutants of L. (L.) chagasi by homologous recombination. As an alternative approach, antisense mRNA expression was also employed. Results obtained from both, gene disruption and antisense mRNA expression shows that Ldccys2 plays an important role in the survival of amastigotes within the U937 macrophages. Results Ldccys2 is a single copy gene From our earlier studies it is known that Ldccys2 is a single copy gene in L. (L.) chagasi . In order to compare the genomic organization of Ldccys2 gene in L. (L.) donovani and L. (L.) chagasi , the members of L. (L.) donovani complex, Southern analysis of genomic DNA was performed. Genomic DNA was digested with different restriction enzymes as shown in Figure 1A and probed with Ldccys2 coding region as a probe. Identical hybridizing bands were present in both L. (L.) chagasi and L. (L.) donovani (Figure 1A ). PCR amplification and sequencing of Ldccys2 coding region from L. (L.) donovani genomic DNA showed identical nucleotide sequences, indicating that L. (L.) chagasi and L. (L.) donovani had identical Ldccys2 cysteine protease genes. Ldccys2 gene is differentially expressed To study the expression of Ldccys2 cysteine protease genes, Northern blot analysis was performed using L. (L.) chagasi total RNA from log and stationary phase promastigotes and amastigotes. The 3'UTR region of Ldccys2 used as a probe, hybridized to a 2.5 kb transcript in amastigote (Figure 1 BI) unlike the cysteine protease, Ldccys1 , which is predominantly expressed in promastigote stages of the parasite (Figure 1 BII). The α- tubulin gene expression was used as an internal control (Figure 1 BIII). Ldccys2 showed similar pattern of expression in L. (L.) donovani (data not shown). It is important to note that this result is in contrast to the earlier report (17), wherein Ldccys2 was identified as promastigote specific cysteine protease gene and Ldccys1 was thought to be amastigote specific. We have confirmed the expression pattern and rectified this error in the following publication (18). To further confirm the amastigote expression of Ldccys2, Western blot analysis was carried out using extracts from both promastigotes and U937 infected amastigotes of L. (L.) chagasi . The anti-Ldccys2 antibody detected a 30 kDa band in amastigotes alone (Figure 1C , lane 3). However, a single 46 kDa band seen in log and stationary phase promastigotes (Figure 1C , lanes 1 and 2) represents promastigote specific cysteine protease Ldccys1, due to the cross-reactivity of the antibody. Taken together, the above results show that Ldccys2 is expressed only in the amastigote stage of L. (L.) chagasi . Recombinant Ldccys2 expressed in insect cells is functionally active In order to determine the cysteine protease activity encoded by Ldccys2 , the predicted coding region was cloned into pIE1/153A.jhe (6hep) vector and transfected into High Five insect cells (invitrogen). The supernatant was harvested and analyzed for the presence of recombinant protein by Western blot analysis using JHE and Ldccys2 antibodies. Western blot analyses with α-JHE antibody (Figure 2A , lane3) detected a band of 96 kDa, corresponding to the fusion protein. The lower molecular mass bands are due to the processed products as a result of autocatalytic activity. Anti- Ldccys2 antibody however, could not detect the recombinant fusion protein (Figure 2B lane 3). The processed products were detected instead, which may be because of inaccessibility of the epitope in the fusion protein. Both the antibodies did not give any background with the negative control (cell extract, Figure 2B and 2C , lane1). Protease activity determined by gelatin SDS-PAGE, showed a band of approximately 30 kDa (Figure 2C , lane 3), confirming the protease activity. Cysteine protease activity was completely inhibited by a specific inhibitor, E-64 at 20 μM concentration (data not shown). Ldccys2 single allele gene replacement by homologous recombination In order to investigate the function Ldccys2 gene, we have performed gene replacement by disrupting a 500 bp coding region of the gene with hyg and dhfr-ts cassette as shown in figure 3A . Following transfection, hygromycin B resistant parasites ( Ldccys2KO , heterozygous knockout mutants) were selected for further analysis. The Pst I restriction enzyme digested genomic DNA from wild type L. (L.) chagasi and Ldccys2 KO were subjected to Southern analysis and results are presented in figure 3B . The disrupted Ldccys2 allele had an additional Pst I site that is present within the hyg gene. As shown in the figure 3B , 5' and 3'probes (probes a and b) hybridized to a single 3.5 kb band with wild type genomic DNA. In Ldccys2KO , probe-a hybridized to two bands (3.5 kb and 2.8 kb) and probe-b hybridized to 3.5 kb and 3.0 kb bands. As expected, one of the bands corresponds to the wild type allele whereas the other band corresponded to the disrupted allele. The hyg specific probe-c resulted in no hybridizing band with Pst I digested wild type genomic DNA, while in Ldccys2KO , two bands (2.8 kb and 3.0 kb) corresponding to the disrupted allele were detected (Figure 3B , probe c). These results clearly indicate that one of the alleles of the Ldccys2 gene has been replaced by hygromycin gene. Similarly, a second cassette using neomycin gene was used in order to replace the second allele of Ldccys2 gene. Despite numerous attempts, we were unable to obtain a null mutant of Ldccys2 . We then checked for any events of rearrangements or change in ploidy levels in the heterozygous mutants. No such detectable events were detected (data not shown). Ldccys2 single allele knock-out mutants had reduced mRNA and protein levels In order to compare the expression of Ldccys2 gene in wild type and heterozygous knockout mutants ( Ldccys2 KO ), amastigote RNA was isolated from U937 cells infected with wild type as well as Ldccys2KO mutant L. (L.) chagasi promastigotes. Northern blot analysis with Ldccys2 coding region probe detected a single transcript of 2.5 kb from wild type as well as Ldccys2 KO amastigotes (Figure 4A , panel 1). However, the transcript level of Ldccys2 in the Ldccys2KO amastigotes was reduced by 64% compared to that in the wild type L. (L.) chagasi . There was no change in the level of α-tubulin expression either in the Ldccys2KO or in the wild type amastigotes (Figure 4A , panel 2). The effect of Ldccys2 single allele disruption at protein level was studied using western blot analysis. Equal amounts of lysates from wild type and Ldccys2KO L. (L.) chagasi amastigotes were analyzed using antiserum raised against Ldccys2 protein. Wild type L. (L.) chagasi amastigotes showed a major band of 30 kDa (Figure 4B , panel 1), while Ldccys2KO amastigotes showed an identical sized band, but there was a reduction in amounts by 45%, based on the quantitation of bands. The antibody did not pick up any signal from the negative control (U937 cells). Figure 4B , panel 2 indicates the equal loading of proteins. These results once again, confirm the deletion of one of the alleles of Ldccys2 . Ldccys2 heterozygous knockout mutants exhibit reduced infectivity and intra-macrophage survival of amastigotes In order to study the effect of Ldccys2 heterozygous knockout mutants on parasite growth, a growth curve analysis was performed. No significant differences were observed between the wildtype parasites and the Ldccys2KO (data not shown). Next, we performed an infection assay with differentiated U937 macrophage cells using with wild type and Ldccys2KO L. (L.) chagasi promastigotes. Interestingly, twelve hours after infection, the number of amastigotes in Ldccys2KO infected macrophages were reduced by half as compared to that of wild type infected macrophage cells. This trend remained same at all the time intervals tested (24, 36 and 48 hours post infection), indicating that the intracellular amastigotes were not able to multiply efficiently (Figure 4C ). The percentage of infected macrophages in the wildtype and Ldccys2KO were similar at 12, 24, 36 and 48 hours post infection (Figure 4D ), however, there was a clear decrease in the initial infection in case of Ldccys2KO , as compared to the wildtype parasites. The above data suggests that Ldccys2 heterozygous knockout has resulted in reduced levels of Ldccys2 expression and furthermore, has decreased the ability of the amastigotes to infect the macrophages and survive inside the macrophage cells. Taken together, these results suggest that Ldccys2 gene has a significant role in the infection and intramacrophage survival of amastigotes. Antisense mRNA inhibition of Ldccys2 resulted in reduced macrophage infection and poor amastigote survival within the macrophages We were unable to generate a complete knock out mutant of Ldccys2 and the heterozygous knockout mutants only partially inhibited the gene expression. Therefore, we used an alternative approach of antisense mRNA inhibition to silence the Ldccys2 gene and to confirm the conclusions derived from Ldccys2 heterozygous knockout mutants studies. Plasmid constructs containing the Ldccys2 antisense and sense sequences were transfected into L. (L.) chagasi promastigotes. The presence of plasmids in these transfectants was confirmed by Southern analysis with Ldccys2 probes (data not shown). Northern blot analyses of amastigotes expressing different plasmid constructs were carried out using specific sense and antisense probes. As expected, Ldccys2 antisense mRNA was expressed only in amastigotes with antisense constructs (Figure 5A , panel 1), and these amastigotes showed a 74% decrease in the Ldccys2 mRNA levels (Figure 5A , panel 2). The amastigotes containing either the sense plasmid or the control plasmid did not show any change in the levels of Ldccys2 mRNA (Figure 5A , panel 2). The stained agarose gel indicates the equal amounts of RNA used in the Northern blots (Figure 5A , panel 3). These results clearly demonstrate that the episomal expression of Ldccys2 antisense mRNA reduces the levels of endogenous Ldccys2 mRNA. We next performed Western blot analysis of transfectants expressing sense and antisense Ldccys2 mRNA using α-Ldccys2 antibody. Amastigotes expressing antisense mRNA showed a significant reduction (by 76%) in the levels of Ldccys2, however, the expression of sense mRNA did not increase the levels of Ldccys2 compared to amastigotes containing the control plasmid (Figure 5B , panel1). Equal loading of the protein is depicted by the Coomassie stained gel (Figure 5B , panel 2). The results further confirm that Ldccys2 antisense mRNA expression does inhibit the endogenous levels of Ldccys2. Survival of L. (L.) chagasi amastigotes within the macrophages was assessed in vitro , using U937 cells. Results from three individual experiments are pooled and the collective data is presented in figure 5C . Results showed that the amastigotes carrying either the control plasmid (P6.5) or expressing Ldccys2 sense mRNA did not have any effect on their capacity to infect and survive inside the macrophages. The number of amastigotes per macrophage cell remained the same at initial stages of infection (data not shown). However, amastigotes expressing Ldccys2 antisense mRNA showed significant reduction (by almost 50%) in terms of the initial infection and intra macrophage survival starting from 12 hours post infection and remained consistent up to 48 hours after infection (Figure 5D ). The percentage of total infected macrophages is shown in figure 5D . Consistent with the results from the heterozygous knockout mutants, the Ldccys2 antisense transfectants had reduced levels of infection when compared to that of the wildtype and control plasmids. These results further underline the importance of Ldccys2 in infection of macrophages and subsequent survival of amastigotes within the macrophage cells. Taken together, these results indicate that the amastigote specific cysteine protease, Ldccys2 plays an important role in intramacrophage survival of L. (L.) chagasi amastigotes. Discussion In this study, we have shown that amastigote cysteine protease, Ldccys2 necessary for macrophage infection and for survival of L. (L.) chagasi amastigotes inside the macrophage cells using two different approaches, gene disruption and antisense mRNA expression. Leishmania (L.) chagasi cysteine protease, Ldccys2 is a single copy gene [ 18 ] and is expressed only in amastigote stage of the parasite. L. (L.) donovani also possesses an identical gene. Expression studies performed demonstrates that Ldccys2 is expressed only in amastigotes of L. (L.) chagasi and L. (L.) donovani . A similar single copy gene in L. (L.) mexicana ( Lmcpa) has a significant sequence identity (80%) to Ldccys2 . Lmcpa gene of L. (L.) mexicana which is the closest homolog of Ldccys2 is differentially expressed in amastigotes [ 23 ], whereas the Ldccys2 gene of L. (L.) chagasi is expressed specifically in the amastigote stage. In addition, the major amastigote specific cysteine proteinase gene of L. (L.) mexicana , cpb2.8 is a multicopy gene, located within the lmcpb cluster [ 24 ] and there is no significant sequence identity (34%) between Ldccys2 and lmcpb2.8 . Therefore, gene organization and expression of amastigote specific cysteine proteases in L. (L.) mexicana and L. (L.) donovani complex are not identical. Interestingly, Llacys1 , an amastigote specific cysteine protease from L. (L.) amazonensis was identified recently, which has high homology (88%) with Ldccys2 and like Ldccys2 , is only expressed in amastigote stages [ 25 ]. In order to confirm the protease activity of amastigote cysteine protease, full-length Ldccys2 was expressed as JHE-fusion protein in an insect cell expression system. The recombinant protein hydrolysed gelatin in activity gels. Recently, functions of many genes have been established in Leishmania by following gene disruption approach [ 26 , 27 ]. To study the biological role of Ldccys2 cysteine protease, we have followed targeted gene disruption using hyg gene as the selectable marker. Genomic Southern analyses of the hygromycin B resistant clones using specific probes confirmed the replacement of 500 bp region within the ORF of the Ldccys2 by hyg/dhfr-ts . Northern and Western blot analyses Ldccys2KO amastigotes showed a clear decrease in the endogenous levels of Ldccys2 compared to that of wild type L. (L.) chagasi amastigotes (Figure 4 ), demonstrating the disruption of one of the Ldccys2 alleles. Despite our repeated effort using different selectable marker genes, we failed to create a null mutant of Ldccys2 in L. (L.) chagasi . Since gene amplification and increase in ploidy level has been reported for Leishmania species [ 28 ], we analyzed for changes in ploidy level and or gene amplification and found no such events occurring in these heterozygous knockout mutants. This suggests that Ldccys2 is an important gene necessary for the initiation of macrophage infection and survival of amastigotes within the macrophages. Therefore we used the Ldccys2 heterozygous knockout parasites for further functional studies. The U937 macrophage cells infected with Ldccys2 heterozygous knockout mutants showed a significant decrease in the initial infection and survival within the macrophage cells, compared to wild type (Figure 4 ). This trend remained consistent from 12 hours up to 48 hours after infection. The decrease in initiation of infection and subsequent survival of amastigotes may explain the role of Ldccys2 in amastigote metabolism and its survival inside the macrophages [ 4 ]. Consistent with our results, null mutants of the lmcpb cluster in L. (L.) mexicana exhibited reduction in macrophage infectivity and survival [ 23 ]. However, lack of Ldccys2 null mutant prevented us from a detailed study on the role of Ldccys2 in survival and pathogenesis, using animal models. The lack of Ldccys2 null mutants prompted us to use an alternative approach of antisense mRNA inhibtion in order to confirm the results obtained from Ldccys2 heterozygous knockout studies. Using antisense RNA technique it has become possible to investigate the contribution of cysteine proteases to virulence mechanism of E. histolytica [ 19 , 20 ] and A2- protein in the survival of L. (L.) donovani amastigotes in macrophages [ 21 ]. Furthermore, episomal expression of sense and anti sense mRNA of L. (L.) amazonensis gp63using Leishmania - specific P6.5 vector provided evidence for a role both in binding of macrophages and the intracellular survival and replication [ 22 ]. Not surprisingly, L. (L.) chagasi amastigotes expressing Ldccys2 antisense mRNA has clearly down regulated the endogenous Ldccys2 mRNA levels (Figure 5 ). However, a small amount of Ldccys2 mRNA is still present in antisense transfectants, which may be due to the incomplete suppression, which is inherent to this approach. Infection of U937 macrophage cells with transfectants expressing Ldccys2 antisense mRNA showed a significant decrease in amastigote survival inside the macrophage cells, as compared to amastigotes expressing either sense Ldccys2 mRNA or the control plasmid (Figure 5 ). This is in agreement with the results obtained from Ldccys2 heterozygous knockout mutants, reaffirming the role of Ldccys2 in initiation of intracellular infection, and subsequent survival and multiplication of amastigotes. However, the actual effect of antisense mRNA expression may be greater, since reversal of antisense effects is expected from the release of the selective pressure necessary to avoid drug toxicity to macrophages [ 21 ]. This is especially true in the assay for intracellular survival and replication, since it requires prolonged period of incubation. The transient nature of the antisense suppression in the absence of selection prevented us from taking up animal studies that also require prolonged infection with these transfectants. The out come of this study shows that Ldccys2 is expressed only in amastigote stage of the parasite. Both the heterozygous knockout mutants and antisense mRNA expression shows that Ldccys2 is important for infection of macrophages and survival of L. (L.) chagasi amastigotes inside the macrophage cells. Based on the data obtained in this study, there is a clear justification for targeting the Ldccys2 gene in L (L.). chagasi and carrying out further studies in animal models. Currently, we are working towards generating null mutants of Ldccys2 in L. (L.) chagasi and to further understand the role of this protein in the pathogenesis of L. (L.) chagasi parasites. Methods Parasite culture L. (L.) chagasi strain MHOM/BR/74/PP75 promastigotes were grown in HOMEM, pH 7.4 (Gibco BRL) supplemented with 10% (v/v) heat inactivated calf serum, at 26°C. Promastigote cultures were seeded at 1 × 10 6 parasites/ml and harvested in logarithmic or stationary growth phase, as defined by cell concentration. L. (L.) chagasi promastigotes were transformed to axenic amastigotes using the standardized protocol as described earlier [ 29 ]. All the transfectants were periodically converted to amastigotes by infecting U937 macrophage cells in order to ensure the property of virulence. RNA isolation and northern hybridization Total RNA was isolated using trizol (GibcoBRL) as recommended by the manufacturer. Promastigote RNA was obtained from logarithmic and stationary growth phase parasites of L. (L.) chagasi . To isolate amastigote RNA, human macrophage cell line (U937, ATCC) was infected with promastigote parasites, which were subsequently converted into intracellular amastigotes [ 29 ]. RNA (10 μg/lane) was separated on 1.2% (w/v) formaldehyde agarose gels and transferred onto Hybond N+ membrane (Amersham Pharmacia). Standard procedures were followed to perform Southern and Northern blot hybridizations. Using Quick Prime Labeling Kit (Amersham Pharmacia Biotech), probes representing specific region of Ldccys2 cDNA clone was labeled with [α 32 P] dCTP. The sense and antisense probes were generated by 5'end-labeling the specific sense and antisense oligonuecleotide primers with γ- 32 P ATP, using T4 polynucleotide kinase according to the standard protocol. Polyclonal antibody production against Ldccys2 Direct DNA immunization [ 30 ] was used to generate antibody against Ldccys2. Briefly, the pro-mature region of the gene was PCR amplified from the Ldccys2 cDNA clone using specific 5' (5'CAGACAGGATCCG CGGCCGCCATGGACGACTTCATTGCC 3') and 3' primers (5' GGCCGCGGATCCGCG GCCGCCTATGAGGTGTTGGAGTCGTC 3') and the PCR product was sub-cloned into BamH I site of pcDNA 3.1 (Invitrogen). After confirmation by sequencing the plasmid constructs were amplified in E. coli and DNA was isolated using endotoxin free maxi prep kit (Qiagen). 50 μg of plasmid DNA in 100 μl of saline was injected into the hind limb quadriceps muscles of BALB/c mice. All the mice were boosted after 14 days and tested for antibody response seven days after the boost. Mice with positive response were boosted one more time and after a week sera was collected and stored at -20°C for further use. Western blot analysis Denaturing gel electrophoresis was carried out according to the standard protocols. Following SDS-PAGE, proteins were transferred onto to Hybond P membrane (Amersham Pharmacia Biotech) using semi-dry transblot apparatus (BioRad). Membranes were blocked for 2 hours with 5% (w/v) skim milk solution and probed overnight at 4°C with 1:200 dilution of anti-Ldccys2 antiserum. Membranes were washed three times with phosphate buffered saline containing 0.1% (v/v) Tween and incubated with 1:5000 dilution of anti-mouse horseradish peroxidase linked secondary antibody (Molecular Probes). Enhanced chemiluminescent kit (Amersham Pharmacia Biotech) was used for detection. Expression of cysteine protease genes In order to express Ldccys2 cysteine protease gene, the coding region was amplified by PCR. The digested PCR product was cloned in-frame into unique NotI site of pIE1/153A.jhe (6hep) vector [ 31 ] and confirmed by sequencing. The fusion of the Cysteine proteases to the c-terminus of Juvenile hormone esterase (JHE) directs the secretion into the culture supernatant of transfected insect cells. Culture of insect cells and procurement of samples containing recombinant protein from transfected insect cells are described earlier [ 32 ]. Briefly, High Five insect cells (Invitrogen) were seeded into6-well culture plates (35 mm diameter) at a density of 5 × 10 5 cells/ml (2 ml/well) and transfected for 5 h with 0.5 ml of transfection solution containing 30 μg/ml lipofectin (Life Technologies) and 6 μg/ml total plasmid DNA in basal IPL-41 medium. Then the transfection solution was removed, the cells were rinsed with basal media and 2 ml IPL-41 + 10% FBS was added and incubated for sixty hours. The supernatant was then harvested and stored at -20°C for analysis. Enzyme activity was analyzed on gelatin SDS-PAGE containing 0.2% gelatin and 8% (w/v) acrylamide as described previously [ 17 ]. For inhibition studies, gels were pre-incubated with 20 μM E-64 (Sigma), a Cysteine protease specific inhibitor. Gels were stained with Coomassie Brilliant Blue for half hour and destained with 30% Methanol for 2 hours, to visualize the bands. Bands were photographed using Diamed apparatus (Bio-Rad). Ldccys2 gene disruption by homologous recombination Gene replacement vector, pXLdccys2-HygKO was designed to replace 500 bp mature region of Ldccys2 coding region and constructed as outlined below (Figure 2b ). The sense primer Ldccys2 Hind III (5' GCTCTCAAGCTTGCTCACGCATCCGCCGC-3') and the antisense primer Ldccys2 Xho I (5'GAAGGCCTCGAGCGAGCCGCACATTCCCTG-3') were used to amplify a 500 bp region from Ldccys2 cDNA and cloned between Hind III – Xho I sites in pX63-Hyg (kindly provided by Dr. S. M. Beverley) for the 5' end homologous recombination. A 450 bp 3' end homologous recombination fragment was amplified using the sense primer Ldccys2 Sma I (5'CCGGAGCCCGGGCCCACGGCGCTTGTGCAG-3') and antisense primer Ldccys2 Bgl II (5' ACTGTCAGATCTGCTGTGCGCCAGATCGCG-3') and cloned between Sma I and Bgl II sites of pX63-Hyg . The plasmid construct pXLdccys2-HygKO was confirmed by sequencing and digested with Hind III and Bgl II. The resulting 3.8 kb linear fragment containing hygromycin phospotransferase/dihydrofolate reductase-thymidylate synthase ( hyg/dhfr-ts ) gene sequence and the homologous recombination fragment was purified and used for transfection. Episomal expression of Ldccys2 specific sense and antisense mRNA The Ldccys2 cDNA clone was isolated from an amastigote specific cDNA library of L. (L.) chagasi . Ldccys2 was earlier reported as promastigote specific cDNA clone [ 17 ]. However, it was further verified to be amastigote specific gene (18). The coding region (1 kb) of Ldccys2 was inserted at BamH I site of Leishmania specific vector P6.5 [ 22 ], a kind gift from Dr. K.P.Chang. Cloning was done in both sense and antisense orientations with reference to that of N-acetylglucosamine 1-phosphate transferase (nagt) , which was used as the selective marker for tunicamycin resistance. Plasmid constructs were confirmed by sequencing. P6.5 plasmid and the sense and antisense plasmid constructs were amplified in E. coli and isolated with a DNA maxi prep kit (Qiagen) and the purified DNA samples were used for transfection. DNA transfection All the transfections were carried out using L. (L.) chagasi promastigotes. Briefly, 4 × 10 7 log phase promastigotes were pelleted, washed twice with phosphate buffered saline and resuspended in 0.4 ml of electroporation buffer in a 0.2 cm cuvette. Twenty-five microgram of P6.5 plasmid DNA constructs was used for episomal expression, whereas for homologous recombination 5 μg of linearized DNA fragment was transfected. DNA was electroporated using BioRad Gene Pulser Unit at 0.45 kV, capacitance of 25 μF and resistance of 250 ohms. Following electroporation, cells were resuspended in drug-free media for 24 hours. Parasites with P6.5 plasmid DNA constructs were selected for tunicamycin resistance at 5, 10 and 20 μg/ml tunicamycin (Sigma). Parasites transfected with DNA fragment for homologous recombination were selected for hygromycin B (GibcoBRL) resistance at 100 μg/ml. In vitro assay for intra macrophage survival The U937 macrophage cells were infected with L. (L.) chagasi stationary promastigotes at a host to parasite ratio of 1:10 [ 33 ]. Briefly, U937 suspension cells (10 6 cells/ml) in 10 ml RPMI 1640 medium supplemented with 10% (v/v) heat inactivated fetal bovine serum, 2 mM L-glutamine and 50 μg gentamycin (Gibco BRL) were allowed to grow at 37°C for 3 days in 10% (v/v) CO 2 . The cells were then induced to adhere with 7.5 ng/ml of phorbol myristate acetate (PMA) and allowed to recover for 72 hours. The macrophage cells were then infected with stationary phase promastigotes and allowed to infect for 6 hours. The unbound parasites were washed away with RPMI medium and the infected macrophage cells were fed with fresh RPMI medium and left for 4 days. At every 12 hours interval up to 72 hours, cells were scraped and Diff-Quick stained slides were prepared for microscopic counting, Abbreviations Hyg, hygromycin; JHE, juvenile hormone esterase; kb, kilobase; KO, knock-out; ORF, open reading frame; UTR, untranslated region Authors' contributions VM carried out all the plasmid constructs for gene disruption and antisense mRNA inhibiton studies and performed the Southern and Northern blot analyses shown in Figures 1 , 3 , 4 & 6; wrote parts of the manuscript. ASK performed all the transfections, western blots and macrophage assays shown in Figures 1 , 2 , 4 , 5 , 6 and 7; wrote parts of the manuscript. LG supervised the experiments and gave laboratory support, and critically read the manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC549197.xml
526220	Rank Difference Analysis of Microarrays (RDAM), a novel approach to statistical analysis of microarray expression profiling data	Background A key step in the analysis of microarray expression profiling data is the identification of genes that display statistically significant changes in expression signals between two biological conditions. Results We describe a new method, Rank Difference Analysis of Microarrays (RDAM), which estimates the total number of truly varying genes and assigns a p-value to each signal variation. Information on a group of differentially expressed genes includes the sensitivity and the false discovery rate. We demonstrate the feasibility and efficiency of our approach by applying it to a large synthetic expression data set and to a biological data set obtained by comparing vegetatively-growing wild type and tor2-mutant yeast strains. In both cases we observed a significant improvement of the power of analysis when our method is compared to another popular nonparametric method. Conclusions This study provided a valuable new statistical method to analyze microarray data. We conclude that the good quality of the results obtained by RDAM is mainly due to the quasi-perfect equalization of variation distribution, which is related to the standardization procedure used and to the measurement of variation by rank difference.	Background In a typical microarray experiment, thousands of genes have their relative expression levels measured in parallel under different biological states [ 1 , 2 ]. To identify differentially-abundant genes, most published methods [ 3 - 5 ] progress through a similar sequence of elementary steps. First, a normalizing procedure is applied to make data sets comparable. If certain experimental conditions comprise several replicates, methods based either on parametric or nonparametric tests usually reduce the number of values generated by using their means. Then, gene variation is quantified by a statistic derived from intensity measurements. Knowledge of the null distribution of the gene variation, which is the distribution of its statistic when only random fluctuations occur, allows p-values to be assigned to observed variations and genes to be ranked according to the significance of their variation. As the test is repeated as many times as there are genes, p-values are corrected accordingly, and the false discovery rate (FDR, "the expected proportion of false positives among all genes declared significantly differentially expressed"[ 6 ]) is estimated. We describe here, in detail, a new analysis method that has been used to analyze the transcriptome in yeast [ 7 ]. This method is original in several respects. First, Rank Difference Analysis of Microarrays (RDAM) replaces raw signal by its rank (R), expressed on a 0–100 scale, and we show that this simple transformation is a powerful normalizing procedure. Also, RDAM does not reduce replicated signals to their means, but instead only considers variations, expressed as rank differences (RD), between individual experimental points. An essential step is the standardization of RD observed between two replicates, permitting easy access to the empirical null distribution and allowing accurate and precise p-values to be assigned to observed standardized RD (zRD). When dealing with replicated points, RDAM uses a random variable, the product of p-values (ppv), for which the null distribution is straightforward to compute in a manner that is independent of the experimental conditions. Finally, RDAM estimates the total number of truly varying genes (TV), assigns a p-value to each gene variation, characterizes the selection of a gene using the FDR and the percentage of truly varying genes included in the selection (sensitivity, S). Analysis of synthetic data sets allowed us to specify the error distribution of all the estimators used (FDR, TV and S), and to demonstrate the strong predominance that the number of varying genes and the distribution of their variation have on the quality of the results. We also analysed the transcriptional effects of the TOR2-controlled signaling function using a genome-wide microarray approach in yeast. In S. cerevisiae , TOR2 has two essential signaling functions. One, shared with TOR1, is required for translation initiation, transcription, and cell growth in response to the presence of nutrients [ 8 - 10 ]. The second is unique to TOR2, and functions in cell-cycle-dependent actin polarization and possibly in transcription [ 8 , 11 ]. A previous genetic screen for mutants defective in the TOR-shared and the TOR2-unique functions identified several TOR2 temperature-sensitive alleles [ 12 ]. In this study, we compared total transcription profiles for strain SH121, which is specifically defective in the TOR2-unique function, and its isogenic wild type counterpart SH100 [ 12 ]. Results Standardization of positive variations The simplest system to which our method can be applied comprises three experimental points, of which two are replicates, as described by the expression {Exp1A, Exp1B, Exp2A}, where the number refers to the biological condition and the final letter refers to the replicates. To identify significant variations in the comparison Exp2A vs. Exp1A, we have to first calculate the variation of gene i, VARi. From among several possibilities, we tested three different variation units: the fold change (FC), corresponding to the ratio of signals, the signal difference (SD), and the rank difference (RD). The RD uses a standardized signal measure that is independent of the scanner settings, because the signal is replaced by its rank, expressed on a 0–100 scale. This normalizing procedure consists of first calculating the absolute rank (AR) of each gene by ordering their signals from 0 to N (with the signals of all genes having a negative signal being set to zero, and N representing the number of non-null and non-negative signals) and then transforming the absolute rank value into a relative one (R = AR*100/N). In this way, all the signals are expressed on the same scale and are directly comparable. We studied the variation distribution between the two replicates, i.e. Exp1A and Exp1B, reasoning that the observed empirical variation distribution would be an excellent approximation of the null distribution corresponding to the null hypothesis. The null hypothesis we have in mind states that all observed mRNA changes occurring under replicated conditions are due to a combination of biological and technological noise, and are not the result of any biologically significant process. We first restricted our study to positive variations. As the distribution of positive variations should be the same in both comparisons – Exp1B vs Exp1A and Exp1A vs Exp1B – we plotted the positive variations against rank for both comparisons on a single graph. This revealed that the most salient property of variation distribution, common to all tested measures of variation, is its dependence on the signal rank. This is exemplified for the RD in figure 1A , which shows the absolute value of RD against the minimum of the ranks (i.e., \|R i (Exp1A) - R i (Exp1B)\| vs min{R i (Exp1A), R i (Exp1B)} for gene i). This mode of presentation can be interpreted in either of two ways: as a plot of the positive variations of both comparisons, or as the plot of the positive and the negative variations of a single comparison, with both of variation being represented by a positive number. Whichever interpretation is prefered, it should be underlined that this presentation allows all the gene variations to be taken in account, and ensures the uniqueness of the resulting variation distribution. We tried to eliminate dependency of positive variation distribution on the signal rank by standardizing variations according to the general formula: where VAR is to be replaced by any one of the variation units tested (SD, FC or RD). Using this expression, the sample mean and standard deviation (μVAR and stdVAR) were calculated for all genes having a rank within a given neighbourhood of R i , the rank of gene i. This notation reflects the fact that the VAR distribution is not gene specific, but rank dependent. Figure 1B shows the results obtained by applying this procedure to the distribution of zRD. In concrete terms, we traced, as shown in figure 1A , two standardization curves, μRD and stdRD, which provide for a given rank the local mean and standard deviation of all the genes with a similar rank. Then, each RD i was standardized according to (1). Figure 1B illustrates the beneficial effect of this standardization procedure: first, the mean and the std are no longer dependant upon the rank. Second, the distribution is equalized all along the rank scale, as shown by QQ-plots in figure 2A . Comparable results are obtained if the standardization is applied to the FC or the SD, but the zRD gives the best results in terms of distribution equalization: figure 2B shows, for example, that QQplots derived from zFC are more erratic than those derived from zRD, which are almost identical to the first diagonal up to the 99th percentile. The fact that the zRD distribution is independent of rank can be explained by the fact that each gene's zRD i follows the same zRD distribution. Therefore, we reasoned that the empirical cumulative frequency distribution, ecfd(zRD), approximates the distribution of zRD for any gene i under the null hypothesis, and we used Fo = 1 - ecfd(zRD), based on a comparison between two replicated experiments, to assign a p-value to any zRD calculated in a comparison between two different biological conditions (the p-value is defined as the probability of zRD of an unchanged gene i to be equal to or greater than the observed zRD i under the null-distribution Fo). Because of the very large number of genes present on a chip, the null distribution is sampled a great number of times, generating a quasi continuous set of points that spans a wide range of values. This improves the precision of the Fo curve and allows accurate p-values to be assigned for even large variations. The entire procedure can then be applied to the comparison Exp2A vs. Exp1A. Because standardization curves constructed on the basis of the two replicates are used in the standardization process, the calculated zRD can be justifiably compared to the null distribution and interpolated on the Fo curve in order to assign a p-value to each gene variation. At this step positive and negative variations can be processed together, although it is necessary to keep track of the actual type of variation, i.e. positive or negative, in order to conduct subsequent analysis. Standardization of negative variations We also tested to see if it is possible to apply the same standardization techniques to negative variations. In figure 3A , we have plotted the opposite of absolute RD value against the maximum of the ranks (i.e., -\|R i (Exp1A) - R i (Exp1B)\| vs max{R i (Exp1A), R i (Exp1B)} for gene i). It is clear that the weak signals are characterized by a truncation of their variation distribution, as evidenced by the clear alignment of points between ranks 0 and 20. This explains why the standardization procedure applied in figure 3B fails to equalize the variation distribution, and also why the power of the test is lower for down-regulated genes than it is for up-regulated genes (see Discussion). False Discovery Rate (FDR), Total Variation (TV) and Sensitivity (S) Because the test is repeated N times, issues related to multitesting must be considered: the more tests that are performed, the more an outlier outcome becomes probable. In view of this, we first compute the observed distribution of zRD in the comparison Exp2A vs. Exp1A for increased and decreased genes, giving two curves: F INC for the positive variations and F DEC for the negative variations (in both cases F = 1 - ecfd (zRD)). Then we plot F INC (or F DEC ) and F 0 on the same graph, corresponding to the observed positive (or negative) variation distributions and to the expected variation distribution according to the null hypothesis, respectively (figure 4 ). In the following discussion, the rationale is the same for increased and decreased variations, and F stands for either F INC or F DEC (same remark for TV, FDR, S, K and N). In most encountered situations, F is on top of F 0 . F(x) gives the probability, for any variant or invariant gene i, of observing a zRD i that is at least as high as x, and NF(x) gives the corresponding number of genes. If 0 <= K <= 1 is the fraction of invariant genes, then KNF 0 (x) is the number of invariant genes that have a zRD i at least as high as x. As a consequence, NF(x) - KNF 0 (x) is an estimate of the number of variant genes with zRD equal to or greater than x. We can call k the value of x that gives the maximum number of variant genes, such as NF(k) - KNF 0 (k) = max(NF(x) - KNF 0 (x)), and use this value as an estimate of the total variation (TV), that is, the number of truly varying genes. As these quantities must verify the equation N = KN + TV, i.e., N = KN + NF(k) - KNF 0 (k), we can deduce that . For each value of x, we estimate the False Discovery Rate, and the sensitivity, In the context of a transcriptome analysis, p-values reflect how probable it is that a variation reaches or exceeds an observed value. P-values can always be used to rank genes, but the selection of significant variations in the context of multiple testing requires defining significance levels that are far more stringent than 0.01 or 0.05, as used in single testing. FDR and S parameters allow this difficulty to be overcome: for each c used as a potential critical value, S(c) reflects the fraction of truly-varying genes that are selected by zRD>c, and FDR(c) estimates the fraction of selected genes that are likely to be invariant genes. It is therefore possible to plot FDR and S against c, and to construct, for positive and negative variations, what we call a selection abacus (figure 4 ). Analysis of replicates The simplest example of a replicated experimental scheme is the system {Exp1A, Exp1B, Exp2A, Exp2B}. While it would be tempting to average signals or ranks for each experimental condition and apply the method described above, this is not possible because averaging changes statistics and we have no practical way of obtaining the corresponding empirical null distribution. In a first round of comparison, we conducted two analyses in parallel by applying RDAM to the first comparison Exp2A vs. Exp1A and to the second comparison Exp2B vs. Exp1B. Based on this first round of comparison, we obtained two p-values for gene i: p1 i and p2 i . It could occur that gene i is detected as an increasing variation in the first comparison and as a decreasing variation in the second comparison. In this case, we apply a direction rule to decide on the final direction of variation. We consider simply that the lowest p-value is in favor of its corresponding variation direction, and we set the p-value of the discordant comparison to one. Once we have calculated and possibly corrected the p-values, we construct a new random variable, the product of p-values, ppv i = p1 i × p2 i . To obtain an unbiased value for ppv, we apply the same procedure to a second round of comparison by exchanging Exp2A and Exp2B between the two comparisons, giving a second ppv value. The direction rule is applied to the two ppv before obtaining the final, averaged ppv. The advantage of the random variable ppv is the ease of constructing its null distribution. In fact, the cfd(p i \|H 0 ) is a uniform distribution over the interval [0,1]. Therefore, for cases in which two independent comparisons between two sets of duplicates were to be considered, we constructed two sets, U1 and U2, of 100000 points uniformly distributed over the interval [-1,1], to take into account the possibility of increased and decreased variation for each point. These sets were randomized to make them independent in order to model the independence of measurement according to the null hypothesis. This hypothesis states that all variations are due to noise, and that for a particular gene all corresponding p values must be independent. Then, we apply the direction rule to the pair U1, U2 and calculate ppv for genes that are detected as increased. Thus, the F 0 = 1 - cfd(log10(1/ppv)) curve allows the significance of any value for ppv to be tested. The significance of ppv combines the significance of variation within each individual comparison and the significance of the correlation between these variations. F curve is, as usual, the observed 1 - ecfd(log10(1/ppv)), and we get exactly the same kind of selector abacus, as shown in figure 4 . Simulation of the ppv null distribution used exactly the same steps that the analysis process follows, i.e. application of the direction rule and construction of the product of p-values, and resulted in a null distribution model we found appropriate seeing, both with experimental and synthetic data sets, that the observed distribution of ppv matches the null distribution when no variation occurs (data not shown). The system {Epx1A, Exp1B, Exp2A, Exp2B} allows the construction of two sets of sandardizing curves, one from Exp1 replicates and the other from Exp2 replicates. As these curves are not equivalent, it is necessary to carry out both analyses and then use the more conservative one, whichever has the lower F curve. The generalization of the entire procedure to more than two replicates is straightforward. For example, with three replicates {Exp1A, Exp1B, Exp1C, Exp2A, Exp2B, Exp2C}, there are 3 × 2 ways of arranging the experiments in order to obtain different sets of comparisons. Each round of comparison gives three p-values for the gene i - p1 i , p2 i and p3 i – and the direction rule is applied the following way: in case the gene i is detected as an increasing variation in the first comparison and as a decreasing variation in the two other comparisons, we compare p1 i to p2 i × p3 i and determine the variation direction. Once we have calculated and possibly corrected the p-values, we obtain the product of p-values for gene i, ppv i = p1 i × p2 i × p3 i . As the number of comparison rounds increases very rapidly with the number n of replicates (n!), we simply apply a circular permutation – the circular permutation of ABC consists in the subset of permutations ABC, BCA and CAB – to the replicates inside one of the biological condition which allows the number of rounds of comparison to be restricted to the number of replicates (n). Generation of synthetic data In order to test our method, we devised a way of generating synthetic data having similar statistical properties as real biological data. We selected two replicated experiments, Exp1A and Exp1B, as seeds for generating synthetic data, traced standardization curves, and calculated ecfd(zRD) for (Exp1A, Exp1B). We randomly selected half of the genes and exchanged the signals for Exp1A and Exp1B, giving two new data sets Exp1A' and Exp1B'. Random numbers uniformly distributed over interval [0,1] are generated for each gene. Each random number is interpolated on the inverse of the cfd of zRD to assign a random standardized variation zRD i to each gene i. New values R i are obtained by adding or subtracting to the rank R' i of Exp1A' the rank difference RD i calculated by applying to zRD i the inverse of the normalization function (1). Rank values are finally back-converted into signal values by interpolation of the rank on the graph of signal vs. rank constructed with one of the original data sets. This procedure allows two data sets to be obtained, Exp1C and Exp1D, which are statistically indistinguishable from the original data. To obtain a synthetic data set in which a pre-determined subset of genes receives a significant variation value we can possibly add a second step. We selected in Exp1C and Exp1D a random subset of genes, for example 500 increasing genes and 500 decreasing genes. To these genes, a second random variation value is applied, but instead of drawing random numbers on the interval [0,1], we limit the selection to the interval [0,p]. If we set the limiting p to 0.10, then the variation applied to the subset will have a p-value <= 0.1. For the genes receiving an additional variation contribution, the mean magnitude of zRD that is calculated between the synthetic and the original data is proportional to the magnitude of the applied p-value, as shown in figure 5 . This entire procedure, composed of two successive steps, results in synthetic data sets of high quality, because the generation of data mimics the observed variation of genes. We compared the synthetic data sets Exp1C and Exp1D to the same natural data set Exp1A, and plotted the corresponding zRD of one comparison against the other, as shown in figure 5 . We observed that these zRD were independent for genes that had not received an additional variation contribution, but were correlated for genes that had been changed. In general, however, this correlation was not absolute, except for some decreased genes. This phenomenon is explained by the high noise that characterizes weak signals: for such genes, there is a high probability that negative variation makes them reach the minimal rank value (zero). For high signals, there also exists a limit for the rank variation, but the noise is very small, and the truncation effect is not visible. We also observed that a small proportion of genes receiving an increased (decreased) variation contribution could be detected as increased (decreased) in one comparison, but decreased (increased) in another. All of these properties support the realistic nature of the synthetic data generated by our algorithm. RDAM performances We generated several synthetic data sets from the two experiments (Wt_t0a and Wt_t0b) by letting the number of increased and decreased genes equal 0, 100 or 500, and letting the maximum p-value of extra variation equal 0.3, 0.1, 0.05 or 0.01. The first question we addressed relates to the effectiveness of our scoring method in discriminating among real variations: does the overall process rank the genes correctly? To address this point, genes were ranked according to their ppv, and the number of hits was computed in a series of sublists of increasing length, selected from the top. From this number of hits, and the number of genes in the sublist, we calculated the real S and FDR. Plotting FDR against S allowed us to visualize the respective effects of the number of replicates, the magnitude of the variations, the number of varying genes, and the direction of variation on the performance of our rank difference method. Figure 6 shows the effect of the first two parameters on the ranking of increasing genes, in this case 100 varying genes. The FDR 50 , defined as the FDR observed when S = 50%, can be used to demonstrate these effects. Increasing the number of replicates improves the scoring performance, but this improvement is strongly modulated by the magnitude of the applied variation. For example, when the number of replicates equals, successively, 2, 3 and 5, the FDR 50 equals, respectively, 85%, 74% and 41% for small variations (p <= 0.3) and 3%, 0% and 0% for large variations (p <= 0.01). For variations that we consider from our experience to be realistic, i.e., p <= 0.10, the FDR 50 is equal to 58%, 30% and 5%, respectively. The number of varying genes also has an important impact, since under the same variation conditions, but with 500 increased genes instead of 100, we measured FDR 50 values of 24%, 8% and 1%, respectively, which represents a mean decrease in the FDR 50 of 20 percentage points. The second question we addressed is the quality of the FDR and S estimators. The genes were ranked according to their estimated FDR (or S), and the number of hits computed in a series of sublists of increasing length, selected from the top. From this number of hits, and from the number of genes in the sublist, we calculated the mean real FDR and S (figure 7 ). Both FDR and S are overestimated in this case, except for the point at 5% FDR in the groups with two replicates. If we consider individual comparisons, the distribution of errors has a higher variance for small estimator values. Despite this dispersion of errors in the low FDR range, the absolute number of genes attributed to a faulty category is always negligible and mainly conservative (overestimation of false positives), as shown in Table 1 . Figure 7 shows that the ratio between real and estimated S is rather constant, and we found that this ratio was close to the ratio between the estimated and real TV. Finally, we tested to see whether the independent analysis of positive and negative variations subsequent to standardization was dispensable, or if a one-step procedure could be used instead. In order to illustrate this point, we have constructed, from the experimental replicates Exp1A and Exp1B (sh100 at t = 0 h), a first group of two synthetic replicates having 500 increased and no decreased genes and a second group of two synthetic replicates without changed genes in order to reveal any clear differences that may exist between the one-step and the two-step procedures. Table 2 shows the number of genes selected at several FDR levels when the two competing methods were applied to the comparison between the two groups of synthetic data We can see that with the one-step analysis the number of true positives is lower and the estimate of FDR is largely biased toward higher values relative to the two-step analysis. Comparison with SAM on synthetic data The generation of synthetic data is also a powerful tool for comparing different methods of analysis. As an example, we conducted a systematic comparison between RDAM and SAM. We selected this method because it is popular, easy to use (there exists an Excel add-in), and can be considered as representative of numerous other nonparametric methods, which apply Monte Carlo procedures to estimate the distribution of the statistics used to quantify the relative difference of gene expression. Figure 6 shows that for two and three replicates our scoring procedure generates less FDR than SAM does across the entire sensitivity scale. For five replicates, the scoring procedure of SAM is better only in the low sensitivity (<20%) range. In terms of practical gain, and particularly when experimental costs are considered, the improvement obtained with RDAM is important because we have the same overall ranking quality as SAM but with one replicate less. We also compared the errors made on the estimation of FDR by the two methods, when the nominal FDR equals 20% (Table 3 ). We concluded that in this particular case FDR estimation was as good in RDAM (mean error of -4 percentage points and extreme error values of -7 and -1 percentage points) than in SAM (mean error of 0 percentage point and extreme error values of -6 and +5 percentage points). Other comparisons show that this conclusion holds true for all other conditions used to generate synthetic data sets (data not shown). However, we observed that SAM estimator was unable to reach the nominal level of FDR detection in case of few replicates and/or small extra variations (for example in case of two replicates and extra variation with p >= 0.10, the smallest FDR estimation delivered by SAM is greater than 20%). In conclusion the large differences in the number of true and false positives found between the RDAM and SAM methods (Table 3 ) are mainly explained by difference of scoring procedure efficiency between the two methods. Analysis of the TOR experiment When strains SH121 and SH100 were shifted to 37°C, RDAM detected roughly 2300–2500 genes as being either increased or decreased in each strain (column TV, Table 4 ). Most of these gene variations were caused by the temperature shift and were common to both strains, as shown by the differential analysis which detected only up-regulation of genes as a consequence of TOR2 temperature inactivation: 106 genes at 2 h and 92 genes at 6 h (column TV, Table 4 ). After 2 hours at 37°C, 19 annotated genes showed significant induction with a 10% FDR, whereas after 6 hours 39 genes were induced (supporting Tables 5,6 and 7 [see additional file 1 , 2 and 3 ]). However, these two groups of genes do not overlap, i.e. the shift to the nonpermissive temperature leads to a subsequent and transient increase in transcription of a small set of defined genes. We note that among these 39 genes, 2 are known to be regulated by the amino-acid-responsive transcriptional activator Gcn4 ([ 13 ], see Table 7 in additional file 3 ). With a selection criterion of 20% FDR, five other Gcn4 regulated genes are detected (CPA2, THI11, SNO1, SNZ1 and PRB1). Therefore, it seems that inhibition of the TOR2-unique function leads to an significant increase in the transcription of known Gcn4 target genes. It is still unclear, however, how the TOR2-unique pathway is connected to nutrient sensing or, vice versa, how nutrient sensing interferes with actin polarization. Comparison with SAM on the TOR experiment We ran our method in parallel with SAM in two situations displaying contrasting transcriptional responses. We tested a first comparison, Wt-t1 vs Wt-t0, which is characterized by a high number of varying genes and a good reproducibility between replicates, facilitating the detection of changes as reflected by the results of RDAM analysis which selected 620 increased genes at S = 50% and FDR = 6%. As SAM does not detect any increased genes at this selection level, we compared results obtained by the two methods at FDR = 10% and observed that among the 817 and 804 genes selected repectively by RDAM and SAM, only 426 were found in common. These results match what we found with synthetic data sets in case of high strength of variation (e.g. p <= 0.01 in figure 6D ). We then considered comparisons of biological interest, i.e Mu-t1 vs. Wt-t1 and Mu-t2 vs. Wt-t2, and in this situation RDAM did not select any decreased genes and found only a few increased genes (see Discussion and Table 4 ). On the contrary SAM failed to detect any genes, either increased or increased. Discussion Analysis of the TOR experiment RDAM is a method for identifying genes with changing expression levels using the user-determined FDR and/or S selection parameters. This method was used to study the effects of a thermosensitive mutation of TOR2 in yeast. RDAM succeeded in identifying the few genes that are differentially regulated by the TOR mutation from among the entire mass of genes perturbed by the temperature shift. Recently it has been shown that TOR controls the translation of Gcn4 via the eIF4alpha kinase Gcn2 [ 14 ]. Under conditions of TOR inactivation by rapamycin, Gcn4 translation is enhanced, leading to the activation of Gcn4-mediated transcription. Our data also demonstrate that TOR2 inactivation leads to enhanced transcription of Gcn4-controlled target genes (biological results based on RDAM analysis are discussed in a forthcoming paper). Further experiments may show how the TOR2-unique function is integrated into nutrient- (or amino acid-) responsive signaling pathways. Normalizing of signal Apart from randomly-distributed noise, microarrays are also prone to systematic effects that can bias the measurement of signal. All analysis methods are sensitive to systematic bias and include a preliminary normalizing step to make chips comparable. This is a limitation of this kind of approach, because the final result depends on the normalizing procedure used. Considering that all normalizing procedures rely on monotonous transformations that do not change the rank of raw data, we reasoned that if we used a statistics based on rank there would be no need to optimize the normalizing procedure. The rank unit we describe is similar to quantile normalization [ 15 ], but does not depend on the signal values of a particular chip as a reference: it can therefore be considered as an invariant. For example, if we focus specifically on the Affymetrix platform, we observe that the signal distribution changes with the different versions of the software: in MAS5, the 50th percentile is around 100, as compared to 1000 in MAS4. In our system, the rank of the genes at the same position in the signal distribution would not change, and would always be roughly equal to 50. This rank unit allows the drawing of plots in which all data are evenly distributed alongdimensions representing a signal. In addition, the linear density of points on the corresponding axes is constant, and the skewness of signal distribution has no effect on the graphical representation. In our system, all values different from 0 or 100 are assigned to one and only one gene, because ordering of signals always delivers a series of contiguous rank values, even in cases of equivalent signal values. 0 is assigned to all unexpressed genes, as long as a robust method is available to detect them, and 100 to all genes for which the signal is saturated. In Affymetrix technology, especially with the scanner setup presently used, saturation is not a matter of concern and in our analysis, the value 100 is simply assigned to the highest signal. It is a complex problem to identify genes that are not expressed in a given experiment, and we decided to consider as absent only genes having a signal less than zero, as they occur in results delivered by MAS4. Rank normalization results in transformation of the original signal distribution which is heavily skewed towards low values into a uniform distribution. As a consequence high rank variation could be assigned to small signal variations of weakly expressed genes, and it could be argued that our rank normalization method may bias variation detection towards genes with low signal. By using comparison between synthetic data sets we found no evidence of such a bias (data not shown). Systematic usage of duplicates and standardization of variation Our method has been developed within the framework of hypothesis testing and requires knowledge of the variation distribution for each gene when the null hypothesis is verified. The rationale of our approach considers replicated experiments as precisely representing a system in which all genes follow this hypothesis. However, it has long been recognized that the variation distribution expressed as a ratio or fold change is dependent upon the level of gene expression [ 16 ], and we show here that this property subsists when difference of signals or difference of ranks is used to measure variation. In theory it could be possible to use numerous replicates to obtain the empirical variation distribution of each gene. This is not possible for practical reasons, however, and we found that the classical centered-reduced standardization procedure can render variation distribution totally independent of gene expression level, as demonstrated by the QQplot analysis of figure 2A , and allow us to use duplicates to obtain the null variation distribution. Independent analysis of positive and negative variations In the algorithmic implementation of our method, we chose to proceed in two steps and deal with increased and decreased variations independently. First, this ensures that symmetrical comparisons (e.g. Exp1 vs. Exp2 and Exp2 vs. Exp1) give perfectly symmetrical results. Second, even if it were possible to devise another method that would allow one to proceed in one step, it seems more logical to consider increased and decreased variation separately. To clarify this point, a clear distinction must be made between up- or down-regulated mRNAs and increased and decreased variation. Regardless of the experimental points that are being compared, one always observes increased and decreased variations, but these variations have no absolute meaning because one only has to reverse the comparison to change the direction of variation. On the contrary, we can speak of up- or down- regulated mRNAs only when a causal effect exists, such as in a differentiation process or a kinetics or drug assay. In other words, a positive variation observed, for example, between two successive time points in a kinetic can be considered as an up-regulation whatever its mechanism – gene or post-transcriptional regulation – but it is meaningless to invoke any particular form of regulation when comparing, for example, two unrelated cancer tissues. In the case of down-regulated genes, the variation distribution of all weakly-expressed genes is truncated, due to the impossibility of a decreasing signal crossing the zero line. In the case of up-regulated genes, we do not observe the same effect for increasing variation of highly-expressed genes, first because the signal distribution is heavily skewed towards low values, and second because the variance of highly expressed genes is very small (figure 1A ). The observation that the reproducibility of variation was lower for down-regulated genes than it was for up-regulated genes [ 17 ] is partly explained by this reason, and we were able to demonstrate the statistical difference between up- and down-regulated genes by using synthetic data and observing that all FDR vs. S curves (figure 6 ) constructed with down-regulated genes were lower than the corresponding curves for up-regulated genes (data not shown). Moreover, we conducted a test showing that the joint analysis of increased and decreased variations degrades the quality of FDR estimation and reduces the number of true positives detected. Replicates We did not try to reduce the amount of raw data when using replicates, and devised a two-step method. A first statistics, the standardized rank difference zRD is constructed on each independent comparison, and p-values are assigned by considering an empirical distribution that matches the null hypothesis. Then a second statistics, the product of p-values ppv, is calculated and p-values are assigned from the null distribution obtained by simulation. Simulation of the null distribution used exactly the same steps that the analysis process follows, i.e. application of the direction rule and construction of the product of p-values, and resulted in a null distribution model we found appropriate seeing, both with experimental and synthetic data sets, that the observed distribution of ppv matches the null distribution when no variation occurs (data not shown). It turned out that this scheme is very flexible and of general applicability: because the second step is rooted in a rigorous statistical method that uses only p-values as input data, it is possible to adapt or to improve the entire process simply by focusing on the first step of p-value estimation. For example, to apply our method to cDNA glass arrays, the only step to be modified would be the variation standardization. Alternatively we could use the segmental approach proposed by Yang and colleagues [ 18 ], which is claimed to equalize log ratio distribution, or the variance stabilization method of Huber et al [ 19 ], which is efficient in equalizing variation distribution of transformed intensity measurements in both cDNA and oligonucleotide platforms. FDR, Total Variation and Sensitivity Estimation The way in which we estimated FDR is exactly the same as that suggested by B. Efron et al. in their demonstration of the equivalence of empirical Bayes and frequentist approaches (Efron B, Storey J. D. and Tibshirani R. , see equation 3.8 and [ 20 ]). We did, however, use another heuristic approach to estimate TV because we observed that the estimator proposed by Storey et al [ 20 ] could be very difficult or impossible to calculate when the expression of a small fraction of genes changes. We demonstrated using synthetic data that our estimator was not prone to this type of instability (not shown), and that under realistic conditions (additional variation of p <= 0.10) our estimate was 60%, 65% and 80% of the true TV in the case of two, three and five replicates, respectively. The accuracy of this estimator is obviously dependent on the power of the test, which is itself under the control of the number of replicates. We have also shown that estimated sensitivity was biased by a constant factor that was mostly determined by the error made in TV estimation. Finally, it must be emphasized that the error made in TV estimation has little effect on FDR estimation, as demonstrated by forcing RDAM to use the true TV and K values during the process of synthetic data analysis (data not shown). Synthetic data sets Using the empirical noise distribution observed between two replicates, we devised a method for constructing synthetic data sets. Most published methods add noise to a signal that is supposed to represent the true signal of the gene. We showed here that raw signals without denoising could be used and gave excellent result as judged both by the final distribution of signals and by indirect controls such as the preservation of variation distribution and the possibility of successfully analyzing synthetic data substituted for the original data. Synthetic data sets are well adapted for judging the respective performances of different analysis methods. To characterize the scoring procedure of a particular method, we used a new type of diagram that plots FDR vs. S, two quantities that relate to the subset of selected genes and that seem better adapted than Receiver Operator Characteristic (ROC, [ 21 ]), which relates to both selected and rejected genes (FDR vs. specificity). We proposed using FDR 50 , the FDR at S = 50%, as a comparative index between different methods and showed that RDAM has an FDR 50 that is 30 percentage points smaller than SAM in the case of three replicates and applied changes with p <= 0.10 (figure 6 ). Conclusions RDAM is a new statistical method whose performances have been precisely evaluated through extensive analysis of synthetic data sets. When applied to TOR experiment, our method succeeded in finding the few genes of biological interest which were concealed in the mass of varying genes induced by the temperature shift. Comparison with SAM showed that our method obtained the same (if not better) results but with a smaller consumption of chips We conclude that the good quality of the results obtained by RDAM is mostly due to the use of replicates to calibrate the noise and to the quasi-perfect equalization of variation distribution, which is related to the standardization procedure used and to the measurement of variation by rank difference. Methods Preparation of RNA Saccharomyces cerevisiae strains SH100 and SH121 [ 12 ] were grown overnight in yeast extract peptone glucose (YPD), diluted to an optical density measured at 600 nm of 0.05 (OD600 = 0.05), and grown for an additional 4 hours the next day. The main cultures were then inoculated in YPD medium and grown at 25°C or shifted to 37°C for 2 or 6 hours. All cultures were grown as independent duplicates and were harvested at a final OD600 of 0.8 to 0.9 to minimize the influence of differences in growth phase. Upon harvesting by centrifugation (2 min, 3000 × g) at 4°C, cells were washed once in ice-cold water, centrifuged again, and the cell pellet was flash frozen in liquid nitrogen. Total RNA was extracted using a hot phenol method essentially as described by Schmitt, M.E. et al. [ 22 ]. Microarray hybridization Affymetrix™ S98 Yeast Genome GeneChips, containing 6,400 S. cerevisiae (S288C strain) genes and 600 additional probe sets representing putative open reading frames [ 23 ], were used throughout this study. Synthesis of cDNA and in vitro transcription of biotin-labeled cRNA, as well as microarray hybridisation, washing and staining procedures, were carried out according to standard protocols as recommended by the manufacturer. Two independent preparations were used for each experimental point. Data processing The scanned microarray images were analysed using the algorithm implemented in MAS 5.0 (Affymetrix, Santa Clara, CA) and the generated raw data were further processed by scripts written in Matlab language (MathWorks, Natick, MA.). SAM analysis [ 3 ] of synthetic data was made using version 1.21 of the program [ 24 ] with the following default parameters: unlogged data, number of permutations set to 100 and "K-Nearest Neighbors Imputer" used. Raw data files were uploaded to NCBI's GEO repository under the series number GSE1814 . Abbreviations RD, rank difference; zRD, standardized rank difference; FDR, false discovery rate, S, sensitivity; TV; total variation Authors' contributions DM and MH initiated and conducted the TOR experiment. DM and PD processed probe preparation and chip hybridization. MB developped and tested RDAM method. All authors read and approved the final manuscript. Supplementary Material Additional File 1 Table 5 - Genes found decreased in the comparison sh121 vs sh 100 at 0 h and selected at FDR = 10% Click here for file Additional File 2 Table 6 - Genes found increased in the comparison sh121 vs sh 100 at 2 h and selected at FDR = 10% Click here for file Additional File 3 Table 7 - Genes found increased in the comparison sh121 vs sh 100 at 6 h and selected at FDR = 10% Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526220.xml
555594	Epidemiology of acute coronary syndromes in a Mediterranean country; aims, design and baseline characteristics of the Greek study of acute coronary syndromes (GREECS)	Background The present study GREECS was conducted in order to evaluate the annual incidence of acute coronary syndromes (ACS) and to delineate the role of clinical, biochemical, lifestyle and behavioral characteristics on the severity of disease. In this work we present the design, methodology of the study and various baseline characteristics of people with ACS. Methods/Design A sample of 6 hospitals located in Greek urban and rural regions was selected. In these hospitals we recorded almost all admissions due to ACS, from October 2003 to September 2004. Socio-demographic, clinical, dietary, psychological and other lifestyle characteristics were recorded. 2172 patients were included in the study (76% were men and 24% women). The crude annual incidence rate was 22.6 per 10,000 people and the highest frequency of events was observed in winter. The in-hospital mortality rate was 4.3%. The most common discharged diagnosis for men was Q-wave MI, while for women it was unstable angina. Discussion This study aims to demonstrate current information about the epidemiology of patients who suffer from ACS, in Greece.	Background During the past decades epidemiological investigations have provided a portrait of the potential candidate for acute coronary syndromes, in the US, as well as in many other parts of the world, especially in European countries [ 1 - 9 ]. However, many investigators claimed for differences in the profile, i.e. socio-demographic characteristics, prevalence of risk factors, dietary habits, etc. of people who suffer from coronary heart disease, between populations as well as among individuals within populations [ 10 - 12 ]. A potential explanation can be attributed to the gene-environment interactions, as well as various cultural and behavioural particularities. The profile of cardiovascular disease patients in Greece has been, mainly, investigated from few large scale, population-based studies, like the Seven Countries Study in the early 1960s [ 13 ], the Hellenic study of acute myocardial infarction, which recruited about 7500 patients with myocardial infarction from almost all hospitals countrywidethe in early 1990s [ 14 ], the CARDIO2000 case-control study of acute coronary syndromes [ 15 ], as well as some small case-control or observational studies that included patients from specific areas [ 16 , 17 ]. However, the prevalence, annual incidence and management of patients with coronary heart disease in Greece is unknown, since all these studies recruited patients during certain time periods, and information relating cardiovascular events with treatment, as well as lifestyle habits, like exercise, diet, and psychological stress and depression, are lacking. Additionally, during the past decade, Greece has experienced marked but uneven socio-economic development, with the average income increasing by about 20-fold [ 18 ]. Consequently, the lifestyle of people throughout the country has changed dramatically, as well as the incidence of cardiovascular disease. The aim of the GREEk study of acute Coronary Syndromes (GREECS) is to evaluate the prevalence and annual incidence of acute coronary syndromes (ACS), as well as the characteristics, and management of these patients, in a sample of six Greek urban and rural regions. Secondary goals are to examine the role of adoption to the Mediterranean diet and other lifestyle habits, as well as various clinical, biochemical, psychological and personal characteristics of these patients on the severity and the short (30 days) – long (6, 12 months) – term prognosis. In this work we present the methodology used, and the baseline characteristics of the studied population. Methods / Design Population of the study Between October 1, 2003 and September 30, 2004 (12 months) we enrolled almost all consecutive patients (participation rate = 98%) that entered in the cardiology clinics or the emergency units of six major General Hospitals, in Greece (Hippokration hospital in Athens, and the general hospitals in Lamia, Karditsa, Halkida, Kalamata and Zakynthos island). By the exception of Athens – where there are several other hospitals -, all the other hospitals cover the whole population of the aforementioned regions, including urban and rural areas. The studied regions were randomly selected from all Greek regions in order to cover a wide range of the county (Figure 1 ). Figure 1 Regions covered by the study. During the study period 2,172 patients were admitted for ACS in the selected hospitals, 1649 (76%) of them were men and 523 (24%) were women. Power analysis showed that the number of enrolled participants is adequate to evaluate two – sided differences between groups of the study and the investigated parameters greater than 20% (± 5%), achieving statistical power greater than 0.80 at 5% probability level (P-value). Diagnosis of ACS At entry a 12-lead electrocardiogram was performed and clinical symptoms were evaluated in all patients, by a cardiologist of the Study. Based on the electrocardiographic findings patients were classified as having ST-segment elevations, non-ST segment elevations or other electrocardiographic abnormalities. Moreover, blood tests were performed to detect evidence of myocardial cell death. We measured troponin I levels and the MB fraction of total creatinine posphokinase (CPK). According to the Joint European Society of Cardiology and American College of Cardiology Committee, blood samples were obtained on hospital admission, at 6 to 9 h, and again at 12 to 24 h if earlier samples were negative and the clinical index of suspicion was high [ 19 ]. We included only cases with discharge diagnoses of ACS (acute myocardial infarction (MI) or unstable angina (UA)). In particular, acute myocardial infarction was defined by at least two of the following features: (a) electrocardiographic changes (patients with or without ST segment elevations), (b) compatible clinical symptoms, and (c) specific diagnostic sensitive biomarkers elevations (troponin I > 0.4 ng/ml and the MB fraction of CPK > 8.8 ng/ml). UA was defined by the occurrence of one or more angina episodes, at rest, within the preceding 48-hours, corresponding to class III of the Braunwald classification [ 20 ]. The study was approved by the Medical Research Ethics Committee of our Institution and was carried out in accordance with the Declaration of Helsinki (1989) of the World Medical Association. Other clinical and biochemical characteristics In all patients a detailed medical history was recorded, including previous hospitalization for cardiovascular disease (i.e. coronary heart disease, stroke or other cardiovascular disease), presence and management of hypertension, hypercholesterolemia, renal failure and diabetes mellitus. Moreover, we recorded patients' medical family history. In particular, we asked information concerning first-degree relatives (biological parent, or brother, or sister) about presence of coronary heart disease, hypertension, dyslipidemias and diabetes. Premature history (<55 years old for males and < 65 years old for females) of myocardial infarction, sudden death, coronary arteries bypass grafting procedure or percutaneous coronary angioplasty in first-degree relatives classified the participants in the positive family history group for coronary heart disease [ 20 ]. Premature history (<55 years old for males and < 65 years old for females) of hypertension, hypercholesterolemia, hypertriglyceridemia, and diabetes, defined as the use of special medication or known, but untreated, condition classified the participants in the positive family history group for these co-morbidities. In addition to troponin I and the MB fraction of creatinine mentioned above, we also measured white blood cell counts, urea, and uric acid. Demographic, anthropometric and lifestyle characteristics Socio-demographic characteristics included: age, sex, marital status and number of children, years of school, type of occupation and occupational skills, which they were evaluated through a ten-point scale from unskilled – hand workers (lower values) to executive – skilled workers (higher values) that has been developed for the purposes of the Study; and mean annual income of the family (through self reports) during the last three years. Regarding people in the family who were not working, we used the average family income, while for unemployed individuals we used the basic monthly allowance they take from the Social Service Office. Height and weight was measured, to the nearest 0.5 cm and 100 g respectively. Body mass index (BMI) was then calculated as weight (in kilograms) divided by height (in meters) squared. Based on the World Health Organization [ 21 ], overweight was defined as BMI between 25 and 29.9 kg/m 2 , while obesity as BMI greater than 29.9 kg / m 2 . To evaluate physical activity status of the patients during the past year we used a modified version of a self-reported questionnaire provided by the American College of Sports Medicine [ 22 ]. Based on this questionnaire we assessed the frequency (times per week), duration (in minutes per time) and intensity of sports or occupation related physical activity. Participants who did not report any physical activities were defined as sedentary. For the rest of the participants we calculated a combined score by multiplying the weekly frequency, duration and intensity of physical activity. Current smokers were defined as those who smoked at least one cigarette per day or have stopped cigarette smoking during the past 12 months. Former smokers were defined as those who had stopped smoking more than one year previously. The rest of them were defined as never smokers or rare smokers. Exposure to environmental cigarette smoke for at least 30 minutes per day and three days per week at workplace, home or other public places was recorded in all patients, too. Nutritional habits and dietary ascertainment The evaluation of the nutritional habits was based on a validated semi-quantitative food frequency questionnaire [ 23 ]. The consumption of certain food items and the portion size as an average per week, during the past year, was recorded. Then, the frequency of consumption was quantified approximately in terms of the number of times a month the food was consumed. In order to describe total diet composite scores were applied, which are necessary for the evaluation of epidemiological associations. According to a dietary pyramid that has been developed to describe the Mediterranean dietary pattern [ 24 ] we calculated a special diet score for each participant that assessed adherence to the Mediterranean diet (range 0 – 55). In particular, for the consumption of 11 items presumed to be close to this pattern (i.e. those suggested on daily basis or more than 4 servings per week) we assigned score 0 when a participant reported no consumption, 1 when reported consumption of 1 to 4 times / month, 2 for 5 to 8 times, 3 for 9 to 12 times / month, 4 for 13 to 18 times / month and 5 for more than 18 times per month. On the other hand, for the consumption of foods presumed to be away from this diet (like meat and meat products) we assigned the opposite scores (i.e. 0 when a participant reported almost daily consumption to 5 for rare or no consumption). Especially for alcohol we assigned score 5 for consumption of less than 3 wineglasses per day, score 0 for consumption of more than 7 wineglasses per day and scores 1 to 4 for consumption of 3, 4 to 5, 6 and 7 wineglasses per day. Higher values of this diet score indicates greater adherence to the Mediterranean diet, while lower values indicate adherence to the "Westernized" diet. Psychological evaluation Depressive symptomatology was assessed using a translated and validated version of the Center of Epidemiological Studies Depression Scale (CES-D) [ 25 ]. The CES-D is a well known and world-widely used self-rating scale for the measurement of depression. It is a self-reporting instrument and was originally developed in order to assess depression symptoms without the bias of an administrator affecting the results. Higher scores on this scale are indicative of more severe depression [ 25 ]. CES-D consists of 20 items that cover affective, psychological, and somatic symptoms. The patient specifies the frequency with which the symptom is experienced (that is: a little = 1, some = 2, a good part of the time = 3, or most of the time = 4). Previous investigations indicate that the CES-D score is a valid and sensitive measure of clinical severity in depressed patients and support its continued use as a research instrument [ 26 ]. Clinical symptomatology of occupational stress was determined by a special self-reported questionnaire based on the survey obtainable by the Job Stress Help Line [ 27 ]. By this questionnaire the presence of occupational stress and insecurity was evaluated and summarized for every patient by scoring each item of the questionnaire (1 = yes, 0 = no). Total score ranged from 0 to 10. Statistical analysis In this, brief, baseline report continuous variables are presented as mean values ± standard deviation. The categorical variables are presented as absolute and relative (%) frequencies. Future analyses will follow with appropriate statistical techniques. Associations between continuous variables and group of patients will be evaluated through the analysis of variance, after controlling for equality of variances (homoscedacity) using the F-test. Due to multiple comparisons we will apply the Bonferroni rule to correct for the inflation of type – I error. Associations between categorical variables will be tested by the use of the chi-squared test, without the correction of continuity. Correlations between continuous variables will be tested by the use of Pearson's correlation coefficient for the normally distributed, and by the use of Sperman's rho coefficient for the ordinal or skewed variables. The association between the investigated socio-demographic, clinical and biochemical characteristics on the short term outcome (i.e. 1 month) will be tested by the development of multiple logistic regression models, while the association of the aforementioned characteristics on the long term outcome (6 and 12 months) will be tested by the use of Cox proportional hazards models. Appropriate tests for goodness-of-fit (i.e. deviance and Pearson's residuals) will be applied in all models. All statistical calculations will be performed on the SPSS version 12.0 software (SPSS Inc, Texas, U.S.A.). Baseline characteristics From October 2003 to September 2004, 2172 patients with discharge diagnosis of ACS were enrolled into the study (1649 men, 65 ± 13 years old and 523 women, 62 ± 11 years old, p < 0.001). The men-to-women ratio was 3- to -1. The annual incidence of ACS was 22.6 per 10000 of people (34.0 per 10000 men and 10.9 per 10,000 women). The annual incidence was calculated with the exception of events at Hippokration Hospital in Athens, because it was very difficult to define the referent population since there are many hospitals in the area. The mean number of daily admissions was 6 ± 3 persons per day, averaging 5 ± 4 men and 2 ± 1 women. Table 1 illustrates the age-sex distribution of the patients. Figure 2 illustrates the series of the monthly number of admissions for ACS in the selected hospitals during the 12 months period. A seasonal variability was observed in the counts of hospital admissions due to ACS (Figure 2 ). Table 1 Age-sex distribution of the patients Age (years) Men (n = 1609) Women (508) < 30 5 (0.3%) 0 (0.0%) 30–39 28 (1.7%) 5 (1.0%) 40–49 220 (13.3%) 31 (5.9%) 50–59 335 (20.3%) 52 (9.9%) 60–69 388 (23.5%) 115 (22.0%) >70 673 (40.8%) 320 (61.2%) Figure 2 Monthly distribution of hospital admissions for ACS in the centers of the study (dotted line represents the smoothed 12 month average) Of the 2172 patients enrolled into the study, 38% had ST segment elevations, 27% had non-ST segment elevations and the rest of them, i.e. 35% patients, had other electrocardiographic findings. According to the discharge diagnosis, 764 (35%) patients were diagnosed as having unstable angina, 699 (32%) patients as having non-Q-wave MI and 709 (33%) patients as having Q-wave MI. Table 2 presents the cross-tabulation of electrocardiographic (ECG) findings and discharge diagnosis. The majority of patients with ST segment elevations had Q-wave MI, while about 61% of patients with an undetermined electrocardiographic pattern had UA. It is of interest that a considerable proportion of patients without ST segment elevations or other ECG findings had Q-wave MI. Moreover, 8% of patients with ST segment elevations were defined as having UA at discharge. Table 2 Electrocardiographic (ECG) findings and discharge diagnosis Diagnosis at discharge ECG changes Q-wave MI Non-Q-wave MI Unstable angina ST-segment elevations 88% 18% 8% Without ST-segment elevations 5% 46% 31% Other ECG findings 7% 36% 61% % values are by diagnosis group Table 3 illustrates various baseline clinical characteristics of the patients by discharge diagnosis. We also found that 24% of ACS patients had family history of diabetes, 24% of patients had family history of dyslipidemias and 43% of patients reported family history of hypertension. Table 3 Clinical characteristics by discharge diagnosis Unstable angina Non-Q-wave MI Q-wave MI Men Number 544 (33%) 523 (32%) 582 (35%) Age (years) 65 ± 12 67 ± 13 63 ± 13 Body mass index (kg/m 2 ) 28 ± 4.7 27 ± 3.5 27 ± 37 Obesity (%) 116 (24%) 86 (20%) 113 (22%) Hypertension (%) 257 (51%) 215 (41%) 215 (41%) Hypercholesterolemia (%) 245 (49%) 151 (42%) 212 (46%) Diabetes mellitus (%) 146 (30%) 145 (34%) 134 (27%) Renal failure (%) 22 (5%) 31 (8%) 23 (6%) Prior coronary heart disease 299 (60%) 212 (51%) 153 (29%) Women Number 220 (42%) 176 (34%) 127 (24%) Age (years) 70 ± 10 74 ± 11 72 ± 13 Body mass index (kg/m 2 ) 28 ± 5.1 28 ± 4.2 28 ± 4.2 Obesity (%) 52 (26%) 27 (20%) 37 (31%) Hypertension (%) 160 (76%) 106 (71%) 76 (63%) Hypercholesterolemia (%) 105 (50%) 61 (50%) 52 (46%) Diabetes mellitus (%) 80 (40%) 63 (45%) 32 (28%) Renal failure (%) 13 (7%) 17 (13%) 3 (3%) Prior coronary heart disease 115 (56%) 63 (44%) 29 (25%) P < 0.05 and *P < 0.01 between diagnosis group after correcting for multiple comparisons through the Bonferroni adjustment Table 4 illustrates various lifestyle and behavioral characteristics of the patients by discharge diagnosis. Table 4 Lifestyle and behavioral characteristics by discharge diagnosis Unstable angina Non-Q-wave MI Q-wave MI Men Number 544 (33%) 523 (32%) 582 (35%) Diet score (0–55) 27 ± 2.6 26 ± 2.7 24 ± 2.4 CES-depression score (0–60) 18.5 ± 11 22.5 ± 10 21 ± 10 Physical inactivity (%) 20% 17% 22% Former smoking (%) 50% 41% 32% Current smoking (%) 37% 47% 57% Passive smoking (years) 16 ± 15 17 ± 15 20 ± 16 Financial status Low 7% 5% 8% Medium 53% 65% 55% High 37% 28% 33% Very high 3% 2% 4% Years of school 8 ± 4 7.7 ± 4 9.0 ± 4.6 Occupational skills (0–10) 4.2 ± 1.7 4.1 ± 1.4 4.6 ± 1.6 Women Number 220 (42%) 176 (34%) 127 (24%) Diet score (0–55) 26 ± 2.6 25 ± 2.4 23 ± 2.3 CES-depression score (0–60) 24 ± 11 23 ± 11 20 ± 9 Physical inactivity (%) 32% 28% 38% Former smoking (%) 11% 5% 11% Current smoking (%) 49% 71% 55% Passive smoking (years) 23 ± 15 22 ± 15 21 ± 16 Financial status Low 15% 9% 13% Medium 58% 73% 66% High 23% 18% 16% Very high 4% 1% 5% Years of school 6 ± 4.3 5.5 ± 3.6 6 ± 3.5 Occupational skills (0–10) 4.0 ± 1.3 3.5 ± 1.2 3.8 ± 1.2 P < 0.05 and **P < 0.01 between diagnosis group after correcting for multiple comparisons through the Bonferroni correction The median (and 25 th , 75 th percentiles) time between the overt of symptoms and the time medical care was sought, was 4 (2, 10) hours. Based on the discharge diagnosis the duration of hospitalization was 7 (5, 8) days for patients with Q-wave MI, 6 (5, 8) days for non-Q-wave patients and 5 (3, 7) days for UA patients. Furthermore, 60% of patients with ST-elevation received trombolytic therapy. The in-hospital mortality rate was 36 deaths per 1000 male patients and 63 deaths per 1000 female patients (i.e. overall 82 deaths). The in-hospital mortality rate of patients with ST segment elevation was 74 deaths per 1000 patients, for non-ST segment elevation was 34 deaths per 1000 and for undetermined electrocardiographic findings was 2 deaths per 1000 patients. Discussion In this brief report we presented the design, and the aims of an epidemiological study of ACS, that has been conducted in six major general hospitals, in Greece (i.e. the GREECS). We also presented the baseline characteristics of almost all patients who hospitalized in these hospitals for ACS during the study period (i.e. October 2003 to September 2004). The overall incidence rate of ACS observed in our survey was 22.6 events per 10000 of population. Based on this figure it could be speculated that the prevalence of ACS in the investigated Greek areas is 2.6%. In a similar recent study located only in northwestern Greece, which also included sudden cardiac death before hospital admission, the incidence rate of ACS was much higher (i.e. 39 events per 10000 people) [ 17 ]. Based on a review paper by Chimonas [ 28 ] that evaluated the prevalence and incidence of ACS in Greece during 1988, we observe that the current rates are slightly higher in men compare to the annual incidence in late 1980s (i.e. 29.7 events per 10000 of people), but the observed incidence rate in women is almost two-fold as compared to 1988 (i.e. 5.2 events per 10000 people). Future analyses of our study will answer to the questions whether the observed difference in incidence rates could attribute to various lifestyle and behavioral changes, like dietary and smoking habits, or other environmental particularities, occurred in Greece the last years. We also observed that the annual incidence for men was 34 per 10,000 of people, while the incidence for women was significantly lower, i.e. 11 per 10,000 of people. This finding will be tested whether can be attribute to the increased smoking habits observed in women during the past decades, as well as to various other lifestyle habits and work-related conditions, like physical inactivity, unhealthy diet, social insecurity and job stress, observed in female population from late 1960s to present [ 18 ]. Furthermore, we observed that the in-hospital mortality rate in was 4.3%, which was similar to the in-hospital mortality rate that has been calculated based on a sample of 25 European countries, i.e. 4.9% [ 9 ]. In this work we have also presented various baseline clinical and lifestyle characteristics of the enrolled patients. Future analyses will evaluate the role of the investigated clinical characteristics as well as family history of the common cardiovascular risk factors on the severity and prognosis of the ACS patients. Regarding dietary and other lifestyle related habits, we aim to evaluate whether the adherence to a Mediterranean dietary pattern, the adoption a physically active lifestyle, the abstinence of smoking and the control of psychological factors is associated with less severe disease and a better short and long term outcome. We anticipate that the completion of the follow up and the analysis of the results will provide current, novel and valuable information about the epidemiology of ACS, in Greece, as well as the role of clinical, psychosocial and lifestyle characteristics on the prognosis of cardiac patients. Abbreviations ACS = acute coronary syndromes CPK = creatinine posphokinase MI = acute myocardial infarction UA = unstable angina ECG = electrocardiographic CES-D = Center of Epidemiological Studies depression BMI = body mass index Competing interests The author(s) declare that they have no competing interests. Authors' contributions CP, DBP = are the principal investigators of the study, had the concept and the design of the study, and wrote the paper, AA, YK, YM, SZ, PS, CS = contributed to the design of the study, GK = contributed to the data management and analysis Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555594.xml
519007	A Relay-Signal Model of Nematode Vulval Development	null	A fundamental question in developmental biology is, how does a multicellular organism develop from a single cell? It's clear that one cell begets two, two beget four, and so on, but how do the newly created cells know which developmental fate to pick? Major insights into this question have come from identifying genes, molecules, and intercellular signaling pathways involved in a wide range of developmental processes. Operating in labyrinthine, often overlapping pathways, intercellular signals determine whether a cell divides, differentiates, migrates, and even lives or dies. Scientists prefer to work out such problems in organisms with a manageable number of cells for obvious reasons, making the 959-cell soil nematode Caenorhabditis elegans a popular developmental model. C. elegans can exist as either a male or a hermaphrodite, and for some biologists, the hermaphrodite vulva—which consists of just 22 cells—is the perfect system for working out key aspects of intercellular signaling and cell fate. In a new study, Alex Hajnal and colleagues challenge conventional thinking about vulval cell specification by identifying an enzyme that can amplify a signal's range and help turn three non-vulval precursors into vulval cells. Surprisingly, the enzyme, called ROM-1, accomplishes this feat by acting in the signal-receiving vulval precursor cells, rather than in the signal-sending cell that instructs the vulval cell fates. The worm vulva forms a bridge between its gonad and the opening to the outer epidermal layer, called the cuticle. In the current model of vulval formation, a group of twelve epidermal cells, called Pn.p cells, lines the ventral surface of the worm. Six of these cells, P3.p–P8.p, the vulval precursor cells (VPCs), have the potential to become vulval cells. During postembryonic development, the anchor cell in the larval gonad secretes an epidermal growth factor (called LIN-3) that activates the EGFR/RAS/MAPK signaling pathway and induces just three of the precursors to differentiate into vulval cells. The VPC closest to the anchor cell, P6.p receives most of the signal, and differentiates into eight vulval cells that form the tube linking the uterus to the gonad. Positioned on either side of P6.p, P5.p and P7.p receive a slightly attenuated signal, which, combined with a lateral signal from P6.p, gives rise to seven vulval cells that form vulval structures. The other three vulval precursors, P3.p, P4.p, and P8.p, it was thought, are too far away to receive the vulval induction signal and fuse into the surrounding epidermis. Vulval precursor cells in C. elegans The LIN-3 epidermal growth factors sit nestled within the cell membrane and must be “processed” to become active, prompting Hajnal and colleagues to look for candidate enzymes that could be doing the processing. They investigated the Rhomboid family of proteases, which are known activators of epidermal growth factor transmembrane proteins, and found one, ROM-1, with the amino acid profile required for catalytic protease activity. After showing that rom genes were not required for normal vulval development, the authors had a closer look at their role in vulval cell fate specification. Since loss of ROM-1 reduces the severity of a defect (in this case, multiple vulvas) caused by hyperactivation of the EGFR/RAS/MAPK pathway but has no effect on the precursors closest to the anchor cell, the authors conclude that ROM-1 enhances the EGFR/RAS/MAPK pathway, allowing it to reach the distant P3.p, P4.p, and P8.p precursors. LIN-3 exists in two variant forms of different lengths, the longer one carrying a stretch of 15 extra amino acids in the region that is cleaved off to yield an active growth factor. Hajnal and colleagues show that ROM-1 only acts on the longer form to regulate the EGFR/RAS/MAPK pathway—and that the ROM-1/LIN-3 interaction occurs in the VPCs, independently of the anchor cell. They go on to propose a two-step model of vulval cell specification in which ROM-1 “extends the range” of the anchor signal, relaying it from the proximal to the more distant precursor cells by promoting the secretion of the long version of LIN-3. In normal development, LIN-3 secretion by the VPCs may serve initially to maintain the differentiation potential of all the precursors, while the anchor cell signal may seal their fates at a later phase.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC519007.xml
526208	Use of polyethyleneimine polymer in cell culture as attachment factor and lipofection enhancer	Background Several cell lines and primary cultures benefit from the use of positively charged extracellular matrix proteins or polymers that enhance their ability to attach to culture plates. Polyethyleneimine is a positively charged polymer that has gained recent attention as a transfection reagent. A less known use of this cationic polymer as an attachment factor was explored with several cell lines. Results Polyethyleneimine compared favorably to traditional attachment factors such as collagen and polylysine. PC-12 and HEK-293 cells plated on dishes coated with polyethyleneimine showed a homogeneous distribution of cells in the plate, demonstrating strong cell adhesion that survived washing procedures. The polymer could also be used to enhance the adherence and allow axonal outgrowth from zebrafish retinal explants. The effects of this coating agent on the transfection of loosely attaching cell lines were studied. Pre-coating with polyethyleneimine had the effect of enhancing the transfection yield in procedures using lipofection reagents. Conclusion Polyethyleneimine is an effective attachment factor for weakly anchoring cell lines and primary cells. Its use in lipofection protocols makes the procedures more reliable and increases the yield of expressed products with commonly used cell lines such as PC-12 and HEK-293 cells.	Background Molecular cell biology experimentation often requires the culture of primary cells or immortalized cell lines. The most common substratum used in cell culture consists of a plastic dish that offers a negatively charged surface. A drawback of this technology is that some anchorage-dependent cell types do not produce sufficient amounts of positively charged extracellular matrix proteins, adhering only weakly to the plastic substratum. Pre-coating of the plastic surface with extracellular matrix proteins such as collagen, fibronectin, laminin, etc., usually enhances the attachment of these cell types [ 1 , 2 ]. Synthetic polymeric cations such as polylysine or polyornithine have also been used as attachment promoting factors in numerous studies [ 3 , 4 ]. Polyethyleneimine (PEI) is an organic polymer that has a high density of amino groups that can be protonated. At physiological pH, the polycation is very effective in binding to DNA and can mediate the transfection of eukaryotic cells [ 5 ]. The original PEI-based protocol has been used in our laboratory successfully for transfections with HEK-293 (human embryonic kidney) and PC-12 (rat pheochromocytoma) cells. However, lipofection using some of the last generation cationinc lipids, LipofectAMINE-2000 or LipofectAMINE-Plus (Invitrogen), yielded higher efficiencies of gene delivery, particularly with PC-12 cells (M. González-García and R.P. Ballestero, unpublished observations). Several modifications of the original PEI-based protocol have been described that improve the efficiency of transfection significantly, rivaling the lipofection protocols [ 6 - 10 ], but they have not been attempted yet in our laboratory. A notable difference, related to the attachment of the cells to the substratum, was observed during the transfection experiments. Whenever the cells were transfected by the original PEI-based protocol, the cells remained firmly anchored to the plates throughout histochemical procedures. HEK-293 and PC-12 cells are considered weakly anchoring cells [ 11 - 14 ], and many were detached from the plates during histochemical stainings of cells transfected by lipofection. This initial observation prompted us to study the attachment-enhancing properties of PEI and its effects on lipofection of eukaryotic cells. PEI has been used previously to coat surfaces to promote the attachment of neurons in primary cell cultures [ 15 , 16 ]. A modified version of PEI that incorporates hydrophobic groups was demonstrated to be highly effective in the attachment of several cell lines, allowing differentiation of neurons and preventing cell loses during multiple washing steps [ 17 ]. This report presents a comparison of PEI with other traditional coating agents as attachment factors for several cell lines and primary cultures of retina tissue. Additionally, an enhancement of the transfection yield of weakly anchoring cell lines is shown by the combination of PEI coating with lipofection. Results PEI promotes the attachment of weakly anchoring cells and primary tissues PC-12 cells are used as a model of neuronal differentiation in the laboratory, as treatment of these cells with nerve growth factor (NGF) induces neurite extension and the expression of biochemical markers of the sympathetic neuronal phenotype [ 18 ]. They grow as weakly anchoring cells and collagen or polylysine polymers are frequently used for pre-coating plates for PC-12 cell culture [ 13 , 19 , 20 ]. PC-12 cells were cultured in the laboratory in naive wells or in wells pre-coated with diverse attachment factors (multiwell-12 tissue culture dishes), to observe the effect on the anchorage of the cells to the substratum. In the absence of any coating agent, the cells showed a characteristic tendency to form clusters of cells that accumulate towards the center of the well; these cells are firmly attached among themselves, but very weakly attached to the tissue culture dish (Figure 1D ). This results in a very heterogeneous distribution of cells (very few cells remain towards the edge of the wells) and significant loss of cells during washing procedures or changes of medium. Pretreatment of the plates with PEI resulted in a significantly more homogeneous distribution of cells in the well, with cells attaching to the plate firmly, showing a much lower tendency to clustering (Figure 1A ). For comparison, plates were also pretreated with other commonly used coating agents, collagen (Figure 1B ) and poly-D-Lysine (PDL; Figure 1C ), resulting also in firmer attachment of the cells to the plates and a more homogeneous distribution of cells. To test further the anchoring enhancement property observed with the PEI pretreatment of the culture dishes, a second system was utilized. Retinal explants from teleost fish have been used in the study of nerve regeneration [ 21 , 22 ]. When a lesion is applied to the optic nerve of the fish, a regeneration response is initiated and the retinal ganglion cells (RGCs) of the retina re-extend their axon towards the tectal target tissue. If the retina from such a "primed" fish is explanted and cultured, the regeneration response is observed in vitro by the accelerated extension of long neurites. However, this phenomenon requires the use of extracellular matrix or attachment factors (for example collagen or PDL coating), as the explants show very low affinity for the uncoated plastic surface. Experiments were conducted to observe if PEI could act as an attachment factor conducive to axonal outgrowth from zebrafish retinal explants. Retinal explants from control zebrafish eyes were able to attach to PEI-pretreated culture dishes (Figure 1E ). Furthermore, retinal explants from fish that received a conditioning lesion were able to extend axons vigorously when cultured using PEI as attachment factor (Figure 1F ). The results with the retina explants suggested that neuronal cells attached to PEI-coated dishes can differentiate, generating neurites that attach well to the substrate. To test this hypothesis with the pro-neuronal PC-12 cells, differentiation experiments by NGF treatment were performed with cells attached to dishes coated with PEI. The PC-12 cells treated with NGF remained firmly attached to the plate over several days, generating networks of neurites (Figure 2 ). The results indicate that PEI is permissive for the differentiation process and for the attachment of the neurites to the dish. The differentiated cells remained firmly anchored throughout immunocytochemistry experiments (M. Challa, G. R. Chapa, M. González-García and R. P. Ballestero, unpublished results). Strength of anchoring of eukaryotic cells to culture dishes pretreated with PEI and other attachment factors Three different cell lines were selected to test the ability of PEI to promote strong anchoring to plastic culture dishes. PC-12 and HEK-293 cells have been described above as weakly anchoring. On the other hand, MYS cells are primary fibroblasts that adhere strongly to plastic culture dishes, growing to form a monolayer of cells on the surface. To test the strength of anchoring of the diverse cells to plates, a protocol was performed in which 4 consecutive washes with isotonic buffer were performed, followed by the use of a colorimetric protocol for counting the cells that remained in the plate (based on the vital dye neutral red). Plates pretreated with the diverse attachment factors were compared with plates that received no pretreatment (untreated). Experiments were performed in triplicate, and the averages of dye retained in the plates that received each treatment were calculated. Those averages were normalized to the average obtained with the PEI pretreatment, which was assigned the arbitrary value of 100.0% in all the experiments. The results from representative experiments with the 3 cell lines are shown in Figure 3 (at least three independent experiments were conducted with each cell line). PC-12 cells attached almost equally well to plates coated with PEI, collagen or PDL (relative cell counts of 100.0% ± 5.3%, 89.3% ± 5.0% and 96.3% ± 5.8%, for PEI, collagen and PDL respectively). However, a significant loss of cells was observed when comparing untreated plates with PEI-pretreated plates, with a relative cell count of 43.9% ± 5.8% (Figure 3A ). In the case of HEK-293, the cells attached strongly to both PEI- and PDL-pretreated wells. In the representative experiment shown in Figure 3B , the relative counts with those two treatments were 100.0% ± 0.4% and 96.8% ± 1.7% respectively. However, a large number of cells were lost when plated in the untreated wells (relative count of 8.3% ± 0.6%), or in the wells that were pretreated with collagen (11.5% ± 1.2%), suggesting that these cells attach fairly loosely to plastic or to collagen-coated wells. Finally, when the fibroblast cells (MYS cells) were utilized, the cells seemed to attach rather well to all four surfaces, including to the untreated wells (Figure 3C ). Figure 3C shows a representative plot, displaying relative MYS cell counts of 100.0% ± 2.2%, 85.9% ± 7.8%, 73.7% ± 6.6%, and 77.1% ± 1.4%, with wells pretreated with PEI, collagen, or PDL, or untreated wells respectively. This cell line therefore attaches fairly well to the untreated plastic surface comparatively to the PC-12 and HEK-293 cell lines. To provide an indication of the variation among experiments, the averages ± standard deviations of the relative cell counts obtained in the independent experiments were calculated for each cell line and treatment (note that since in all the experiments the count for PEI-pretreated wells was set at 100.0%, the value for the global average with this coating agent is exactly 100.0% for all the cell lines). The comparative results for the other treatments are indicated below. For the PC-12 cells, the relative counts were 81.5% ± 8.8% for collagen-pretreated wells (n = 4), 93.9% ± 21.2% for PDL-pretreated wells (n = 4), and 52.1% ± 13.7% for untreated wells (n = 4). For the HEK-293 cells, the relative counts were 16.3% ± 12.7% for collagen-pretreated wells (n = 3), 99.6% ± 2.5% for PDL-pretreated wells (n = 3), and 9.0% ± 4.1% for untreated wells (n = 3). In the experiments with MYS cells, the average of the relative counts with collagen-pretreated wells was 92.7% ± 16.2% (n = 3), 71.1% ± 6.7% for PDL-pretreated wells (n = 3), and 78.4% ± 11.5% for untreated wells (n = 3). Statistical analysis indicates that the enhancement of adhesion of both PC-12 and HEK-293 cells to PEI-pretreated dishes versus untreated plastic is significant (p < 0.05 and p < 0.01 respectively, t-test analysis), while it is not statistically significant for MYS cells (p > 0.05). To further characterize the properties of PEI as an attachment factor for weakly anchoring cells, experiments were carried out to analyze the dosage of PEI that can provide optimal cell anchoring, the range of cell numbers that can benefit from the presence of PEI, and the stability of PEI as attachment factor. Figure 4A shows the results from a representative experiment testing the strength of attachment of HEK-293 cells to dishes coated with various doses of PEI. The cell counts were normalized to the value obtained for the treatment with 25 μg/ml of PEI, which was set at 100.0%. The results indicate that concentrations of PEI of 2.5 μg/ml or higher resulted in maximal attachment enhancement, suggesting that the surface of the plastic dish is fully coated with the polymer at these concentrations. The higher concentrations of the polymer did not seem to have toxic effects on the cells if the excess of PEI remaining in solution was removed thoroughly (slight toxic effects were observed at the 250 μg/ml concentration if the solution was not fully removed after the treatment). The experiment shown is representative of 4 independent experiments. Figure 4B represents the results obtained with various numbers of PC-12 and HEK-293 cells. The figure shows relative cell counts, where an arbitrary value of 100.0% was assigned to the average count obtained from the PEI-treated wells with the highest number of each cell line (3.2 × 10 5 HEK-293 cells and 1.5 × 10 6 PC-12 cells). The results indicate that PEI worked well as an attachment factor over a wide range of cell numbers for both PC-12 and HEK-293 cells. Lower cell numbers could not be tested reliably because they were close to the limit of sensitivity of the neutral red assay. The results shown are representative of at least 4 independent experiments performed with each cell line. Figure 4C shows the results of an experiment conducted to test the stability of PEI as an attachment factor. In this experiment, a set of plates were treated with PEI and then kept in PBS at 4°C for 10 days. In a second test, plates were treated with PEI and kept (with medium) in the 37°C CO 2 incubator for 3 days, with a medium change performed every 24 h. The performance of PEI in these plates was then compared with plates treated with PEI using the standard protocol described for previous experiments. The results show relative cell counts normalized to the average of the absorbances obtained with the standard PEI-treated plates, which was given the arbitrary value of 100.0%. The results indicate that the PEI coating remains stable on the surface of the plastic dish for at least 10 days of refrigeration, and that it is not removed by incubation at 37°C in medium or even by repeated medium changes. Results are representative of 4 independent experiments performed in triplicate. PEI pretreatment enhances lipofection of weakly anchoring cells Transfection of eukaryotic cells by lipofection involves several steps, media additions and replacements, which can take a toll in weakly anchoring cells. Even if care is practiced to avoid cell loses, the weakly anchored cells may be less efficient in the uptake of the transfection complexes. Since PEI pretreatment of cell culture dishes increased the strength of attachment of cells to such surfaces, it was hypothesized that the attachment factor may have a positive effect in transfection yields obtained by lipofection. The three cell lines described above were used to test this hypothesis utilizing the coating agents previously examined. Transfections were performed with a plasmid encoding the reporter enzyme β-galactosidase. The transfection yield was monitored by determination of reporter enzyme activity in lysates from the transfected cells. At least two independent assays in triplicate were performed with each cell line. Representative plots are shown in Figure 5 . The yields (β-galactosidase activities) were normalized to the activity obtained with PEI pre-coating, which was set as a reference at 100.0%. Generally, it was observed that the transfection yields were improved in the weakly anchoring cells by the pre-coating with agents that promote cell attachment to the plate. In the case of PC-12 cells, pre-coating of wells with PEI, collagen or PDL resulted in a 2- to 3-fold enhancement of transfection yield related to the untreated wells: relative yields of 100.0% ± 1.4% for PEI, 87.0% ± 8.7% for collagen, and 100.1% ± 2.4% for PDL, versus 42.9% ± 2.2% for the untreated wells (Figure 5A ). The improvement of transfection yield by PEI pretreatment was more pronounced for HEK-293 cells. The experiment in Figure 5B shows a relative activity of 21.1% ± 3.4% for the cells in the untreated wells, when compared with 100.0% ± 8.4% for the PEI-pretreated wells (approximately 5-fold increase). Pretreatment with collagen or PDL resulted in more modest inductions (approximately 2- to 3-fold in Figure 5B ). The lowest improvement was observed with collagen, similarly to the lower enhancement of attachment of HEK-293 cells that was previously observed in Figure 3B . Finally, for the strongly attaching MYS fibroblasts, there was not a significant positive effect observed by using attachment factors in the transfection procedure. Variations were usually less than 20% for all the experimental conditions. The representative experiment shown in Figure 5C in fact shows a slightly higher transfection yield for the untreated wells than for the PEI-pretreated plates (relative activity of 117.7% ± 3.2% for the untreated wells versus 100.0% ± 4.8% for the PEI-pretreated wells, or approximately 1.2-fold higher transfection yield in the control). Compiling the results from all the experiments performed, the average fold-induction (± standard deviation) observed in the transfection yield of PC-12 cells anchored to PEI as compared with cells on untreated wells was 2.4-fold (± 0.1-fold; n = 3), while the enhancement with HEK-293 cells was 6.3-fold (± 0.5 fold; n = 2). t-test statistical analysis indicates that both increases are significant (p < 0.05). No significant transfection enhancement was observed in the experiments with MYS cells, with an average fold-change of 1.0-fold (± 0.3-fold; n = 2) by the pretreatment with PEI (basically identical to the yield without treatment). Discussion While some cell lines are able to produce extracellular matrix components in sufficient quantities to allow them to attach well to plastic culture dishes, some others have a more limited capacity, resulting in cells that are loosely attached to the dish. Natural extracellular matrix factors such as collagen, laminin, fibronectin, etc., can be utilized in the culture of these cells [ 2 ] and in the performance of multiple laboratory procedures that require the cells to remain attached to the culture vessel, for example immunocytochemical procedures. Other more economical alternatives have been developed, such as the use of polylysine or polyornithine [ 3 , 4 ]. A much less utilized polymer, PEI, offers a successful alternative. The most economical of all the mentioned factors, PEI seems to perform the task of attachment factor efficiently with primary neurons (see Figure 1 and references [ 15 , 16 ]), and with weakly anchoring cell lines (see Figures 1 , 2 and 3 , and reference [ 17 ]). The PEI solution is very easy to prepare, it can be frozen for several months without any loss of activity, and remains active on the plastic surface for numerous days after the treatment (Figure 4C ). The anchoring effect of PEI seemed to work well for a wide range of cell numbers for both PC-12 and HEK-293 cells (see Figure 4B ). PEI was efficient with high numbers of PC-12 cells probably by preventing cell-cell interactions (that result in the "clumping" of the cells), favoring instead the attachment to the dish surface. PEI worked well with HEK-293 cells in the range of 4 × 10 4 to 3.2 × 10 5 cells per well. When larger numbers of HEK-293 cells were utilized the results were more variable, suggesting that when these cells are near confluency they may be able to form monolayers that attach more firmly to the plate, however further experiments are necessary to study this hypothesis. The effects of PEI and the other cationic attachment factors on PC-12 and HEK-293 cells suggest that these cells have a limited capacity to produce an effective extracellular matrix to allow them to attach to the plastic surface, resulting in weak anchoring to the dishes on their own. Cationic polymers such as PEI and PDL seem to work very effectively to coat the surface of the dish and substitute for the absence of the extracellular matrix components. On the other hand, fibroblast cells (such as the MYS cells) are very efficient producing extracellular matrix proteins, which allows them to attach firmly to the culture dishes, therefore no benefit was observed by the use of attachment factors with these cells (see Figure 3 ). Lipofection is an efficient transfection procedure that requires culture dish manipulations and media changes. This fact causes difficulties when using cells that are not well anchored to the culture dish. The experiments presented support the idea that attachment factors can enhance the yield of lipofection of weakly anchoring cells. PEI provided very good results with PC-12 and HEK-293 cells, while the other factors tested showed more specific and limited effects (see Figure 5 ). Based on the results regarding the anchoring promotion effect of PEI and the other factors, it is likely that the increases in the yields of β-galactosidase observed are due to a larger number of cells remaining in the dishes that were treated with attachment factors. However, the transfection procedures were performed trying to minimize the cell loses, therefore it is also possible that part of the effect observed could be derived from a higher transfection efficiency of the cells that are more firmly attached versus the cells that are loosely attached or forming clusters. Cells in suspension are often harder to transfect than cells anchored to the substratum [ 9 , 10 ]. Preliminary experiments in which the transfection yield was normalized to the amount of protein in the lysate suggested that the increased yield is due primarily to the larger number of cells rather than an increase in efficiency (S. Govindaraju, K.V.L. Parsa, M. González-García and R.P. Ballestero, unpublished results). Future experiments will analyze these possibilities in more detail. In any case, the PEI pretreatment offers additional advantages, for example the firmly attached cells can be used in immunocytochemical and immunofluorescence analysis with great ease and virtually no cell loss (data not shown). Most protocols of transfection of PC-12 cells suggest the use of an attachment factor, however this is usually not the case for protocols involving the widely used HEK-293 cells [ 19 , 20 , 23 , 24 ]. Our experiments suggest that pretreatment of culture surfaces with PEI is advantageous in lipofection protocols with both cell lines. Conclusions PEI is used frequently as a transfection reagent. A second application of this reagent as an attachment factor is much less recognized. A comparison of PEI with two very commonly used attachment factors for cell and tissue culture showed that PEI is highly efficient and convenient. Two commonly used cell lines, PC-12 and HEK-293, attached firmly to plastic dishes coated with PEI. Although the anchoring properties of PEI had been previously recognized, its use with these two cell lines had not been characterized in detail in comparison with other frequently used coating agents. PEI will likely work with a variety of weakly anchoring cells, for example PEI was shown to be effective with primary retinal explants in this report. Additionally, the results presented indicate that the use of coating agents can enhance lipofection protocols. Cell culture and transfection protocols using PC-12 cells frequently involve the use of coating agents (commonly collagen or polylysine), but this is not so for protocols with HEK-293 cells. PEI was very effective in improving lipofection protocols with both cell lines. The firm attachment afforded by PEI allowed transfections with cationic lipids (lipofection) to provide higher yields and more consistent results. Methods Tissue culture and microscopy PC-12 (rat pheochromocytoma, ATCC CRL-1721) cells were cultured with RPMI complete medium: RPMI medium (Invitrogen) supplemented with 10% horse serum and 5% fetal bovine serum (both from Hyclone), 100 u/ml of penicillin, 100 μg/ml of streptomycin, 10 mM HEPES, 2 mM glutamax and 1 mM sodium pyruvate (all from Invitrogen). HEK-293 (293, human embryonic kidney, ATCC CRL-1573) and MYS (MYS-Cl-2-BCF1, mouse yolk sac, ATCC CRL-9292) cells were cultured in DMEM complete medium: DMEM medium (Invitrogen) supplemented with 10% fetal bovine serum (Hyclone), 100 u/ml of penicillin, 100 μg/ml of streptomycin and 2 mM glutamax (all from Invitrogen). Cell lines were maintained at 37°C in a 5% CO 2 atmosphere in a tissue culture incubator (NuAire). Zebrafish retinal explants were maintained in L-15 complete medium: L15 medium (Sigma) supplemented with 1% fetal bovine serum (Hyclone), 100 u/ml penicillin, 100 μg/ml of streptomycin and 2 mM glutamax (all from Invitrogen). The explants were incubated at 28°C at normal atmospheric CO 2 concentrations. For the culture and microscopic observation of PC-12 cells, the wells of a multiwell-12 tissue culture dish (MW12; Nunclon) were pretreated for 20 min with 100–200 μl of the different solutions of attachment factors: PEI-800 kDa, from Fluka, at 25 μg/ml in 150 mM NaCl; PDL, from ICN, at 100 μg/ml in deionized water; bovine dermal collagen, from ICN, 3 mg/ml aqueous solution. Usually, each column of wells in the MW12 was pretreated with one of the attachment factors, while the last column was left untreated. After removal of the different attachment factor solutions, 1 ml of a PC-12 cell suspension (approximately 0.5 × 10 6 cells) was added to each well of the MW12 and the plate was incubated at 37°C in a 5% CO 2 atmosphere for 48 h. The cells were then examined under a phase contrast Olympus-CK40 microscope and photomicrographs were obtained with a Pixera Penguin 150 CL camera. The experiment was performed in triplicate and pictures were taken at the center, near the edge and midway between center and edge of the plate, to evaluate the distribution of cells on the wells. For the microscopic observation of retinal explants, optic nerve crush (conditioning lesion) was performed retro-orbitally with anesthetized zebrafish 10 days prior to explantation, using protocols previously described [ 21 , 25 ]. Control explants were obtained from unoperated fish. The retinal pieces were laid on the wells of a MW12 that were pretreated with PEI as described above, and cultured in 200 μl of L15 complete medium at 28°C for 3 days. Photomicrographs were obtained as previously described. PC-12 differentiation with NGF The wells of a MW12 were pretreated with PEI as described above. The wells were then seeded with 2.5 × 10 5 PC-12 cells and cultured with RPMI complete medium as previously described. After 24 h, the medium was replaced by RPMI medium with low serum (same as RPMI complete medium, except for lower concentrations of sera: 1% horse serum and 0.5% fetal bovine serum), which was supplemented with NGF (Sigma) at 100 ng/ml to induce differentiation. The culture medium was replaced by new RPMI low serum medium with 100 ng/ml of NGF every two days. At different times during the differentiation process (0, 24, 48, and 96 h after the initial NGF addition), photomicrographs were obtained as previously described. Cell number measurement with neutral red dye Pretreatment of MW12 dishes with the attachment factors was performed as described above. PC-12, HEK-293 and MYS cells were counted using a Neubauer hemocytometer, and the MW12 plates were seeded with 1 × 10 6 cells per well for the experiments with PC-12 cells, 1.5 × 10 5 cells for the experiments with HEK-293 cells, or 3 × 10 4 cells for the experiments with MYS cells. The MW12 dishes were incubated at 37°C in a 5% CO 2 atmosphere for 48 h in the case of PC-12 cells, or 24 h for HEK-293 and MYS cells. In all the experiments, the wells were washed 4 times with phosphate buffered saline (PBS) to test the strength of anchorage to the substratum. To estimate the number of cells remaining after the multiple washings, a procedure that utilizes the vital dye neutral red was employed [ 26 ]. Briefly, the cells were incubated for 90 min at 37°C with 750 μl of a solution of 0.01% neutral red in PBS, and then the excess dye was removed with two washes in PBS. The dye retained by the cells was extracted with 900 μl of ethanol-citrate solution (1:1 mixture of 0.1 M citrate pH 4.2 solution and ethanol) for 20 min with gentle agitation. Relative cell counts were determined by measuring the amount of dye spectrophotometrically at 540 nm (Beckman DU500 spectrophotometer). All experiments were performed in triplicate. To compare the experiments, absorbances were normalized to the values obtained from the PEI-pretreated wells. Statistical analyses Two-sample (unpaired) t-test analyses were performed with a Microsoft Excel worksheet designed for this purpose. The worksheet calculates the t-factor by comparing 2 independent sets of data and estimates the probability (p) of obtaining that result assuming that the two samples came from the same population (null hypothesis: mean-1 = mean-2; alternate hypothesis: mean-1 < > mean-2) based on the Student's t-distribution (two-tailed). The averages of the two samples from the same experiment were considered statistically different whenever p was lower than 0.01. For comparisons of the independent experiments, one-sample t-test analyses were utilized to estimate the p values and assess the statistical significance of the difference observed between the PEI-treated wells and the untreated controls. The effect of PEI was considered significant whenever p was lower than 0.05. Determination of effective PEI dosages For the determination of optimal concentrations of PEI for the pretreatment procedure, a series of solutions of PEI were prepared so that their final concentrations of PEI polymer (w/v) were 0.025, 0.25, 2.5, 25 and 250 μg/ml. They were used to coat MW12 tissue culture dishes as previously described for the 25 μg/ml solution. To test the efficacy of coating of the solutions, 2 × 10 5 HEK-293 cells were seeded onto the dishes and incubated at 37°C for 24 h. Untreated wells were seeded in the same manner as controls. The dishes were then subjected to the multiple washing procedure described above and the number of cells remaining in the dishes was measured by the neutral red dye procedure as described previously. All experiments were performed in triplicate and the relative cell counts were calculated by normalization to the values obtained with 25 μg/ml of PEI, which was assigned a value of 100.0%. Determination of effective cell number ranges for PC-12 and HEK-293 cells The wells of MW12 culture dishes were pretreated with 25 μg/ml of PEI as described above. The wells were then seeded with various numbers of either HEK-293 cells (4 × 10 4 , 8 × 10 4 , 1.6 × 10 5 and 3.2 × 10 5 ) or PC-12 cells (2.5 × 10 5 , 5 × 10 5 , 1 × 10 6 and 1.5 × 10 6 ). Control untreated MW12 culture dishes were seeded with the same numbers of cells in parallel with the treated plates. After 24 h for the HEK-293 cells or 48 h for the PC-12 cells, the strength of anchoring of the cells to the dishes was measured as described before and relative cell numbers were determined (normalized to the raw absorbance values obtained from the PEI-pretreated wells with the highest numbers of each cell line tested, 3.2 × 10 5 for HEK-293 cells and 1.5 × 10 6 for PC-12 cells, which were assigned arbitrarily a value of 100.0%). All experiments were performed in triplicate. Stability of PEI coating experiments To analyze the stability of the PEI attachment factor onto the plastic surface of the tissue culture dish, two tests were performed. One MW12 dish was coated with 25 μg/ml of PEI as previously described, then the PEI solution was replaced by PBS and the dish was kept at 4°C for 10 days. A second MW12 dish was coated with 25 μg/ml of PEI, then the PEI solution was replaced by DMEM complete medium and the dish was kept at 37°C in the CO 2 incubator for 24 h. Afterwards, the medium was replaced by fresh DMEM complete medium and incubated for a further 24 h. This process was repeated for a third time, so that the dish was incubated for a total of 96 h with 3 medium changes. These two dishes were then seeded with 2 × 10 5 HEK-293 cells, in parallel with a third MW12 in which a column of wells was pretreated with PEI using the standard protocol described earlier (an untreated column of wells in this dish was used for the controls). The strength of attachment of cells to these dishes was determined as mentioned before, and relative cell counts were calculated by normalization to the value obtained with the standard PEI-treated wells, which was assigned arbitrarily a value of 100.0%. All experiments were performed in triplicate. Transfection yield determinations MW12 culture dishes were pretreated and seeded with cells as indicated above for the cell number measurements. The cells were subjected to lipofection with 1 μg per well of the plasmid pcDNA3-βgal [ 27 ] and LipofectAMINE-2000 (PC-12 and HEK-293 cells) or LipofectAMINE-Plus reagent (MYS cells; both reagents from Invitrogen), following the protocols recommended by the manufacturer. After incubation at 37°C in a 5% CO 2 atmosphere for 48 h for the PC-12 cells, or 24 h for the HEK-293 and MYS cells, the transfection yield was determined by measuring β-galactosidase activity in cell extracts [ 28 ]. The cells were lysed using reporter lysis buffer (Promega). The reaction mixture was made of 50 μl of lysate supernatant, 100 μl of reporter lysis buffer, and 150 μl of 2X-ONPG substrate solution (1.33 mg/ml O-nitro phenyl β-D-Galactopyranoside in 164 mM Na 2 HPO 4 , 36 mM NaH 2 PO 4 , 2 mM MgCl 2 and 100 mM β-mercaptoethanol). The reaction mixture was incubated at 37°C until development of yellow coloration was apparent (usually within a few hours). Reactions were stopped by the addition of 500 μl of 1 M Na 2 CO 3 . The relative levels of β-galactosidase activity were obtained by determination of the absorbance at 420 nm in a spectrophotometer (Beckman DU500). Parallel assays with cells transfected with pcDNA3 plasmid (Invitrogen) were conducted as negative controls in every experiment, and subtracted from the experimental absorbances to correct for endogenous activities. The experiments were conducted in triplicate. Absorbances were normalized to the values obtained with PEI-pretreated wells. Competing interests The authors declare that they have no competing interests. Authors' contributions A.R.V. performed the experiments comparing the various attachment factors. S.G. and K.L.V.P. performed the PEI dosage, cell number and PEI stability analysis experiments, as well as the NGF differentiation experiments with PC-12 cells. M.J. participated in the experiments with MYS cells. R.P.B. and M.G.G. collaborated as Principal Investigators in this project, participating in the design of the experiments and the writing of the manuscript. R.P.B. performed the experiments with zebrafish retinal explants.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526208.xml
516446	How do drug users define their progress in harm reduction programs? Qualitative research to develop user-generated outcomes	Background Harm reduction is a relatively new and controversial model for treating drug users, with little formal research on its operation and effectiveness. In order to advance the study of harm reduction programs and our understanding of how drug users define their progress, qualitative research was conducted to develop outcomes of harm reduction programming that are culturally relevant, incremental, (i.e., capable of measuring change), and hierarchical (i.e., capable of showing how clients improve over time). Methods The study used nominal group technique (NGT) to develop the outcomes (phase 1) and focus group interviews to help validate the findings (phase 2). Study participants were recruited from a large harm-reduction program in New York City and involved approximately 120 clients in 10 groups in phase 1 and 120 clients in 10 focus groups in phase 2. Results Outcomes of 10 life areas important to drug users were developed that included between 10 to 15 incremental measures per outcome. The outcomes included ways of 1) making money; 2) getting something good to eat; 3) being housed/homeless; 4) relating to families; 5) getting needed programs/benefits/services; 6) handling health problems; 7) handling negative emotions; 8) handling legal problems; 9) improving oneself; and 10) handling drug-use problems. Findings also provided insights into drug users' lives and values, as well as a window into understanding how this population envisions a better quality of life. Results challenged traditional ways of measuring drug users based solely on quantity used and frequency of use. They suggest that more appropriate measures are based on the extent to which drug users organize their lives around drug use and how much drug use is integrated into their lives and negatively impacts other aspects of their lives. Conclusions Harm reduction and other programs serving active drug users and other marginalized people should not rely on institutionalized, provider-defined solutions to problems in living faced by their clients.	Background Harm reduction programs operate with the assumption that some people who engage in high-risk behaviors are unwilling or unable to abstain. Using a "low-threshold approach," they do not require that clients abstain from drug use in order to gain access to services, nor expect adherence to one service to be eligible for another. Rather than having abstinence goals set for them, clients in such programs take part in a goal-setting process, an approach that has been shown to correlate consistently with retention and success [ 1 - 3 ]. Providers help clients make connections among their complex attitudes, behaviors, and the change they are trying to pursue as a result of an interactive process – not a dogmatic format. Behavior change is regarded as incremental and based on the premise that people are more likely to initiate and maintain behavior changes if they have the power both to shape behavioral goals and enact them. Research on harm-reduction/syringe-exchange programs has been limited largely to demonstrating their success with reducing the transmission of HIV/AIDS among drug users [ 4 - 11 ]. While this is an important accomplishment, little is known about their impact in assisting drug users in making changes in life conditions, circumstances, and quality of life. This is partially because few efforts have been made to establish appropriate measures of client and program progress in these areas. The traditional field of drug treatment has generated many assessment tools including the Addiction Severity Index used extensively for treatment planning and outcome evaluation [ 12 ]. This tool, and others like it, such as the Chemical Dependency Assessment Profile (CDAP) [ 13 ] and the assessment forms created at the Institute of Behavioral Research at Texas Christian University (TCU/DATAR), generate severity ratings that are subjective ratings of the client's need for treatment derived by the clinician. The ASI interview asks questions related to domains or "problem areas" in substance abusing patients that have been determined by clinicians, not the patients themselves. Thus, despite the formally established validity and reliability of the tool, and others like it, the measures are developed from the perspective of the clinician and researcher and are designed to generate information that is consistent with their view of the world, not the world of drug users. Given the tenets of harm reduction in which drug users participate in their own goal setting, such tools lack cultural sensitivity and relevance. Denning's work on harm reduction psychotherapy (2000) is considerably more grounded in the life circumstances of drug users. Her Multidisciplinary Assessment Profile (MAP), a baseline assessment tool to use with chemical dependency clients, however, was not designed to generate information about what drug users consider to be realistic goals and progress towards these goals. If service providers are to guide an effective interactive process of goal setting, it is important that they understand the parameters of realistic, incremental behavior change from the perspective of the client. Since the development of the ASI, there has been growing movement to acknowledge the value of participatory research in which the "subjects" of research become directly involved with shaping the research agenda and designing data collection tools. Such an approach empowers the community participating in the research so that members are not objects acted upon but rather partners in an endeavor to improve their circumstances. This approach increases the cultural appropriateness of the way the research is conducted, the potential validity and reliability of the data that are generated, and the utility of the results. In order to advance the study of harm reduction programs and our understanding of how drug users define their own progress, we conducted participatory research to develop outcomes of harm reduction programming. The goal of the research was to involve program clients in a process that would generate valid measures that are 1) culturally relevant to the way they see the world and live their lives; 2) incremental – i.e., capable of measuring small changes, and 3) hierarchical – i.e., capable of showing how clients improve over time. This article summarizes information on the research methods used and the outcomes that were generated. Methods The research study was conducted in two phases. In the first phase, drug users participated in a group process using nominal group technique (NGT) [ 15 ] to develop the outcome measures. In the second phase, other drug users participated in focus group interviews to reflect on the measures developed and their validity for the drug-using population. Below, the sample and methods of the two phases are described in more detail. Sample Study participants were recruited from the New York Harm Reduction Educators, Inc. program, which has delivered comprehensive services to over 40,000 enrollees at six sites located in East Harlem and the Bronx, New York. To involve recipients in the study, clients were recruited by program staff and given a $10 incentive for participation. The study was advertised at the main program site, and a stratified convenience sample of approximately 120 clients was recruited for phase one of the study, and 120 for phase two, with some duplicate count of clients who participated in both phases of research. The sample was stratified by neighborhood, by time in the harm-reduction program, and by whether participants took part in the syringe exchange only or accessed a fuller range of services. The demographics of the sample closely represented the larger program and included 26 % African American, 50% Latino, and 24% white; 72% male and 28% female; 17% ≤ 29 years of age; 36% between 30–39, and 46% ≥ 40 years of age. Methodology In the first phase of the research, a facilitator independent of the program used NGT with 10 groups of approximately 12 individuals per group. This group process provides a structure for small group meetings so that client participation is maximized and judgments effectively pooled. The technique was especially helpful in establishing priorities in that it neutralized differences in status and verbal dominance among participants. Before the present study, NGT was used successfully with clients in the program to identify areas of life functioning that people like themselves (i.e., drug users) deemed to be the most important and meaningful in their lives. The top 10 "life" areas were money (income); housing; food (nutrition); family relations; self-improvement; connectedness to services/benefits/programs; dealing with negative feelings (mental health); health problems (physical health); and legal and drug use problems. To generate items in a scale for every outcome listed above, 12 people were recruited for a group that used several NGT process steps. First, the facilitator asked the group to contemplate a question related to the outcome of interest. For example, "What ways do people in your circumstance make money?" was asked to generate the outcome of source of income; "What ways or places do people in your circumstance get something good to eat?" was asked to generate the outcome of source of nutrition; and "What are the types of places that people in your circumstance live?" was asked to generate the outcome of housing. Next, the members of the group brainstormed their answer(s) to the question posed to them and the facilitator recorded these answers on a large chart. The facilitator continued to call upon the members until all ideas were recorded. Next, all group members received a packet of 10 index cards numbered 1 to 10. The facilitator engaged the group in ranking the ideas according to every individual's order of preference. This step started with the facilitator asking individuals to take out the index card marked #1. She then asked every group member to select from the list that idea that he/she considered the best (i.e., the best way to make money, the best way/place to get something good to eat, the best place to live, etc.) and write it down on card #1. She then asked everyone to take out card #10 and select the idea that he/she considered the worst (i.e., the worst way to make money, the worst way/place to get something good to eat, the worst place to live, etc.) and write it down on card #10. Finally, she led them through a similar ranking process with cards #2 – #5, used by the individuals to record their "next best" preferences and cards #6 – #9 to record their "next worst" preferences. This group work resulted in 10 scaled outcomes. In the second part of the research, 10 focus group interviews were conducted to allow more of the target population to reflect on the validity of the measures and further explore the meanings of the scaled items based on the lives of drug users. In most cases, a completely different group of drug users who had participated in the NGT process for a certain outcome were selected to participate in the focus group related to that same outcome. Data Analysis The analysis of the data from each NGT group for every outcome was done by first eliminating all ideas that received no votes. The remaining data were then analyzed by: 1) determining the total score for all remaining ideas (the individual score of each idea was based on the card number on which it was recorded – i.e., 1–10 – and the total was the sum of individual scores); 2) determining the mean score for every idea (dividing the total score for an idea by the total number of votes – i.e., cards received for the idea; 3) rank-ordering the ideas by mean score; and 4) grouping the 40 or more individual ideas with similar mean scores by larger concepts so that every outcome had about 12 hierarchical, scaled items from best to worst, with mean scores from low to high. Results The results of the preliminary research are displayed in Table 1 (see Additional file: 1 ) showing every individual outcome and the hierarchical scale of items that measure incremental change from better to worse. "Better" items in every outcome, near the top of the scale, were those items that had received a mean score of 5 or better while "worse" items, near the bottom of the scale, were those items that received a mean score of 6 or higher. A summary of results for every outcome is presented below. Money (income) The outcome 'money (income)' includes a hierarchical scale of 11 items measuring better to worse 'ways of making money'. According to the clients, better ways of making money were from entitlements; a legitimate job; from family (i.e., through marriage, inheritance); or borrowing from others. Worse ways of making money were from hustling (i.e., conning or informing to police); stealing (i.e.,"boosting" – stealing); drug trade (e.g., selling, holding, transporting); panhandling; more serious criminal activity (i.e., credit card fraud, robbery, hitman); sex work; and selling blood or body parts. When other program clients reviewed this outcome scale, they felt it reflected their lives overall. One exception was that many felt that "job" (employment, peddling, odd jobs, volunteer) should be at the top of the scale rather than "entitlements." Some saw entitlements as "an easy way to make money – "it's a way of survival," – while having a job offered more independence and feelings of self-esteem. Other changes suggested by individuals were that "stealing," "drug trade" and "hustling, police informant" should be further down on the scale while "panhandling" should be further up. Strong negative feelings were expressed about being a police informant;" "Where I come from, when out on the street..., you don't inform the police! Police will lie to you and use you, then throw you back to the dogs and you're dead." Focus group participants spoke about what money meant to them and what life was like when one made money via drug trade, sex work or selling blood or body parts. Like most people, they felt having money offered a sense of freedom and independence. In addition, many felt that having it contributed to self-esteem and allowed you to help others you cared for: "It lets you help a family member who needs help"; "It helps support my spouse... and give some money to my son when I can." Regarding money making from drug trade, study participants felt, overall, that it was unstable and risky. It's like addiction. It's an adrenaline pump ... that keeps you in it for so long. You do it to keep your habit going... If you sell what you use, it's not good. You wind up doing all the drugs, then you have to run for your life from the dealer. It's an unstable life...and the consequences are great if you get caught. You go to jail, get clean, come back out, and start all over again. It's a never-ending cycle.... Eventually your luck runs out and you either go to jail or get killed. Sex work was also described in negative terms: That's the last thing you want to have to do, man or woman. Selling your body is the worst thing you could do for yourself. You can get HIV, hepatitis C, or even get killed. You don't know who this person is that you are engaging in this sex at with. He could be a serial killer. Selling blood and body parts were described as both limited and, again, risky: Now you are put on a computer and once they have a record of your blood test, you can't do it because they can screen you from sight to sight... If you sell a body part, they might not wait for you to die to get it, they might hunt you down and kill you. Housing The outcome 'housing' is made up of a hierarchical scale of 11 items measuring the better to worse places to live. Based on client input, better places to live were a house or apartment that you rented or owned; a friend's home; a drug program; a family member's home; housing provided through a social program; institutionalized housing, such as a shelter or hospital; and homelessness that was considered "least severe (e.g., sleeping on the subways or at a bus station). Clients considered worse places to live to be jail, homelessness considered to represent an intermediate level of severity (sleeping in cars, bathrooms, hallways, abandoned buildings) or the most severe level (in tunnels, caves, sewers, parks, on a roof). When other program clients reviewed this outcome scale they agreed, overall, that it reflected the reality of their lives. A few suggested minor changes in the scale could be made such as moving "jail" as a housing option to the very end of the scale because, as one said, "I don't want my freedom taken away – it's degrading and lowers your self-esteem being subject to a strip search at any time." A few others suggested putting "apt/room that you rent or own" before "friend's home." Most felt that it was better to have your own place than to live with family or friends because "it makes you feel independent, feel human – it's an accomplishment." They also offered some insight on what it was like being a drug user and living with a family member vs. a friend. A family member would let you get away with more than a friend would. With a friend, you would have to be on time, with all of your part of the rent money, food money, and clean up after yourself. With family you might dib and dab a little with the rent money or get out of doing some things around the house. However, when you're on drugs, going to your family is not good because they can give you the boot. Other "windows" of insight into how drug users deal with housing came in participants' discussion of being in a drug program, institutionalized housing (shelters, hotels, hospitals, SROs), and living on the street. Participants thought that drug programs offered an important option for housing but also saw them as a last resort: They give you structure and are stable – you can get food, clothes and confidence. Last resort when you have no place to stay and have no money... It's the place to go if you want to change your life around. Some program clients saw institutionalized housing as a crutch, while others discussed the advantages and disadvantages of various types of institutionalized housing: I became "shelterized" after being in a shelter for several years. I got locked into a routine where you don't want to take care of yourself because the basics are provided for you. Life was sweet, too sweet – I had no responsibility. Some hotels, depending on where they are located, have a high level of theft with no security. Hospitals will take you in depending on the weather and your state of mind. The homeless quarters and the psyc ward are often connected. Sometimes you have to fake it, like you need psychiatric care, if you want to get off the street for a while. If you act like you are going to harm yourself, they give you a bed fast. SROs are like apartments but they have rules... Some don't allow company but others do. You have to sign visitors in and out at the desk. Tier 1 and 2 housing is for families. But it's only temporary – 30 to 60 days. Then you have to pack up everything and move somewhere else. Section 8 housing has a lot of limitations. You have to be a family that is homeless living in a shelter, a victim of domestic violence, or in the witness-protection program. You can't have any felonies on your record. Some of section 8 is only available if you are HIV positive. Generally, clients felt that living on the street was a last resort but an option that could work: If you bum everything out by not following rules, stealing, looking for fights or taking drugs, then the street will become your home... But you can make the street work for you if you know how to survive. I did. You pick your area and take a claim for it. I had a half car that was my roof... I even evicted some people from my neighborhood because they didn't act right. We had our own rules. Food (nutrition) The outcome 'food (nutrition)' is made up a hierarchical scale of 11 items measuring the better to worse 'ways/places to get something good to eat'. According to the clients, better ways/places to get something good to eat were to cook food yourself; from friends or family; from a supermarket; from place that gave out free food (from soup kitchen, shelter, pantry, social gathering); buying food (with food stamps or from money earned); or from a restaurant. Clients considered worse ways/places of getting food were from begging or stealing; from institutions (hospital, jail), from trying to provide for yourself (hunting, fishing); and, lastly, from the garbage. After other program clients reviewed the developed scale, they generally agreed that it reflected their lives. The exceptions were that some felt that "buying food" should be placed higher in the scale, preferably at the top. The rationale was that before you could cook food yourself, you needed to buy it. In addition, others felt that "food from facilities" and "providing your own food" should be placed before stealing food in that stealing food involved risk and possible repercussions. Study participants spoke about what they considered "something good to eat." They most often mentioned that food needed to be tasteful, although not necessarily nutritious. The feelings that they associated with getting something good to eat were "feeling good," "wanting to act civil," and "wanting to treat people better." Feelings associated with not getting something good to eat were "feeling cheated" and "developing an instant attitude." They also discussed why "cooking food yourself" was higher on the scale than "going to a restaurant" or having others prepare the food for you. The importance of self-sufficiency emerged when participants spoke about the value of "cooking it the way you want it" and the feeling of comfort that came from "doing it yourself" and being able to "be at home with my girl and be able to afford a full-course meal." Participants had much to say about the topic of "free food" and the better/worse places of getting. Although "the price was right," and they were all aware of "street sheets" listing several places to go for free food, they also spoke about traveling long distances, waiting on long lines, walking up several flights of stairs, and having to have a referral and register with a program to get food. The factors that affected their decision about where to get food were the attitudes of the staff, the quality of food offered (i.e., brand names were preferred over generic, U.S.D.A. – grade foods), and whether the program also offered other needed services (e.g., some pantries offered services like showers and laundry facilities). Some participants contended that some program staff in pantries "pick through the groceries and bag up the best stuff for themselves and friends." Other revealing insights that drug users had about food were that they did not feel it was ever necessary to steal or beg for food: "There are plenty of places to get food. Anyone you see stealing food or begging is doing it for a profit, to be able to purchase something else." A number of participants referred to "dumpster divers" (people who eat street or building garbage) as people who were mentally ill and took great risk of eating contaminated food. Most felt more comfortable eating leftover, pre-wrapped food from fast-food restaurants than resorting to street dumpsters. Family relations The outcome family relations includes a hierarchical scale of 15 items measuring the better to worse 'types of family relations'. At the top of the scale clients considered better ways of relating to family to include loving your family; taking part in special family gatherings; having positive communication (open, honest, tolerant); interacting directly with members (playing games, picnics, talking about family history); arguing; showing support and respect, spending high quality time; and engaging in passive contact (movies, TV, reading the Bible). Clients felt the worse ways of relating to family were showing a lack of respect between members (being stigmatized/disrespected for who you are, talking about drugs around children); members' having negative attitudes toward one another (envy, judgment, alienation); conflicting lifestyles; engaging in abusive relations (incest, sexual abuse, violence); having difficulties with financial support; abandoning a family member; or being deceitful (stealing, gossiping, lying). Upon reflecting on the scaled outcome later, other program clients generally agreed that the scale reflected their lives, with a few exceptions. Several felt that "arguing," in the middle of the scale, should be placed further down on the scale as they saw it as a way of relating that often leads to abuse. In addition, there was some disagreement on the order of the items considered the worse types of family relations. Some clients felt that "abusive relations" should be listed last. These individuals spoke painfully about how abusive relations in their childhood had damaged them throughout life, and others spoke about how exposure of the abuse within their family had created lasting division between members. Client input on the items "love," "respect," and "negative attitudes" illuminates the meaning of these terms in the lives of drug users. "Love" was seen as a building block and foundation for family relations. It was equated with respect among family members, with one client asserting that "Love for my family may mean not spending time with them so that I do not expose them to my drug use." Clients felt that indicators of family respect were listening, letting people have their say, giving people the benefit of the doubt, and living life your way without interfering with others. Clients reflected on the negative attitudes they had experienced around family members. Along with outwardly judgmental remarks, clients also experienced a great deal of nonverbal behavior that they interpreted as negative attitudes. Examples included when they walked into the room where family members were conversing, and people suddenly stopped speaking or hid their purses. Overall, clients felt that a family member's drug use should not necessarily engender negative attitudes among other family members and that their families needed to learn more about understanding the harm reduction approach. Connection to services, programs and benefits The outcome 'connection to services, programs and benefits' consists of a hierarchical scale of 12 items measuring the better to worse 'types of services/ programs/benefits available to drug users'. Clients felt better types of services to connect with were those related to housing; HIV/AIDS assistance; mental health; drug treatment; entitlements (i.e., public assistance, SSI, and social services); and harm reduction (outreach, needle exchange, condoms). They considered the less preferred (i.e., worse) available services to be those in mainstream institutions (churches, library, legal services); "getting-connected" services (escort services, resource directories); support services (12 step, women's groups); family-prevention services (parenting skills, domestic violence services); stress reduction (acupuncture, field trips); and work (WEP) programs. Later, when other clients reviewed the developed outcome, most felt that it overall reflected their lives. The one change that a sizable number of clients called for was to put harm reduction services farther up on the scale. The value they placed on this type of service was shown in a number of comments: Harm reduction has taught us a lot about taking care of yourself physically, mentally and emotionally. If you are using drugs, it teaches you how to use drugs safely and in a safe environment. If you want to stop using, there are places to go to get the help you need. If you are out on the street hustling, selling your body it teaches you about using protection. Harm reduction is very important because it taught me a lot about how to take care of myself, manage my drug use, use my needles properly, and reduce my stress. In addition, certain individuals, based on their circumstances, made other suggested changes. One client who disclosed himself as HIV sero-positive said that "AIDS-related services" was the best service on the list for him. Another client remarked that "drug treatment" would need to be listed before housing since you are required to be drug-free to get housing. Still another felt that all the listed services were important "because they can assist me in preparing for my future." The clients discussed why "support services" (12-step programs, advocacy groups) were fairly far down on the scale. Overall, they felt this was because participation in these programs was dependent on giving up drugs, which some people are not ready to do. They also felt some people do not agree with the philosophy of the programs nor are they ready to be in a group environment. Clients also spoke at length about why the "WEP (Work Related) program" was listed last on the scale. Although they thought it might benefit people who have no skills, they felt, overall, that the program was degrading. For those who may have skills, it's kind of degrading in a way because you are working for a check that you are receiving from public assistance. For a single person you might get $68.00 every 2 weeks and you are doing the same kind of work that the people earning above minimum wage are getting. You can be working in the Parks department, cleaning people's toilets or picking up paper in the street for the sanitation dept. Some of it can be real degrading and discouraging. Self-improvement The outcome 'self-improvement' consists of a hierarchical scale of 12 items measuring the better to worse 'ways of improving yourself'. Study participants felt that better ways of improving yourself were having a better relationship with yourself (self-love, respect) and with others; getting and staying clean from drugs; being spiritual; taking part in self-help groups (12-step programs, support groups); working or developing work skills; and engaging in stress-reduction activities. They considered less preferred or "worse" ways of improving yourself to be helping others; taking care of yourself (i.e., going to dentist, taking medications, dieting, going to gym); being more responsible (i.e., living on a budget, accomplishing goals); behaving yourself (stop lying, stay out of trouble); and having a hobby (i.e., art work, boating, fishing, hunting). After a different group of program clients had reviewed this outcome, most felt that it adequately reflected their lives. A few individuals suggested that "caring for self" and "being more responsible" (i.e., items #9 and #10) should be listed further up on the scale. In addition, one individual felt that "becoming more spiritual" should be first on the scale, "because if you have God in your life, everything else will fall into place." Program clients were asked about the meaning of "self-improvement," "self-respect," "relating to others" and other items as they appeared on the scale. Concerning self-improvement, clients often thought of the topic as one that involved personal goal attainment. Setting goals that are positive and reaching them. Then setting another and reaching it, one step at a time.... Setting up a network that will help me to build a foundation of positive aspects in my life that I can follow. The clients described self-respect as requiring self-esteem, as being linked with showing respect for others and with how you physically appear to others, and as dependent on managing your drug use. If you have self-esteem and care for yourself..., respect will come. By you respecting yourself and wanting to be treated a certain way, you know you have to respect others to get it back in return. If I looked better, I would feel better about myself. A lot of time when people are drugging, they get caught up in a lot of things and before they know it, they have done some things that have cost them their self-respect, so getting it back is important to be able to get on with your life. In clients' discussion of the meaning of "relating better to others," several indicators emerged such as honest communication; holding an intelligent conversation about yourself; being comfortable relating your feeling to others; and listening. Their thoughts on "getting/staying clean" (item #3) demonstrated the challenges they face and the degree to which their lives must change to stay clean. It was a hard process for me because I would always fool myself that it wasn't the right time.... You can't do it for someone else, it has to be for you. It took me becoming homeless to decide that I had to make some changes in my life. Now that I have a new apartment, I want to keep it. My budget won't allow me to get high and keep my rent paid. Once I got out (of jail), my body was clean but my mind was still dirty. Mentally I still wanted to do drugs... I had to leave people, places and things alone because I feel powerless over the influence of others. Being around positive people and getting the support of groups helped me stay clean. Clients also provided rich detail on what they meant by "behaving myself" (item #11), including this response: It's the whole package. Your attitude, the way you talk, the language you use. When you start to change your life for the better, everything changes. You don't use a lot of 4-letter words. You want to socialize with different people in a different atmosphere. Not getting high where you work at; being more responsible. Finally, the clients were asked why "working/developing work skills" (item #6) was as far down on the scale as it was. Most acknowledged that this was a goal that many drug users are not yet ready to achieve, given their difficulty functioning in an environment that they are not used to. If you are coming into work and you are in this other world where you are not sick (to others), but you are not well either, it is hard to function. You have to have a functional mind that is able to concentrate on work...and for a lot of people, they are not there yet. Alternatively, their discussion suggested that volunteering was a better way to approach the world of work: "I started volunteering here at NYHRE, and I intend to go to computer school so I can get a better job." Mental health The outcome 'mental health' consists of a hierarchical scale of 13 items measuring the better to worse 'ways of handling negative feelings'. Study participants felt that better ways of handling negative feelings were getting informal support; (from friends, support groups), spiritual help (going to church, praying); or professional help (from a doctor, counselor); working; engaging in diversions (interacting with children; going to ball game or the beach, singing), or in stress reduction (smoking, massage, sex) and physical activities (exercise, cooking, sports). They considered worse ways of dealing with negative feelings to be engaging in violence against self (suicide, bulimia, anorexia); outward violence (hurting others, breaking things); bringing negative feelings into social relations (into marriage, when visiting someone in jail); withdrawing ("isolating"); and engaging in illicit activity (working the streets, using drugs, gambling). When another group of program clients reviewed the scaled outcome, they saw the relevance of all the items and agreed on the general order of the items in the scale. A few clients suggested some minor adjustments in the scale, however. For example, a few felt that "professional help" (item #3) should be considered the best way of handling negative feelings, rather than "get support" (from friends, support groups). For the most part, however, the majority of the clients agreed that getting support from friends and support groups was more functional for people in their circumstance than going to a professional because of issues of availability. As one person put it, "The drug man never sleeps" and people involved in this culture need easy access to those who can help them with their negative feelings. When you are out in the street drinking and drugging, there is something going on at every corner 24 hours of the day. Support from friends and groups are available to you on those off hours when "professional help" is not. This was also the rationale for few clients as to why "spiritual help" should be placed before professional help – you can pray at any time. Other individual clients felt that "social relationships" should be further up on the scale because peers and loved ones were often the most understanding. You need communication with someone that understands you and is willing to put up with your shortcomings. Problems do arise if one gets high and the other does not, but you can usually work this out. Things get tough sometimes but she helps to keep the balance in the relationship. My spouse helped me with my addiction. In regard to how drug users resorted to abuse in deal with negative feelings, the clients often referred to circumstances involving drugs. When verbal abuse did not work, they often resorted to physical abuse. A spouse will be abused when you want your drugs and you don't have the money. You know that she has the money but she won't give it to you. When selling drugs and someone comes to you with short money, even if it is only $1, he might get his butt whipped. Dealing with health problems The outcome 'dealing with health problems' consists of a hierarchical scale of 12 items measuring the better to worse 'ways of handling health problems'. The clients in the study considered better ways of handling health problems to be using home remedies (external and internal cleansing, praying); stress reduction; drug treatment/therapy; "clean living" (i.e., reduced drug use, taking meds, stopping smoking); seeing the doctor; and getting health screenings. They felt less preferred (i.e., worse) ways of dealing with health problems were maintaining a good diet, getting health education information, exercising, using alternative therapies (i.e., fasting, herbs, psychic readings, witch doctors), exhibiting negative emotions (depression, denial, suicide, anger); and using illicit drugs. After another group of program clients reviewed this outcome scale, they generally felt that it reflected their lives, with a few exceptions. Several felt that "see a doctor," "educate yourself" and "alternative therapies" should be higher on the scale. Most clients felt that "home remedies" should stay at the top of the scale because "they work the best." When they spoke about their experiences with doctors, it often was not positive. The waiting is horrible. As an inpatient, you could die before you see a doctor. Once you are identified as an addict, whether on methadone or still using drugs, you're discriminated against. Sometimes I am too leery to go to see a doctor. I may wait for someone else to go to the doctor first and then get their opinion. When asked about the health problems they encountered, client usually mentioned serious conditions (cancer, STDs, HIV, pneumonia), indicating that ailments were not a health program unless they had become serious. Regarding drug treatment, clients saw it as a positive way to deal with health problems, with certain parameters. "Drug treatment is not going to help you if you are not ready to stop using... It won't help you unless you have a follow-up plan like a support network at a church, family or groups, and being around positive people." Clients also discussed how other items on the scale were related to their drug use. Several felt that "using illegal drugs" should be at the very bottom of the scale, but opinions about this varied based on level of drug use. Clients knew that drugs could eventually bring about bad health but were often so out of touch with their feelings while doing drugs that they thought they were healthy: When I was on a constant run (doing drugs), I didn't get sick. Thought I had a wonder drug. Didn't feel anything; drugs preserved me. I didn't get headaches, toothaches or colds. If I was sick, the drugs controlled my inner body, I couldn't feel a thing. They felt that the item on the scale "educate yourself about health" was especially important for drug users who are often controlled by their substance: No one used to take vitamins because your drug controlled your mind. You couldn't eat properly because you had to get high first. Education about my health has helped me make some changes. Before I didn't go to a doctor. Now I make an effort to go on a regular basis. Dealing with drug use problems The outcome 'dealing with drug-use problems' consists of a hierarchical scale of 17 items measuring the better to worse 'ways of handling drug-use problems'. Study participants felt that better ways to handle drug-use problems were to admit the problem (and make amends with family); engage in religious activity (go to church, pray); get social support (from support groups, asking for help, making new positive friends); go into drug treatment; quit using drugs; get professional help (therapy, education about drugs, medications); stay distracted (keep busy, play with kids); and avoid the drug culture (avoid places that trigger drug use, drug paraphernalia). Clients considered the less helpful (i.e., worse) ways of handling drug-use problems were to follow a treatment plan (go to the hospital, take and not sell medications), get family support or spiritual guidance (from 12-step programs, minister); be in jail; be honest with yourself (reflect on past behaviors and pain associated with use); be deceitful (lie, manipulate others); engage in illegal activity (i.e., deal drugs, steal, prostitution); "isolate"; and continue to binge. After reviewing the developed outcome, another group of program clients generally agreed that the scale reflected their lives on the better to worse ways of handling drug-use problems. Several people commented that the items that were near the bottom of list, or the worse ways of handling drug use, were not ways of handling the problem but were, in fact, the kinds of things that went on when your drug use was out of control. They described in graphic terms what this looked like: To be focused every minute of every day on just getting the next bag of dope. My life is non-functional... I am a zombie. Wake up in the morning, get dressed and head straight for the corner to hustle up enough money to get that bag of dope. Binging is like being on a mission. You go all the way out until everything is gone.... It can be one hour, a day or longer. It is when you have used all your resources and there is no more to be had. There is no one left for you to use or manipulate. They also talked about why it was important to handle problems with drug use. One person admitted, "Your drug use is like a marriage, something you live with for life," and several clients talked about what their life looked like when they were able to handle their drug-use problems. I need to have something with structure in my life to keep going... so you can function better... go forward... handle you apartment, raise your kids, keep yourself clean... stay out of jail and live a longer life. My everyday life is my life now. The clients made several other insightful remarks about various items in this outcome. They commented on how "praying" helped them to function: "Praying helps me get things straight in my head." "It makes me strong and gives me more confidence,"; "Praying makes me more humble"; and "When I pray I feel more positive in my thinking." Regarding the item "follow a treatment plan," some people felt it was farther down on the scale because of the coercive aspect they associated with it: "Sometimes following a treatment plan is what you have to do because you have to see your parole officer every week, so you are forced to do it." In describing their experience in jail, many felt it did not help with problems with their drug use because it is very easy to get drugs in jail. The clients did feel, however, that it was something to be avoided at all costs: What you experience in jail makes you never want to go through that again. It takes your freedom away. It changed me. Now I don't even steal a Hershey bar. Dealing with legal problems The outcome 'dealing with legal problems' consists of a hierarchical scale of 11 items measuring the better to worse 'ways of handling legal problems'. The participants in the study considered better ways to handle legal problems were to pay, go see, and speak with a legal professional; address the problem yourself (go to law library, represent yourself, write to the judge); speak to a non-legal person (employer, counselor, parole officer, case manager); respect the law (by serving time, making court appearances); and learn from mistakes. The clients in the study considered worse ways of dealing with legal problems to be disrespecting the law (breaking the law, not respecting authority); facing the consequences of one's actions (serve time in prison or drug program, give up parental rights); avoiding legal responsibility (run from parole, leave the state, not show up in court, jump bail, ignore bench warrants); and relying on support from friends. When a different group of program clients reviewed the outcome they agreed overall with the order of the items in the scale. They offered rich detail on specific items on the scale and insight into how drug users experience legal problems. Drug users confront a wide variety of legal problems, including being arrested for various drug-related charges; police harassment related to petty crimes like loitering or suspicion of a crime; legal problems related to one's children and the Bureau of Child Welfare; and taking part in a hearing to qualify for SSI. Clients talked about their experience with legal professionals and items near the top of the scale. Many agreed that it was best when you could pay for an attorney, or, as one client put it, "Money talks and bullshit walks." However, they also realized that the steps in dealing with a professional first involved seeing and talking to one to find out the fee for services. Clients had varied experiences with professional attorneys, with several agreeing that legal aid lawyers were most helpful. I prefer legal aid lawyers because they work from the heart and not by what you put in their pockets. In housing I had a legal aid lawyer who helped me in a very positive way. A private, paid lawyer to help me keep my kids did not do what he was supposed to do. It was an SSI case and I had to pay and I got very little help or feedback from the lawyer at all. As with other outcomes that have been reviewed, the participants in the study spoke favorably about trying to solve the problem themselves (item #4 on the scale). It is good to do everything you can to help yourself first before you pay a legal professional or seek out their help. You might be homeless, out there on the streets...and ready to come in and get your life together.... You need to investigate how to clean up any legal problems that may be lingering. Sometimes friends may have gone through a similar experience and can tell you some of the steps they took to avoid jail or paying fines. Clients talked about how they "learned from legal mistakes" (item #7) and what "disrespect the law" (item #8) meant to them. Learning from legal mistakes often involved experiencing the consequences when the police caught up with you: I use to smoke my pipe out on the street and didn't care about the cops or anybody...When I saw the cops, I would run and hide and thought I got away from them. But when I came out of hiding, they were waiting for me and I got arrested. Clients associated several different acts with "disrespecting the law" and often spoke of "testing" the authorities by jumping the subway turnstile, going to the bathroom in the street, jay-walking, and cursing out the cops. Another interesting insight into the lives of drug users was how the clients felt that the law did not understand their ability to manage their drug use and lead a responsible life. Overall, they knew that when their drug use was out of control, they realized their respect for the law was "the farthest thing from your mind." One woman described how during a chaotic period she became suicidal and her children were taken away by the authorities. However, after she had effectively managed her drug use for some time and had sought out legal representation, even then she had not been allowed access to her children for the past six years. Discussion The methodology used in the study contributed to the field of harm reduction and how to work more effectively with drug users. Meaningful outcomes for active drug users cannot be accurately measured in "either/or" terms (i.e., drug use vs. abstinence) or reflect a yardstick of achievement that is not culturally based in the lives of program clients. In that harm reduction programs work with drug users "where they are" and strive for incremental change that can be realistically accomplished, the results of the study represent the field's first attempt to establish relevant, culturally sensitive outcomes for measuring client and program success. Rather than using measures/standards developed by researchers in a different cultural world, the generated outcomes seek to more closely represent the lives of the population we are trying to understand and serve. The ethics and benefits of this participatory research/evaluation approach have been acknowledged by many [ 16 - 19 ]. This said, it should be noted that that the major limitation of the research was that it was done with a representative sample of only one harm reduction program in an urban area. And, although the sample included a variety of program clients whose drug use was characterized on a continuum from stable to chaotic, the external validity of the measures may be questionable for use with harm reduction programs having a different client population. The methodology is one that can be used by other harm reduction programs to help them identify ranked goals in each of the life areas for their particular population. In addition to its methodological contribution, the study provided some important insights into drug users' lives and values, and an increased understanding of how this population envisions a better quality of life. Based on a number of items in the scaled outcomes, results showed that drug users often cited traditional measures in defining their life progress. Examples of this include having a legitimate job as a positive measure of source of income; being homeless and sleeping on the street as a measure of an undesirable housing situation; having an open and honest relationship with one's family as a measure of positive family relations; avoiding legal responsibility (i.e., jumping bail, running from parole) as a measure of an ineffective way of dealing with legal problems; and getting/staying clean of drugs as a positive measure of how to deal with drug use. It is generally felt that drug users' low ratings of certain activities as solutions to their problems (i.e., crime, prostitution) does not reflect a belief that these activities are inherently bad or immoral. Rather, these ratings reflect drug users' pragmatic nature and recognition that these activities can be impractical and dangerous. Other findings from the qualitative research were somewhat counterintuitive to those outside the drug-use culture and reflect the realities of poverty, racism, social isolation, past trauma, and discrimination faced by individuals in this stigmatized population. For example, the study sample considered the very best way of making money was through entitlements (i.e., welfare, disability). This finding reflects the fact that since drug users struggle for day-to-day survival and often have a criminal record, many do not have the confidence, skills or opportunities to make a legitimate living in the dominant culture. It also reflects the reality that getting needed resources via entitlements is relatively easy and dependable for drug users while getting pay from a part-time job can be problematic when an employer in the drug trade can disappear on pay day. The stark realities of the drug user world were apparent in the fact that sex work and selling blood or body parts were ways that some individuals survived. Findings related to the housing and nutrition outcomes also provided a window into the lives of many drug users. For example, being in jail was considered a more preferred housing arrangement than many forms of being homeless; eating out of the garbage – "dumpster diving" – was considered the worst way to try to get something good to eat. Interestingly, when it came to family relations, violence among family members was considered less problematic than being deceitful (stealing, gossip, lies) and may partially reflect the culture's shared value of openly expressing feelings. In regard to the outcome of connectedness to services and programs, it was not that surprising that the sample valued those services that addressed their day-to-day survival needs (entitlement benefits and programs related to housing, AIDS, and drug treatment), more than those that addressed less immediate needs (prevention, stress reduction, nutrition, and employment services). Certain consistencies across the outcomes and items in the scales shed light on other values of drug users and the barriers they face. Overall, the sample indicated that the most preferred way of living was one in which they could try to work things out for themselves and remain independent of the dominant culture. This was seen in such examples as "having your own place to live"; "cooking food yourself as a way of getting good food"; "developing a better relationship with self as a way of improving yourself"; "praying and getting support from friends in order to deal with negative feelings"; "relying on home remedies to address health problems"; "admitting you have a problem and making amends as a way of handling problems with drug use"; and "going to a law library and doing research as a way of dealing with legal problems." These findings are not surprising, considering the fact that drug users often experience stigma and discrimination when they try to rely on the traditional service delivery system and, as a result, remain isolated and marginalized. It is also a finding that challenges the dominant view of drug users as lazy, dependent, and not wanting to change. Implications for the field This preliminary research needs to be expanded in order to develop more valid and reliable outcome measures for the field of harm reduction. Other harm reduction programs are encouraged to conduct similar research with drug users in their locales and share their results. In addition, to validate that the outcomes are not simply research artifacts but reflective of the target population, field research can be conducted. For example, "hanging out" with the homeless and watching them as they spend the day looking for the best place to be homeless would help validate the outcome on housing. With measures that are science-based and evaluation results that can demonstrate the effectiveness of this approach to working with drug users, harm reduction programs will stand a better chance of receiving funding from potential donors. In addition to their use, these measures for program outcome evaluation can also benefit the clinical component of programs. Case managers could collaboratively use the methodology in sessions with individual clients to develop relevant and realistic treatment plans. Future plans for behavior change developed in one-on-one sessions and using user-generated measures of progress have more potential for achievement than plans that do not consider how clients define progress. Conclusion Clearly drug users have a set of workable solutions for meeting their own survival needs. Results of the present study show that they can often view institutions in the existing culture as irrelevant when addressing day-to-day living problems. The scales and rankings of many of the socially approved ways of solving life's problems show that just as there is an underground economy, there is a whole underground subculture where those marginalized from the mainstream have developed a culture with its own set of relevant structures, informal relationships, and home-grown recipes for addressing life's challenges. These study findings convey an understanding of drug users as people who are interested in positive change in all areas of their lives, and that by empowering them in a process to identify their own goals, they may be more motivated and engaged in the program. The findings also provide some initial empirical evidence that challenges traditional ways of measuring drug use based solely on quantity and frequency. Results suggest that more appropriate measures may be the extent to which the drug users organize their lives around drug use, the extent to which drug use is integrated into their lives, and the extent to which drug use negatively impacts other aspects of their lives. They also suggest that harm reduction and other programs serving active drug users and other marginalized people should not rely on institutionalized, provider-defined solutions to the problems in living faced by their clients. Rather, drug users should be assisted with problem solving by being encouraged to consult their own set of culturally shared solutions. Competing interests None declared. Authors' Contributions T.R. conceived the study, designed the research, and provided input on the manuscript. S.R. coordinated the study, provided oversight on data collection, analyzed the data and wrote the manuscript. Both authors read and approved the final manuscript. Supplementary Material Additional file 1 Outcomes of Harm Reduction Programming to Measure Incremental Change from Better to Worse. This is a tabular form of all categories of behavior discussed in the paper. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC516446.xml
538750	Interaction among general practitioners age and patient load in the prediction of job strain, decision latitude and perception of job demands. A Cross-sectional study	Background It is widely recognized and accepted that job strain adversely impacts the workforce. Individual responses to stressful situations can vary greatly and it has been shown that certain people are more likely to experience high levels of stress in their job than others. Studies highlighted that there can be age differences in job strain perception. Methods Cross-sectional postal survey of 300 Lithuanian general practitioners. Psychosocial stress was investigated with a questionnaire based on the Reeder scale. Job demands were investigated with the Karasek scale. The analysis included descriptive statistics; logistic regression beta coefficients to find out predictors and interactions between characteristics and predictors. Results Response rate was 66% (N = 197). Logistic regression as significant predictors for job strain assigned – duration of work in primary care; for job demands- age and duration of working in primary care; for decision latitude- age and patient load. The interactions with regard to job strain showed that GP's age and job strain are negatively associated to a low patient load. Lower decision latitude for older GP age is strongly related to higher patient load. Job demands and GP age are slightly positively related at low patient load. Conclusions Lithuanian GP's have high patient load and are at risk of stress, they have high job demands and low decision latitude. Older GP's perceive less strain, lower job demands and higher decision latitude in case of low patient load. Young GP's decision latitude has week association to patient load. Regarding to the changes in patient load younger GP's perceive it more sensitively as changes in job demands.	Background The issue of job stress is of utmost important to the public health community and working people because it adversely impacts the workforce. Strain has been considered as an environmental condition, as an appraisal of an environmental condition, as a response to an environmental condition, and as a form of relationship between environmental demands and a person's abilities to meet these demands. Although there are a lot of controversies about the epistemology of job strain, there is an agreement about it as a complex phenomenon related to health. In considering workplace-related stress, it should be recognized that stressors may occur because of individual characteristics of the worker as well as the work environment [ 1 - 5 ]. In general, physicians are at risk of stress. The main experienced pressures at work were uncertainty and insecurity, isolation, poor relationships with other doctors, disillusion with the role of the general practitioner and awareness of changing demands [ 6 , 7 ]. It has been demonstrated that negative feelings of tension, lack of time, excessive paper work among physicians take turnover to quality of care and was associated with poor clinical performance and patient's dissatisfaction [ 8 - 10 ]. The importance of job strain understanding as a problem for the general practitioners (GP's) was yielded by Appleton [ 11 ] in a study among 406 GP's. There was found that the prevalence of stress was 52%. Other studies also showed, that general practice is one of the most stressful workplaces among health care workers [ 12 - 15 ]. The specific characteristics that make general practice stressful are largely unknown. Sociodemographic factors such as age were depicted as independent predictors of vulnerability to GP's [ 16 - 21 ]. The personal and social conditions have influences on the relationship between age and stress. Continuing problems at work and job strain mostly affects young GP's [ 20 , 22 ]. On the contrary some studies showed that as a result of the age interaction, the total effects on job strain are twice larger in the sample of old persons as in the sample of young persons [ 21 ] and the age impact on job strain increases in successively in older age groups until retirement age [ 23 ]. The results of different studies showed that age also attribute to stress, anxiety, job satisfaction and quality of life for GP's [ 22 - 24 ]. It is shown that GP age and patient load have additive effects and increase vulnerability to stress [ 25 ] but still unknown how it interact with decision latitude and perception of job demands in general practice? The aim of this study was to investigate physician's age, duration of work in primary care and patient load interactions with job strain, decision latitude and perception of job demands. Methods Target group Lithuanian GP's. Study design Cross – sectional study. A mailed survey of random national samples. Computerized random sampling was performed from the registry of Lithuanian physicians. The data collected through the questionnaires filled-in by the GP's. Sample size Total number of GP's in Lithuania at the time was 1007 GP's. Sample size was calculated using EpiInfo 2000 Statcalc software which argued the sample size of 192 GP's with the 95% confidence level. From the previous studies the expected response rate was 63%. Therefore, it was decided to send questionnaires to 300 Lithuanian GP's. Our observed response rate was 66%. We collected 197 filled-in questionnaires. Assessment of Psychosocial Stress Psychosocial stress in this study was investigated by a questionnaire based on the Reeder scale [ 26 , 27 ]. The Reeder scale uses four statements experienced in everyday stressful situations as "usually tense or nervous", "daily activities are extremely trying and stressful". The respondents should indicate whether each of the statements describe them. Each question has four alternative responses, which were coded using Likert-like scale. A simple inversion of the Coulson scoring system (table 1 ) was used, giving a score of between 0 and 8 [ 28 ]. We have previously found analyses based on the Coulson approach to give very similar results to analyses based on the simple summation of scores [ 29 ]. Table 1 Coulson scoring system Score Description 0 No response on one or more statements. 1 Not at all' for all four statements. 2 'Not at all' for any three statements with any other response on the fourth. 3 'Not at all' for any two statements with 'Not very accurately' for the other two. 4 'Not at all' for any one or two statements with any other responses for the remainder but not those for a score of 3. 5 All other response sets not specified under 0, 1, 2, 3, 4, 6, 7, or 8. 6 'To some extent' to all four statements, or 'To some extent' for three statements with 'Exactly' for the fourth. 7 'Exactly' for any three statements with 'To some extent' or 'Not very accurately' for the fourth. Or 'Exactly' for two statements with 'To some extent' for two. 8 'Exactly' in response to all statements. Assessment of stressful work characteristics Work characteristics were measured by the Karasek's Job Content Questionnaire. This instrument has two scales that measure stressful job character – job decision latitude and psychological workload demands. This model, also known as the "job strain" model [ 30 - 32 ]. Psychological workload demands were defined by questions such as "working very fast," "working very hard," "doing so many things". Job decisions latitude was measured within questions as: "always must learn for new skills", "working a lot". A four point Likert – like scale was used with the coding from 4 to 1 for series, so that the responses were summarised to give a score [ 33 ]. Statistical analysis Data were computed, coded and analyzed using Statistical Package for the Social Sciences for Windows version 11.0 (SPSS Inc) and Microsoft Excel 2000. The analysis included descriptive statistics; logistic regression beta coefficients were used to assess physician's age, duration of work in primary care and patient load impact on job strain, job demands and decision latitude. Results differences at the p = 0.05 level were considered as statistically significant. Results Descriptive statistics Of the 197 respondents, 162 (82.2%) GP's were female, and 35 (17.8%) male. This is very similar to whole GP population in Lithuania. The GP ages ranged from 31 to 66 years (mean 44.2 years, 95% CI 42.9 – 45.4). GP's were investigated in 3 age groups: < 44 yr – N = 90 (45.7%); 45–54 yr – N = 85 (43.1%); 55 and > – N = 22 (11.2%). Regarding to our data in general Lithuanian GP's have high patient load and are at risk of stress, they have high job demands and low decision latitude (table 2 ). Table 2 Descriptive analysis of measured characteristics Characteristics Values Mean SD 95% CI Age 44.2 9.0 42.9–45.4 Patient load 23.8 6.7 22.8–24.7 Duration of work in primary care 17.6 10.0 16.2–19.0 Job demands 37.1 6.8 36.2–38.1 Decision latitude 23.5 6.5 22.6–24.4 Psychosocial stress 5.0 1.2 4.8–5.2 Logistic regression The logistic regression beta coefficients showed that job strain development and higher job demands could be predicted by the shorter duration of GP practice. Otherwise older age for GP's can predict lower job demands and higher decision latitude. We found that lower decision latitude can be predicted by high patient load (table 3 ). Table 3 Predicting coefficients of psychosocial stress, job demands and decision latitude Predictor Psychosocial stress Psychological workload demands Job decisions latitude Beta p-value Beta p-value Beta p-value Age 0.009 0.13 0.008 0.05 -0.008 0.01 Duration of work in primary care -0.012 0.03 -0.009 0.02 0.004 0,14 Patient load -0.003 0.40 0.003 0.21 -0.003 0.05 In figures the interactions are graphically presented according to the method described by Aiken [ 34 ] and recognized in psychological research [ 35 ]. In terms of interactions we analysed job strain, job demands and decision latitude with respect to age and patients load. Values of the predictor variables were chosen one standard deviation below and above the mean. The interactions with regard to job strain (fig. 1 ) shows that GP's age and job strain are negatively associated to a low patient load. In other words, for older GP's job strain development have stronger associations with high patient load than young GP's. Figure 1 Interaction among general practitioner age and patient load in the prediction of job strain. The age interactions with respect to psychological job demands (fig. 2 ) shows that job demands and GP age are slightly positively related at low numbers of patients per day. It shows that young GP's in terms of job demands more sensitively perceive increase in patient load that those in older age group. Figure 2 Interaction among general practitioner age and patient load in the prediction of job demands. Regarding to job decision latitude (fig. 3 ), the interaction terms shows that higher decision latitude and older general practitioner's age are strongly related to a lower patient load, which means that these variables are positively but inversely associated with patient load. Decision latitude and patient load for younger GP's has week associations. Figure 3 Interaction among general practitioner age and patient load in the prediction of decision latitude. Discussion In the current social and political climate Lithuanian GPs face many stressors that are peculiar to the medical profession. However there are many stressors that are also attributed to the personality. GPs are the professionals who are at the forefront of helping patients to manage urgent health problems, and as gatekeepers they have to make decisions on patient's health; whether to send them to hospitals. Sometimes it can interfere with personal life that can cause negative feelings about work, frustration, tension and lack of time to make appropriate decisions [ 23 ]. Our study has highlighted a matrix of issues contributing to elevated levels of job strain. These issues are rarely attributable to a simple cause and effect formula but there are complex problems with the many linkages. Lithuanian GP's has indicated twofold age interaction with job strain because it depends on patient load. Work related stress development was hardly related to duration of working in primary care. GP's perceive higher job strain and higher job demands when they have shorter duration of GP practice. Older GP's are more vulnerable to job strain, when age interaction compared among low and high patient load groups. This also means different workload and job demands. It seems to be the confirmation of Cox definition of work related stress, where the concept includes an external demand and an internal perception that the response to the demand is uncomfortable: "Work related stress is a person's recognition of his/her inability to cope with demands relating to work, and his/her subsequent experience of discomfort" [ 34 ]. We found differences in perceived job demands and in objectively measured workload units. It can be explained within growing psychological adaptation to working environment with increasing duration of GP practice. We can see the same in fig 2 . younger GP's are more vulnerable in perception of the increase in workload. Peterson's substantial review found that detrimental work environments had social and psychological consequences for all [ 35 ]. He mentioned that the extent of decision-making power, decisions latitude, as well as overwork is related to job strain development. We can say more, namely that higher patients load can be a predictor of lower decision latitude and it seems also to be related to GP age. Our results highlighted that high patient load can cause decrease in decision latitude for the older age GP's and has only week associations to younger GP's. Several weaknesses of the present study have to be mentioned. As main weakness of our study we see its cross-sectional nature, which precludes an evaluation of temporal precedence and causality of the observed associations. Karasek Job Strain model guided our hypothesis about causal relationships between age, patient load and work characteristics, explored causal relations should be interpreted carefully and longitudinal studies should be carried out in the future research. Another limitation is the Karasek's Job Content Questionnaire it self. It was designed to be broadly applicable to a wide range of occupations. However, this generalisability inevitably means that factors that are specific to particular occupations may be overlooked. For example, job demands as it has been conceptualized and operationalised in this survey would not take into account some emotional demands that could be source of stress to general practitioners such as dealing with difficult patients or caring for the dying patients [ 35 , 36 ]. Third limitation is our exclusive reliance on self-reported rating scales, which raises the issue of systematic positive or negative response tendencies. Furthermore, as no scale is perfectly reliable, the associations between self-reported measures and self-reported workload appear to be weaker than they could be in reality. Several authors have argued that this phenomenon is not a major threat if interactions has been found [ 7 , 37 ]. On the positive side, our results were obtained among a sample of people working in general practice. Respondents were with similar education level that can be seen as strength of the investigation. The sample was sufficient regarding to sample size calculation and allow exploration of tendencies. The participation rate was acceptable, and the scales we used were previously validated instruments that retained their psychometric properties in our population [ 26 ]. Otherwise it is important to mention that generalisability of Karasek's model allow to us comparisons among different medical and non medical occupational groups and this is important factor selecting job strain model. One of the principal outputs of this article is a categorization of the characteristics into a series of domains, in order to provide consistent information on the prediction of job strain, job demands and decision latitude perception. Findings from this research have hopefully emphasized the importance of examining changes and associations between work characteristics and job strain among GP's before health care reform in Lithuania will be definitely implemented. Conclusions Lithuanian GP's have high patient load and are at risk of stress, they have high job demands and low decision latitude. Job strain development and higher job demands can be influenced by shorter duration of general practice. Older GP's perceive less strain, lower job demands and higher decision latitude in case of low patient load. Young GP's decision latitude has week association to patient load. Regarding to changes in patient load younger GP's perceive it more sensitively as changes in job demands. Competing interests The author(s) declare that they have no competing interests. Authors' contributions GV designed the study, abstracted data, made data analysis, drafted and revised the manuscript. SBA participated in initial study design, participated in data analysis and revised the manuscript. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538750.xml
545963	Identifying spatially similar gene expression patterns in early stage fruit fly embryo images: binary feature versus invariant moment digital representations	Background Modern developmental biology relies heavily on the analysis of embryonic gene expression patterns. Investigators manually inspect hundreds or thousands of expression patterns to identify those that are spatially similar and to ultimately infer potential gene interactions. However, the rapid accumulation of gene expression pattern data over the last two decades, facilitated by high-throughput techniques, has produced a need for the development of efficient approaches for direct comparison of images, rather than their textual descriptions, to identify spatially similar expression patterns. Results The effectiveness of the Binary Feature Vector (BFV) and Invariant Moment Vector (IMV) based digital representations of the gene expression patterns in finding biologically meaningful patterns was compared for a small (226 images) and a large (1819 images) dataset. For each dataset, an ordered list of images, with respect to a query image, was generated to identify overlapping and similar gene expression patterns, in a manner comparable to what a developmental biologist might do. The results showed that the BFV representation consistently outperforms the IMV representation in finding biologically meaningful matches when spatial overlap of the gene expression pattern and the genes involved are considered. Furthermore, we explored the value of conducting image-content based searches in a dataset where individual expression components (or domains) of multi-domain expression patterns were also included separately. We found that this technique improves performance of both IMV and BFV based searches. Conclusions We conclude that the BFV representation consistently produces a more extensive and better list of biologically useful patterns than the IMV representation. The high quality of results obtained scales well as the search database becomes larger, which encourages efforts to build automated image query and retrieval systems for spatial gene expression patterns.	Background The complexity of animal body form arises from a single fertilized egg cell in an odyssey of gene expression and regulation that controls the multiplication and differentiation of cells [ 1 - 3 ]. For over two decades, Drosophila melanogaster (the fruit fly) has been a canonical model animal for understanding this developmental process in the laboratory. The raw data from experiments consist of photographs (two dimensional images) of the Drosophila embryo showing a particular gene expression pattern revealed by a gene-specific probe in wildtype and mutant backgrounds. Manual, visual comparison of these spatial gene expressions is usually carried out to identify overlaps in gene expression and to infer interactions [ 4 - 6 ]. Whole fruit fly embryo and other related gene expression patterns have been published in a wide variety of research journals since late 1980's. These efforts have now entered a high-throughput phase with the systematic determination of patterns of gene expression [e.g., [ 7 ]]. As a result, the amount of data currently available has doubled leading to the imminent availability of multiple expression patterns of every gene in the Drosophila genome [ 7 ]. In addition, the use of micro-array technology to study Drosophila development has revealed additional and important insights into changes in gene expression levels over time and under different conditions at a genomic scale [ 8 , 9 ]. With this rapid increase in the amount of available primary gene expression images, searchable textual descriptions of images have become available [ 7 , 10 , 11 ]. However, a direct comparison of the gene expression patterns depicted in the images is also desirable to find biologically similar expression patterns, because textual descriptions (even using a highly structured and controlled vocabulary) cannot fully capture all aspects of an expression pattern. In fact, there is a need for automated identification of images containing overlapping or similar gene expression patterns [ 6 , 12 ] in order to assist researchers in the evaluation of similarity between a given expression pattern and all other existing (comparable) patterns in the same way that the BLAST [ 13 ] technique functions for DNA and protein sequences. Of course, unlike the genomes with four letters and proteomes with 20 letters, all gene expression anatomies cannot be easily reduced to, and thus represented by, a small number of components. We previously proposed a binary coded bit stream pattern to represent gene expression pattern images [ 6 ]. In this digital representation, referred to as the Binary Feature Vector (BFV; BSV in [ 6 ]), the unstained pixels in the images (white regions and background) were denoted by a value of 0 and the stained areas (colored and foreground: gene expression) were denoted by a value of 1. Based on the BFV representations of the expression pattern, we proposed a Basic Expression Search Tool for Images (BESTi) [ 6 ] with an aim to produce biologically significant gene expression pattern matches using image content alone, without any reference to textual descriptions. We found that the BESTi approach generated biologically meaningful matches to query expression patterns [ 6 ]. In this paper, we explore how a more sophisticated Invariant Moment Vectors (IMV, [ 14 ]) based digital representation of gene expression patterns performs in generating an ordered list of best-matching images that contain similar/overlapping gene expression patterns to that depicted in a query image. IMV are frequently used in natural image processing (e.g., optical character recognition [ 15 ]) and have a number of desirable properties, including the compensation for variations of scale, translation, and rotation. If successful, IMV representations hold the promise of producing significantly shorter computing times for image-to-image matching compared to BFV. Previously, we had examined the performance of the BFV representation for a limited dataset of early stage images [ 6 ]. Here we compare the relative performances of BFV and IMV first using a dataset containing 226 images (from 13 research papers). Then we test for scalability of the BESTi search by using a seven times larger dataset containing 1819 (1593 new + 226 previous) images from 262 additional research papers (list available upon request from the authors). Both datasets contained lateral views of early stage (1–8) embryos. During these investigations, we also developed another measure of image-to-image similarity for the BFV representation. This measure is aimed at finding images that contain as much of the query image expression pattern as possible, but without penalizing for the presence of any expression outside the overlap region in the target image. In addition, we examined whether partitioning a multi-domain expression pattern into multiple BFV representations, each containing only one domain, yields a better result set. Recently, Peng and Myers [ 16 ] have proposed a different procedure involving the global and local Gaussian Mixture Model (GMM) of the pixel intensities (of expression) to identify images with similar patterns. This GMM method is expected to find images with intensity and spatial similarities. This is different from the BFV and IMV methods examined here, which are intended to find only spatially similar patterns. This focus is important because, as mentioned in [ 6 ], the differences in gene expression intensity among images in published literature can arise simply due to use of different techniques, illumination conditions, or biological reasons. However, Peng and Myers method [ 16 ] appears to be promising and we plan to examine its effectiveness in a separate paper. Results and discussion Data set generation An image database of 226 gene expression pattern images was initially generated using data from the literature [ 17 - 29 ]. All were lateral images and exhibited early stage (1–8) expression patterns. These images were selected because they had some commonality of gene expression (as seen by the human eye), which allowed us to evaluate the performance of the BESTi in finding correct as well as false matches under controlled conditions. BESTi was also tested for scalability on a larger dataset containing 1819 (1593 plus the 226) lateral views of early stage embryos. These 1593 images were obtained from 262 articles. In order to present comprehensible result sets in this paper, we have primarily discussed the findings from the dataset of 226 and provided information on how those queries scaled when they were conducted for the larger dataset. In general, our focus was to show the retrieval of biologically significant matches based on both the visual overlap of the spatial gene expression pattern and the genes associated with the pattern retrieved. Each image was standardized and the binary expression pattern extracted following the procedures described previously [ 6 ]. These extracted patterns, their invariant moments ( φ 1 through φ 7 ), and binary feature representations were stored in a database. We also calculated and stored the expression area (the count of the number of 1's in the binary feature represented image), the X and Y coordinates of the centroid ( , ), and the principal angle ( θ ) for each extracted pattern. To quantify the similarity of gene expressions in two images, we computed two measures ( S S , S C ) based on the BFV representation (See equations 2 and 3 in Methods ). S S is designed to find gene expression patterns with overall similarity to the query image, whereas S C is for finding images that contain as much of the query image expression pattern as possible without penalizing for the presence of any expression outside the overlap region in the target image. For a given pair of gene expression patterns (A and B), S S is the same irrespective of which image in the pair is the query image. That is, S S (A,B) = S S (B,A). This is not so for S C , because S C measures how much of the query gene expression pattern is contained in the image. Therefore, S C (A,B) ≠ S C (B,A). For IMV representation, we computed one dissimilarity measure ( D φ , equation 13 in Methods ). Results from D φ should be compared to that from S S , as both of these measurements do not depend on the reference image, i.e. , D φ (A,B) = D φ (B,A) and, also they capture overall similarity or dissimilarity. Matches and their biological significance The effectiveness of the BESTi in finding biologically similar expression patterns was geared towards determining the biological validity of the results obtained from the image matching procedure. All results were based solely on quantitative similarities between images without using any textual descriptions. All images were lateral views from the early stages of fruit fly embryogenesis and were oriented anterior end to the left and dorsal to the top. We refer to the images retrieved as the BESTi-matches. Performance of BFV- S S search Figure 1A shows the query image with gene expression restricted to the anterior (left) portion of the embryo, except that the expression is absent at the anterior terminus [ 22 ]. The query image depicts the expression of the sloppy paired ( slp1 ) gene in a wildtype embryo. The BESTi-matches based on the S S measure for the representations are given in Figure 1A1–A8 . BESTi retrieves images showing similar expression patterns, all of which are from same research article as the query image [ 22 ]. These images depict the expression patterns of sloppy paired genes ( slp1 and slp2 ) in a variety of genetic backgrounds or in combination with a head gap gene orthodentical ( otd ); all of these genes are essential for the pattern formation in Drosophila head development [ 30 ]. In fact, slp1 and slp2 are tightly linked genes found in the slp locus of the Drosophila genome. They are not only closely related in their primary sequence structure, but also significantly similar in their expression pattern (compare Figure 1A7 and 1A8 ). Figure 1 BESTi search results with smaller dataset. Results from the BESTi-search for the same query image [22] based on (A) BFV [ S S ], (B) IMV [ D φ ] and (C) BFV [ S C ] representations in the original dataset (226 images); and based on (D) BFV [ S S ] and (E) IMV [ D φ ] representations in the domain database (in which distinct domains of the multi-domain expression patterns were added to the original dataset as additional data points). The search argument and the results retrieved are shown on the left and right of the arrow, respectively. The original data used to generate these expression patterns are shown above this row. BESTi-matches are arranged in descending order starting with the best hit for the given search image. Values of difference in centroids (Δ C XY ) and principal angles (Δ θ ) are also given. Each image is identified by the last name of the first author of the original research article and the figure number with the following abbreviations: Ashe [19]; Casares [20]; Gaul1 [28]; Grossniklaus [22]; Hartmann [24]; Hulskamp1 [27]; Hulskamp3 [26]. A search was conducted using the same query image and same distance measure ( S S ) on the larger dataset. Figure 2 shows the top-35 matches, which contain all 8 matches shown in Figure 1A (images with blue colored legends). This allowed us to directly compare the quality of matches between the two datasets. Analysis of larger database of images yields more matches for the same S S cut-off value, as expected. A visual inspection reveals that these are all relevant images (Figure 2 ), with the larger dataset yielding more images for otd (20 images, Figure 2C ). Images with expression patterns from slp1 , slp2 and combined otd expression are found in Figure 2A,B , and 2D . More importantly, searches in the larger dataset provide images containing expression patterns of additional genes: Kruppel (Kr), hunchback ( hb ), bicoid ( bcd ), nanos , snail , hu-li tai shao ( hts ) and hairy (Figure 2E–K ). Since these images did not exist in the smaller dataset, they were not included in the search results in Figure 1A . All are biologically useful matches because combinatorial input from gap genes (Kr, hb ) along with slp1 establishes the domains of segment polarity genes in the head [ 22 ]. As for the snail , hts and hairy genes, there are no known interaction between them and slp1 (gene in the query image) in the wildtype embryo, but the images show overlap in gene expression due to the genetic backgrounds used [ 31 - 33 ]. Therefore, they are also biologically relevant matches. Figure 2 BESTi search results for S S with larger dataset. Comparison of search results from the small (226 images) and large (1819 images) dataset using the S S measure for the same query image (Figure 1A) [22]. Panels (A-K) are based on the genes whose expression patterns were retrieved as follows (A) slp1 , (B) slp1 and otd , (C) otd , (D) slp2 , (E) Kr, (F) hb , (G) hb and bcd , (H) Hb, bcd and nanos , (I) snail , (J) hts and (K) hairy . Images are referenced with the last name of the first author of the original article and its figure number: Grossniklaus [22]; Zhao [43]; Gao [44]; Wimmer [45]; Schulz1 [46]; Tsai [47]; Janody [48]; Stathopoulos [31]; Brent [32]; Zhang [33]. Common search results between the small and large dataset are indicated with dark blue image names. Performance of IMV search We used the same query image for the IMV method applied to the smaller dataset ( D φ , results in Figure 1B ) and compared the results to the BFV- S S search. In this case, we obtain images containing expressions of hb , Kr, tailless ( tll ), slp1 , hairy and infra-abdominal ( iab ) (type I transcript). It is clear that IMV search produces some biologically disconnected matches. For example, Figures 1B2, 1B4–B7 exhibit no visual overlap in gene expression pattern with the query. Furthermore, even the biologically significant matches were retrieved out of order (Figure 1B1 before 1B3 ). This happens because D φ retrieves expression patterns that are of similar shape and/or size, regardless of the translation or rotation with respect to the query image. A comparison of the results from the smaller and larger dataset for the IMV measure is given in Figure 3 . Twenty-six images were retrieved from the larger dataset when we used the same maximum distance value for the same query image. Of these, only two images were with expression pattern from slp1 (Figure 3 A1–A2 ). The expression of bcd was found in two of the results (Figures 3 B1–B2 ). 13 images containing gap gene expression patterns of Kr, hb , tll , giant ( gt ) and knirps ( kni ) (Figures 3 C1–C4, D1–D3, E1–E2, F1–F2, I1 and 3J ) were also retrieved. Images with expression patterns of hairy , achaete-scute complex (AS-C), iab (type I transcript), IAB5 enhancer, ventral nervous system defective ( vnd ), short gastrulation ( sog ) and a combined expression of bcd , nanos and cap 'n' collar ( cnc ) accounted for the remaining nine (Figures 3 G1–G2, H1–H2, K1, L1, M1, N1 and 3O1 ). We see that the new results also suffer from the same problems as before. For example, images in Figure 3 C,E,K and 3L have no common expression pattern with the query image. Hence these are not biologically significant results even though few of them (Figures 3 C1–C4, E1–E2 ) contain expression patterns of developmentally connected genes (Kr and tll with slp1 ). Figure 3 BESTi search results for D φ with larger dataset. Comparison of search results from the small (226 images) and large (1819 images) dataset using the D φ measure for the same query image (Figure 1A) [22]. Panels (A-O) are based on the genes whose expression patterns were retrieved as follows (A) slp1 , (B) bcd , (C) Kr, (D) hb (D1,D3) and Hb(D2), (E) tll , (F) gt , (G) hairy , (H) AS-C, (I) hb and Kr, (J) kni , (K) iab (type I transcript), (L) IAB5 enhancer, (M) vnd , (N) sog and (O) nanos , bcd and cnc . Images are referenced with the last name of the first author of the original article and its figure number: Grossniklaus [22]; Sauer[49]; Tsai[47]; Hulskamp1[27]; Gaul1[28]; Strunk[50]; Colas[51]; Wu[52]; Ghiglione[53]; Pankratz[54]; Melnick[55]; Janody[48]; Zhang[33]; Parkhurst[56]; Zhou[57]; Stathopoulos[31]. Common search results between the small and large datasets are indicated with dark blue image names. Since both S S and D φ measures capture the overall similarity or dissimilarity, we can use Figures 2 and 3 to compare the relative effectiveness of the BFV and IMV methods on the larger dataset. We clearly see that the BFV method performs much better in retrieving both overlapping and similar expression patterns that are also biologically significant. In addition to the Hu moments, one could also compute Zernike moments, which are based on the polar coordinate system. Both Hu moments and Zernike moments are susceptible to the same problem namely expression patterns showing a similar shape but translated to different locations in the embryo would be in the same result set. We chose to study the Hu Invariant Moment Vectors mainly because the centroid of the image can be used to distinguish between similarly shaped but translated expression patterns. With Zernike moments, the image must be inherently contained within a unit circle anchored at the centroid [ 34 ]. Thus, there is no straightforward method to eliminate the translational problem. Using the Hu moments, the spatial location problem can be corrected by considering the Euclidean difference in the centroid location expressed in pixels (Δ C XY ) of the query and results. In the case of BFV- S S search results in Figure 1 (A1–A8) , the maximum Δ C XY is less than or only slightly greater than the minimum Δ C XY for the IMV search results (Figure 1 B1–B8 ). Therefore, in the present case, the IMV-based BESTi search results need to be pared down using the centroid location difference. For example, if we consider results based on a Δ C XY lesser than or equal to 50 pixels, images shown in Figure 1 B2, B4–B7 would be removed producing a more meaningful result set. Performance of BFV- S C search Figure 1C shows the result for the same query image as used in Figure 1A , but using the newly devised S C distance for the BFV representation (BFV- S C search). This is expected to retrieve images with gene expression patterns that contain the largest amount of the overlap with the expression pattern in the query image. The top eight hits shown (Figure 1C1–C8 ) all contain over 93% of the query expression pattern: five of the matches are to the expression of hunchback ( hb ; C1, C3–C6) and the remaining three are from slp1 under different genetic backgrounds. As mentioned above, the combinatorial input from gap genes (including hb ) along with slp1 establishes the domains of segment polarity genes in the head [ 22 ]. Therefore, gene expression patterns found by BFV- S C search are for developmentally connected genes. However, using the same query image, BFV- S C search yielded only two images in common with the BFV- S S results (Figure 1 ; C7 and C8 are the same as A5 and A4, respectively). This difference occurs because S S is designed to find gene expression patterns with overall similarity to the query image (Figure 1A ), whereas S C is intended for finding images that contain as much of the query image expression pattern as possible and exclusive of the presence of the gene expression in the result image outside the region of overlap with the query image. Therefore, BFV- S S and BFV- S C have the capability of finding gene expression patterns from different biological perspectives. Using the same minimum similarity value for the BFV- S C in the larger dataset resulted in 55 images, given in Figure 4 . Gene expression patterns of slp1 and otd accounted for 8 of these images (Figure 4A and 4B ). 22 images contained expression patterns of the various gap genes hb , Kr, kni and tll (Figure 4C, 4E–F, 4I–L ) that were co-expressed with bcd and nanos (Figure 4E and 4J ) or with en (Figure 4I ). Five other genes, developmentally connected to the gene, slp1 , in the query image were also retrieved in this result set ( eve , twist , dpp ( decapentaplegic ) [ 35 ]; en ( engrailed ) [ 36 ]; arm ( armadillo ) [ 37 ]; Figure 4M–Q ). These images were not found in the top-35 of S S result set, which accentuates the different capabilities of the two BFV similarity measures in retrieving biologically relevant matches. The remaining images had expression patterns of AS-C, sc (s cute ), snail , hairy , zen ( zerknullt ), run , Hsp83, nmo ( nemo ), Tc'hb, iab , hts and sog (Figure 4D, 4G–H, 4R–Z ) which are not known to be directly related to the gene slp1 . All but seven of these images (Figures 4 D3–D4, H1–H2, R1, X1 and 4Y1 ) were from a different developmental stage than the query image. Hence, by limiting the results to those from a specific stage, extraneous matches can be removed. The seven images having the same stage as the query image were retrieved because of their significant overlap (more than 94%) with the query gene expression pattern. Thus, we observe that the new distance measure S C has the potential to identify images containing expression patterns of developmentally connected genes, other than those retrieved by S S , thus improving the overall performance of the BFV method and the BESTi tool. Figure 4 BESTi search results for S C with larger dataset. Comparison of search results from the small (226 images) and large (1819 images) dataset using the D φ measure for the same query image (Figure 1A) [22]. Panels (A-Z) are based on the genes whose expression patterns were retrieved as follows (A) slp1 , (B) otd , (C) hb , (D) AS-C, (E) nanos , bcd and Hb, (F) Kr, (G) sc , (H) snail , (I) en and hb , (J) bcd and hb , (K) kni and hb , (L) tll , (M) eve , (N) twist , (O) dpp , (P) en , (Q) arm , (R) hairy , (S) zen , (T) run , (U) Hsp83, (V) nmo , (W) Tc'hb, (X) iab , (Y) hts and (Z) sog . Images are referenced with the last name of the first author of the original article and its figure number: Grossniklaus [22]; Gao [44]; Hulskamp1 [27]; Hulskamp3 [26]; Zhao [43]; Gaul1 [28]; Tsai [47]; Niessing [58]; Sauer [49]; Parkhurst [56]; Janody [48]; Schulz2 [46]; Yagi [59] Cowden [60]; Stathopoulos [31]; Miskiewicz [61]; Schulz1 [62]; Goff [63]; Sackerson [64]; Rusch [65]; Steingrimsson [66]; Hamada [67]; Zhang [33]; Klingler [68]; Bashirullah [69]; Verheyen [70]; Wolff [71]; Casares [20]; Brent [32]. Common search results between the small and large dataset are indicated with dark blue image names. Analysis of multi-domain gene expression patterns Due to the presence of multiple areas of expression, some patterns in the database that appeared to contain much better matches (by eye and biologically) to the query image were not found or ranked very high. Hence, we also analyzed multi-domain expression patterns separately for the smaller dataset. Developmental biologists are also interested in finding such patterns as they contain overlaps with the expression domains in the query image. In fact, a large number of the expression patterns available today contain multiple isolated domains of expressions since more than one topologically distinct region of expression may be produced by many genes, transgenic constructs, probes or experimental techniques (multiple staining). In such cases, we need to consider each of these regions individually as well as in the context of the composite pattern. Biologically, it is important to consider them separately because different regions of expression may be under the control of distinct cis -regulatory sequences [e.g., [ 28 , 38 ]] or may represent the expression of different genes in a multiply-stained embryo. Separating multi-domain gene expression patterns into individual components was straightforward; we simply generated multiple images from the same initial image and included them in the target dataset. This resulted in 192 additional images (418 total) in the database all of which were components of the initial gene expression patterns. The images were separated into expression regions horizontally and/or vertically depending on the gene expression. For this new set of images, the IMV as well as BFV representations were re-calculated and the BESTi query constructed as above. Results from BFV- S S and IMV queries for this data set are given in Figures 1D and 1E , respectively. Now, many images with multiple regions of expression are retrieved in the result set (Figure 1D: D1–D8 ) and many of them show an even better match with the query pattern than those in Figure 1A for the BFV-based BESTi search. For instance, gene expression patterns are now retrieved (with more than 55% pattern similarity) from embryos with the expression of tailless ( tll ), which is known to interact with slp1 in defining the embryonic head [ 22 ], and with a composite expression of race (related to angiotensin converting enzyme) , sog ( short gastrulation ) and eve ( even-skipped ) due to enhanced race expression in the anterior domain caused by a transgenic construct causing ectopic expression of sog [ 19 ]. Therefore, the strategy of dividing multi-domain expression data into individual domains provides additional flexibility to query individual components or sub-sets of complex expression patterns. Results also improved for IMV (Figure 1E ), but again the outcome reinforced the need to use the difference in centroid to limit the result set. Next we examine the performance of S S , S C and D φ in finding BESTi matches for a query pattern with multiple regions of expression (Figure 5A ). This complex expression pattern consists of anterior and posterior domains caused by enhanced race expression resulting from dosage alteration of dpp in a gastrulation defective ( gd ) mutant background, and a middle stripe due to misexpressed sog using an eve stripe-2 enhancer [Figure 2d in [ 19 ]]. The results from this query are shown in Figure 5A1–A8 (only the original image set (226) was used as the target database in this case). We again find that S S finds many images from the same paper as well as some images from other research articles with similar expression patterns. The results correctly include expression pattern of eve (Figure 5A4 ), of another pair-rule gene ( ftz : fushi tarazu ; Figure 5A6 ), and of two other developmentally related genes [ 39 , 40 ]. Figure 5 BESTi search results with multiple domains of expression using smaller database. Results from BESTi-search for a query image with multiple domains of expression. (A) BFV [ S S ], (B) IMV [ D φ ] and (C) BFV [ S C ] searches for the same expression pattern in the original database (226 images). (D) BFV [ S S ] search using the complete multi-domain expression in the original database and (E) BFV [ S S ] search using only the pattern on the left in the domain database. Search argument and the results retrieved are shown on the left and right of the arrow, respectively. Original data used to generate these expression patterns are shown above this row. BESTi-matches are arranged in descending order starting with the best hit for the given search statistic. Values of difference in centroids (Δ C XY ) and principal angles (Δ θ ) are also given for panels A, B and C. Each image is identified by the last name of the first author of the original research article and the figure number; with the abbreviations as follows: Ashe [19]; Arnosti [17]; Borggreve [18]; Casares [20]; Gaul1 [28]; Gaul2 [29]; Grossniklaus [22]; Hartmann [24]; Hulskamp1 [27]; Hulskamp2 [25]; Hulskamp3 [26]. When D φ is used as a search criterion, it produces some correct matches in the result set (Figure 5B1–B8 ). However, it generally fails to rank biologically meaningful matches as the best matches. Use of the centroid in this case is also not productive, as most of the matches show very close centroids. The principal angle ( θ ) value calculated does not show a significant difference in the early stage embryos used in this study. The results using the S C based search are given in Figure 5C1–C8 . They show a number of images in common with the S S results. However, as expected, there are significant differences between the two searches. The results in Figures 5D and 5E demonstrate the power of the BESTi-search when the multi-domain expression data are represented in their component patterns (domain database). In this case, all the BESTi searches are based on the use of S S as the search criterion. These searches are based on the complete expression (Figure 5D ) and on one of its components (bottom-left domain, Figure 5E ). All, but one, BESTi-matches in Figure 5D contain both domains of expression. In contrast, the use of only the left, anterior, domain (Figure 5E ) in the BESTi search produces many other images in which the gene expression pattern is similar to only the anterior-ventral query pattern. Therefore, the use of individual expression components as search arguments increases the potential of directly identifying different overlapping expression patterns. Conclusions We have found that it is possible to identify biologically significant gene expression patterns from a dataset by first extracting numeric signature descriptors and then using those descriptors in a computerized search of the database for expression patterns with similar signatures or maximum pattern similarities. We find that the BFV methodologies provide a longer and more biologically meaningful set of expression pattern matches than IMV. Even though IMV representations will produce much faster retrieval speeds for large collections of embryogenesis images, the lack of biological validity of BESTi-matches retrieved makes IMV undesirable for the present problem. Instead, investigations and strategies aimed at improving the real time performance of the BFV representation will better serve the developmental biological research. Methods The wide variety of input methodologies, illumination conditions, equipment, and publication venues involved in the acquisition and presentation of gene expression patterns makes the available gene expression pattern data rather diverse. Extracting a gene expression pattern from its background requires the use of a combination of manual and automatic techniques. Each image is first standardized into a binary image as described in [ 6 ]. The standardized images are then represented using the Binary Feature Vector (BFV) [ 6 ], and the Invariant Moment Vectors (IMV) [ 14 ]. Similarity measures S S and S C are derived from BFV of which, S S is the one's complement of the distance metric D E presented in [ 6 ] and S C is a new measure introduced in this paper. The third metric D φ is deduced from the invariant moment vectors. Binary Sequence Vector analysis The binary coded bit stream pattern, in which the two possible states indicate staining over or under a threshold value, is called as Binary Feature Vector (BFV). This is referred to as the Binary Sequence Vector (BSV) in [ 6 ]. In other words, we represent each image as a sequence of 1's and 0's, where the black pixels (stained areas) are denoted by a value of 1 and the white pixels (unstained and background) are denoted by a value of 0. This BFV holds the gene expression and localization pattern information of each image. The expression patterns are ordered by evaluating a set of difference values, D E , between the binary feature vectors of every possible pair of images in the dataset. D E was introduced in [ 6 ] and is formally given as, D E = Count (A XOR B)/ Count (A OR B) (1) The term Count (A XOR B) corresponds to the number of pixels not spatially common to the two images and the term Count (A OR B) provides the normalizing factor, as it refers to the total number of stained pixels (expression area) depicted in either of the two images being compared. For simplicity, we use the one's complement of D E , as a measure of similarity of gene expression patterns between two images, S S , is given by the equation S S = (1 - D E ). (2) S S quantifies the amount of similarity based on the overlap between two expression patterns. S S is equal to 1 when the two expression patterns are identical ( D E = 0). We introduce a new similarity measure in this paper that does not penalize for any non-overlapping region. The measure S C quantifies the amount of similarity based on the containment of one expression pattern in the other given by S C = Count (A AND B)/ Count (A) (3) If the entire query image is contained within the result set images found in the database, i.e. , there is complete overlap (with respect to the query image) S C is equal to 1. Note that, S C (A,B) ≠ S C (B,A), because the denominator corresponds to the gene expression area of the query image. Invariant Moment Vector (IMV) analysis Some methodologies of image analysis produce numeric descriptors that compensate for variations of scale, translation and rotation. In the following section, we describe the invariant moment analysis of gene expression data. Invariant moment calculations have been used in optical character recognition and other applications for many years [ 15 ]. To calculate these invariant moment descriptors the standardized binary image [ 6 ] is converted to a binary representation of the same pattern (BFV). From this binary sequence of the image, the invariant moments and other descriptors are extracted using the following method [ 14 , 41 ]. The continuous scale equation used is M pq = ∬ x p y q f ( x , y ) dxdy , (4) where M pq is the two-dimensional moment of the function of the gene expression pattern, f ( x , y ). The order of the moment is defined as ( p + q ), where both p and q are positive natural numbers. When implemented in a digital or discrete form this equation becomes We then normalize for image translation using and which are the coordinates of the center of gravity, centroid, of the area showing expression. They are calculated as Discrete representations of the central moments are then defined as follows: A further normalization for variations in scale can be implemented using the formula, and is the normalization factor. From the central moments, the following values are calculated: where φ 7 is a skew invariant to distinguish mirror images. In the above, φ 1 and φ 2 are second order moments and φ 3 through φ 7 are third order moments. φ 1 (the sum of the second order moments) may be thought of as the "spread" of the gene expression pattern; whereas the square root of φ 2 (the difference of the second order moments) may be interpreted as the "slenderness" of the pattern. Moments φ 3 through φ 7 do not have any direct physical meaning, but include the spatial frequencies and ranges of the image. In order to provide a discriminator for image inversion (and rotation), sometimes called the "6", "9" problem, it has been suggested [ 14 , 42 ] that the principal angle be used to determine "which way is up". This is extremely important in embryo images because gene expression at the anterior and posterior regions may simply appear to be mirror images of each other to the invariant moments, but biologically they are completely distinct. The principal axis of the gene expression pattern f ( x , y ) is the angular displacement of the minimum rotational inertia line that passes through the centroid ( , ) and is given as: The slope of the principal axis is called the principal angle θ . It is calculated knowing that the moment of inertia of f around the line is a line through ( , ) with slope θ . We can find the θ value at which the momentum is minimum by differentiating this equation with respect to θ and setting the results equal to zero. This produces the following equation: Using the condition \| θ \| < 45° one can distinguish the "6" from the "9" and rotationally similar gene expression patterns. In invariant moment analysis, our initial method of image comparison calculates the Euclidean distance between the images using all moments ( φ 1 through φ 7 ) and combinations of these moments. For example, if the first two invariant moments are used, then and the distance D ij , between a pair of images i and j where i , j = 1, 2,...n is given by This can be expanded to use all of the moment variables. Here, the Euclidean distance, D φ , between any two images is calculated as where i and q designate images whose distance is being calculated and j designates the parameters used in the distance calculation and j = 1, 2, ..., 7. This assumes that all moments have the same dimensions or that they are dimensionless. Using this method, it is possible to rank each of the images in order of their similarity based on, for example, the first two invariant moments that have clear-cut physical meanings. Expansion to include additional moments or parameters can be performed in a number of ways. It is possible to add additional parameters to the distance calculation making sure that each of the parameters has the same dimension. For example, φ 1 has the dimension of distance squared, while φ 2 has the dimension of the fourth power of distance, thus requiring the square root function to equalize dimensions for comparable distance calculation purposes. In general, the greater number of invariant moments used in the distance calculation, the more selective the ranking. We have also allowed for the use of the centroids and principal angle as a means of list limiting. Authors' contributions SK originally conceived the project, developed the image distance measures based on the BFV representation, wrote an early version of the manuscript, and edited it until the final version. RG was responsible for writing new and using pre-existing programs to perform the image distance and parameter calculations, helped prepare the figures, searched the literature for gene expression data, maintained the database of gene expression pattern images, and helped in writing the manuscript. BVE provided the IMV method description, managed the day-to-day activities in the project, and did significant editing to produce the manuscript in the desired format for the journal. SP originally proposed the use of invariant moment vectors for biological image analysis, contributed significantly for the image distance and parameter calculations and provided critical feedback during the later stages of revision.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545963.xml
350674	Exploring Small RNA Function	null	Regulation of gene expression—deciding how much of what proteins are produced in the cell—is controlled by a myriad of different molecules. One type of naturally occurring regulatory molecule is small interfering RNA (siRNA), which selectively disrupts the production of a protein it is programmed to recognize, a process called RNA interference. These short stretches of nucleotides combine with other cellular proteins to form an RNA-induced silencing complex, called RISC, which locates and destroys a targeted messenger RNA—the molecule that carries a protein recipe from the nucleus to the site of production in the cytoplasm. While RNA interference has been widely exploited by researchers as a tool to knock out gene expression and infer the function of missing proteins, very little is known about the mechanisms behind this regulatory process. Recently, biologists have discovered hundreds of other short pieces of regulatory RNA, called microRNAs, in both plants and animals. Like siRNA, they also affect gene expression, through similar, possibly even identical RISC molecules. Animal microRNAs, however, target messenger RNA at a different stage in protein production. Though researchers have determined the sequences of these microRNAs, uncovering their function—that is, which protein they interrupt and, in turn, what the interrupted protein does—has progressed slowly and sporadically without any decisive tool to study them. Only four animal microRNAs have known biological functions, despite the intense level of work going on in this field. In this issue, György Hutvágner and colleagues report a rapid and reliable method for knocking out both siRNAs and microRNAs and thereby exploring their functions. The authors found that a short stretch of nucleotides, called a 2′- O -methyl oligonucleotide, whose sequence mirrors the targeted siRNA or microRNA, could bind and inhibit their function, allowing researchers an unprecedented glimpse at the regulatory roles and mechanisms behind RNA interference. The authors first tested their oligonucleotide design against an siRNA known to interfere with production of the firefly protein, luciferase—this luminescent protein is often used as a “reporter,” lighting up when cells successfully produce the protein. Any interference means the glow is gone. Using extracts from fruitfly embryos as the test-bed, the researchers mixed in the luciferase-associated siRNA and the sequence-specific oligonucleotide. What holds these two molecules together is complementary base-pairing, the same force that holds two molecules of DNA together. As predicted, the oligonucleotide inhibited RISC activity—it could no longer silence the production of luciferase. Because the authors could easily control the concentration of both the siRNA and the oligonucleotide inhibitor in these fly extract experiments, they were able to answer several questions about how these two molecules interact. They found that adding greater and greater concentrations of siRNA molecules did not result in equally great numbers of RISC; the process became saturated, indicating that a protein in the RISC assembly pathway limits production. Furthermore, the authors saw a marked 1:1 relationship between the concentration of the oligonucleotide and the concentration of RISC, indicating that each inhibitor binds to one RISC molecule in order to inactivate, a binding that appears to be irreversible. The results also showed that, though RISC molecules bind to the inhibitor through complementary base-pairing, a very different and more complex interaction is used by RISC molecules to find and bind their natural interference targets. The authors then went on to use the luciferase siRNA to test the function of their oligonucleotide inhibitor in cultured human cells, which had been engineered to contain the luciferase gene. This in vivo experiment, using living and metabolizing cells, showed results similar to those with fruitfly extracts. But the real test for these inhibitors was to use them in a whole animal against a previously identified microRNA where the outcome of its inactivation was already known. Hutvágner and colleagues constructed an oligonucleotide inhibitor based on the sequence of a microRNA called let-7 , which blocks the production of the protein Lin-41 and is important for proper developmental timing in roundworm larvae. Larvae injected with the oligonucleotide had the exact features of a let-7 deficient worm, showing that the inhibitor did indeed block this microRNA's function. The authors also used the oligonucleotides to provide evidence that two proteins, previously suggested to be involved with let-7 , were directly associated with its interfering activity. Using the technique described here, scientists could make rapid headway toward uncovering the biological functions of hundreds of microRNAs, their accessory RISC proteins, and even the proteins and genes they are programmed to interrupt. Furthermore, finding that RISC production is saturable could have significant implications for genetic studies that use RNA interference to uncover the function of sequenced, but unknown, genes; knowing the minimum required concentration of siRNA, researchers can avoid a buildup and any unwanted cell activity that goes along with it.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC350674.xml
555580	Daily rhythm of cerebral blood flow velocity	Background CBFV (cerebral blood flow velocity) is lower in the morning than in the afternoon and evening. Two hypotheses have been proposed to explain the time of day changes in CBFV: 1) CBFV changes are due to sleep-associated processes or 2) time of day changes in CBFV are due to an endogenous circadian rhythm independent of sleep. The aim of this study was to examine CBFV over 30 hours of sustained wakefulness to determine whether CBFV exhibits fluctuations associated with time of day. Methods Eleven subjects underwent a modified constant routine protocol. CBFV from the middle cerebral artery was monitored by chronic recording of Transcranial Doppler (TCD) ultrasonography. Other variables included core body temperature (CBT), end-tidal carbon dioxide (EtCO2), blood pressure, and heart rate. Salivary dim light melatonin onset (DLMO) served as a measure of endogenous circadian phase position. Results A non-linear multiple regression, cosine fit analysis revealed that both the CBT and CBFV rhythm fit a 24 hour rhythm (R 2 = 0.62 and R 2 = 0.68, respectively). Circadian phase position of CBT occurred at 6:05 am while CBFV occurred at 12:02 pm, revealing a six hour, or 90 degree difference between these two rhythms (t = 4.9, df = 10, p < 0.01). Once aligned, the rhythm of CBFV closely tracked the rhythm of CBT as demonstrated by the substantial correlation between these two measures (r = 0.77, p < 0.01). Conclusion In conclusion, time of day variations in CBFV have an approximately 24 hour rhythm under constant conditions, suggesting regulation by a circadian oscillator. The 90 degree-phase angle difference between the CBT and CBFV rhythms may help explain previous findings of lower CBFV values in the morning. The phase difference occurs at a time period during which cognitive performance decrements have been observed and when both cardiovascular and cerebrovascular events occur more frequently. The mechanisms underlying this phase angle difference require further exploration.	Background It has been well documented that cerebral blood flow velocity (CBFV) is lower in sleep [ 1 - 7 ] and in the morning shortly after awakening [ 8 - 10 ] than in the afternoon or evening. Generally accepted theories about the time of day changes in CBFV attribute the fall in CBFV to the physiological processes of the sleep period and the increase during the day to waking processes. The low CBFV in the morning is thought to be a consequence of the fall in the overall reduced metabolic level [ 8 , 10 , 11 ] and reduced cognitive processing [ 12 ]. Additionally, the reduced physical activity [ 13 ], reduced body temperature, and the recumbent sleeping position have also been proposed as contributors [ 14 ] to the decline in CBFV and analogous brain processes. An alternative to these explanations that attribute changes in CBFV to sleep and wake dependent processes is that this pattern of fluctuation reflects an endogenous process with circadian rhythmicity. The decline of CBFV across the sleep period and rise after subjects are awakened in the morning resemble the endogenous circadian changes in core body temperature (CBT), a reliable index of endogenous circadian rhythmicity. Both patterns are low during sleep, start to rise in the morning, reach their peak in the late afternoon, and then drop during the sleep period. The aim of this study was to examine CBFV over ~30 hours of sustained wakefulness to unmask and quantify contributions of the endogenous circadian system. By not permitting sleep, the evoked changes dependent on this change of state will not contribute to the observed CBFV changes. We hypothesized that time of day changes in CBFV are due to endogenous circadian regulation. Previous studies have been limited by several factors. First, the environmental conditions (light level) and the behavior of the subject (sleep, meals, and caffeine intake) were not controlled [ 15 , 13 , 1 , 16 ]. Second, CBFV measurements were obtained at only a few circadian points. For example, Ameriso et al. [ 15 ] and Qureshi et al. [ 16 ] assessed CBFV between 6–8 am, 1–3 pm, and 7–9 pm. Diamant et al [ 13 ] assessed CBFV during the first 15 minutes of every hour across a 24 hour period. Given these brief time periods, the findings are only a schematic of the 24 hour profile. Third, primary output markers of the endogenous circadian pacemaker (such as core body temperature and melatonin production) were not assessed. We employed the "constant routine" protocol, which was designed specifically to unmask underlying circadian rhythms in constant conditions [ 17 ]. CBFV was collected by Transcranial Doppler (TCD) ultrasonography for the entire study period. Core body temperature and salivary dim-light melatonin onset (DLMO) were measured for determination of circadian phase. Continuous electroencephalography (EEG) was performed to ensure wakefulness across the study. Additionally, measurements of blood pressure, heart rate, and end tidal carbon dioxide (Et CO2 ), three of the main regulators of CBFV, were collected every half hour. Methods Subject selection Twelve subjects (10 men and 2 women; ages 19–38, mean 28 years) agreed to participate. One subject discontinued her participation because of a headache 15 hours into the study. Subjects were in good health, as assessed by medical history, semi-structured clinical interview, and physical exam. Information regarding menstrual cycle was not obtained from female subjects. Subjects also underwent an independent standard cerebrovascular assessment and were determined to be normal. They reported no symptoms of sleep problems (such as insomnia, obstructive sleep apnea, narcolepsy, or restless legs syndrome). Subjects that were selected to participate kept to a designated sleep-wake schedule (that was negotiated from the subject's typical pattern) and filled out a sleep diary for the two weeks prior to the time in the laboratory. According to sleep diary reports, bedtimes ranged from 10:30 pm to 1:00 am and waketimes ranged from 6:00 am to 10:00 am. Alcohol and caffeine intake was discontinued for the entire week before the study. During the data collection, subjects were not permitted either alcohol or caffeine. All subjects were non-smokers. Laboratory constant routine protocol The study protocol was approved by the Institutional Review Boards of New York Presbyterian Hospital – Weill Medical College of Cornell University and The City College of New York. Subjects gave written and informed consent before participating. Subjects arrived at the sleep laboratory between 9:30 am and 10:00 am. They were oriented to the study procedures and to their bedroom. Electrodes were placed on the subject's head and face as they sat in a chair next to the bed. Data collection began at 11 am. Subjects remained in bed and awake in a semi recumbent position for 30 hours in an established "constant routine" (CR) protocol. Subjects remained in low (<25 lux) light levels which have been shown to have little or no entraining effect on the circadian pacemaker [ 18 ]. They were not allowed to get out of bed to urinate. Instead they urinated in private in a urinal or bedpan. Subjects remained awake from 11:00 a.m. on Day 1 until 5 p.m. on Day 2. Throughout the study, subjects were provided small meals (Ensure ® liquid formula plus one-quarter nutritional food bar) every 2 hours. Subject's typical total food and liquid intake for a day and a quarter were divided into 15 relatively equal portions. Only one subject participated in the CR per 30-hour period. This protocol represents a modified CR in two ways. First, subjects were allowed to watch television and were therefore were not in "time isolation." Television content was monitored so that subjects were not exposed to programs with highly emotional themes. Second, subjects needing to defecate were allowed to go to the bathroom, which was located a few steps away from the bedside. We chose this method as an alternative to using the bedpan to ensure subject's comfort and study compliance. Three subjects (subjects 05, 06, and 10) got out of bed once at 3:30, 21:30, and 15:30, respectively, to defecate. One subject, subject 12, got out of bed twice, at 22:30 and 6:35. Subject 10 used the bathroom only during the adaptation period. A paired-samples t-test was conducted to evaluate the impact of getting out of bed to defecate on subject's CBT and CBFV values. The CBT and CBFV values in the two hours before getting up were compared to the two hours after the subject got up. Subjects 5 showed a slight decrease in CBT from before (M = 98.12, SD = 0.14) to after the subject returned to the bed (M = 97.91, SD = 0.08), t(3) = -5.17, p = .014). Subject 6 showed a decline in CBFV from before (M = 56.14, SD = 2.3) to after the subject returned to the bed (M = 45.67, SD = 3.7), t(3) = 5.49, p = 0.012). There were no other significant differences detected between these two time periods for subject 5's CBFV, subject 6's CBT, or for both times subject 12 got out of the bed. By visual inspection, the overall shape of the curves in these subjects was not affected and therefore these subject's data were included in subsequent analyses. Transcranial Doppler ultrasound recordings The current study utilized TCD ultrasonography to measure cerebral blood flow velocity. TCD is a non-invasive instrument (consisting of one or two 2-Mhz transducers fitted to a headband, MARC500, Spencer Technologies, Nicolet Biomedical Inc) that is used predominantly as a diagnostic tool to assess cerebral hemodynamics in normal and pathological conditions. TCD ultrasonography is predicated on a theory that involves the measurement of moving objects when combined with radar. When the instrument emits the sound wave, it is reflected by the blood cells that are moving in the vector of the sound wave [ 19 ]. CBFV was measured using either the right or left middle cerebral artery (MCA) using TCD sonography (TCD: DWL Multidop X-2, DWL Elektronische Systeme GmbH, D-78354 Sipplingen/Germany) through the temporal window. An observer who was present continuously during the recordings evaluated the quality of the signal. This enabled long-term recording of CBFV throughout the study. Fast Fourier Transformation (FFT) of the signal was used to analyze the velocity spectra. The mean velocity of the MCA was obtained from the integral of the maximal TCD frequency shifts over one beat divided by the corresponding beat interval and expressed in cm/sec. Analysis was conducted off line. Measurement of standard markers of the circadian pacemaker Body temperature recordings Core body temperature was recorded at 1-minute intervals with an indwelling rectal probe (MiniMitter, Co. Bend, OR). A wire lead connected the sensor out of the rectum to a data collection system worn on the belt. Temperature readings were collected and saved into the device and monitored at hourly intervals by the investigator. After the study, the recordings were visually inspected and artifacts resulting from removal or malfunction of the probe were excluded from further analysis. Salivary melatonin Salivary samples of 3 ml were collected every hour from 11:00 a.m. on Day 1 to 4:00 p.m. on Day 2. Ten of these samples were used only for the determination of the timing of the salivary dim light melatonin onset (DLMO). For nine subjects, salivary DLMO was assessed across a ten-hour time window that included the ten hours before the CBT minimum. Immediately after collection, each saliva sample was frozen and stored at -20°C. Saliva samples were assayed using Bühlmann Melatonin Radio Immunoassay (RIA) test kit for direct melatonin in human saliva (American Laboratory Products Co., Windham, NH). Analysis was conducted at New York State Institute for Basic Research. Salivary DLMO time was selected based on two criteria. The saliva sample needed to have melatonin concentration 3 pg/ml or above and later samples needed to show higher levels (Bühlmann laboratories). Second, the 3 pg/ml threshold needed to occur within 6–10 hours before core body temperature minimum [ 20 ]. Polygraphic recordings Electroencephalography (EEG) was continually assessed across the 30 hours to ensure that subjects maintained wakefulness. The following montage was used according to the international 10–20 system: C3-A2, C4-A1, O1-A2, O2-A1, ROC-A1, LOC-A2, and submentalis electromyogram (EMG). One channel of electrocardiogram was continuously recorded by monitoring from two electrodes (one on each side of the body at the shoulder chest junction). The EEG software (Rembrant Sleep Collection Software Version 7.0) was used for data acquisition and display of the signals on a personal computer. Throughout the CR, the investigator (DAC) monitored the quality of the recordings. The recordings were scored by RQS and DAC. Blood pressure, heart rate, and end-tidal CO2 An automated blood pressure cuff was placed on the bicep of the subject and inflated two times each hour in order to determine changes in blood pressure and heart rate over time. Blood pressure and heart rate in one subject (02) was recorded via a finger blood pressure monitor (Omron Marshall Products, Model F-88). Blood pressure and heart rate in subjects 03, 04, 05, 06, and 07 were recorded with Omron Healthcare, Inc, Vernon Hills, Illinois 60061 Model # HEM-705CP Rating: DC 6V 4W Serial No: 2301182L. Blood pressure and heart rate for subjects 08, 09 and 10 was recorded with a similar blood pressure monitor (CVS Pharmacy Inc, Woonsocket, RI 02895 Model # 1086CVS). Blood pressure and heart rate recordings were not measured in subjects 11 and 12. Et CO2 was continuously obtained. A nasal cannula for monitoring expired gases was placed under the nose. Relative changes in carbon dioxide content were measured by an Ohmeda 4700 Oxicap (BOC healthcare). Mean Et CO2 levels were analyzed off-line. Et CO2 recordings were not measured in subjects 11 and 12. Data Analyses Data reduction and statistical procedures CBT and CBFV values were first subjected to data rejection. All CBT values less than 96 degrees were determined to be artifact and were rejected. All CBFV values less than 20 cm/sec were determined to be artifact according to the clinical criteria set by the staff neurologist. Data reduction was accomplished by averaging into one minute, 30 minute or hourly bins. Correlations presented here were performed on mean values in 30 minute bins. To ensure that circadian measurements were made under basal conditions, the first five hours of the constant routine were excluded from all analyses to eliminate effects of study adaptation. The last hour was excluded to eliminate confounding effects such as expectation effects. The data are presented in this article in three ways. First, CBT and CBFV values were plotted according to time of day (Figures 1 and 2 ). Second, CBFV values were aligned according to the CBT nadir (Figure 3 ) and third, the CBFV nadir was aligned to the CBT nadir (Figure 4 ). To align CBFV to the CBT circadian nadir as shown in Figure 3 , the CBT nadir of each individual subject was set to circadian time 0, or 0°. The CBFV value that corresponded to the CBT nadir was then also set to 0. Each half hour data point after the temperature nadir and corresponding CBFV values were then set to a circadian degree. There were a total of 48 data points across the 24 hour period. Therefore, each data point was equal to 7.5 degrees so that each data point would accumulate to 360°. Lastly, mean values were obtained for CBT and CBFV at each circadian degree. Figure 1 24-hour Cosine Curve fit to Mean Core Body Temperature (°F). Time course of CBT according to time of day. Shown is a double plot of the group (n = 11) mean levels (+/- SEM) of CBT (blue diamonds) fit with a 24-hour cosine curve (purple squares). Time of day is shown on the abscissa. The ordinate shows CBT values (degrees F). The vertical line indicates where the data was double plotted. Also displayed in the upper right corner is the non-linear cosine curve fit for mean CBT, R 2 = 0.62. The overall mean circadian phase position of the minimum was 6:05 am. Figure 2 24-hour Cosine Curve fit to Mean Cerebral Blood Flow Velocity (cm/sec). Time course of CBFV according to time of day. Shown is a double plot of the group (n = 11) mean levels (+/- SEM) of CBFV (blue diamonds) fit with a 24-hour cosine curve (purple squares). Time of day is shown on the abscissa. The ordinate shows CBFV values (cm/sec). The vertical line indicates where the data was double plotted. Also displayed in the upper right corner is the non-linear cosine curve fit for mean CBFV, R 2 = 0.67. The overall mean circadian phase position of the minimum was 12:02 pm. Figure 3 Mean CBT and CBFV Aligned to CBT Nadir. Time course of mean CBFV and mean CBT aligned to the nadir of CBT and then averaged. Shown is a double plot of the group (n = 11) mean levels (+/-SEM) of CBT (purple squares) and CBFV (blue circles) aligned to the phase of the circadian temperature cycle. Circadian time in degrees is shown on the abscissa. The ordinate on the left shows CBT values (degrees F) and CBFV (cm/sec) on the right. The vertical line indicates the CBT nadir. Figure 4 Mean CBT and CBFV Aligned to Their Respective Nadir. Time course of mean CBFV and mean CBT aligned to each of their respective nadirs and then averaged. Shown is a double plot of the group (n = 11) mean levels (+/-SEM) of CBT (purple squares) and CBFV (blue circles) aligned to the phase of the circadian temperature cycle. Circadian time in degrees is shown on the abscissa. The ordinate on the left shows CBT values (degrees F) and CBFV (cm/sec) on the right. The vertical line indicates both the CBT nadir and the CBFV nadir. The correlation coefficient between the aligned rhythms is 0.77 (p < 0.01). To align the CBFV nadir to the CBT nadir, first, the lowest value of CBT and the lowest value of CBFV were identified and set to circadian time 0, or 0°. Each half hour data point after the CBT nadir and CBFV nadir were then set to a circadian degree. There were a total of 48 data points across the 24 hour period. Therefore, each data point was equal to 7.5 degrees so that each data point would accumulate to 360°. Lastly, mean values were obtained for CBT and CBFV at each circadian degree. Estimation of circadian phase A 24-hour non-linear multiple regression -cosine curve fit analysis was performed on the CBT and CBFV data (SAS Institute, Cary, NC). This technique constrains the circadian period of CBT and CBFV to be within 24 hours. This technique used the following equations: model cbt = &avg_cbt + r * cos((2 * 3.1415) * (hours-&max_cbt)/24; model cbfv = &avg_cbt + r * cos((2 * 3.1415) * (hours-&max_cbfv)/24, where & = constants that center the curve at the actual average for each series (vertical centering) and the predicted maximum at the actual maximum (horizontal centering); r = the amplitude of the cosine wave. An additional analysis was performed which also yielded the estimated clock time for the CBT nadir and CBFV nadir (Synergy software, Kaleidagraph Version 3.6). Third, the minimum of the circadian rhythm of CBT and salivary DLMO were also used as markers of the endogenous circadian phase. A paired t-test was used to determine the overall phase difference between CBT and CBFV. Results Eleven subjects completed the protocol. The TCD probe was placed on either the right or left temple, whichever gave the better signal. Mean isonation depth of the TCD signal was 56.5 mm for the right MCA and 55.6 mm for the left MCA (range 53–60 mm). The constant routine ranged from 28 to 30 hours in duration. Polygraphic recordings confirmed sustained wakefulness across essentially the entire protocol in all but one subject. Subjects that had difficulty remaining awake were monitored closely and aroused when needed by engagement in conversation. Results from the polygraphic recordings are not presented here. We do not present the results of the polygraphic recordings because, for the purposes of this study, these recordings were used solely to monitor whether subjects were awake or asleep. The first five hours and the final hour of data from the constant routine were excluded from analysis. Core body temperature, cerebral blood flow velocity and the 24-hour day A 24 hour non-linear multiple regression, cosine fit analysis revealed that the overall mean CBT rhythm (n = 11) fit a 24 hour cosine rhythm (R 2 = 0.62, p < 0.01), Figure 1 . The mean CBT across all subjects was 98.6 °F (+/- 0.03 °F). Figure 2 shows that a 24-hour non-linear multiple regression, cosine analysis fit a 24 hour cosine rhythm (R 2 = 0.67, p < 0.01), Figure 2 . The mean CBFV across subjects was 40.6 cm/sec (+/- 0.54 cm/sec). Salivary DLMO occurred 7.7 hours prior to the CBT nadir in nine subjects, which served only as a secondary measure of endogenous circadian phase position in those subjects. The mean salivary melatonin concentration across the ten hour window was 15.3 pg/ml (+/-3.05 pg/ml). CBFV rhythm is 90 degrees out of phase with the CBT rhythm The overall mean circadian position of CBT occurred at 6:05 am and the mean position of CBFV occurred at 12:02 pm (Figure 3 ), yielding a 6 hour or 90 degree statistically significant difference (t = 4.9, DF = 10, p < 0.01). In individual subject data, the differences ranged from 0 to 8.5 hours. In eight subjects, the CBFV phase occurred later than the respective CBT phase, with mean difference of 5.2 hours. In two subjects, the CBFV nadir occurred earlier than the respective CBT nadir, with a mean difference of 6 hours. In one subject, there was no difference between the phase of CBT and CBFV. However, this subject's CBT rhythm was highly unusual, with the nadir occurring at 11:35 am on Day 2. Nevertheless, we felt the most appropriate way to present the data was to include this subject in the overall analysis. When the phase of CBFV was shifted so that the lowest value was aligned to the lowest CBT value, the two parameters were highly correlated (see Figure 4 ; r = 0.77, n = 98, p < 0.01). While the difference in the two rhythms variability was large, Fisher's z-transformed values revealed that the amplitudes of the two parameters were similar. The amplitude of CBFV yielded a z score of 4.25 and CBT yielded a z score of 3.06. Blood pressure recordings and systemic hemodynamic variables A Pearson correlation revealed a positive relationship between CBT and heart rate (r = 0.40, p < 0.01) across the 24 hour period. Diastolic blood pressure (DBP) and CBT showed a negative correlation (r = -0.30, p < 0.05). Et CO2 showed a trend towards a direct relationship with CBFV (r = 0.24, p = 0.10). Blood pressure, heart rate, and Et CO2 served only as regulators of CBFV and were not analyzed according to circadian phase. Discussion This study is the first to use the constant routine (CR) protocol to determine whether the endogenous circadian pacemaker contributes to the previously reported diurnal changes in CBFV. The current work demonstrates that, with limited periodic external stimuli and a constant posture, there is 24-hour rhythmicity in CBFV. Subjects showed a cycle of approximately 24 hours in CBT, which has been previously demonstrated with the CR [ 21 ]. Figure 3 illustrates the intricate relationship between the rhythms across the study period. At approximately the CBT acrophase, the relationship between the two rhythms undergoes a transition. Between 180 and 240 degrees, CBFV is still rising and CBT is changing directions (first rising, reaching its peak and then falling). This period between 180 and 240 has been described as a "wake maintenance zone", a time in the circadian cycle during which humans are less likely to fall asleep [ 22 ]. In our subjects, the CBT is near its zenith or just starting to fall at this time and CBFV is still steadily rising. Higher values in CBT and CBFV are associated with activation and therefore these two endogenous rhythms may be promoting wakefulness during this "wake maintenance zone". However, at the end of this transition period, CBT is falling and CBFV is still rising, perhaps reflecting continued activation of the cerebral cortex. Whereas the two-process model predicts increased tendency to sleep as CBT falls [ 23 ], our finding may provide the mechanism by which wakefulness is effortlessly maintained before bedtime. Figure 3 further illustrates that as wakefulness is extended past the subject's habitual bedtime (approximately 270 degrees), the two rhythms decline together. Between 0 and 60 degrees, CBFV steadily declines and CBT is steadily rising. The lower CBFV values in the morning may play a role in cognitive performance impairments [ 24 ], particularly the 3–4.5 hour phase difference in neurobehavioral functioning relative to the CBT rhythm that has been previously demonstrated in constant routine protocols [ 25 ]. Earlier studies using simultaneous EEG and TCD to continuously measure CBFV across the sleep period have concluded that, except for periods of REM sleep, [ 26 , 27 ], there is a linear decline in CBFV across the night during periods of non-REM sleep [ 1 , 28 ]. Other groups utilizing these techniques simultaneously speculated that the decline in CBFV through the night was a "decoupling" of cerebral electrical activity and cerebral perfusion during non-REM sleep [ 8 - 10 ]. In all studies [ 1 , 8 - 10 , 28 ], CBFV values were lower in the morning during wakefulness than during wakefulness prior to sleep at night. The current findings show that the decline in CBFV is present during wakefulness in the night time hours and therefore may not be attributed solely to sleep and associated changes that normally influence CBFV (including factors such as the shift to recumbency, and reduced activity, metabolic rate and respiratory rate). Moreover, our interaction with the subjects and the monitoring of EEG for signs of sleep resulted in no sleep in all but one subject. The one exception was in a subject who lapsed into brief periods of sleep. Therefore, the fall in CBFV in 10 out of 11 subjects cannot be explained by the occurrence of non-REM sleep. It is possible, however, that the decline of CBFV across the night and early morning may be secondary to the sleep deprivation that is part of the constant routine. Brain imaging studies across sustained periods of wakefulness have shown significant decreases in absolute regional cerebral glucose metabolic rate in several areas of the brain [ 29 - 34 ]. The drop in CBT which preceded the parallel fall in CBFV needs to be considered as a possible explanation for the CBFV changes. The fall in CBT during sleeping hours is attributed in part to sleep-associated changes and in part to strong regular circadian forces independent of the sleep period. CBT is, in fact, one of the key and most extensively studied indices of the circadian phase. It is also known that CBT is highly correlated with brain temperature and brain metabolic rate [ 35 ]. Imaging studies have documented the intimate relation between brain activity and increased metabolic rate and oxygen delivery through perfusion. Therefore, it is plausible that CBT is a direct influence on CBFV or an index of decreased metabolic need for blood flow. The prevailing hypothesis that there is tight coupling of normal neuronal activity and blood flow was formulated over 100 years ago [ 36 ]. The drop in CBFV may be a consequence of the lowered cerebral activity secondary to lowered brain temperature. In contrast, two studies of exercise-induced hyperthermia showing decreased global and middle cerebral artery CBFV [ 37 , 38 ] do not support this hypothesized direct relationship between the two variables. However, one of the main purported mechanisms for the fall in CBFV in these exercise studies, the hyperventilation induced lowering of Pa CO2 , is unlikely present during waking while lying in bed at night. Therefore, CBT declines remain a plausible explanation for the portion of the 24 hours when CBFV declined. Mechanisms of CBFV regulation This protocol allowed the unique opportunity to evaluate blood pressure, heart rate, and Et CO2 in the absence of sleep, in subjects with constant posture, and highly restricted movements. While blood pressure clearly falls during sleep in normal individuals, the absence of sleep in the current study obviates the explanation that CBFV declines are secondary to lowered blood pressure. Furthermore, we sampled blood pressure throughout the day and night and found a weak inverse relationship between DBP and CBT. This finding is in contrast to a careful study of circadian influence on blood pressure in the absence of sleep which showed no change in blood pressure during the descending portion of the body temperature curve [ 39 ]. Nevertheless, our finding was weak and likely does not provide the explanation for the CBFV changes. The small-inverse relationship between Et CO2 and CBT is similar to that found by Spengler et al. [ 40 ], who showed a consistent but small amplitude circadian rhythm in mean end-tidal Et CO2 on a CR protocol. Et CO2 showed a trend towards a direct relationship with CBFV, which is consistent with previous studies showing that changes in Et CO2 are associated with changes in CBFV [ 41 , 42 ]. Heart rate was correlated with CBT, consistent with the findings of Van Dongen et al [ 39 ]. Clinical correlation The approximate 6 hour (90 degree) phase angle difference between the CBFV and CBT suggests that CBFV continues to decline into the early to mid-morning hours. This finding is consistent with a time window in the morning during which several physiological changes have been observed. For example, cerebral vasomotor reactivity to hypocapnia, hypercapnia, and normoventilation has been found to be most reduced in the morning [ 15 , 16 ]. It is tempting to suggest that the the low CBFV values in the morning may also help explain the well established diurnal variation of the onset of cerebrovascular accidents (CVAs) [ 43 ]. A meta-analyses of 11,816 publications between 1966 to 1997 found that there was a 49% increased risk of strokes between 6 am and 12 pm [ 44 ]. This time period is in agreement with studies on myocardial infarction (MI) and sudden death [ 45 ]. The increased incidence of these events has been attributed, in part, to the surge of blood pressure [ 13 , 46 , 47 ] and platelet aggregability [ 48 , 49 ] in the morning when patients are getting out of bed. Our results demonstrate that even in the absence of surges in blood pressure, the phase of CBFV reaches its lowest values during the hours before 12 pm. This further suggests that the endogenous rhythm of CBFV may be associated with the risk of CVAs in the late morning hours even without changes in posture or activity. Conclusion Overall, the results demonstrate that CBFV, in the absence of sleep, exhibits properties of a circadian rhythm, as it rises and falls across a 24 hour period. The 6 hour (90 degree) phase angle difference in the CBFV rhythm with respect to the CBT rhythm may help explain previous findings of lower CBFV values in the morning. The phase difference occurs at a time period during which cognitive performance decrements have been observed and when both cardiovascular and cerebrovascular events occur more frequently. The mechanisms underlying this phase angle difference require further exploration. List of abbreviations CBFV Cerebral Blood Flow Velocity CBT Core Body Temperature TCD Transcranial Doppler EtCO2 End tidal Carbon Dioxide DLMO Dim Light Melatonin Onset EEG Electroencephalogram MCA Middle Cerebral Artery FFT Fast Fourier Transformation CR Constant routine EMG Electromyogram SBP Systolic Blood Pressure DBP Diastolic Blood Pressure CVA Cerebrovascular accident MI Myocardial infarction Competing interests The author(s) declare that they have no competing interests. Authors' contributions DAC coordinated, carried out, analyzed, and interpreted the study. AJS participated in the analysis and interpretation of the findings. DAC drafted the manuscript and AJS provided final approval of this version. RQS participated in data collection and data analysis. DAC and AJS co-designed the study. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555580.xml
554107	Possible role of eclosion rhythm in mediating the effects of light-dark environments on pre-adult development in Drosophila melanogaster	Background In insects, circadian clocks have been implicated in affecting life history traits such as pre-adult development time and adult lifespan. Studies on the period (per) mutants of Drosophila melanogaster , and laboratory-selected lines of Bactrocera cucurbitae suggested a close link between circadian clocks and development time. There is a possibility of clock genes having pleiotropic effects on clock period and pre-adult development time. In order to avoid such pleiotropic effects we have used wild type flies of same genotype under environments of different periodicities, which phenotypically either speeded up or slowed down the eclosion clock of D. melanogaster . Results We assayed pre-adult development time and pre-adult survivorship of four laboratory populations of D. melanogaster , under five different light regimes, continuous light (LL), continuous darkness (DD), and light-dark (LD) cycles of 10:10 h ( T20 ), 12:12 h ( T24 ), and 14:14 h ( T28 ). Although the development time was significantly different in most light regimes, except for females under T24 & T28 , pre-adult survivorship remained largely unaffected. The development time was shortest under LL, followed by T20 , DD, T24 and T28 regimes, in that order. Interestingly the development time showed a positive correlation with the period of eclosion rhythm, i.e., faster oscillations were associated with faster development, and slower oscillations with slower development. Conclusion Based on these results we conclude that periodicity of imposed LD cycles, and/or of eclosion rhythm plays a key role in regulating the duration of pre-adult development in D. melanogaster in a manner that does not involve direct pleiotropic effects of clock genes on both clock period and development time.	Background Circadian (Latin: circa = about, dies = day) clocks regulate a number of physiological and metabolic processes in organisms as diverse as unicellular bacteria, fungi, fruit flies and humans [ 1 , 2 ]. The core molecular mechanisms underlying these rhythms are conserved across a range of taxa, and involve the expression of several clock genes, interlocked in transcriptional – translational auto-regulatory feedback loops [ 2 ]. Circadian clocks have been implicated in affecting life history traits such as pre-adult development time, and adult lifespan [ 3 - 5 ]. It is generally believed that faster clocks speed up pre-adult development, and shorten adult lifespan, while slower clocks slow down development and lengthen adult lifespan. The role of circadian clocks in the development of Drosophila melanogaster has been quite elegantly addressed in an exhaustive study on the period ( per ) mutants, which display circadian rhythms with widely different periods [ 3 ]. The pre-adult development time of the different per mutants under continuous dim light (LL) and continuous darkness (DD) was positively correlated with the free-running period (τ) of their circadian clocks, i.e. per S mutants (τ = 19 h) developed faster than wild type flies (τ = 24 h), which in turn developed faster than per L mutants (τ = 28 h) [ 3 ]. The correlation between development time and clock period remained unchanged even under very bright continuous light (VLL), wherein flies are rendered arrhythmic [ 6 , 7 ]. Moreover, the development time and clock period showed positive correlation even under light-dark (LD) cycles of 12:12 h, and LD 12:12 h superimposed with temperature cycles (LD 12:12 T), wherein flies of different genotypes were entrained to a common 24 h periodicity. A positive correlation between development time and clock period under LD cycles is difficult to explain, unless one considers pleiotropic effects of clock genes that are not mediated by a direct causal relationship between clock period and development time. In a separate study on the melon fly, Bactrocera cucurbitae , which involved selection for faster and slower pre-adult development, the selection regimes yielded faster developing lines with faster circadian clocks (τ ~ 22.6 h), and slower developing lines with slower circadian clocks (τ ~ 30.9 h)[ 8 , 9 ]. A positive correlation between development time and clock period in the above studies suggests pleiotropic effects of clock genes on period of circadian rhythms and pre-adult development time. Therefore, it appears that the genotype, which enables the fly clocks to run faster or slower also aids faster and slower pre-adult development thus leaving the primary question of the role of circadian clocks in regulating pre-adult development in D. melanogaster unresolved. In order to investigate the role of circadian clocks in development time without confounding pleiotropic effects of clock genes one would need to assay development time of flies from similar genetic background under short and long day lengths, wherein their clocks would entrain by speeding up or slowing down oscillations. In the present study we assayed pre-adult development time and pre-adult survivorship of four large out-bred laboratory populations of D. melanogaster (LL1.. 4) under five different light regimes: LL, DD, LD cycles of 10:10 h ( T20 ), 12:12 h ( T24 ), and 14:14 h ( T28 ). The average periodicity of eclosion rhythm in T20 , DD, T24 , and T28 was, 20 h, 23.5 h, 24 h and 28 h, respectively, while eclosion was arrhythmic under LL [ 10 ]. The results suggest that periodicity of LD cycles and/or of eclosion rhythm play an important role in determining the duration of pre-adult development in D. melanogaster . Results ANOVA on pre-adult development time data revealed a significant main effect of light regime ( F 4,12 = 2411.97, p < 0.001) (Figures 1 , 2 ), while pre-adult survivorship remained largely unaffected ( F 4,12 = 1.06, p = 0.42). The development time of males and females was shortest under LL, followed by T20 , DD, and T24 & T28 , in that order (Figures 1 , 2 ; see table 1 ). Multiple comparisons using 95% Confidence Interval (CI) around mean showed that development time of flies under different light regimes was significantly different from each other, except for T24 and T28 regimes. ANOVA also revealed a significant main effect of sex ( F 1,3 = 607.85, p < 0.001), and light regime × sex interaction ( F 4,12 = 6.56, p < 0.05). Multiple comparisons using 95% CI showed that females developed faster than males under all five light regimes and the difference in male-female development time was greatest under T28 regime, followed by DD, T24 , LL and T20 regimes. In addition, eclosion appeared to be bimodal under T28 , and the pattern of bimodality was more prominent in females than in males (Figure 1 ). In order to compare the waveform of eclosion under five light regimes, Kruskal-Wallis test was performed, and the test revealed a significant main effect of light regime on the eclosion profile of males [ H (4, N = 2566) = 1923.295, p < 0.001] and females [ H (4, N = 2709) = 2061.568, p < 0.001]. The Kruskal-Wallis test thus confirmed the results obtained in ANOVA. Although eclosion waveforms under no two light regimes were similar, the dissimilarity was even more striking under T24 and T28 . The test revealed that mean development time of males was shortest under LL, followed by T20 , DD, T24 , and T28 , in that order. While the mean development time of females followed a similar trend, it did not differ significantly between T24 and T28 . Figure 1 Eclosion profile of fruit flies D. melanogaster in five different light regimes. Pre-adult development time in hours is plotted along x-axis, and the percentage of eclosing flies is plotted along y-axis. Eclosion profile of males in five different light regimes is shown in the left panels, while that of females is illustrated in the right panels. Graphs were plotted using data pooled over four populations. Figure 2 Mean pre-adult development time of D. melanogaster populations in five different light regimes. Light regime is plotted along x-axis, and pre-adult developmental time in hours along y-axis. Mean development time of males is shown in the left panels, while those for females in the right panels. The error bars represent 95% CI around the mean. ANOVA revealed significant effect of light regime ( F 4,12 = 2411.97, p < 0.001), sex ( F 1,3 = 607.85, p < 0.001), and light regime × sex interaction ( F 4,12 = 6.56, p < 0.05). The mean pre-adult development time values are provided in table1. Table 1 Mean pre-adult development time of four laboratory populations of D. melanogaster assayed under five different light regimes. Light regime Sex Pre-adult development time (hours) LL M 224.37 LL F 221.91 DD M 258.32 DD F 253.63 T20 M 248.63 T20 F 246.72 T24 M 263.96 T24 F 261.07 T28 M 265.32 T28 F 258.83 The periodicity of eclosion rhythm under DD, T20 , T24 and T28 , as reported in one of our previous studies on the same populations were 23.5 h, 20 h, 24 h and 28 h, respectively, while eclosion was arrhythmic under LL [ 10 ]. In addition, the peak of eclosion rhythm under different light regimes [ 10 ] matched closely the peak eclosion in the development time assay (Figure 1 ). The mean development time of males and females under four light regimes (DD, T20 , T24 and T28 ) showed a significant positive correlation with the mean period of eclosion rhythm under the corresponding environments ( r = +0.83 & +0.71, p < 0.001; Figure 3 ). Figure 3 Mean development time and the period of the eclosion rhythm show a significant positive correlation. The mean development time in hours under T20 , DD, T24 and T28 is plotted along y-axis, and the period of eclosion rhythm under the corresponding light regime is plotted along x-axis. The correlation between mean development time and period of eclosion rhythm for males is shown in left panel while that for females is illustrated in the right panel. Discussion In several insect species adult eclosion is gated in a manner that it occurs only during a narrow window of time, generally around dawn when environmental humidity is the highest [ 11 ]. In D. melanogaster the clocks that gate adult eclosion are located in the prothoracic gland and ventral lateral neurons [ 12 ], and it is believed that these clocks also play a key role in the regulation of pre-adult development [ 13 ]. Development time under environments wherein eclosion is arrhythmic, such as bright LL, is solely determined by the developmental state of a fly, and under such a situation pre-adult development time would reflect the minimum time required by flies to complete pre-adult development. On the other hand, environments such as DD and LD cycles, wherein eclosion is rhythmic, the interaction between developmental state and eclosion clock would determine the duration of pre-adult development, and the developmental time would then be expected to be greater than those under LL. In a previous study on four Drosophila populations maintained under LL (JB1..4), the ancestral populations of the flies used in the present study, we had reported shortest development time under LL regime, followed by LD 12:12 h, and DD [ 14 ]. In the present study too development time was shortest under LL, followed by T20 , DD, and T24 and T28 , in that order. As opposed to LL, eclosion under DD is gated in a circadian manner, and as a result flies took longer to develop compared to that in LL. We believe that the slight discrepancy in the results of our two studies could be due to the fact that different set of flies (about 100 generations apart) with different clock periods were used in the two experiments. The periodicity of JB populations under DD was greater than 24 h, whereas those of LL populations was lesser than 24 h. Thus, consistent with our proposal, development time of JB populations was greater under DD compared to development time in LD 12:12 h regime, whereas those of LL populations was shorter in DD compared to those under LD 12:12 h. The mean development time under four different light regimes ( T20 , DD, T24 , and T28 ) showed a significant positive correlation with the mean period of eclosion rhythm under the corresponding light regimes; i.e. shorter period of eclosion rhythm under T20 was associated with faster pre-adult development, followed closely by DD, whereas flies took longest to develop under the light regimes wherein their eclosion periodicities were 24 h and 28 h, suggesting that development in D. melanogaster is regulated by eclosion rhythm. The peak of eclosion under three periodic light regimes closely matched phase relationships of eclosion rhythm relative to LD cycles, suggesting that phases of eclosion rhythm also play a key role in the regulation of development time [ 10 ]. Finally, flies took different number of environmental cycles to develop under three periodic light regimes. The average number of cycles taken to develop under LD 10:10 h, 12:12 h, and 14: 14 h were approximately 12.5, 11 and 9.5, respectively, which suggests that pre-adult development of D. melanogaster is not entirely regulated by the periodicity of the environment and/or the periodicity of eclosion rhythm. These results are in agreement with the findings of previous studies, where the duration of pre-adult development was positively correlated with clock period [ 3 , 9 , 15 ]. A subtle but important difference between the outcome of present study and that of most of previous studies is that, the correlation between developmental time and eclosion period in our study is clearly mediated via the periodicity of LD cycles and/or of eclosion rhythm, whereas in previous studies the correlation was independent of external environment and the eclosion rhythm, instead was dependent upon the genotype of the flies [ 3 ]. Although development time in the present study was greater under T28 compared to T24 , the differences did not reach levels of statistical significance in ANOVA, possibly due to complex interactions between the developmental states, phase of the LD cycles, and eclosion profiles under T24 and T28 regimes. A careful analysis of development time of flies under these two regimes revealed that eclosion under T24 and T28 was bimodal, and bimodality was more prominent under T28 than in T24 . To complicate the matter further the eclosion patterns of females had a greater propensity towards bimodality compared to males. The Kruskal-Wallis test revealed that eclosion profiles of flies under T24 and T28 were indeed significantly different, and mean development time of males under T28 was greater than in T24 regime, whereas those of females did not differ between the two light regimes. According to the gating hypothesis ([ 13 ]) bimodality could arise when flies are exposed to LD cycles of non-24 periodicity, perhaps due to a mismatch between developmental time and eclosion gate. Indeed, the same phenomenon also occurs under T20 , where a small, statistically insignificant eclosion peak appears between 260 h and 270 h in both males and females (fig 1 ). Conclusion Pre-adult development time and circadian rhythm are both multigenic traits, and genes involved in regulating development time as well as circadian rhythms are known to have pleiotropic effects [ 3 ]. Our study pre-designed to bypass such pleiotropic effects demonstrates a possible role of the periodicity of light-dark environment and/or of eclosion rhythm in determining the duration of pre-adult development. Taking into account the results of our experiments and those of the per mutant experiments, it appears that the duration of pre-adult development in D. melanogaster is determined by several factors such as the circadian rhythm, developmental state, and the interaction between the phase of eclosion rhythm, and the phase of the LD cycles. Methods Fly stock maintenance The four populations of D. melanogaster used in this study were maintained under constant light (~ 100 lux), at constant temperature of 25°C (± 1°C), and constant humidity of 70 ± 5%, on a 21 day discrete generation cycle (henceforth will be referred as LL1, 2, 3, 4 populations). These populations were maintained at moderate larval densities of ~ 60–80 larvae per 8 dram vial (9.0 mm height × 2.4 mm diameter) containing banana-jaggery food medium (henceforth banana food). The ancestry and maintenance of these populations has been described in detail in an earlier paper [ 16 ]. In brief, at every generation, adults of each population are allowed to lay eggs for about 18 hours on petri plates of fresh banana food placed in a plexiglass cage (25 × 20 × 15 cm 3 ). From these petri plates, 60–80 eggs are collected into each of 40 vials in which larvae then develop into adults. Adults eclosing from these vials are transferred to plexiglass cages on 12 th day after egg lay. On the 18th day after egg lay, adult flies are supplied with banana food supplemented with live yeast paste for two days, after which eggs are collected to initiate the next generation and the adults discarded. The breeding population typically consists of about 1500 flies. Pre-adult development time and survivorship assay From the running culture of each population (LL1..4), eggs laid on banana medium over a 2 h window were collected for the assay. Exactly 30 eggs were dispensed into 8 dram vials containing ~ 6 ml banana food and 50 such vials were set up from each population. Ten vials from each population were introduced into continuous light (LL), continuous dark (DD), light-dark (LD) cycles of 10:10 h ( T20 ), 12:12 h ( T24 ), and 14:14 h ( T28 ). Thus a total of 200 vials were set up for the assay (10 vials × 4 populations × 5 light regimes). These vials were introduced into five different light regimes at 20:00 h, when lights went off simultaneously in all LD regimes. Fluorescent white light of intensity ~ 100 lux was used during light phase, whereas in dark phase red light of λ >650 nm was used. Temperature and relative humidity in the five light environments monitored continuously using a Quartz Precision Thermo-Hygrograph, Isuzu Seisakusho Co, LTD, were found to be comparable. The vials were monitored for eclosion of adult flies after the pupae became dark. Eclosing adults were collected every 2 h, sexed, and counted until all the flies eclosed. Statistical analyses Pre-adult development time in hours was calculated as the duration between the midpoint of the egg collection window and the midpoint of the 2-h period during which eclosion occurred. The mean pre-adult development time for a particular sex in a particular light regime was used as data in a mixed model analysis of variance (ANOVA), in which replicate populations were treated as random blocks, and light regime and sex were treated as fixed factors. In order to detect differences in the eclosion profiles of the flies under different light regime, development time data of individual flies from all four replicate populations were pooled and compared using Kruskal-Wallis test. Pre-adult survivorship was calculated as the fraction of eggs that successfully developed into adults in each vial. The mean pre-adult survivorship values of each population in each light regime were used as data in a mixed model ANOVA, with replicate populations as random blocks, and light regime as a fixed factor. Authors' contributions DAP and AD were involved in collecting eggs, collecting and counting flies for pre-adult development time, estimating pre-adult survivorship, data entry and analyses. VKS conceived of the study and participated in its design and coordination. MKC and AJ gave valuable comments and suggestions throughout the study. All authors read and approved the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC554107.xml
543453	Ablation lesions in Koch's triangle assessed by three-dimensional myocardial contrast echocardiography	Background Myocardial contrast echocardiography (MCE) allows visualization of radiofrequency (RF) ablation lesions in the left ventricle in an animal model. Aim: To test whether MCE allows visualization of RF and cryo ablation lesions in the human right atrium using three-dimensional echocardiography. Methods 18 patients underwent catheter ablation of a supraventricular tachycardia and were included in this prospective single-blind study. Twelve patients were ablated inside Koch's triangle and 6, who served as controls, outside this area. Three-dimensional echocardiography of Koch's triangle was performed before and after the ablation procedure in all patients, using respiration and ECG gated pullback of a 9 MHz ICE transducer, with and without continuous intravenous echocontrast infusion (SonoVue, Bracco). Two independent observers analyzed the data off-line. Results MCE identified ablation lesions as a low contrast area within the normal atrial myocardial tissue. Craters on the endocardial surface were seen in 10 (83%) patients after ablation. Lesions were identified in 11 out of 12 patients (92%). None of the control patients were recognized as having been ablated. The confidence score of the independent echo reviewer tended to be higher when the number of applications increased. Conclusions 1. MCE allows direct visualization of ablation lesions in the human atrial myocardium. 2. Both RF and cryo energy lesions can be identified using MCE.	Introduction Catheter ablation is a curative treatment for most patients with arrhythmias. In some patients, the results are still suboptimal[ 1 ] because of inadequate lesion formation during ablation. Therefore, in these patients direct visualization of ablation lesions may have significant impact on the outcome of the ablation procedures. Direct visualization can also provide additional information for both the development and testing of new dedicated ablation tools. Intracardiac echocardiography (ICE) has been extensively investigated for this purpose [ 2 - 6 ], but the reported results are disappointing[ 5 , 7 ]. Recently, myocardial contrast echocardiography (MCE) has been tested for visualization of ablation lesions in animals in the left ventricle during continuous intracoronary echocontrast infusion[ 8 ]. The aim of the present study was to assess the potential use of MCE to demonstrate ablation lesions in human atrial myocardial tissue with continuous venous echocontrast administration. Methods Patients and study protocol 18 patients were included into this study. The clinical characteristics of the patients are shown in Table 1 . All patients underwent EP study and subsequent ablation procedures for supraventricular tachycardia. The Ethics Committee of Erasmus MC, Rotterdam, The Netherlands approved this study. Written, informed consent was obtained. Regardless of the final diagnosis, Koch's triangle was visualized with ICE in all patients at baseline without echocontrast and immediately thereafter, during continuous echocontrast infusion (SonoVue, Bracco). After the ablation procedure, which was either inside or outside Koch's triangle, the ICE procedure was repeated in six patients without echocontrast and in all patients using echocontrast. All ICE procedures were performed using a respiration and ECG gated and triggered pullback technique allowing three dimensional (3D) reconstruction of Koch's triangle. All 2D recordings were analyzed off-line by two independent echocardiographers and they provided confidence scores using a 1–10 grade scale. They were not aware that 6 patients were not ablated in Koch's triangle. This protocol implies that this was a prospective single blind study. 3D reconstruction of the ablation lesions was performed in patients where ablation lesions were seen. Table 1 Clinical characteristics and procedural outcome of the study patients Overall group Number of patients (n) 18 Gender (F/M) 10/8 Age (years ± SD) 49.3 ± 15.7 AVNRT (n) 12 AP (n) 6 Successful ablation (n) 17 Procedure time (min ± SD) 177.1 ± 68.9 Fluoroscopy time (min ± SD) 38.8 ± 27.5 RF/cryo applications (n) 4.2 ± 5.2 n = number, F = female, M = male, AVNRT = atrio-ventricular nodal reentry tachycardia, AP = accessory pathway, min = minutes, RF = radiofrequency, cryo = cryothermy, SD = standard deviation Electrophysiology testing and ablation Standard electrophysiology (EP) and ablation procedures were undertaken using quadripolar electrode catheters in the high right atrium, to record the His bundle electrogram, in the right ventricle and a decapolar diagnostic catheter was inserted into the coronary sinus (CS). The initial portion of the EP procedure was directed at determining the presence of dual AV nodal physiology or accessory pathways, measuring the conduction properties and refractory periods of the fast and slow AV nodal pathways (if present), and determining the mechanism of the paroxysmal SVT. Mapping was performed and after the target site was identified, ablation was applied. Cryothermy and radiofrequency energy were used alternately during the study period. Myocardial contrast echocardiography MCE was performed using SonoVue (Bracco), which is a second generation contrast agent made of microbubbles stabilized by phospholipids and containing sulphur hexafluoride. The mean bubble diameter is 2.5 μm and more than 90% of the bubbles are smaller than 8 μm. The blood level curve shows a distribution half-life of about 1 minute and an elimination half-life of about 6 minutes[ 9 ]. In this study we administered SonoVue by continuous intravenous infusion through the cubital vein at a rate of 100 ml/hour. Gain settings were not changed throughout the study. Intracardiac echocardiography (ICE) The ClearView™ system (CardioVascular Imaging Systems Inc, Fremont, CA) was used with an 8F sheath-based ICE imaging catheter that incorporates a 9 MHz beveled single-element transducer rotating at 1800 rpm (model 9900, EP Technologies, Boston Scientific Corp., San Jose, CA, USA). ECG- and respiration-gated image acquisition and 3-D image processing A custom-designed ECG- and respiratory-gated pullback device and a 3D-ultrasound workstation (EchoScan, TomTec GmbH, Munich, Germany) were used to acquire and process the ICE images using a technique described elsewhere[ 10 ]. The pullback device is controlled by the 3D workstation and uses a stepping motor to move the catheter stepwise and linearly through the right atrium. The workstation receives video input from the ICE system and an ECG- and respiration-signal (impedance measurement) from the patient. Prior to the acquisition run, the range of RR- and breathing-intervals are measured to define the upper- and lower-limits. The workstation starts acquisition of 2D images after detecting the peak of the R-wave and in the same phase of respiration, at a speed of 25 images/ sec (image interval 40 ms). After acquiring one cardiac cycle, the workstation stores the images, and the catheter is then pulled back by a 0.5-mm axial increment. This process is repeated until the inferior vena cava (IVC) is reached. The acquisition time is much shortened when all cardiac cycles are of the same length, therefore, the right ventricular apex is paced at 100 bpm. In accordance with their timing in the cardiac cycle, all images are formatted in volumetric data sets (256256256 pixels/each 8 bits). During post-processing, several algorithms are applied to reduce noise, enhance edges, and reduce spatial artifacts (ROSA filter). Statistical analysis Continuous variables are expressed as mean ± standard deviation. Correlation analysis between the confidence scores and number of ablation lesions were performed using Pearson's test. The level of significance was set at a p value of 0.05. Results Ablation results (Table 1 ) Of the 18 patients undergoing catheter ablation of supraventricular arrhythmias 12 had AVNRT tachycardia. 6 out of these 12 patients were ablated using cryothermy. All but one patient were successfully ablated (one patient with AVNRT). The number of applications in Koch's triangle was 6 ± 4.9, ranging from 1 to 15 applications. The fluoroscopy and procedure times were 45.7 ± 30.8 min and 196 ± 80.2 min, respectively. Myocardial contrast echocardiography (Table 2 ) MCE identified ablation lesions as a low contrast area within the normal atrial myocardial tissue (Figure 1 ). Lesions were identified in 11 out of 12 patients (91%) ablated in Koch's triangle. In only one patient with a single radiofrequency application, was the lesion not recognized. None of the control patients were recognized as having been ablated. The average confidence scores of the independent echo reviewers were 8.5 ± 2.4 and 8.1 ± 2.4, respectively. Both reviewer's confidence scores ranged from 3–10. The confidence score of the independent echo reviewer tended to be higher when the number of applications increased (Reviewer 1: r = 0.697, p = 0.025; Reviewer 2: r = 0.748, p = 0.013). Craters on the endocardial surface were seen in all 12 patients after ablation, by both echo reviewers (Figures 1 and 2 ). Table 2 Results of myocardial contrast echocardiography in patients ablated in Koch's triangle Reviewer 1 Reviewer 2 Pt. No Ablation energy Lesion before ablation Lesion after ablation Crater after ablation Lesion before ablation Lesion after ablation Crater after ablation NC C NC C NC C NC C 1. RF - - NA + + - - NA + + 2. Cryo - - NA + + - - NA + + 3. RF - - NA - + - - NA - + 4. Cryo - - NA + + - - NA + + 5. RF - - NA + + - - NA + + 6. Cryo - - NA + + - - NA + + 7. RF - - NA + + - - NA + + 8. Cryo - - NA + + - - NA + + 9. RF - - - + + - - - + + 10. Cryo - - - + + - - - + + 11. RF - - - + + - - - + + 12. Cryo - - - + + - - - + + RF = radiofrequency, Cryo = cryothermy, No = number, NC = no contrast (without contrast), C = with contrast, NA = not applicable, Pt = patient Figure 1 Two-dimensional intracardiac echocardiography images showing part of Koch's triangle between the tricuspid valve and the ostium of the coronary sinus under four different conditions. A: Native 2D horizontal cross-sectional echocardiography image before ablation. B: The same region before ablation with use of echocontrast. C: The same region after radiofrequency energy ablation without echocontrast infusion. A crater as an indirect sign of the ablation lesion (arrow) can be seen on the endocardial surface at the atrial side adjacent to the tricuspid valve. D: The same region after radiofrequency energy ablation and during echocontrast infusion. The ablation lesion (arrow) is visualized as a low contrast area within the atrial myocardial tissue. A crater can be seen on the atrial side adjacent to the tricuspid valve. In both C and D situations (post-ablation) there is significant swelling of the ablated region compared with pre-ablation situations (A and B). ICE = central artifact of the intracardiac echocardiography catheter, TV = tricuspid valve, RA = right atrium, CSos = ostium of the coronary sinus Figure 2 Three-dimensional reconstruction of Koch's triangle: "En face" view of a radiofrequency ablation lesion (arrow). The crater on the right atrial endocardial surface is also well visualized directly to the right. RA = right atrium, TV = tricuspid valve, SUP = superior, INF = inferior 3D reconstruction of the lesion Koch's triangle was successfully reconstructed in 3D in all patients in whom the ablation lesion was previously identified. Lesions could be easily found by 3D echocardiography. An "en face" view of the lesion could be reconstructed in all of these patients (Figure 2 ). Discussion This study demonstrates that MCE allows visualization of ablation lesions in the human right atrial myocardium during continuous venous echocontrast infusion. This potentially opens a new avenue for objective, and easily accessible evaluation of ablation lesions. Since transmural lesion formation is critical for successful ablation[ 11 ], the MCE method may have a significant impact on the outcome of ablation procedures. The role of echocardiography in assessment of ablation lesions Crater formation and tissue changes as increased echo-density have been detected using ICE immediately after RF ablation[ 12 , 13 ]. However, the lesions are not always seen after ablation and indirect signs are being searched for such as changes in local wall thickness[ 3 , 4 ]. The magnitude of the observed changes showed a certain level of correlation with the lesion size. These indirect signs may indicate appropriate lesion formation, but there is an obvious need for direct visualization of the lesions. MCE offers this potential and has allowed visualization of ablation lesions in animals in the left ventricle during continuous intracoronary echocontrast perfusion[ 8 ]. The investigators demonstrated high accuracy and reliability in visualizing tissue damage. Human use does not seem to be practical in this way and neither was safety assessed. In the present study we used continuous peripheral venous echocontrast infusion and we screened lesions in Koch's triangle in humans. We showed in atrial myocardial tissue that the lesions can be visualized and continuous venous infusion provides sufficient differences in echo contrast intensity to directly visualize ablation lesions after focal ablation. In this study we used cryothermy as well as radiofrequency energy for creating lesions in Koch's triangle. We do not have a sufficient number of patients for volumetric comparison of ablation lesions using this method, but it seems that both types of ablation lesion can be reliably visualized using this technique. Limitations of the study This single blind controlled study allowed two independent reviewers to examine the 2D ICE recordings. Although in none of the control patients a lesion in the area of interest was recognized, there was one patient with a single ablation lesion who was not identified by the two independent reviewers. This may suggest that the sensitivity of the technique is still suboptimal. One possible reason is that the concentration of the echocontrast infusion was titrated too low. This is also reflected in the results of the confidence scores. These showed a linear correlation with the number of applications. Furthermore, a correlative study to gross inspection and histopathology would be advantageous. Obviously, this cannot be done in humans. The role of 3D reconstruction and future clinical implications 3D reconstruction and a creation of a volumetric data set allow visualization of an "en face" view of the ablation lesion. Without 3D reconstruction determination of the depth and the shape of the ablation lesion was not possible. Therefore the technique could be used to assess whether the ablation lesion was transmural. Furthermore, with 3D reconstruction, the volume of the lesion can be determined. This could potentially be a major asset for clinical electrophysiology since on line 3D echocardiography would allow real-time assessment of ablation lesions. Most importantly, the appropriateness of such lesions could be judged during linear ablations and the continuity of the lines could be checked. In conclusion, MCE is a safe and promising new method to detect ablation lesion in the human atrial tissue.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC543453.xml
479042	An Integrin-Dependent Role of Pouch Endoderm in Hyoid Cartilage Development	Pharyngeal endoderm is essential for and can reprogram development of the head skeleton. Here we investigate the roles of specific endodermal structures in regulating craniofacial development. We have isolated an integrinα5 mutant in zebrafish that has region-specific losses of facial cartilages derived from hyoid neural crest cells. In addition, the cranial muscles that normally attach to the affected cartilage region and their associated nerve are secondarily reduced in integrinα5 − animals. Earlier in development, integrinα5 mutants also have specific defects in the formation of the first pouch, an outpocketing of the pharyngeal endoderm. By fate mapping, we show that the cartilage regions that are lost in integrinα5 mutants develop from neural crest cells directly adjacent to the first pouch in wild-type animals. Furthermore, we demonstrate that Integrinα5 functions in the endoderm to control pouch formation and cartilage development. Time-lapse recordings suggest that the first pouch promotes region-specific cartilage development by regulating the local compaction and survival of skeletogenic neural crest cells. Thus, our results reveal a hierarchy of tissue interactions, at the top of which is the first endodermal pouch, which locally coordinates the development of multiple tissues in a specific region of the vertebrate face. Lastly, we discuss the implications of a mosaic assembly of the facial skeleton for the evolution of ray-finned fish.	Introduction The skeletal elements that form and support the vertebrate jaw and gills are derived from a specialized population of ectomesenchyme cells, the cranial neural crest ( Platt 1893 ; Le Douarin 1982 ; Schilling and Kimmel 1994 ; but see Weston et al. 2004 ). In the larval zebrafish, crest cells of the first, or mandibular, arch give rise to Meckel's and palatoquadrate cartilages that constitute the lower and upper jaws, respectively. Several cartilages are derived from the second, or hyoid, arch, including the ceratohyal (CH) and hyosymplectic (HS) cartilages that support the jaw. In particular, the HS cartilage serves to connect the upper jaw to the skull by means of a hyomandibula (HM) plate and a symplectic (SY) anterior rod-like extension. In addition, the HM plate supports the overlying opercular apparatus that helps to ventilate the gills ( Hughes and Shelton 1958 ). The tetrapod stapes is homologous to HM. During vertebrate development, cranial neural crest cells delaminate from near the dorsal neural primordium and migrate to ventrolateral positions where they populate a series of pharyngeal arches (reviewed in Le Douarin 1982 ). Once the pharyngeal arches are established, skeletal elements develop from cylinders of neural crest whose mesodermal cores undergo stereotypic divisions to form the cranial muscles ( Edgeworth 1935 ; Schilling and Kimmel 1997 ; Kimmel et al. 2001 ). The segmentally organized pharyngeal arches are separated from one another by reiterative outpocketings of the pharyngeal endoderm called pouches. Recent work in chicken has demonstrated an important role for endoderm in patterning cartilages of all the pharyngeal arches ( Couly et al. 2002 ; Ruhin et al. 2003 ). Using grafting and ablation experiments, these researchers divided the pharyngeal endoderm into anterior-posterior (A-P) and mediolateral domains that are required for and have the ability to induce segment-specific pharyngeal cartilages. In zebrafish, experiments have demonstrated a genetic requirement for endoderm in pharyngeal cartilage development. casanova mutant embryos make no endoderm and pharyngeal cartilages fail to form, and transplantation experiments show that wild-type endoderm is sufficient to rescue cartilage development ( Alexander et al. 1999 ; David et al. 2002 ). In addition, a role for pharyngeal pouches in segmentation and survival of cranial neural crest has been shown. In zebrafish tbx1 (vgo) mutant embryos, most pouches fail to develop and posterior pharyngeal cartilages are reduced and fused together ( Piotrowski and Nusslein-Volhard 2000 ; Piotrowski et al. 2003 ). However, though endoderm is clearly critical for pharyngeal cartilage development, it is not well understood how interactions between neural crest–derived cells and endoderm produce segment-specific patterns of cartilage. In this work, we isolate and characterize a zebrafish integrinα5 loss-of-function mutant. Integrins are a family of heterodimeric receptors, composed of α and β subunits, that bind to ligands in the extracellular matrix such as fibronectin and laminin. Integrins have structural roles in adhesion that promote tissue integrity and cell migration, and signaling functions important for cell differentiation and survival (reviewed in Bokel and Brown 2002 ). Various studies in mouse and chick have shown a role for integrins and their ligands in neural crest migration. Integrins α4, α1, and αV are expressed early in neural crest development, and function-blocking antibodies against these integrins perturb crest migration in vitro ( Delannet et al. 1994 ; Desban and Duband 1997 ; Kil et al. 1998 ; Testaz and Duband 2001 ). The in vivo roles of specific integrins in neural crest development are less clear ( Yang et al. 1995 ; Gardner et al. 1996 ; Bader et al. 1998 ). Mice lacking Integrinα5, which in complex with primarily β1 forms the major fibronectin receptor, die early during embryogenesis because of mesodermal defects ( Yang et al. 1993 , 1999 ). Analysis of integrinα5 −/− mouse embryos showed that Integrinα5 is required for the survival of a subset of hyoid crest ( Goh et al. 1997 ). However, it was not known where Integrinα5 functions to control hyoid crest development. Here we report that, in zebrafish, Integrinα5 functions in the pharyngeal endoderm to control hyoid crest development. In integrinα5 − embryos, the first pharyngeal pouch fails to develop, and the lack of a first pouch correlates with reductions in specific regions of the HS cartilage. integrinα5 mutants also have defects in a subset of dorsal first and second arch muscles and facial motor nerve VII, suggesting that Integrinα5 is required for region-specific development of multiple pharyngeal tissues. However, both expression and penetrance data suggest that the muscle and nerve defects are likely secondary to the cartilage and pouch defects. integrinα5 is expressed in pharyngeal endoderm during pouch formation and is required in endoderm for both first pouch and hyoid cartilage development. In order to understand the remarkable specificity of the integrinα5 − cartilage phenotype, we fate mapped regions of the HS cartilage in the hyoid arch. We found that the regions of the HS cartilage that are lost in integrinα5 mutants develop from anterior crest–derived cells immediately adjacent to the first pouch. Analysis of integrinα5 mutants suggests that the first endodermal pouch specifies a portion of the hyoid cartilage pattern by locally stabilizing hyoid crest. Lastly, we present a model in which new local interactions of endodermal structures with hyoid crest underlie the elaboration of the jaw support apparatus during the evolution of ray-finned fish. Results Isolation of a Mutation in Zebrafish integrinα5 In a genetic screen for zebrafish with altered pharyngeal cartilages, we isolated a single recessive mutant allele, b926, with specific defects in the hyoid cartilage pattern. Using polymorphism mapping, we placed b926 on Linkage Group (LG) 23 between the markers Z5141 and Z20492, a region containing a zebrafish integrinα5 homolog ( Figure 1 A). We performed reverse transcription polymerase chain reaction (RT-PCR) to obtain a full-length cDNA encoding a protein product with 54% identity to mouse Integrinα5. Sequencing of integrinα5 in b926 revealed a T to A nucleotide substitution that segregated with the mutant phenotype. The b926 mutation converts a conserved tyrosine residue to an asparagine in the third beta-propeller repeat of the extracellular domain, a region known to be important for ligand binding ( Figure 1 B and 1 C) ( Springer 1997 ; Mould et al. 2000 ). In addition, a morpholino designed against the exon13-intron splice site of integrinα5 ( integrinα5 -MO; Figure 1 D) phenocopies both the cartilage and first pouch defects of b926 . We confirmed by RT-PCR that itga5 -MO effectively inhibits the splicing of integrinα5 ( Figure 1 E and 1 F). We conclude that b926 is a loss-of-function mutation in zebrafish integrinα5 . Figure 1 Identification of a Zebrafish integrinα5 Mutant (A) Based on the hyoid cartilage phenotype the b926 allele was mapped to LG23. Using polymorphism mapping, we placed b926 between the markers Z5141 (2/362 recombinants/meioses) and Z20494 (3/362 recombinants/meioses). Databases of the partial zebrafish genomic sequence revealed that a homolog of integrinα5 mapped to this region. (B) Zebrafish Integrinα5 protein is predicted to have seven extracellular beta-propeller repeats (stippled), a transmembrane domain, and a short intracellular cytoplasmic tail. Integrinα5 forms heterodimeric complexes with Integrin β chains, primarily β1, and binds extracellular matrix ligands containing the RGD motif. Sequencing of integrinα 5 in b926 revealed a T to A nucleotide substitution at position 952 of the cDNA that converts a tyrosine to an asparagine at amino acid 218, a residue within the third beta-propeller repeat. (C) Alignment of the third beta-propeller repeat of the Integrinα5 proteins of zebrafish, human, mouse, Xenopus, and Fugu . Y218 is absolutely conserved among all five species and is mutated to N in b926 . (D) The genomic locus of integrinα5 consists of 30 exons (not drawn to scale). The integrinα 5 morpholino, itga5 -MO, was designed against the exon13-intron splice site (arrow). (E) The primers GC145 and GC147 were used to amplify a 497-bp fragment from wild-type cDNA. PCR amplification was then performed using cDNA from 24-hpf embryos that had been injected with 10 ng of itga5 -MO at the one-cell stage, the same concentration of morpholino used for our phenotypic analysis. The resultant product was 587 bp long, and sequencing confirmed that the increased length was due to the failure to splice out the intron between exons 13 and 14. The left lane of the agarose gel contains standard size markers in basepairs. (F) A schematic of the wild-type Integrinα5 protein shows seven extracellular repeats and the transmembrane (TM) domain. Inhibition of splicing by itga5 -MO would be predicted to result in a nonfunctional protein that is truncated in the seventh repeat of the extracellular domain. Region-Specific Pharyngeal Defects in integrinα5 Mutants integrinα5 mutant zebrafish showed partially penetrant and variably expressive losses of specific hyoid cartilage regions at 4 d ( Figure 2 ; Table 1 ). The most frequent phenotype seen in integrinα5 − animals was a specific loss of the anterior half of the HM plate (anterior HM [aHM]) ( Figure 2 B). In other integrinα5 mutants, we saw variable reductions of the SY element in addition to aHM ( Figure 2 C). In what we interpret as the most severe integrinα5 (b926) phenotype, HS was reduced to a rod that variably fused with CH; rarely, the first arch joint was fused as well ( Figure 2 D). However, even in the most severe class of integrinα5 mutants, the posterior half of HM (posterior HM [pHM]), the CH, and the hyoid opercle bone remained unaffected. In addition, integrinα5 mutants had variably reduced numbers of ceratobranchial (CB) cartilages and rare fusions of adjacent CBs ( Figure 2 G; Table 1 ). Animals injected with itga5 -MO displayed a similar range and spectrum of cartilage phenotypes ( Figure 2 E; Table 1 ). Figure 2 Region-Specific Pharyngeal Defects in integrinα5 Mutants (A–E) Flat mount dissections of hyoid and mandibular cartilages from fixed, 4-d-old wild-type (A), integrinα5 − (B–D), and itga5 -MO (E) animals. Meckel's (M) and palatoquadrate (PQ) cartilages are derived from the mandibular arch (1), and CH, SY, and HM cartilages and the opercle (Op) bone are derived from the hyoid arch (2). A phenotypic series (B–D) shows that the anterior half of HM (arrows) is absent and SY is progressively reduced in integrinα5 − animals. Rarely, mandibular and hyoid joints are also missing in integrinα5 − animals (asterisks in D). (E) Animals treated with itga5 -MO display similar reductions of HM (arrow) and SY. (F and G) Flat-mount dissections of the pharyngeal cartilages of 4-d-old wild-type (F) and integrinα5 − (G) animals. In addition to the mandibular and hyoid cartilages, the five CB cartilages (CB1–CB5) that are derived from the third through seventh arches are shown. Note the teeth on CB5 (dots in F). In integrinα5 − embryos we see rare fusions of CB cartilages (arrow in G). (H–J) Confocal micrographs of the pharyngeal arches of wild-type fli1 -GFP (H) and integrinα5 − ; fli1 -GFP (I and J) embryos stained with anti-GFP and Zn8 antibodies at 38 hpf. Neural crest cells of the pharyngeal arches are labeled with fli1 -GFP (green, numbered in [H]), and the pharyngeal pouches are labeled by the Zn8 antibody (red, numbered p1–p5 in [H]). In integrinα5 − ; fli1 -GFP embryos, the first pouch is absent or very reduced at 38 hpf (arrows in I and J). Less frequently, we also see reductions in more posterior pouches in integrinα5 − ; fli1 -GFP embryos (arrowhead in J shows a single endodermal mass where p3–p5 would be in wild-type embryos). The Zn8 antibody also recognizes cranial sensory ganglia (dots). (K and L) In situ hybridizations of wild-type (K) and integrinα5 − (L) embryos stained with the pharyngeal pouch marker pea3 at 36 hpf (arrowhead denotes first pouch). The first pouch of integrinα5 − embryos is very reduced, but still expresses pea3 . Sensory ganglia also stain with pea3 (dots). (M and N) Cranial muscles of 4-d-old wild-type fli1 -GFP (M) and integrinα5 − ; fli1 -GFP (N) embryos stained with MF20 antibody. Mandibular muscles (intermandibularis posterior [imp], adductor mandibulae [am], levator arcus palatine [lap], and do) and hyoid muscles (interhyal , hyohyal [hh], ah, ao, and lo) are labeled in wild-type. integrinα5 − embryos have a selective reduction of do and ah muscles (arrow in [N]). Confocal projections of integrinα5 − animals did not include ocular muscles (asterisks in M). (O and P) Cranial motor nerves of wild-type islet1 -GFP (O) and integrinα5 − ; islet1 -GFP (P) live embryos at 54 hpf. islet1 -GFP-expressing cranial motor neurons innervate muscles of the pharyngeal arches with the following strict segmental correspondence: trigeminal (V)—mandibular; facial (VII)—hyoid; glossopharyngeal (IX)—third; and vagus (X)—fourth through seventh. In integrinα5 − ; islet1 -GFP embryos, facial nerve VII (arrowhead in P) is reduced and/or fails to branch. (Q and R) Summary of integrinα5 regional pharyngeal defects extrapolated to a 4-d-old embryo and color-coded for cartilage (blue), muscle (red), and nerve (green). Shown in black are the eye (filled circle within larger circle), ear (two dots within oval), and opercle bone (mushroom). In wild-type animals, facial nerve VII innervates and passes by do and ah muscles that are in close association with the aHM cartilage region (enlarged in Q′). In integrinα5 mutants, we see specific reductions of the first pouch (not shown), the aHM cartilage region, do and ah muscles, and facial nerve VII (enlarged in R′). Scale bars: 50 μm. Table 1 Pharyngeal Defects in Animals Reduced for integrinα5 Percentage of sides with each phenotype is listed for integrinα5 − (b926); fli1 -GFP animals and fli1 -GFP animals injected with 10 ng of itga5 -MO. In 100 wild-type animals, none of these phenotypes was seen. For cartilage defects, percentage of mutant clutch showing any hyoid cartilage defect is listed. Hyoid cartilage defects are divided into categories based on the morphology of the HS element: HS rod-shaped with SY extension (“aHM lost”), HS rod-shaped without SY extension (“aHM and SY lost”), and HS rod-shaped and fused to CH (“losses and fusion to CH”). Average number of CB cartilages per side is listed; wild-type animals invariably have five CBs per side. Fusions of CBs in mutant clutches were rare and not quantified. Pharyngeal endoderm defects were scored on a fluorescence dissecting microscope as percentage of mutant clutch displaying any pouch defect at 34 hpf. Pouch defects were divided into two classes based on whether more posterior pouches were lost in addition to the first pouch. Cranial muscle defects were scored as percentage of mutant clutch with reductions of do and ah muscles. In addition, we found five animals from integrinα5 ; islet1 -GFP mutant clutches ( n = 200) with defects in the facial (VII) motor nerve. Animals with potential nerve defects were identified in the dissecting microscope and subsequently confirmed on the confocal microscope. As such, we cannot give an absolute percentage of animals with nerve defects. All animals with muscle and nerve defects also had hyoid cartilage defects. In addition to the phenotypes listed, some integrinα5 mutants developed heart edema and had smaller eyes by 4 d. Animals injected with higher doses of itga5 -MO (15 ng) also developed severe heart edema, had shorter bodies, and often stained poorly for cartilage In addition to cartilage defects, we found partially penetrant reductions of the first pharyngeal pouch, an endodermal outpocketing, in integrinα5 mutants ( Figure 2 I and 2 J). Loss of the first pouch was apparent as early as 24 hours post fertilization (hpf), and similar first pouch reductions were seen in animals injected with itga5 -MO (data not shown; Table 1 ). Although the majority of integrinα5 − embryos showed defects restricted to the first pouch, a few integrinα5 − embryos had graded reductions of more posterior pouches as well ( Figure 2 J; Table 1 ). In order to examine whether the reduced first pouches in integrinα5 − embryos retained pouch identity, we examined expression of the pharyngeal pouch marker pea3, a downstream effector of Fgf signaling ( Figure 2 K) ( Roehl and Nusslein-Volhard 2001 ). We found that the reduced first pouches of integrinα5 mutants still expressed pea3 at 36 hpf ( Figure 2 L). As both the pouch and cartilage phenotypes were incompletely penetrant in integrinα5 mutants, often with one side of an embryo showing defects and the other side not, we examined how tightly coupled the phenotypes were in individual sides. In order to visualize early pharyngeal arch structure in live animals, we made use of a fli1– green fluorescent protein (GFP) transgenic line ( Lawson and Weinstein 2002 ). fli1 -GFP expression initiates in neural crest cells of the pharyngeal arches shortly after crest migration (ca. 18 hpf) and persists as pharyngeal cartilages and bones develop. Pouches are evident as nonexpressing regions between the GFP-labeled crest-derived cells of the arches. We sorted live integrinα5 − animals for first pouch defects at 36 hpf and then grew these animals to 4 d to analyze pharyngeal cartilages. Strikingly, we found a strong correlation between reductions of the first pouch and later hyoid cartilage defects in integrinα5 mutants. integrinα5 − sides that lacked the first pouch early had hyoid cartilage defects 93% of the time ( n = 116), whereas only 7% of integrinα5 − sides with a normal first pouch early developed hyoid cartilage defects ( n = 108). The correlation was highly significant ( p < 0.0001). We next asked whether Integrinα5 was required for the development of cranial muscles. By 4 d of development, the mesoderm of the first pharyngeal arch has undergone stereotypic subdivisions to form, ventrally, intermandibularis anterior and intermandibularis posterior; medially, adductor mandibulae; and, dorsally, levator arcus palatini and dilatator operculi (do) cranial muscles. Second arch mesoderm gives rise to interhyal and hyohyal muscles, ventrally, and adductor hyomandibulae (ah), adductor operculi (ao), and levator operculi (lo) muscles, dorsally ( Edgeworth 1935 ; Schilling et al. 1997 ; Figure 2 M). In a few integrinα5 mutants, the dorsal first and second arch muscles, do and ah, were selectively reduced, whereas levator arcus palatini, ao, and lo were present but appeared closer together ( Figure 2 N; Table 1 ). The muscles that were disrupted in integrinα5 mutants correspond to those that associate most closely with the aHM cartilage (schematized in Figure 2 Q and 2 R). Lastly, we found that integrinα5 mutants had specific defects in the nerve that innervates second arch muscles. Facial motor neurons send a ventral-directed nerve (VII) that innervates dorsal second arch muscles ah and ao, passes through the foramen of HM, and subsequently branches to innervate ventral muscles interhyal and hyohyal ( Higashijima et al. 2000 ; Figure 2 O). In a small fraction of integrinα5 mutants, facial nerve VII failed to branch and/or was hypotrophic ( Figure 2 P). Both nerve and muscle defects were only seen in integrinα5 mutants that also displayed cartilage defects. Moreover, the lower penetrance of nerve and muscle defects suggests that they are secondary to endodermal and/or crest defects. In conclusion, we have found a requirement for Integrinα5 in the development of multiple pharyngeal tissues in the vicinity of the first pouch and aHM cartilage region. integrinα5 Is Expressed Dynamically in Pharyngeal Endoderm and Cranial Neural Crest In order to understand where Integrinα5 might be acting to control pharyngeal arch development, we examined the expression of integrinα5 mRNA by in situ hybridization. The integrinα5 expression domains were complex and dynamic, and we do not give an exhaustive description of domains other than the pharynx here. We observed integrinα5 expression as early as the 32-cell stage, indicating a maternal integrinα5 contribution ( Figure 3 A). Maternal expression of integrinα5 also has been reported in frog ( Whittaker and DeSimone 1993 ). By gastrulation stage (60% epiboly), the mesendoderm broadly expressed integrinα5 ( Figure 3 B). At the 1-somite (s) stage, integrinα5 expression was in ectoderm adjacent to the anterior neural plate, a domain consistent with cranial neural crest and placode precursors ( Figure 3 C). In addition, we saw strong expression in the first somite and posterior mesoderm and weaker expression in scattered, large cells lateral and anterior to the notochord that we interpret as early endoderm ( Warga and Nusslein-Volhard 1999 ). By 5-s stages, cranial neural crest and the otic placode expressed integrinα5 ( Figure 3 D), and pharyngeal endoderm expression was seen ventrally along the surface of the yolk ( Figure 3 E). From 12-s to 18-s (18 hpf), pharyngeal endoderm continued to express integrinα5, and the ectodermal expression domain became more restricted to hyoid crest–derived tissue and the otic placode ( Figure 3 F and 3 G). Endoderm and ectoderm expression domains of integrinα5 were apparent most clearly in 18 hpf cross-sections ( Figure 3 H– 3 J). In particular, we observed strong integrinα5 expression throughout pharyngeal endoderm, including the first pouch ( Figure 3 H). At later times, we saw dynamic integrinα5 expression in both crest derivatives and pharyngeal endoderm. The six pharyngeal pouches form in an anterior to posterior wave of development. By 26 hpf, the fourth pouch, which was the most posterior pouch forming at this time (J.G.C. and C.B.K., unpublished data), expressed integrinα5, whereas integrinα5 expression was no longer seen in the first pouch ( Figure 3 K). At 38 hpf, the sixth pouch, the last to form, strongly expressed integrinα5, yet the fourth pouch was nonexpressing ( Figure 3 L). In addition, dynamic integrinα5 expression was seen in patchy zones of pharyngeal crest. Finally, integrinα5 expression was not affected by the b926 mutation (examined at 12-s; data not shown). In conclusion, integrinα5 expression in the pharyngeal endoderm is in a pattern that both spatially and temporally corresponds to regions of pouch formation, and expression of integrinα5 in the crest begins during premigratory stages and later becomes refined to patches of crest derivatives within the pharyngeal arches. Figure 3 integrinα5 Expression in Pharyngeal Endoderm and Cranial Neural Crest (A) At the 32-cell stage, strong maternal integrinα5 expression is seen. (B) At 60% epiboly, integrinα5 expresses broadly throughout the mesendoderm. (C) Dorsal view of a 1-s-stage embryo. integrinα5 transcript is concentrated in the ectoderm at the edge of the neural plate (black arrow), in scattered presumptive endodermal cells, and in the first somite (white arrow). (D and E) Dorsal (D) and lateral (E) views of a 5-s-stage embryo show ectodermal (arrows) and pharyngeal endodermal (arrowhead) expression domains of integrinα5 . Ectodermal integrinα5 expression includes migratory hyoid crest, otic placode, and forebrain. (F) At the 12-s stage, integrinα5 continues to be expressed in the pharyngeal endoderm (black arrowhead), postmigratory hyoid crest (arrow), ear (red arrowhead), and forebrain. (G–J) At 18 hpf, a dorsal view of an embryo stained for integrinα5 transcript (G) shows approximate axial levels at which cross-sections were prepared. (H) A cross-section at the level of the first pouch shows strong integrinα5 expression in the pharyngeal endoderm (arrowhead). (I) A cross-section at the level of the hyoid arch shows expression of integrinα5 in neural crest (arrow) and pharyngeal endoderm (arrowhead). (J) A cross-section at the level of the ear shows integrinα5 expression in the otic epithelium (red arrowhead) and pharyngeal endoderm (black arrowhead). (K and L) At 26-hpf (K) and 38-hpf (L) stages, integrinα5 transcript is enriched in the region of the most recent forming pharyngeal pouch (arrowheads) and in patches of crest (arrows). Scale bars: (A–C), (F), and (G): 100 μm; (D), (E), and (H–L): 50 μm. Integrinα5 Is Required in Endoderm but Not Crest for First Pouch and Hyoid Cartilage Development As integrinα5 expression was observed in both pharyngeal endoderm and neural crest, we used transplantation experiments to determine in which tissues Integrinα5 was required for first pouch and hyoid cartilage development ( Figure 4 ). Since it is difficult to transplant large amounts of endoderm from normal zebrafish embryos, we used forced expression of the activated Taram-A receptor (TAR) to generate donor embryos consisting almost entirely of endoderm (see David et al. [2002] for details). This method allows the specific and unilateral introduction of large amounts of endoderm into mutant hosts. Transplants were performed at 40% epiboly (late blastula; ca. 4 hpf), and the first pouch was scored at 38 hpf in live animals ( Figure 4 A). In wild-type to wild-type control transplants, 8% of recipient sides developed first pouch defects, suggesting a low level of toxicity of the TAR construct. We then transplanted wild-type TAR* endoderm into integrinα5 − ; fli1 -GFP hosts. Control mutant sides that did not receive donor endoderm had first pouch defects in 83% of cases. In contrast, first pouch defects were seen in only 17% of mutant sides that received wild-type donor endoderm ( Figure 4 C and 4 D; summarized in Figure 4 I). Thus, wild-type endoderm was able to rescue pouch formation in integrinα5 mutants. Furthermore, this rescue was dependent on donor endoderm contributing to the first pouch. Figure 4 integrinα5 Requirement in Endoderm but Not Crest (A and B) Schematics of endoderm (A) and crest (B) transplant experiments. (C–E) In endoderm transplants, confocal projections at 38 hpf (C) and 4 d (D) of a single integrinα5 − ; fli1 -GFP host animal show that wild-type TAR* donor tissue contributed efficiently to pharyngeal endoderm (red) but not crest (green). (E) Flat-mount dissection of mandibular and hyoid cartilages from the individual in (C) and (D). Wild-type TAR* endoderm rescues first pouch development (arrow in [C]) and partially rescues hyoid cartilage development (arrowhead in [D] and arrow in [E]) in integrinα5 − embryos. (F–H) In crest transplants, confocal projections of a single integrinα5 − ; fli1 -GFP host animal show extensive colocalization (yellow) of donor tissue (red) with crest (green) at 38 hpf (F) and 4 d (G). Donor tissue does not contribute to endoderm or mesoderm. (H) Flat-mount dissection of mandibular and hyoid cartilages from the individual in (F) and (G). Neither the first pouch defects (arrow in [F]) nor hyoid cartilage defects (arrowhead in [G], arrow in [H]) of integrinα5 − animals were rescued by wild-type crest. (I and J) Wild-type and integrinα5 − sides that received wild-type endoderm ( n wt = 39; n itga5 = 12) or crest ( n wt = 30; n itga5 = 12) transplants are plotted against the contralateral integrinα5 − control sides that did not receive transplants. (I) First pouch defects are quantified as percent of sides missing the first pouch. For endoderm transplants, integrinα5 − recipient sides were rescued to wild-type levels. For crest transplants, integrinα5 − recipient sides were not rescued compared to control sides. (J) Hyoid cartilage defects are quantified according to a mutant cartilage index: 0, wild-type; 1, partial aHM reduction; 2, full aHM loss; 3, aHM and SY losses; and 4, aHM and SY losses and fusion to CH. For endoderm transplants, integrinα5 − recipient sides were rescued to wild-type index. For crest transplants, integrinα5 − recipient sides were not rescued compared to control sides. Lowercase letters (a, b) in plots designate statistically significant groupings using Tukey-Kramer HSD test. Scale bars: 50 μm. We next asked whether wild-type endoderm also was able to rescue hyoid cartilage defects in integrinα5 − embryos. In order to quantify the severity of hyoid defects in integrinα5 mutants, we devised a mutant cartilage index that ranged from zero for wild-type to four for the most severe hyoid defects (see legend for Figure 4 ). In wild-type to wild-type control transplants, the mutant cartilage index was 0.46, consistent with TAR* causing a low level of defects on its own. The control nonrecipient sides of integrinα5 − ; fli1 -GFP animals had an index of 2.5. In contrast, the index of mutant sides that received the TAR* endoderm transplant was rescued to 0.92 ( Figure 4 D and 4 E; summarized in Figure 4 J). Hence, wild-type endoderm can nonautonomously rescue hyoid cartilage development in integrinα5 mutants. We also tested whether Integrinα5 was required only in crest for first pouch and hyoid cartilage development. We modified the hindbrain transplantation technique described in Maves et al. (2002) to transplant neural crest precursor cells at shield stages ( Figure 4 B). In wild-type to wild-type controls, transplants resulted in donor cells constituting a large proportion of the crest cells within the pharyngeal arches and resultant cartilages. In contrast to wild-type endoderm rescue, introduction of substantial amounts of wild-type crest failed to rescue first pouch formation in integrinα5 − ; fli1 -GFP animals ( Figure 4 F and 4 G; summarized in Figure 4 I). Furthermore, transplanted wild-type crest did not improve the mutant cartilage index of integrinα5 − ; fli1 -GFP animals (2.50 for recipient sides and 2.29 for nonrecipient sides) ( Figure 4 G and 4 H; summarized in Figure 4 J). Thus, wild-type crest was not able to rescue first pouch and hyoid cartilage development in integrinα5 mutants. Fate Map of Hyoid Cartilages Understanding the developmental basis for the specificity of the integrinα5 − cartilage phenotype requires a fate map of pharyngeal cartilages in wild-type animals. Here we focus on the origins of the SY, aHM, and pHM regions of HS, and a more complete mandibular and hyoid fate map will be published elsewhere. We used in vivo microelectroporation ( Lyons et al. 2003 ) to label crest cells at 24 hpf and later monitor their cartilage fate (see Materials and Methods ; Figure 5 A). Representative examples of 24-hpf in vivo microelectroporations show cells that contributed to SY, aHM, or pHM regions at 4 d ( Figure 5 B– 5 J). We plotted the origins of cells that contribute to each region along normalized A-P and dorsal-ventral (D-V) axes ( Figure 5 K). A comparison of mean distances along the A-P axis showed that cells that contributed to aHM and SY clustered on average 6–7 μm, or 1–2 cell diameters, from the first pouch (the anterior border of the arch). In contrast, cells contributing to pHM were on average 16 μm, or three cell diameters, away from the first pouch. A comparison of mean distances along the D-V axis showed that cells contributing to SY were more ventral than cells contributing to aHM and pHM. No statistically significant differences along the mediolateral axis were seen between cells contributing to different HS regions (data not shown; see legend of Figure 5 ). We conclude that HS cartilage regions most sensitive to loss of Integrinα5 are those developing just beside the first pouch. Figure 5 Fate Map of Hyoid Cartilages (A) In in vivo microelectroporation, a glass needle coupled to a positive electrode and filled with Alexa568 amine dextrans (red) is positioned in the hyoid arch (2) of wild-type fli1 -GFP embryos immobilized adjacent to a negative electrode. A short pulse of current delivers dye into single or pairs of cells. A-P and D-V axes, the mandibular arch (1), and first pouch (p1) are designated in (A) and (B). (B–D) Confocal sections of fli1 -GFP-labeled hyoid arches (2) (green) show the positions of Alexa568-labeled cells (red) shortly after microelectroporation (24 hpf). (E-J) At 4 d, confocal micrographs (E–G) (schematized in [H–J]) show the resultant fate of labeled crest cells (red) in the hyoid cartilage regions (green). Examples shown include labeled hyoid cells that contributed exclusively to SY (B, E, and H), aHM (C, F, and I), and pHM (D, G, and J) cartilage regions. (K) The relative distances (normalized to one) of hyoid crest cells at 24 hpf that contributed to SY (red), aHM (blue), and pHM (green) regions are plotted along A-P and D-V axes. The first pouch and partial outline of the mandibular arch (1) are drawn for reference. One cell gave rise to an aHM/pHM (blue/green) mixed lineage, and another cell gave rise to pHM and unidentified cells (green/light blue). SY and aHM progenitors map to more anterior domains (i.e., closer to the first pouch) than do pHM progenitors (relative distances from anterior: SY, 0.12 ± 0.11; aHM, 0.17 ± 0.06; pHM 0.43 ± 0.05; statistically different using Tukey-Kramer HSD test). SY progenitors map to a more ventral domain than do aHM and pHM progenitors (relative distances from dorsal: SY, 0.68 ± 0.11; aHM, 0.33 ± 0.06; pHM 0.37 ± 0.05; statistically different using Tukey-Kramer HSD test). No significant differences along the mediolateral axis were seen between regions (relative distances from lateral: SY, 0.47 ± 0.12; aHM, 0.53 ± 0.07; pHM 0.43 ± 0.06). (L) For the fate analysis, the 4-d HS cartilage was subdivided into SY (red), aHM (blue), and pHM (green) regions. The outline of the CH cartilage is also shown. Scale bars: 50 μm. Increased Cell Death and Disorganized goosecoid Expression in the Hyoid Arches of integrinα5 − Embryos We next investigated whether the losses of aHM and SY regions in integrinα5 − embryos correlated with increased cell death in the hyoid arch ( Figure 6 ). At 25 hpf, TUNEL staining revealed a greater than 2-fold increase in apoptosis over wild-type in the hyoid arches of integrinα5 − ; fli1 -GFP embryos ( Figure 6 A, 6 B, and 6E). A moderate tendency toward increased apoptosis was also seen in the hyoid arches of integrinα5 − ; fli1 -GFP embryos from 29 to 35 hpf ( Figure 6 D and 6 E). Apoptotic nuclei appeared to cluster in the dorsal anterior portion of the hyoid arch ( Figure 6 B and 6 D), and colocalization with fli1- GFP, a marker of neural crest, in confocal sections showed that some of these nuclei corresponded to dying crest cells ( Figure 6 B′). Interestingly, we observed an increase in hyoid apoptosis only in integrinα5 − ; fli1 -GFP sides in which the first pouch failed to develop ( Figure 6 E). In addition, at 14 hpf and 18 hpf, stages before which the first pouch has normally fully formed, no increase in cell death was seen in the cranial neural crest of integrinα5 − animals (data not shown). These results are consistent with the first pouch being required for the survival of hyoid crest that contributes to aHM and SY. Figure 6 Increased Apoptosis and Disorganized gsc Expression in the Hyoid Arches of integrinα5 Mutants (A–D) TUNEL staining of wild-type fli1- GFP (A and C) and integrinα5 − ; fli1 -GFP (B and D) animals shows apoptotic nuclei (red) relative to the GFP-expressing crest of the pharyngeal arches (green) at 25 hpf (A and B) and 29 hpf (C and D). In wild-type confocal projections arches are numbered. (A′) and (B′) are representative confocal sections taken from the projections in A and B. In integrinα5 − animals lacking the first pouch, increased apoptosis (arrows in [B] and [D]) is seen in the dorsal anterior hyoid arch adjacent to where the first pouch would be in wild-type animals. In mutant sections (B′), TUNEL-positive cells (arrow) colocalize with the fli1 -GFP crest marker. (E) The number of apoptotic nuclei per hyoid arch is plotted versus time for wild-type sides (blue) and integrinα5 − sides without (p1−; red) or with (p1+; green) a normal first pouch. At 25 hpf, integrinα5 − hyoid arches had more apoptotic nuclei than wild-type hyoid arches only when the first pouch was defective ( p < 0.0001). At later time points, integrinα5 − hyoid arches missing the first pouch had a tendency to have more apoptotic nuclei than wild-type or integrinα5 − arches with normal first pouches (only itga5 − with a normal first pouch versus itga5 − without at 35 hpf is statistically significant, p < 0.05). Total sides examined: 25 hpf: n wt = 40, n itga5 = 38; 29 hpf: n wt = 30, n itga5 = 26; 30 hpf: n wt = 30, n itga5 = 20; and 35 hpf: n wt = 30, n itga5 = 14. (F and G) gsc expression at 36 hpf labels dorsal and ventral domains of hyoid crest. Mandibular (1) and hyoid (2) arches are numbered, and the first pouch is denoted by arrowhead. In wild-type animals, dorsal and ventral hyoid gsc domains are well separated. In this integrinα5 − animal, dorsal and ventral hyoid gsc domains are fused, and disorganized gsc -expressing cells envelop the reduced first pouch (arrowhead). Scale bars: 50 μm. We also examined whether hyoid crest was correctly specified in integrinα5 mutant embryos. Hyoid crest expresses Hox class 2 genes, whereas mandibular crest is Hox nonexpressing ( Hunt et al. 1991 ). No defects were seen in the expression of hoxa2 in the hyoid arches of integrinα5 mutants at 36 hpf (data not shown). In 36-hpf wild-type animals, goosecoid (gsc) expression marks dorsal and ventral domains within the hyoid arch ( Figure 6 F). In integrinα5 − embryos, gsc domains were present, although in 12% of mutants they were variably disorganized. In the example shown in Figure 6 G, the dorsal hyoid gsc domain was disorganized and fused to the ventral hyoid gsc domain. However, as the gsc defects were of significantly lower penetrance than the cartilage defects, we conclude that the majority of specific cartilage defects seen in integrinα5 − embryos are not due to altered gsc expression. A Subset of Hyoid Crest Shows Aberrant Behavior and Does Not Contribute to Cartilage in integrinα5 Mutants The first pharyngeal cartilages begin to chondrify around 48 hpf ( Schilling and Kimmel 1997 ). In order to understand neural crest cell behavior during cartilage formation in wild-type animals, we used the fli1 -GFP line to make extended time-lapse recordings of hyoid arch development that began at 38 hpf and ended at 86 hpf, an endpoint when cartilage elements are readily identifiable ( Figure 7 ). In one focal plane, the SY region was observed to form from tightly packed cells adjacent to the ventral tip of the first pouch at 38 hpf ( Video S1 ; Figure 7 A– 7 F). In another focal plane, aHM was observed to form from a tightly packed mass of fli1 -GFP-labeled cells located directly adjacent to the first pouch in the dorsal, anterior portion of the hyoid arch at 38 hpf ( Video S2 ; Figure 7 G– 7 L). We found that crest cells that contributed to aHM remained fairly static during the period of observation, though local rearrangements that contribute to the flattening of the HM plate were not analyzed in detail here (enlarged in Figure 7 G′–7L′). The pHM region formed from cells located posterior to the aHM mass at 38 hpf. In general, our time-lapse recordings of wild-type hyoid development supported and extended the conclusions generated from the 24-hpf fate map. In the hyoid arch, crest cells that will contribute to the aHM and SY regions are tightly packed masses directly adjacent to the first pouch prior to chondrogenesis. Figure 7 Anterior Hyoid Crest Cells Display Aberrant Behavior in integrinα5 Mutants Confocal time-lapse recordings show hyoid cartilage development in wild-type fli1 -GFP (Videos S1 and S2 ) and integrinα5 − ; fli1 -GFP ( Video S3 ) animals from 38 hpf to 86 hpf ( n wt = 3; n itga5 = 4). Videos S1 and S2 are different depths of the same time-lapse recording. Representative imaging stills of Video S1 (A–F), Video S2 (G–L), and Video S3 (M–R) were taken at 38 hpf (A, G, and M), 44 hpf (B, H, and N), 50 hpf (C, I , and O), 56 hpf (D, J, and P), 62 hpf (E, K, and Q), and 86 hpf (F, L, and R). At the beginning of the recordings (A, G, and M), the mandibular (1) and hyoid (2) arches are numbered and an arrow denotes the first pouch (p1). At the end of the recordings (F, L, and R), the cartilage regions are clearly visible as large cells with thick matrix (pseudocolored blue). The outline of the HS cartilage, a composite of SY and HM regions, is shown in (F) and (L). As a reference, the opercle bone and ao/lo hyoid muscle mass are pseudocolored purple and red, respectively, and the eye and ear are labeled. In Video S1 (A–F), red arrowheads denote a cluster of cells adjacent to the first pouch that undergo cellular rearrangements and form the long, anterior SY extension in wild-type animals. (G′–R′) show magnifications of HM-forming regions taken from (G–R) and correspond to areas within white boxes given in (G) and (L) for (G–L) and in (M) and (R) for (M–R). In wild-type development, hyoid crest cells adjacent to the first pouch remain a tightly packed mass as aHM chondrifies (e.g., cells denoted by red arrowheads in G′–L′). In integrinα5 mutants, the first pouch is missing (white arrow in [M]), and anterior hyoid crest cells are disorganized at 38 hpf (e.g., arrowhead in [M′]). Over time, anterior hyoid crest cells migrate out of the region and do not contribute to cartilage (e.g., arrowheads in [N′–Q′]). In contrast, the pHM region and the opercle bone develop normally from more posterior hyoid crest in integrinα5 − animals (R). Scale bar: 50 μm. In order to understand the cellular basis for the losses of the aHM and SY regions in integrinα5 mutants, we made time-lapse recordings of hyoid crest development in integrinα5 − ; fli1 -GFP embryos from 38 hpf to 86 hpf ( Video S3 ; Figure 7 M– 7 R). Whereas in wild-type animals anterior hyoid crest cells were tightly packed masses adjacent to the first pouch at 38 hpf, in integrinα5 − ; fli1 -GFP embryos, crest cells in the dorsal, anterior portion of the hyoid arch were more loosely packed ( Figure 7 M′). Strikingly, over the next day the crest-derived cells migrated out of the dorsal, anterior region of the mutant hyoid arch (enlarged in Figure 7 N′–7Q′). By 86 hpf, the pHM cartilage region had formed, yet no fli1 -GFP-positive cells were seen anterior to pHM ( Figure 7 R). Thus, we found a strong correlation between the lack of compaction and stabilization of dorsal, anterior hyoid crest cells and the loss of the aHM cartilage region in integrinα5 − ; fli1 -GFP embryos. Discussion Isolation of a Zebrafish integrinα5 Mutant In this work, we isolated and characterized a zebrafish mutant allele (b926) that has variably penetrant and expressive reductions of the first pouch and hyoid aHM and SY cartilage regions. By positional mapping, allele segregation, and morpholino phenocopy, we identified the genetic basis of the lesion as a missense mutation in the ligand-binding domain of Integrinα5. As similar pharyngeal phenotypes were observed in mutant and morpholino-treated animals, we conclude that b926 is a loss-of-function allele of integrinα5 . However, we do not know if Integrinα5 activity is completely eliminated in b926 . In addition, we observed strong maternal expression of integrinα5 that could mitigate the zygotic loss of integrinα5 in b926. Thus, the variable penetrance and expressivity of the integrinα5 phenotype could be due to partial activity of mutant Integrinα5 or the presence of maternally supplied Integrinα5. Additionally, other integrins may act redundantly with Integrinα5 in pharyngeal development. A survey of nearly finished genome sequence ( http://www.ensembl.org/Danio_rerio/ ) has uncovered at least 15 additional Integrin α chains, for which no expression or phenotypic data are known in zebrafish. Lastly, integrinα5 is expressed strongly in many tissues, such as the otic placode, for which no overt phenotypes were observed in b926 . Future studies, in particular those using animals in which both maternal and zygotic integrinα5 have been eliminated, may reveal new functions of Integrinα5 in zebrafish development. The First Pouch Is Required for the Development of a Subset of Hyoid Cartilage Our results point to an important role for the first pouch in the development of specific hyoid cartilage regions. We have used the incomplete penetrance of the pouch and cartilage phenotypes of integrinα5 − animals to show that early first pouch defects are strongly predictive of later hyoid cartilage defects. Furthermore, transplantation experiments show that wild-type endoderm, but not crest, rescues first pouch and hyoid cartilage development in integrinα5 mutants. We infer that Integrinα5 functions in the pharyngeal endoderm for the formation of the first pouch, and that the first pouch, in turn, interacts with postmigratory neural crest to promote cartilage development in a region of the hyoid arch. A role for the first endodermal pouch in promoting regional hyoid cartilage development is consistent with work in chicken showing that domains of pharyngeal endoderm specify region-specific cartilage shapes ( Couly et al. 2002 ; Ruhin et al. 2003 ). Our data extend these findings, arguing that the formation of the first pouch is a critical step in allowing pharyngeal endoderm to interact with hyoid crest and promote the development of specific cartilage regions, aHM and SY. It will be interesting to see the extent to which the ability of different pharyngeal endoderm domains to induce cartilage elements of specific shapes depends on their ability to form discrete morphological structures such as the first pouch. How might the first pouch control development of specific cartilage regions within the hyoid arch? The first pouch forms at a time when hyoid crest cells are migrating to ventrolateral positions to form the hyoid arch ( Veitch et al. 1999 ; J.G.C. and C.B.K., unpublished data). Upon reaching the developing arch, crest cells become less motile and form tightly packed masses adjacent to the first pouch. Our wild-type fate map shows that the hyoid cartilage regions that are lost in integrinα5 mutants, aHM and SY, develop from crest cells immediately adjacent to the first pouch ( Figure 8 A). Our time-lapse recordings of wild-type cartilage development show that crest cells that will form aHM remain a tightly packed mass as the aHM region chondrifies ( Figure 8 B and 8 C). In contrast, in integrinα5 mutants, dorsal anterior hyoid crest cells are aberrantly motile and do not contribute to cartilage, whereas more posterior dorsal hyoid crest cells contribute normally to pHM ( Figure 8 D– 8 F). In addition, we observe increased apoptosis in the dorsal, anterior domain of integrinα5 − hyoid arches from 25 to 35 hpf. Importantly, increased death of postmigratory hyoid crest cells was seen only when the first pouch was reduced. It will be interesting to examine whether the increased apoptosis observed in the dorsal anterior hyoid arches of integrinα5 −/− mice ( Goh et al. 1997 ) is a secondary consequence of a missing first pouch as well. In contrast to integrinα5 −/− mice, in which an increase in cell death was seen earlier during crest migration, we found no evidence for increased death of migratory crest in integrinα5 − zebrafish. However, our analysis cannot rule out that increased crest death during migratory stages may contribute to the infrequent, most severe cartilage phenotypes seen in integrinα5 − embryos. Indeed, given the strong expression of integrinα5 in migratory hyoid crest, future studies that further reduce Integrinα5 activity, for example by removing its maternal component, may uncover crest-autonomous functions of zebrafish Integrinα5 in the survival and/or migration of hyoid crest cells. Figure 8 Model for Development and Evolution of Hyoid Cartilage (A–F) Models of hyoid development in wild-type (A–C) and integrinα5 − (D–F) animals show the structure of hyoid arches at 24 hpf (A and D) and 38 hpf (B and E) and mandibular and hyoid cartilages at 4 d (C and F). (A) At 24 hpf of wild-type development, crest that will form aHM (dark green), pHM (medium green), and SY (light green) cartilage regions occupy distinct domains within the hyoid arch. Signals (red arrows) from the first pouch (orange) stabilize adjacent aHM- and SY-producing crest. (B) At 38 hpf of wild-type development, aHM- and SY-producing crest tightly pack along the first pouch. Cranial mesoderm (red) and some mandibular crest (blue) are also shown. (C) At 4 d of wild-type development, the HS cartilage is a composite of aHM, pHM, and SY regions. Also shown are the hyoid CH (yellow) and mandibular Meckel's (light blue) and palatoquadrate (dark blue) cartilages. (D) In integrinα5 − animals, the first pouch is missing or very reduced at 24 hpf. (E) By 38 hpf, as a consequence of the lack of a first pouch, aHM and SY progenitors are disorganized and undergo gradual apoptosis. In contrast, the development of pHM progenitor cells does not require the first pouch. (F) At 4 d, aHM and SY cartilage regions are selectively reduced in integrinα5 − animals. (G–I) The HS element has undergone extensive change during vertebrate evolution. In the illustrations (adapted from De Beer [1937] ), the neurocranium is grey or outlined in black and mandibular and hyoid cartilages are color-coded as described above. Based on relations to morphological landmarks and data presented here on the tripartite mosaic development of HS, an evolutionary scheme is proposed. (G) In the dogfish shark Scyliorhinus canicula, a single rod-shaped element corresponds to pro-aHM/SY regions. (H) In the basal actinopterygian Polypterus senegalus, separate aHM and SY regions are present. (I) As shown for salmon, during actinopterygian evolution a new region, pHM, develops posterior to and fuses with aHM to create a wide HM plate that articulates with the neurocranium and supports an enlarged, overlying opercular apparatus (not shown). Our data show that the first pouch is required for the stabilization and survival of crest cells that will become aHM and SY. Interactions between the first pouch and adjacent hyoid crest could involve direct adhesion and/or diffusible signaling molecules. aHM and SY cartilage regions develop from crest-derived cells immediately adjacent to the first pouch at 24 hpf, whereas the pHM region develops from crest three cell diameters away. The remarkable specificity of the integrinα5 cartilage phenotype suggests that pouch-derived signals act very locally, perhaps through cell–cell contract, to promote development of aHM and SY regions. In support of this, explant studies in the newt show that physical contact between pharyngeal endoderm and neural crest cells is necessary to promote the compaction and differentiation of crest into cartilage ( Epperlein and Lehmann 1975 ). On the other hand, a signaling role for pharyngeal endoderm in crest survival also has been shown. In zebrafish, Fgf3 produced from the pharyngeal pouches acts as a secreted survival factor for neural crest ( David et al. 2002 ; Nissen et al. 2003 ). In addition to promoting stabilization and survival, might endoderm also control local gene expression in hyoid crest? In a small fraction of integrinα5 − embryos, gsc expression domains in the hyoid arches were present but disorganized. However, it is possible that the aberrant gsc expression reflects a disorganization of the hyoid arch and not ectopic gene expression. Although disorganized gsc expression may correlate with more severe integrinα5 phenotypes such as hyoid cartilage fusions (see Figure 2 D), the significantly lower penetrance of gsc phenotypes compared to cartilage phenotypes implies that disorganized gsc expression is not the major cause of the specific cartilage defects. In addition, hoxA2 expression was unaffected in the hyoid arches of integrinα5 − embryos. Thus, although additional markers of hyoid crest need to be examined in integrinα5 − embryos, we have found no strong evidence for the first pouch controlling gene expression in neighboring crest. Instead, our data argue that the first pouch locally controls cartilage development by promoting the compaction and survival of immediately adjacent crest-derived cells. Integrin-Mediated Outgrowth of the First Pharyngeal Pouch We have shown that zebrafish integrinα5 is expressed in pharyngeal endoderm during pouch formation and is required in the endoderm for development of the first pouch. The specificity for the first pouch of the integrinα5 phenotype could be due to either redundancy with other integrins that function preferentially in posterior pouches or greater sensitivity of the first pouch to loss of integrin function. Although the most common phenotype in integrinα5 mutants is loss of just the first pouch, we do occasionally see reductions of more posterior pouches as well, suggesting that Integrinα5 functions in the formation of most or all pouches. Moreover, as the more posterior pouches are required to segment the posterior crest mass into the five branchial arches from which the CB cartilages develop ( Piotrowski and Nusslein-Volhard 2000 ), the variable loss of these pouches likely explains the reductions and rare fusions of CB cartilages seen in some integrinα5 − animals. How might Integrinα5 control pouch formation? The elaboration of a relatively uniform tissue into an organ of more complex curvature and ramifications, termed branching morphogenesis, is a common developmental program in both vertebrates and invertebrates. The formation of an iterative series of outpocketings, or pouches, from the pharyngeal endoderm can be thought of as analogous to branching morphogenesis. Integrins have well-documented roles in cell migration that could promote the outgrowth of branches (reviewed in Bokel and Brown 2002 ). From our unpublished observations in zebrafish, we know that pouches form by the directed lateral migration of pharyngeal endodermal cells (unpublished data). In this work, we find that integrinα5 is expressed transiently in pouch-forming regions of pharyngeal endoderm and is required in endoderm for pouch formation. One possibility is that Integrinα5 adhesion promotes the lateral migration of endodermal cells to form pouches. Alternatively, Integrinα5 may be required for the specification or survival of pharyngeal endoderm that forms pouches. Future time-lapse imaging studies, in which pharyngeal endoderm morphogenesis is analyzed directly in integrinα5 − embryos, will help to clarify the role of Integrinα5 in pouch formation. A Hierarchy of Tissue Interactions Control Regional Development in the Pharyngeal Arches The exquisite functionality of the vertebrate jaw and pharynx requires the precise developmental coordination of their component parts. Arch-specific patterns of muscle connect with pharyngeal skeletal elements and are innervated by motor neurons of appropriate axial levels to orchestrate behaviors such as feeding and gill pumping. In integrinα5 mutants, we see specific defects not only in the endodermal pouches and crest-derived cartilages, but also in cranial muscles and their associated motor nerves (schematized in Figure 2 R). Both a dorsal mandibular (do) and a dorsal hyoid (ah) muscle are reduced in integrinα5 mutants, and facial nerve VII, which innervates ah and other hyoid muscles, is reduced and/or fails to make a characteristic branch into two main fascicles. However, it is likely that muscle and nerve defects in integrinα5 mutants are secondary to pouch and cartilage defects. Whereas integrinα5 is expressed in endoderm and crest during pharyngeal morphogenesis, we were unable to detect integrinα5 expression in cranial mesoderm or hindbrain neurons during axon outgrowth. In addition, muscle and nerve defects in integrinα5 mutants were of significantly lower penetrance than the first pouch and hyoid cartilage defects. The low penetrance of the muscle and nerve defects might be explained by the variably expressive loss, in integrinα5 mutants, of the pouch- and/or crest-derived signals on which muscle and nerve development depend. Unfortunately, due to the low penetrance of both muscle and nerve defects in integrinα5 mutants, we were unable to directly test the tissue autonomy of these defects. Our analysis of the integrinα5 mutant does not distinguish between roles for endoderm and crest in patterning cranial muscle and nerves. The mesodermal cores that give rise to do and ah, the muscles affected in integrinα5 − animals, lie close to and on opposite sides of the first pouch during pharyngeal arch development. The first endodermal pouch could have an early organizing role for cranial mesoderm. However, increasing evidence suggests that crest has a major role in patterning cranial mesoderm. Analysis of the chinless mutation in zebrafish has shown that chinless functions nonautonomously within the crest to promote muscle development ( Schilling et al. 1996 ). In classic experiments in chicken, grafting of mandibular crest into more posterior arches can reprogram both skeletal and muscular fates ( Noden 1983 ), though recent work suggests this effect is mediated by an isthmus-organizing activity included in the grafts ( Trainor et al. 2002 ). In the larval zebrafish, do and ah are found in close association with the aHM region that is lost in integrinα5 mutants. It is possible that loss of dorsal anterior hyoid crest in integrinα5 mutants results in reductions not only of the aHM cartilage region but also of crest-derived signals that support development of do and ah muscles. As has been proposed by others ( Noden 1983 ; Kontges and Lumsden 1996 ), the development of cranial muscles may depend less on their arch origin and more on the crest-derived structures, such as the aHM cartilage region and associated connective tissue, onto which they attach. Likewise, the reduction of facial nerve VII in integrinα5 mutants could be due to either reductions in hyoid muscles and their associated survival signals or reductions in nerve outgrowth–promoting cues normally produced by the crest and/or endoderm. In conclusion, we see evidence for a local hierarchy of interactions that control the development of a specific region of the head encompassing dorsal hyoid and mandibular elements. At the top of the hierarchy is the endoderm-derived first pouch that promotes the development of a subset of hyoid crest into cartilage; in turn, this subset of hyoid crest may control development of neighboring muscles and, directly or indirectly, their associated nerve. Mosaic Assembly of Hyoid Cartilage: Implications for Evolution The shape of the dorsal hyoid cartilage element has undergone extensive modification during actinopterygian evolution. In sharks ( Figure 8 G) and basal ray-finned fishes such as the bichir, Polypterus senegalus ( Figure 8 H), HM is a rod and SY is absent or not well elongated ( De Beer 1937 ). In teleosts, highly derived ray-finned fish, the dorsal hyoid cartilage consists of a wider HM plate and a long SY extension ( Figure 8 I). The elaboration of HM and SY regions before teleosts emerged may have served to more efficiently support the jaw and increase gill pumping. The origin of the HM plate has long been a subject of debate. Allis (1915) proposed that the teleost HM plate consists of two regions that become fused together, whereas Edgeworth (1926) , based on his staging series of the bowfin Amia, a relative of teleosts, concluded that the HM plate develops from a single anterior region that undergoes posterior growth to form a plate. There is ample precedence for differential growth as a mechanism of morphological change. For example, beautiful interspecies mosaic experiments have shown that the difference in beak length between ducks and quails is due to an autonomous growth potential of mandibular crest ( Schneider and Helms 2003 ). However, our data support the composite two-region HM theory of Allis. We see a clear genetic dissociation between the development of aHM and pHM. Whereas aHM is absent in the majority of integrinα5 mutants, pHM and the connecting opercle bone are still present even in the most severe class of integrinα5 mutants. In addition, our fate mapping data show that, although aHM and pHM form a seamless HM plate in the larva, their progenitor cells occupy distinct, albeit contiguous, domains within the hyoid arch at 24 hpf, a result inconsistent with the posterior growth hypothesis of Edgeworth. Thus, aHM and pHM regions develop from spatially distinct domains of crest that depend on different sources of inductive signals. Our data show that the first endodermal pouch is required for the development of aHM, yet in mutants that lack all pharyngeal endoderm, such as casanova, both aHM and pHM are lost ( David et al. 2002 ). Thus, other structures of the pharyngeal endoderm besides the first pouch may be required for the development of pHM. One attractive possibility is that, whereas the first pouch induces aHM in the anterior part of the hyoid arch, the second pouch induces pHM in the posterior part of the arch. Allis (1915) concluded, based on relationships to morphological landmarks, that the rod-shaped HM cartilage in Polypterus represents the anterior portion of the teleost HM plate (i.e., aHM). Although more embryological studies of Polypterus need to be done, the evolution of the HM plate in ray-finned fishes such as teleosts appears to have involved a new induction event that led to a new region, pHM, being added to an older region, aHM. We propose that the de novo addition of regions to the skeletal pattern represents another mechanism, in addition to differential growth, of generating skeletal diversity during evolution. Materials and Methods Zebrafish strains and mutant screen. Zebrafish (Danio rerio) raised at 28.5 °C were staged as previously described ( Kimmel et al. 1995 ; Westerfield 1995 ). The wild-type line used was AB. fli1 -GFP albino transgenic fish are the same as TG(fli1:EGFP) y1 ; alb b4 ( Lawson and Weinstein 2002 ), and islet1 -GFP fish are as described ( Higashijima et al. 2000 ). For the cartilage screen, ENU-mutagenized F2 parthenogenic diploid fish were generated by early pressure treatment ( Streisinger et al. 1981 ; Solnica-Krezel et al. 1994 ) and fixed and stained with Alcian at 4 d. The b926 allele was outcrossed to the AB strain and subsequently crossed onto the fli1 -GFP and islet1 -GFP backgrounds. integrinα5 identification and morpholino. b926 was mapped with respect to polymorphic microsatellites based on the hyoid cartilage phenotype. Initial mapping was performed on an AB background, and fine mapping was performed on an islet1 -GFP background selected for a high degree of LG23 polymorphisms with respect to AB. Full-length integrinα5 cDNA was obtained by 5′- and 3′-RACE. Standard molecular biological techniques were used. For genotyping, primers were designed to turn b926 into a codominant polymorphism digestible with XmnI (GC156, TGACTGTGACCTTCAGCTCAATGTAAACGC; GC158, TGGATCTGGCCACCCACTGAGGTCGAAAAG). A morpholino (Genetools, Philomath, Oregon, United States) was designed against the exon13-intron splice site of integrinα5 ( itga5 -MO) with the following sequence: ATGCTTTCTCACCTGGGTAGCCATT. Embryos were pressure-injected with 5 nl of 2 mg/ml itga5 -MO as previously described ( Maves et al. 2002 ). Phenotypic analysis. Alcian Green staining was performed as described ( Miller et al. 2003 ). For flat mount dissections, Alcian-stained animals were digested for 1 h in 8% trypsin at 37 °C and transferred to 100% glycerol. Cartilages were dissected free from surrounding tissues with fine stainless-steel insect pins and photographed using a Zeiss (Oberkochen, Germany) Axiophot 2 microscope. Image background was cleaned up with Adobe Photoshop. For immunocytochemistry, embryos were prepared as described ( Maves et al. 2002 ). Antibodies were used at the following dilutions: rabbit anti-GFP, 1:1,000 (Molecular Probes, Eugene, Oregon, United States); Zn-8, 1:400 ( Trevarrow et al. 1990 ); MF-20, 1:10 (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, Iowa, United States); goat anti-rabbit Alexa Fluor 488 and anti-mouse Alexa Fluor 568 (Alexa568), both 1:300 (Molecular Probes). TUNEL staining was performed on embryos that were fixed overnight in 4% PFA, MeOH-permeabilized for 20 min, rehydrated, and treated with ProteinaseK (Sigma, St. Louis, Missouri, United States) for 2–20 min at room temperature. TdT/Fluorescein-dUTP reaction (Roche, Basel, Switzerland) was performed for 1 h on ice, followed by 1 h at room temperature. After labeling with Fluorescein-dUTP, immunocytochemistry was performed using rabbit anti-Fluorescein F ab fragments (1:20,000, Molecular Probes) and goat anti-rabbit Alexa Fluor 568 antibodies (1:200, Molecular Probes). GFP fluorescence survived the procedure. Probe syntheses and whole-mount in situ hybridizations were performed as previously described ( Westerfield 1995 ). Embryos were mounted in glycerol and photographed using a Zeiss Axiophot 2 microscope. integrinα5 RNA probes were made from plasmids pINT4150 and pINT4853, constructed by inserting RT-PCR fragments corresponding to nucleotides 254–2,207 and 1,679–2,960 of the integrinα5 cDNA, respectively, into the TA vector (Invitrogen, Carlsbad, California, United States). Plasmids were linearized with BamHI, and T7 RNA polymerase was used for probe synthesis. Both probes gave identical expression patterns, and pINT4853 was used for photographs. pea3 ( Brown et al. 1998 ) and gsc ( Schulte-Merker et al. 1994 ) probes were prepared as described, and mutant embryos were PCR genotyped. Single-cell microelectroporation. The microelectroporation technique was similar to that described by Lyons et al. (2003) . fli1 -GFP embryos, 24 hpf, were dechorionated, anesthetized with tricane solution, and bathed in a solution of 5 mg/ml pronase (Sigma) for 1 min to allow passage of the microelectrode through the skin. Agar mounting of embryos on slides was performed as described in Westerfield (1995) . Under 50× Nomarski optics a micropipette filled with Alexa Fluor 568 dextran amines (Molecular Probes) was positioned next to the cell of interest, and a ground electrode was placed in the bath next to the embryo. Pulses of current between 1 and 4 uA were used to mobilize the dye. Shortly after electroporation, the location of labeled cells relative to the fli1 -GFP-expressing hyoid arch was assessed using confocal microscopy. Only embryos with one or two adjacently labeled cells were used in the analysis. Three-dimensional projections were constructed to determine the position of labeled cells in the arch relative to landmarks. All cell distances were made from the mid point of the cell to the landmark. Distance of the labeled cell from the edge of the first pouch was used to determine A-P position. Similarly, distances of labeled cells from the dorsal and lateral edges of the arch were used to determine D-V and mediolateral positions. In electroporations where two adjacent cells were labeled, the positions of each cell were measured and averaged. To control for variation in arch dimensions among individuals, measurements along the three axes were normalized to total axes lengths. At 4 d, embryos were imaged again to determine the fate of labeled cells in hyoid cartilage. The aHM region was defined as the anterior portion of HM that is characteristically lost in integrinα5 mutants. pHM comprises the rest of the HM region. The SY region begins at the point of attachment to pHM. All graphing and statistical analysis were done using JMP (2002, SAS Institute, Cary, North Carolina, United States). Time-lapse analysis and confocal imaging. Embryos were manually dechorionated, anesthetized with tricane solution, transferred to 0.2% agarose in embryo media with 10 mM HEPES and tricane, and then mounted onto a drop of 3% methylcellulose on a rectangular coverslip with three superglued #1 square coverslips on each side. A ring of vacuum grease was added around the embryo to make an airtight seal upon addition of the top coverslip. A heated stage kept the embryos at 28.5 °C. Approximately 80-μm Z-stacks at 2-μm intervals were captured every 10 min using a Zeiss LSM 5 Pascal confocal fluorescence microscope. Movies of individual Z-sections were made by manually following cells and concatenating sections; further processing was done with Adobe Premiere. For single time point confocal sections, embryos were mounted without vacuum grease. Endoderm and crest transplants. Transplant techniques were as described ( Maves et al. 2002 ). For endoderm transplants, donor embryos were injected at the one-cell stage with a mixture of 2% Alexa Fluor 568 dextran and 3% lysine-fixable biotin dextran (10,000 MW, Molecular Probes) along with TAR* RNA prepared according to David et al. (2002) . At 40% epiboly (ca. 4 hpf), donor TAR* tissue was moved to the margins of fli1 -GFP host embryos. For the crest transplants, donor fli1 -GFP embryos were injected at the one-cell stage with the Alexa Fluor 568 mixture (“Alexa568”). At shield stage (ca. 6 hpf), donor tissue was taken from the animal cap and moved to a position approximately two germ ring widths from the margin and 90° from dorsal in a fli1 -GFP host embryo. Confocal images of host embryos were captured at 38 hpf and 4 d, and embryos were subsequently fixed and Alcian-stained to visualize cartilages. For endoderm transplants, only embryos in which donor tissue contributed to the second and at least one other pouch were included in the analysis. For crest transplants, only embryos in which at least half the hyoid arch was composed of donor tissue were included in the analysis. Supporting Information Video S1 Wild-Type Development of SY Cartilage Confocal time-lapse recording shows hyoid cartilage development in a wild-type fli1 -GFP animal from 38 hpf to 86 hpf. At the beginning of the video, the mandibular (1) and hyoid (2) arches are numbered and an arrow denotes the normal position of the first pouch. At the end of the video (see Figure 7 for representative still images), the SY cartilage is pseudocolored blue. A red arrowhead denotes a cluster of cells adjacent to the first pouch that undergo cellular rearrangements and form the long anterior SY extension in wild-type animals. (6.0 MB MOV). Click here for additional data file. Video S2 Wild-Type Development of HM Cartilage Confocal time-lapse recording shows hyoid cartilage development in a wild-type fli1- GFP animal from 38 hpf to 86 hpf. This video is a different depth of the same time-lapse recording as Video S1 . At the beginning of the video, the mandibular (1) and hyoid (2) arches are numbered and an arrow denotes the normal position of the first pouch. At the end of the video (see Figure 7 for representative still images), the HM cartilage is pseudocolored blue, and the ao/lo muscle quadrant and opercle bone are pseudocolored red and purple, respectively, for reference. A red arrowhead points to hyoid crest–derived cells immediately adjacent to the first pouch that give rise to the aHM cartilage region in wild-type animals. (7.1 mB MOV). Click here for additional data file. Video S3 Development of HM Cartilage in integrinα5 Mutants Confocal time-lapse recording shows hyoid cartilage development in an integrinα5 − ; fli1 -GFP animal from 38 hpf to 86 hpf. At the beginning of the video, the mandibular (1) and hyoid (2) arches are numbered and an arrow denotes the normal position of the first pouch. At the end of the video (see Figure 7 for representative still images), the HM cartilage is pseudocolored blue, and the ao/lo muscle quadrant and opercle bone are pseudocolored red and purple, respectively, for reference. A red arrowhead points to hyoid crest cells that display increased motility in integrinα5 − animals and do not contribute to cartilage as in wild-type. (6.0 KB MOV). Click here for additional data file. Accession Numbers The Genbank ( http://www.ncbi.nlm.nih.gov ) accession number for the zebrafish integrinα5 cDNA is AY550244. Accession numbers for the related Integrinα5 proteins described in Figure 1 are Homo sapiens (Genbank P08648), Mus musculus (Genbank P11688), Xenopus laevis (Genbank Q06274), and Fugu rubripes (manually assembled from Fugublast M002304).	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC479042.xml
548506	A novel heterozygous missense mutation in the UMOD gene responsible for Familial Juvenile Hyperuricemic Nephropathy	Background Familial Juvenile Hyperuricemic Nephropathy is an autosomal dominant nephropathy, characterized by decreased urate excretion and progressive interstitial nephritis. Mutations in the uromodulin coding UMOD gene have been found responsible for the disease in some families. Case presentation We here describe a novel heterozygous p.K307T mutation in an affected female with hyperuricemia, renal cysts and renal failure. The proband's only son is also affected and the mutation was found to segregate with the disease. Conclusions This mutation is the fourth reported in exon 5. Initial studies identified a mutation clustering in exon 4 and it has been recommended that sequencing this exon alone should be the first diagnostic test in patients with chronic interstitial nephritis with gout or hyperuricemia. However, regarding the increasing number of mutations being reported in exon 5, we now suggest that sequencing exon 5 should also be performed.	Background Familial Juvenile Hyperuricemic Nephropathy (FJHN) (McKusick 162000) is an autosomal dominant disorder characterized by hyperuricemia, decreased urinary excretion of urate and the development of progressive chronic interstitial nephritis. Renal impairment usually appears between 15 and 40 years of age, leading to end-stage renal disease (ESRD) within 10 to 20 years [ 1 ]. A candidate gene for FJHN has been positioned in 16p11.2-12, together with evidence for genetic heterogeneity [ 2 , 3 ]. The candidate gene was later found to map within the same genetic interval as MCKD2, a locus responsible for medullary cystic kidney disease, therefore suggesting that FJHN and MCKD2 are 2 facets of the same disease [ 1 ]. The marked thickening of tubular basement membranes observed in FJHN closely resembles the renal histological findings of the nephronophthisis-medullary cystic kidney disease complex (NPH-MCKD). Diseases of this group share the macroscopic feature of cyst development at the corticomedullary border of the kidney and the renal histological triad of tubular basement membrane disintegration, tubular atrophy with cyst development and interstitial cell infiltration with fibrosis [ 4 ]. Within this complex, clinical entities can be distinguished based on the mode of inheritance, the age of onset for ESRD and the presence of extra-renal involvement. For the recessive forms of the disease 4 different genes have been cloned. The NPHP1 gene [ 5 , 6 ] and NPHP4 [ 7 ] are responsible for juvenile forms of NPH, while NPHP2 [ 8 ] and NPHP3 [ 9 ] account for, respectively, the infantile and adolescent forms. MCKD, the autosomal dominant disorder that presents in early adulthood, is usually accompanied by the detection of corticomedullary cysts on imaging studies. MCKD1 maps to 1q21 and remains to be cloned, while MCKD2 has been positioned in 16p12. Mutations in the uromodulin/Tamm-Horsfall protein coding UMOD gene located within the critical interval of FJHN and MCKD2 at 16p11.2-12 were recently identified in FJHN and MCKD2 families [ 10 ], therefore providing definite evidence that MCKD and FJHN are allelic disorders. Meanwhile, a mutation cluster in exon 4 of UMOD was reported for both diseases [ 11 , 12 ]. We here describe a novel heterozygous missense mutation in affected individuals from a Portuguese FJHN family that also displays corticomedullary cysts on ultrasound examination. This mutation, c.920A→C, resides in exon 5 and is the fourth reported outside exon 4. Case presentation Case report The proband is an affected female who was first evaluated at the age of 24, when she presented with a gout attack and hyperuricemia. At age 27 she was told having renal failure. However, a renal biopsy was not performed. At age 44, serum creatinine was 2.8 mg/dl and ultrasound imaging detected numerous renal cysts. Renal disease slowly progressed and the patient reached ESRD when she was 49 years old. Her father died at the age of 55 from ESRD and suffered from hyperuricemia and gout. The proband's only son is also affected. At the age of 18 years he had a gout attack. On the initial evaluation, serum uric acid was 15 mg/dl and serum creatinine 1.6 mg/dl. Renal cysts were also detected on ultrasound examination. Mutation analysis Informed consent was obtained from tested individuals. Genomic DNA was isolated from peripheral blood leucocytes and the coding region of the UMOD gene was screened for mutations by direct sequencing of PCR products. We used a set of primers previously described [ 10 ] except in exons 4 and 5, for which different additional internal sequencing primer were designed based upon sequences from GenBank (accession numbers NT_024776.6 and M17778). Results A novel heterozygous missense mutation, c.920A→C, was detected (Figure 1 ) and found to segregate with the disease in this family. The mutation results in a Lys to Thr at position 307. This mutation was not detected in any of the 100 control chromosomes tested. In fact, no polymorphism affecting the translation of uromodulin was detected in 100 control chromosomes in a previous report [ 10 ] and we are, therefore, excluding the possibility of this allele being a mere polymorphism. In addition, the affected mother was found to be homozygous for the T allele in the common T to C transition synonymous polymorphism at codon C174. Discussion The UMOD gene has 12 exons and codes for the 640 amino-acid uromodulin, a glycsoyl phosphatidylinositol (GPI) anchored protein that accounts for the primary structure of the 85-kD Tamm-Horsfall glycoprotein (THP). THP is the most abundant protein in the urine and, in normal kidney, uromodulin expression is restricted to the thick ascending limb (TAL) and distal convoluted tubule. Urinary excretion occurs by proteolytic cleavage of the GPI counterpart at the luminal surface of TAL. It has been suggested that one of THP major role is of an urinary anti-adherence factor preventing type 1 fimbriated E coli from binding to the urothelial receptors [ 13 ]. The majority of the mutations so far published are clustered in exon 4, between codons 52 and 282, and most are missense mutations affecting cysteine residues (Table 1 ). Exon 4 contains 3 calcium binding epidermal growth factor (cbEGF)-like domains, between residues 31 and 148. A fourth potential cbEGF-like domain extends from amino-acids 281–336, throughout exon 5. They contain 6 conserved cysteine residues responsible for the protein's tertiary structure, as a result of intramolecular disulfide bonding. It has been hypothesized that protein misfolding, consequence of mutations in these cbEGF-like domains, may affect uromodulin intracellular trafficking and lead to cellular protein accumulation and apoptosis [ 14 ]. The release of cells debris and uromodulin aggregates in the interstitium could stimulate an inflammatory response and, in addition, be responsible for tubular obstruction and medullary cyst formation. It has been proposed that hyperuricemia in these patients is secondary to a reduced TAL sodium reabsorption with volume contraction and a compensatory increase in proximal urate reabsorption. The role of hyperuricemia in the chronic interstitial nephritis remains to be clarified, since the deposition of sodium urate crystals in the medullary interstitium does not occur in these patients. The mutation c.920A→C here reported, replaces the basic amino-acid Lys for the uncharged polar Thr at position 307 (p.K307T), being the fourth mutation described in exon 5. The affected residue locates within the fourth cbEGF-like domain and is immediately preceded by a highly conserved cysteine residue. Conclusions It has been referred that exon 4 sequencing should become the first diagnostic test in patients with chronic interstitial nephritis with gout or hyperuricemia, even in the absence of a family history [ 11 ]. In view of the increasing number of mutations in exon 5 being identified, we now recommend that exon 5 should be included in the initial sequencing effort, since otherwise nearly 12% of UMOD mutations can be missed. Competing interests The author(s) declare that they have no competing interests. Authors' contributions JC was responsible for the study design and drafted the manuscript. JC and CC carried out the molecular analysis. AG collected the clinical data. JR participated in the study design. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC548506.xml
514537	Loss of Skeletal Muscle HIF-1α Results in Altered Exercise Endurance	The physiological flux of oxygen is extreme in exercising skeletal muscle. Hypoxia is thus a critical parameter in muscle function, influencing production of ATP, utilization of energy-producing substrates, and manufacture of exhaustion-inducing metabolites. Glycolysis is the central source of anaerobic energy in animals, and this metabolic pathway is regulated under low-oxygen conditions by the transcription factor hypoxia-inducible factor 1α (HIF-1α). To determine the role of HIF-1α in regulating skeletal muscle function, we tissue-specifically deleted the gene encoding the factor in skeletal muscle. Significant exercise-induced changes in expression of genes are decreased or absent in the skeletal-muscle HIF-1α knockout mice (HIF-1α KOs); changes in activities of glycolytic enzymes are seen as well. There is an increase in activity of rate-limiting enzymes of the mitochondria in the muscles of HIF-1α KOs, indicating that the citric acid cycle and increased fatty acid oxidation may be compensating for decreased flow through the glycolytic pathway. This is corroborated by a finding of no significant decreases in muscle ATP, but significantly decreased amounts of lactate in the serum of exercising HIF-1α KOs. This metabolic shift away from glycolysis and toward oxidation has the consequence of increasing exercise times in the HIF-1α KOs. However, repeated exercise trials give rise to extensive muscle damage in HIF-1α KOs, ultimately resulting in greatly reduced exercise times relative to wild-type animals. The muscle damage seen is similar to that detected in humans in diseases caused by deficiencies in skeletal muscle glycogenolysis and glycolysis. Thus, these results demonstrate an important role for the HIF-1 pathway in the metabolic control of muscle function.	Introduction During exercise in normoxia, the partial pressure of oxygen in muscle tissue has been shown to dip to as low as 3.1 mm Hg, whereas in the capillary, it remains at 38 mm Hg ( Hoppeler et al. 2003 ). In order to maintain effort, skeletal muscle exertion must be able to rely on pathways designed to help the tissue cope with oxygen stress after oxygen delivery capacity is exceeded. A switch between aerobic and nonaerobic metabolism during strenuous exertion requires mechanisms to adjust metabolic function, and this need is acute in extended exertion in skeletal muscle. It is clear that the transcription factor hypoxia-inducible factor 1α (HIF-1α) is an essential factor in maintenance of ATP levels in cells ( Seagroves et al. 2001 ). In fact, although HIF-1α is typically thought of as acting only during hypoxia, its loss has an effect on both normoxic and hypoxic ATP levels in a number of tissue types ( Seagroves et al. 2001 ; Cramer et al. 2003 ), and this implicates the factor in regulation of metabolic function even during conditions of normal physiologic oxygenation. In skeletal muscle, signaling of fatigue has been studied extensively, and signaling of exhaustion involves, to some degree, elevated systemic lactic acid, a by-product of the glycolytic pathway of metabolism ( Myers and Ashley 1997 ). Thus, the glycolytic pathway is intrinsically involved in muscle function and fatigue, and this in turn is linked to the response to hypoxia. To understand how the primary hypoxia-responsive transcription factor controls skeletal muscle function, we targeted mouse skeletal muscle for tissue-specific deletion of HIF-1α via the use of a conditionally targeted allele of the gene ( Ryan et al. 2000 ; Schipani et al. 2001 ). This mouse strain was crossed into a strain transgenic for the skeletal-muscle-specific muscle creatine kinase (MCK) promoter, which drives expression of the cre recombinase gene ( Bruning et al. 1998 ; Sauer 1998 ). We found that loss of the regulation of hypoxic response in muscle has a profound effect on the function of the muscle during exertion, with effects that mimic human metabolic myopathies. Results/Discussion In 4-mo–old mice with the skeletal-muscle HIF-1α gene knocked out (HIF-1α KOs), the frequency of excision was evaluated through real-time PCR techniques. We saw deletion frequencies consistent with those described previously for this cre recombinase transgene ( Bruning et al. 1998 ) with some variation in penetration; mean frequency of deletion was 54.9%, with the highest frequency of muscle-specific deletion of HIF-1α being 72% in the gastrocnemius of 4-mo–old mice homozygous for the loxP -flanked allele ( Table 1 ). This transgene is expressed at a lower level in cardiac tissue, and cardiac deletion was detected ( Table 1 ); however, none of the phenotypes described below were seen in cardiac myocyte-specific deletions of HIF-1α ( Figure 1 A). Gross muscle sections were evaluated histologically to evaluate both vascularization and fiber type ( Tables 2 and 3 ), and ultrastructurally to determine number of mitochondria ( Figure 1 B). No changes were detected in any of these features in HIF-1α KOs, except for a slight but statistically significant decrease in type IIA fibers in the soleus muscles ( Table 3 ). Similar hematocrit and blood hemoglobin levels were seen in HIF-1α KOs and wild-type (WT) mice ( Figure 2 ). Figure 1 Exercise Capacity of Cardiac HIF-1α KOs and HIF-1α/MCK/c re Mitochondrial Density (A) Mice lacking cardiac HIF-1α perform no differently in endurance running trials than WT mice, showing that the increase in exercise capacity seen in MCK/Cre mice is due to deletion of HIF-1α in skeletal muscle, not cardiac tissue. (B) Mice lacking skeletal muscle HIF-1α have a slight but nonsignificant increase in mitochondrial density as measured by the number of mitochondria per electron microscope field of view. Figure 2 Hematocrit and Hemoglobin Levels in HIF-1α KOs and WT Mice (A) Hematocrit levels are virtually identical in both HIF-1α KOs ( n = 3) and WT ( n = 4) mice, indicating that loss of HIF-1α in skeletal muscle does not affect oxygen carrying capacity of the blood. (B) In addition to similar hematocrit levels, WT mice and HIF-1α KOs have very close blood hemoglobin levels. Table 1 Excision of HIF-1α in Various Tissues Deletion levels are the average percent of HIF-1α deleted ± SE Table 2 Fiber Typing of Gastrocnemius Muscle Values are percent ± SE Table 3 Fiber Typing of Soleus Muscle Values are percent ± SE * p < 0.05, WT vs. KO As can be seen in Table 4 , significant changes in HIF-1α–dependent gene expression occur in muscle during exercise, including changes in genes involved in glucose transport and metabolism. Vascular endothelial growth factor (VEGF), which increases vascular permeability, and glucose transporter 4 (GLUT4), the muscle-specific glucose transporter, show increased levels in exercise and likely increase the availability of glucose to the muscle. The muscle-specific form of phosphofructokinase (PFK-M), phosphoglycerate kinase (PGK), and lactate dehydrogenase-A (LDH-A) are also up-regulated at the mRNA level by exercise, and this up-regulation is inhibited by the loss of HIF-1α, further demonstrating that HIF-1α is important for transcriptional response during skeletal muscle activity. Table 4 Relative Gene Expression Levels Expression levels are means relative to resting WT for each gene ± SE Percent increase indicates percent increase of postexercise average gene expression over resting average * p < 0.05, rest vs. postexercise; ** p < 0.01, rest vs. postexercise In Table 5 , we show the changes in enzymatic activity in a number of key glycolytic enzymes affected by deletion of HIF-1α. As can be seen from the data, several of the enzymes assayed showed a decrease in activity in response to exercise. In particular, the activity of one of the key rate-limiting enzymes, PFK, was significantly lower following exercise in HIF-1α KOs compared to WT mice, indicating that HIF-1α KOs may have difficulty maintaining optimal PFK activity. The responses of other glycolytic enzymes to exercise were fairly similar between WT mice and HIF-1α KOs. These include no significant changes in phosphoglucose isomerase activity and significant, yet similar, decreases in aldolase, glyceraldehyde 3-phosphate dehydrogenase, and PGK activities. An exception to this is that WT muscles were able to significantly increase pyruvate kinase (PK) activity (see Table 4 ; p < 0.05). LDH activity was also increased in the WT mice, although the level did not reach statistical significance. Activities of both PK and LDH were not significantly changed in HIF-1α KO muscles following exercise. Increased activities of PK, and subsequently LDH, could be expected to lead to increased levels of lactate in the WT mice relative to HIF-1α KOs.In Figure 3 A, it can be seen that the decrease in PFK activity in the HIF-1α KOs is correlated with a trend approaching significance ( p = 0.10) toward an increased amount of hexose monophosphates (HMPs), which are pre-PFK glycolytic metabolites, following stimulation of the HIF-1α KO muscle. This increase was not due to differences in glucose uptake, since animals of both genotypes were able to significantly increase intramuscular glucose to a similar degree ( Figure 3 B). Consistent with decreased flow through the glycolytic pathway, however, the increased amount of HMPs was correlated with increased muscle glycogenolysis ( Figure 3 C) and increased depletion of phosphocreatine (PCr) ( Figure 3 D), with a resultant decrease in the PCr/ATP ratio in HIF-1α KO muscle ( Figure 3 E), although there was only a nonsignificant drop in overall muscle ATP concentrations ( Figure 3 F). Intramuscular levels of lactate did increase in both HIF-1α KOs and WT mice during stimulation, although lactate accumulation did not differ significantly between them ( Figure 3 G). In order to evaluate whether these changes had any effect on overall muscle force, we measured force and calcium release in isolated single fibers; as can be seen in Figure 4 A and 4 B, there were no significant changes in these parameters, indicating that the muscle can compensate at this level for the metabolic changes induced by loss of HIF-1α. Figure 3 Intramuscular Metabolite Levels at Rest and Following Stimulation (A) Glycolytic intermediates were measured from gastrocnemius muscles following the isolated stimulation protocol. Resting values represent levels in the unstimulated gastrocnemius from the same animals. HIF-1α KOs had a trend toward greater accumulated levels of HMPs during the stimulation protocol, although the difference did not reach statistical significance ( p = 0.10). This difference could be indicative of a blockage in the glycolytic pathway at PFK. (B) No significant differences were seen between HIF-1α KOs and WT intramuscular glucose levels at rest or following stimulation. Both HIF-1α KO and WT muscles were able to significantly increase glucose uptake, leading to greater levels of intramuscular glucose in response to stimulation (WT, p < 0.001; KO, p < 0.05). (C) HIF-1α KOs have more stored glycogen than do WT mice. Glycogen levels were measured following the same stimulation protocol as in (B). The change in glycogen from rest to poststimulation was also greater in the HIF-1α KOs, indicating that they metabolized more glycogen in response to stimulation ( p < 0.01; * p < 0.05, WT at rest vs. KO at rest). (D) HIF-1α KOs utilize more PCr in response to stimulation than do WT mice. Similar levels of PCr were seen at rest, but HIF-1α KOs metabolized significantly more during stimulation ( p < 0.05) and had much lower levels following the protocol (** p < 0.01, WT poststimulation vs. KO poststimulation). (E) A trend toward lower PCr/ATP concentration ratios was seen in HIF-1α KOs relative to WT mice following stimulation, although the difference did not quite reach statistical significance ( p < 0.10). A trend toward a greater drop from rest to poststimulation in the PCr/ATP ratio was also seen in HIF-1α KOs following stimulation ( p < 0.10), indicating that they had to rely more heavily on PCr for ATP generation. (F) Slight but nonsignificant differences were seen in whole-muscle ATP levels at rest or following stimulation. Although HIF-1α KOs exhibited altered substrate utilization, they were able to meet their ATP demands during the protocol. (G) Both HIF-1α KOs and WT animals produced significant intramuscular lactate during the stimulation protocol; however, there was no significant difference in the amount produced by either genotype. Resting intramuscular lactate levels were also similar for WTs and HIF-1α KOs. Figure 4 Force Generation and Ca 2+ Release in Isolated Muscle Fibers during Stimulation (A) No differences were seen in total force generation in isolated muscle fibers. Mechanically dissected fibers from the flexor brevis muscle were subjected to a fatiguing protocol. Neither initial nor final forces differed between HIF-1α KO and WT fibers. (B) Ca 2+ release and reuptake in HIF-1α KO and WT fibers was not different during the stimulation protocol. Ca 2+ levels were measured in individual fibers through use of fura-2 Ca 2+ indicator. The altered substrate utilization did not affect the ability of the fibers to maintain proper Ca 2+ flux. Table 5 Glycolytic Enzyme Activity Levels from Gastrocnemius Muscles Activities are in U/mg protein ± SE GAPDH, glyceraldehyde 3-phosphate dehydrogenase; PGI, phosphoglucose isomerase * p < 0.05, WT vs. HIF-1α KO for given exercised or resting state; p < 0.05, rest vs. postexercise within given genotype Given altered levels of glycolytic throughput without significant changes in intramuscular ATP levels, it is likely that there is increased activity of oxidative pathways in the HIF-1α KO muscle. Increased muscle oxidative activity is typical in patients with myopathies involving muscle glycolysis or glycogenolysis, including phosphofructokinase disease (PFKD) and McArdle's disease ( Vissing et al. 1996 ). We analyzed the activity of citrate synthase (CS), a key allosteric enzyme of the citric acid cycle, in WT and HIF-1α KO muscle ( Figure 5 A), and found that it was up-regulated in HIF-1α KOs. CS is a mitochondrial enzyme that responds to decreases in ATP concentration allosterically, allowing for increased oxidative activity in the mitochondria. In addition, significant up-regulation of the mitochondrial enzyme beta-hydroxyacyl CoA dehydrogenase (B-HAD) was seen in HIF-1α KO muscle ( Figure 5 B). B-HAD is also affected by energy levels in the cell, and decreases in NADH/NAD + concentration ratios cause the enzyme to increase mitochondrial oxidation of fatty acids ( Nelson and Cox 2000 ). Increased activity of oxidative pathways in the muscle should result in more rapid lactate clearance, as in fact occurs in PFKD patients during exercise; this phenomenon gives rise to a “second wind” in these patients, and under some circumstances allows for an increase in exercise endurance ( Vissing et al. 1996 ; Haller and Vissing 2002 ), although this was disputed in one recent study ( Haller and Vissing 2004 ). This decreased lactate accumulation postexercise clearly occurs in the HIF-1α KOs, as can be seen in Figure 5 C. This systemically lower level of lactate postexercise indicates that there may be a shift toward a more oxidative metabolism in skeletal muscle. Figure 5 Oxidative Metabolism and Serum Lactate Production in HIF-1α KOs and WT Mice (A) HIF-1α KOs have higher resting levels of CS activity. CS is an enzyme in the Krebs cycle that can be regulated allosterically by ATP levels. Increased CS activity is indicative of increased muscle oxidative capacity, which is common in patients with glycogenolytic or glycolytic myopathies ( # p < 0.10, KO vs. WT). (B) HIF-1α KOs have higher resting levels of B-HAD activity, which is indicative of a greater ability to oxidize fatty acids ( p < 0.01, WT vs. KO). (C) Lower serum lactate levels were seen in HIF-1α KOs following a timed 25-minute run (* p < 0.05, WT vs. KO). As mentioned above, patients with muscle glycolytic deficiencies demonstrate both increased exercise-induced muscle damage and a “second wind”; the latter phenomenon allows them to exercise for extended periods of time at submaximal levels. This is thought to be due to an increase in rates of oxidative ATP production, and a decreased utilization of and need for muscle glycogen ( Vissing et al. 1996 ; Haller and Vissing 2002 ). To assess whether this is also the case in the HIF-1α KOs, both WT mice and HIF-1α KOs were subjected to endurance tests to assess muscle function. To first determine whether HIF-1α KOs were capable of extended activity during exercise, the animals were given a swimming endurance test. As can be seen in Figure 6 A, HIF-1α KOs were capable of significantly longer-duration swimming activity when compared to matched WT controls ( p < 0.05). Figure 6 Endurance Capabilities of Untrained Mice (A) HIF-1α KOs have greater endurance in swimming tests as shown by their ability to swim on average more than 45 min longer than WT (* p < 0.05, WT vs. KO). (B) HIF-1α KOs have greater endurance than WT mice in uphill running tests. Although only a 10-min difference is seen between run times, it is to be noted that because of the protocol, this 10 min included two velocity increases (** p < 0.01, WT vs. KO). (C) HIF-1α KOs have less endurance than WT mice in downhill running tests. The same protocol was used as in Figure 4 A, except the mice were run on a 10° decline (* p < 0.05, WT vs. KO). (D) RER uphill vs. downhill in WT mice. As would be expected from eccentric exercises relying more heavily on glycolytic fibers, the RER values are higher in mice running downhill than in those running uphill. (E) RER uphill vs. downhill in HIF-1α KOs. Once again, higher RER values are observed for mice running downhill than those running uphill. Further testing was done to determine the parameters of this increased endurance. HIF-1α KOs were run on an enclosed treadmill, with a 5° incline and an initial velocity of 10 m/min, with an increase in velocity every 5 min. In their first runs, HIF-1α KOs again had significantly greater endurance, as shown by their consistently longer run times compared to WT controls ( p < 0.01, Figure 6 B). As it has been shown that muscle groups and fibers respond differently to eccentric exercise (i.e., downhill running) than to concentric exercise (i.e., uphill running) ( Nardone and Schieppati 1988 ), mice from both genotypes were run on a 10° decline with the same velocity and time parameters as in the uphill runs. Eccentric exercises have been shown to recruit primarily fast-twitch glycolytic fibers for contraction, as opposed to the traditional recruitment of slower, smaller, oxidative motor units in concentric contraction, where animals with an increased capacity for muscle oxidation would be at an advantage ( Nardone and Schieppati 1988 ). Now, the trend from swimming and uphill running tests was reversed, with WT mice able to run for a significantly longer time than HIF-1α KOs ( p < 0.05, Figure 6 C). Within genotypes, WT mice ran for significantly longer times downhill than uphill ( p < 0.01); HIF-1α KOs did the reverse, and ran for significantly shorter times downhill than uphill ( p < 0.05). Substrate utilization confirms the shift toward glycolytic fibers in downhill running; both genotypes had higher average respiratory exchange ratio (RER) values when running downhill compared with running uphill ( Figure 6 D and 6 E). PFKD and McArdle's disease demonstrate significant myopathic effects in muscle, including soreness and cramping induced by bouts of exercise. After 1 d of recovery from endurance testing, HIF-1α KOs had increased levels of the MM isoform of creatine kinase in their serum (unpublished data), indicative of skeletal muscle damage. To further investigate this finding, mice were run on a treadmill daily for 4 d. By the second day, the trend for increased endurance in the HIF-1α KOs was absent, and by the final day, HIF-1α KOs were running for significantly shorter times than they had on the first day ( p < 0.01, Figure 7 A). In addition, a repeated measures ANOVA performed on run times showed that the response of the HIF-1α KOs to the protocol was significantly different than that of the WT mice ( p < 0.05). Histological examination of gastrocnemius tissue following 1 d of recovery revealed significantly greater amounts of muscle damage in HIF-1α KO tissue than WT tissue ( Figure 7 B). Staining of the tissue for proliferating cellular nuclear antigen (PCNA) and counts of positive nuclei ( Olive et al. 1995 ) also revealed more cell division in HIF-1α KOs than in WTs, another indication that HIF-1α KOs had been subject to greater tissue damage ( Figure 7 C and 7 D). Figure 7 Increased Muscle Damage in HIF-1α KOs Following Repeated Exercise (A) WT mice and HIF-1α KOs underwent a 4-d endurance test, in which animals were run to exhaustion on each of four successive days with a minimum of 22 h rest between trials. HIF-1α KOs demonstrated initially greater endurance under the protocol; however, by the second day, their endurance advantage was eliminated, and by the fourth day, HIF-1α KOs were running for a significantly shorter time (** p < 0.01) than on the first day, while WT animals were running for approximately similar times as on the first day. Repeated measures ANOVA revealed that the decrease in performance on each successive day was unique to HIF-1α KOs ( p < 0.05). (B) Example of hematoxylin and eosin staining of gastrocnemius muscles after 1 d of recovery by mice after the 4-d endurance test. Evidence of greater damage can be seen in HIF-1α KO muscles compared to WT muscles. (C) Example of PCNA staining of gastrocnemius muscles from exercised mice, demonstrating increased levels of muscle regeneration in HIF-1α KOs. (D) Number of PCNA-positive nuclei per square millimeter in gastrocnemius muscles of WT mice ( n = 5) and HIF-1α KOs ( n = 7) that ran repeatedly for 4 d. Although HIF-1α KOs have almost twice as many PCNA-positive nuclei per square millimeter, the difference is not significant, because of wild variations in that population. F-test analysis of the data reveals that the variance is much greater in the HIF-1α KO population than the WT population ( p < 0.05). As noted above, both PFKD and McArdle's disease are marked by increased resting intramuscular levels of glycogen, a failure of serum lactate to rise during exertion, an exercise-induced “second wind,” and signs of muscle damage following exertion, including elevated levels of creatine kinase in the serum ( Tarui et al. 1965 ; Layzer et al. 1967 ). In addition, PFKD is characterized by elevated levels of HMPs ( Tarui et al. 1965 ; Layzer et al. 1967 ; Argov et al. 1987 ; Grehl et al. 1998 ) and greater PCr utilization during contraction ( Argov et al. 1987 ; Grehl et al. 1998 ). We see many of these hallmarks of muscle deficiencies in glycolytic processing in HIF-1α KOs. The effects are not likely due to glucose uptake, as WT and HIF-1α KO intramuscular glucose levels were not different at rest or following stimulation (see Figure 3 B), and both types of mice responded similarly to a glucose tolerance test ( Figure 8 A). Periodic acid–Schiff (PAS) staining of tissue from mice of both genotypes gave further demonstration of increased glycogen levels in resting muscles from HIF-1α KOs ( Figure 8 B). Figure 8 Glucose Tolerance and Glycogen Storage (A) No significant differences were seen in resting blood glucose levels in HIF-1α KOs or WT mice. Following injection of glucose at a dosage of 2 g/kg, no differences were seen in the maximum levels of blood glucose or the rate of glucose disappearance in either genotype. (B) Representative PAS staining of gastrocnemius muscle from WT mice and HIF-1α KOs. HIF-1α KOs demonstrate darker staining, indicating more stored glycogen. Given the differences in performance observed in the HIF-1α KOs in eccentric and concentric exercise, it is clear that the HIF-1 pathway and hypoxic response have a central role in determining the capacity for work and endurance through regulation of glycolysis. It is also clear that these mice will provide an important model system to investigate the physiology of muscle response during work and oxygen depletion, and may be useful as a model for a group of very debilitating myopathic syndromes in humans. Materials and Methods Mouse strains and crosses. Mice were generated from HIF-1α loxP -flanked allele mouse stocks backcrossed into a C57Bl6/J background. These were crossed into a C57Bl6/J strain containing the MCK/cre transgene. Controls were in all cases littermates that were genotyped as containing only the loxP -flanked HIF-1α allele or only the MCK/cre transgene. No phenotypic differences were seen in the two controls, so they were considered interchangeably as WT control animals. Genotyping and real-time PCR for HIF-1α deletion Mice from the above crosses were genotyped using DNA extracted from tail sections. DNA was then extracted from the gastrocnemius, heart, liver, and uterus of eight 4-mo–old, loxP -flanked HIF-1α– positive and MCK/cre –positive mice. HIF-1α levels were measured by real-time PCR analysis using the Universal PCR Master Mix Kit (Applied Biosystems, Foster City, California, United States) and the ABI Prism 7700 Sequence Detector (Applied Biosystems). Conditions for the PCR were one 10-min incubation at 95 °C (polymerase activation), followed by 40 cycles of 15 s at 95 °C (denaturation) and 1 min at 60 °C (anneal/extend). The degree of excision was calculated by comparing HIF-1α DNA levels to c-Jun DNA levels. HIF-1α real-time PCR primers and probe were as follows: forward primer, HIFLOX501/F 5′-CTATGGAGGCCAGAAGAGGGTAT-3′; reverse primer, HIFLOX574/R 5′-CCCACATCAGGTGGCTCATAA-3′; probe, HIFLOX/P 5′-(6FAM)AGATCCCTTGAAGCTAG(MGBNFQ)-3′. Muscle histology and electron microscopy. Paraffined gastrocnemius sections were deparaffinized and stained with Gill II hematoxylin. Sections were then washed successively in water, a bluing agent, water again, and 95% ethanol, and restained with eosin. Hematoxylin and eosin staining was performed by the University of California at San Diego (UCSD) Cancer Center Histology Resource (La Jolla, California, United States). Imaging was performed on sections mounted on slides using Cytoseal 60 (VWR, West Chester, Pennsylvania, United States). Electron microscopy was performed by standard methods on gastrocnemius muscle. Briefly, fixation was by 25.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4). Postfix was in 1% osmium tetroxide. The section was stained in 2% uranyl acetate in sodium maleate buffer (pH 5.2), then placed in Epon resin (VWR, West Chester, Pennsylvania, United States), and cured overnight at 60 °C. Fiber typing was performed using the metachromatic dye ATPase method ( Ogilvie and Feeback 1990 ). PAS staining was performed as has been described ( Bancroft and Stevens 1996 ). Assessment of exhaustion. Untrained, age-matched WT mice and HIF-1α KOs (WT, n = 10; KO, n = 14) were run either on an Omnipacer treadmill (Columbus Instruments, Columbus, Ohio, United States) or on an enclosed-chamber modular treadmill (Columbus Instruments) with a 5° incline at an initial velocity of 10 m/min. Velocity was increased by 2 m/min every 5 min during the assessment. Exhaustion was determined to be the point at which the animal would not resume running when provoked through a low-voltage power grid. Gas flow (O 2 and CO 2 ) into and out of the enclosed chamber treadmill was monitored using the Paramax O 2 sensor and a CO 2 sensor (Columbus Instruments) and analyzed using Oxymax software (Columbus Instruments) to determine metabolic parameters. The downhill running assessment (WT, n = 8; KO, n = 6) was carried out in the enclosed-chamber modular treadmill at a 10° decline using the same protocol as above. In the swimming exhaustion assessment, a second group of WT and HIF-1α KOs ( n = 8 for each class) was placed in a 30 °C water bath with mild turbulence. Exhaustion was determined to be the point at which the animal experienced three successive periods below the surface of more than 3 s. Isolated stimulation and metabolic analysis. The Achilles tendon was surgically freed from live, anesthetized mice (WT, n = 8; KO, n = 6) and attached to a force transducer to record contractile force. Muscles were electrically stimulated through excitation of the sciatic nerve. Stimulation was in the form of 8–10-V direct titanic contractions using 200-ms trains at 70 Hz with 0.2 ms duration. Initial frequency of tetanic contraction was one every 8 s and was increased every 2 minutes to one every 4 s and one every 3 s, up to the end point of 6 min. Isolated muscles were then immediately harvested and snap-frozen for ATP, lactate, phosphocreatine, and glycogen analyses. Samples were freeze-dried and analyzed by enzymatic assay as has been previously described ( Bergmeyer 1974 ). The unstimulated gastrocnemius muscle from each mouse was used as a resting control. Real-time PCR measurement of gene expression. For basal gene expression levels, total RNA was isolated from gastrocnemius tissue from seven WT and five HIF-1α KOs using RNA-Bee (Tel-Test, Friendswood, Texas, United States). Reverse transcription was performed using the Superscript First Stand Synthesis System for RT-PCR (Invitrogen, Carlsbad, California, United States). Amplification was performed using the ABIPrism 7700 as described above. Reverse transcription real-time PCR primers and probes were as follows. For PGK-1: reverse primer, PGK/R 5′-CAGGACCATTCCAAACAATCTG-3′; forward primer, PGK/F 5′-CTGTGGTACTGAGAGCAGCAAGA-3′; probe, PGK/P 5′-(6∼FAM)TAGCTCGACCCA-CAGCCTCGGCATAT(TAMRA)-(phosphate)-3′. For VEGF-A: reverse primer, VEGF/R 5′-ATCCGCATGATCTGCATGG-3′; forward primer, VEGF/F 5′-AGTCCCATGAAGTGATCAAGTTCA-3; probe, VEGF/P (6∼FAM)TGCCCACGTCAGAGAGCAACATCAC(BHQ∼6∼FAM). For GLUT4: reverse primer, GLUT-4/R 5′-CCCATGCCGACAATGAAGTT-3′; forward primer, GLUT-4/F 5′-TGTGGCCTTCTTTGAGATTGG-3′; probe, GLUT-4/P 5′(6-FAM)TGGCCCCATTCCCTGGTTCATT(BHQ1-Q)-3′. For PFK-M: reverse primer, PFK-M/R 5′-AAGTCGTGCAGATGGTGTTCAG-3′; forward primer, PFK-M/F 5′-GCCACGGTTTCCAATAACGT-3′; probe, PFK-M/P 5′-(6-FAM)CCTGGGTCAGACTTCAGCATCGGG(BHQ1-Q)-3′. For LDH-A: reverse primer, LDH-A/R 5′-ATGCACCCGCCTAAGGTTCTT-3′; forward primer, LDH-A/F 5′-TGCCTACGAGGTGATCAAGCT-3′; probe, LDH-A/P 5′-(6- FAM)TGGCAGACTTGGCTGAGAGCAT(BHQ1-Q)-3′. For changes in gene expression due to exercise, age-matched male mice (WT, n = 5; KO, n = 6) were run on a treadmill at 25 m/min for 30 min. Following the run, mice were euthanized and RNA was isolated and analyzed as described above. Analysis of enzyme activity levels For changes in enzyme activity levels with exercise, mice (WT, n = 5; KO, n = 12) were run on a treadmill using the same protocol as for the gene expression analysis. Tissue was harvested after the run and from resting mice (WT, n = 6; KO, n = 10), and enzymes were extracted and analyzed spectrophotometrically as has been described ( Reichmann et al. 1983 ), with the exception that fructose 1,6-bisphosphate was replaced with fructose 2,6-bisphosphate for stabilization of PFK. Units of activity were normalized to milligrams of total protein using a BCA protein quantification kit (Pierce Biotechnology, Rockford, Illinois, United States). Creatine kinase, serum lactate, hematocrit, and hemoglobin levels. Creatine kinase levels were analyzed from serum from WT mice and HIF-1α KOs 24 h after running-induced exhaustion using a kit from Sigma (St. Louis, Missouri, United States). Creatine kinase isoforms were analyzed enzymatically and then fractionated by gel electrophoresis. Serum lactate levels were analyzed by the UCSD Comparative Neuromuscular Laboratory from blood obtained by cardiac puncture from six WT mice and six HIF-1α KOs following 25 min running time on the treadmill ramp at 25 m/min. Hematocrit and hemoglobin levels were measured from resting mice (WT, n = 6; KO, n = 4) by the UCSD Animal Care Program Diagnostic Laboratory. Glucose tolerance curve. Animals were assigned into either a sham (WT, n = 5; KO, n = 4) or glucose tolerance group (WT, n = 8; KO, n = 8). Experimental animals were injected with 0.3 g/ml glucose in PBS to achieve a dosage of 2 g/kg. Sham animals were injected with an equivalent amount of PBS. Blood was drawn from the tail at time intervals of 0, 15, 30, 60, and 120 min. Samples were then centrifuged to isolate plasma. Plasma blood glucose was quantified using the Infinity Glucose Kit (Sigma). Calcium uptake measurements. Intact individual muscle fibers (WT, n = 6; KO, n = 4) were mechanically dissected from the flexor brevis muscle and loaded with fura-2. Fibers were then stimulated while force generation and Ca 2+ release were monitored. Four-day endurance test Endurance was tested by running 24 animals (WT, n = 10; KO, n = 14) on the Omnipacer Treadmill or the enclosed-chamber modular treadmill using the same exhaustion protocol described above. Mice ran according to this protocol every day for 4 d with a minimum of 22 h of rest between trials. Following the fourth trial, mice were given 24 h of rest and then euthanized. Tissue was harvested and stained using hematoxylin-eosin (as described above) and α-PCNA (Pharmingen, San Diego, California, United States) combined with a DAB Kit (Vector Labs, Burlingame, California, United States). Statistical analysis Statistical analyses (unpaired Student's t-test, Mann-Whitney test, ANOVA) were carried out using StatView software (SAS Institute, Cary, North Carolina, United States) or Prism software (GraphPad Software, San Diego, California, United States).	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514537.xml
543447	Seroma formation after surgery for breast cancer	Background Seroma formation is the most frequent postoperative complication after breast cancer surgery. We carried out a study to investigate the effect of various demographic, clinical and therapeutic variables on seroma formation. Patients and methods A retrospective cross sectional study of patients who underwent surgical therapy for breast cancer with either modified radical mastectomy (MRM) or breast preservation (BP) was carried out. The demographic data and clinical information were extracted from case records. Seroma formation was studied in relation to age, type of surgery, tumor size, nodal involvement, preoperative chemotherapy, surgical instrument (electrocautery or scalpel), use of pressure garment, and duration of drainage. The multiple logistic regression analysis was performed to estimate odds ratios. Results A total of 158 patients with breast cancer were studied. The mean age of the patients was 46.3 years (SD ± 11.9). Seventy-three percent underwent modified radical mastectomy and the remaining 27% received breast preservation surgery. Seroma occurred in 35% of patients. In multivariate logistic regression analysis an association of postoperative seroma formation was noted with modified radical mastectomy (OR = 2.83, 95% CI 1.01–7.90, P = 0.04). No other factor studied was found to significantly effect the seroma formation after breast cancer surgery. Conclusion The findings suggest that the type of surgery is a predicting factor for seroma formation in breast cancer patients.	Background Breast cancer is the second leading cause of cancer death among women. The surgical treatment of choice for these patients is either modified radical mastectomy or breast preservation depending upon stage of the disease. Seroma formation is the most frequent postoperative complication after breast cancer surgery. It occurs in most patients after mastectomy and is now increasingly being considered side effect of surgery rather than a complication however, all patients are not clinically symptomatic [ 1 ]. Seroma is defined as a serous fluid collection that develops under the skin flaps during mastectomy or in the axillary dead space after axillary dissection [ 2 ]. Incidence of seroma formation after breast surgery varies between 2.5% and 51% [ 3 - 5 ]. Although seroma is not life threatening, it can lead to significant morbidity (e.g. flap necrosis, wound dehiscence, predisposes to sepsis, prolonged recovery period, multiple physician visits) and may delay adjuvant therapy [ 6 , 7 ]. Fluid collection is ideally managed by repeated needle aspiration to seal the skin flaps against the chest wall. Several factors have been investigated as the cause of seroma formation these include age, duration of wound drainage, use of pressure garment, postoperative arm activity, preoperative chemotherapy, and use of electrocautery [ 3 , 8 - 12 ]. The present study was undertaken to identify risk-factors associated with seroma formation after breast cancer surgery. Patients and methods A cross sectional study of a consecutive sample of 158 patients attending the breast cancer clinic between January 2000 to October 2002 in Tehran, Iran, was carried out. All patients undergoing surgical therapy [modified radical mastectomy (MRM) or breast preservation (BP)] were included. Level II axillary lymph node dissection was performed for both groups. None of the patients underwent immediate reconstruction. The demographic data and clinical information were extracted from case records. Axillary seroma was defined as any clinically apparent fluid collection in the axilla or under the skin flaps and was treated with multiple needle aspirations. Seroma formation was studied in relation to age, type of surgery, tumor size, nodal involvement, preoperative chemotherapy, surgical instrument (electrocautery or scalpel), use of pressure garment, and duration of drainage. To analyze data univariate odds ratio (or relative risk) was calculated using Chi-square tests or regression analysis and this was followed by the multivariate logistic regression analysis to evaluate independent risk factors related to seroma formation. The variables of interest were selected in a single step (enter method), classification cut off was set at 0.5, probably of step for entry into the model was set at 0.05 and removal at 0.1, and the model was set to converge in maximum of 20 iterations. All variables under study were considered as independent predicting factors and seroma formation was considered as dependent variable for multivariate analysis. The study was approved by the institutional ethics committee. Results In all, 158 breast cancer patients were recruited into the study and 55 patients developed seroma, giving an overall incidence of 35% for seroma formation after breast surgery. The mean age of patients was 46.3 years (SD ± 11.9). One hundred and fifteen patients (73%) underwent MRM and BP was performed in 43 patients (27%). The axillary node involvement was significantly different between MRM and BP patients (χ 2 = 4.52, df = 1, P = 0.03) indicating that those who underwent MRM had higher rate of positive axillary nodes compared to those who received BP (78% vs. 21% respectively). Thirty-one mastectomies were performed by scalpel dissection of the skin flap (20%) and 127 by cautery dissection (80%). Two closed suction drains were placed in all patients undergoing surgery. Sixty-six percent of patients (n = 104) were node positive and the remaining 34% (n = 54) were node negative. The patients' characteristics and univariate odds ratios are shown in Table 1 . Table 1 The characteristics of patients in seroma and no seroma groups and univariate odds ratio. Seroma group (n = 55) n. (%) No seroma group (n = 103) n. (%) OR (95% CI)* p value** Age (years) 0.22 <40 12 (21.8) 34 (33.0) 1.00 (ref.) 40–49 20 (36.4) 38 (36.9) 1.49 (0.63–3.49) >50 23 (41.8) 31 (30.1) 2.10 (0.89–4.92) Tumor size (cm) 0.64 <2 21 (38.2) 47 (45.6) 1.00 (ref.) 2–5 21 (38.2) 34 (33.0) 1.42 (0.67–3.01) >5 13 (23.6) 22 (21.3) 1.26 (0.53–2.96) Nodal involvement (n = 152) 0.31 No 14 (26.4) 34 (36.8) 1.00 (ref.) Yes 39 (73.6) 65 (65.7) 1.45 (0.69–3.04) Surgical procedure 0.03 Breast conservation 10 (18.2) 33 (32.0) 1.00 (ref.) Modified radical mastectomy 45 (81.8) 70 (68.0) 2.12 (0.95–4.72) Surgical instrument 0.06 Scalpel 8 (14.5) 23 (22.3) 1.00 (ref.) Cautery 47 (85.5) 80 (77.7) 1.68 (0.70–4.07) Neoadjuvant chemotherapy 0.22 No 46 (83.6) 93 (90.3) 1.00 (ref.) Yes 9 (16.4) 10 (9.7) 1.82 (0.69–4.78) Pressure garment 0.63 Yes 12 (21.8) 26 (25.2) 1.00 (ref.) No 43 (78.2) 77 (74.8) 1.21 (0.55–2.63) Axillary drainage time (days/n= 152) 0.64 >10 8 (15.1) 20 (20.2) 1.00 (ref.) 5–10 30 (56.6) 49 (49.5) 1.53 (0.59–3.90) <5 15 (28.3) 30 (30.3) 1.25 (0.44–3.49) * Odds ratios derived from univariate logistic regression analysis. ** P values derived from the Chi-squared test. The results of multivariate logistic regression analysis indicated that only the surgical type was significantly associated with seroma formation (OR = 2.83, 95% CI 1.01–7.90, P = 0.04). Of patients with BP, 10 of 43 (23%) developed seroma, while those who underwent MRM 45 of 115 (39%) developed seroma. The seroma formation did not show any significant association with any other variables studied. The results of maultivariate analysis are shown in Table 2 . Table 2 Risk factors for seroma formation derived from the multivariate logistic regression analysis β (SE) Wald P OR (95% CI) Age (years) <40 - - 1.00 (ref.) 40–49 0.68 (0.49) 1.92 0.16 1.99 (0.75–5.25) >50 0.75 (0.48) 2.44 0.11 2.12 (0.82–5.44) Tumor size (cm) <2 - - 1.00 (ref.) 2–5 0.38 (0.44) 0.74 0.38 1.46 (0.61–3.49) >5 0.08 (0.52) 0.03 0.88 1.09 (0.39–3.00) Nodal involvement (n = 152) No - - 1.00 (ref.) Yes 0.35 (0.43) 0.66 0.41 1.42 (0.60–3.35) Surgical procedure Breast conservation - - 1.00 (ref.) Modified radical mastectomy 1.04 (0.52) 3.97 0.04 2.83 (1.01–7.90) Surgical instrument Scalpel - - 1.00 (ref.) Cautery 0.60 (0.51) 1.36 0.24 1.83 (0.66–5.07) Neoadjuvant chemotherapy No - - 1.00 (ref.) Yes 0.33 (0.55) 0.37 0.54 1.40 (0.47–4.13) Pressure garment Yes - - 1.00 (ref.) No 0.51 90.46) 1.28 0.26 1.67 (0.67–4.11) Axillary drainage time (days/n= 152) >10 - - 1.00 (ref.) 5–10 0.13 (0.59) 0.04 0.76 1.17 (0.41–3.32) <5 0.16 (0.53) 0.09 0.82 1.14 (0.35–3.66) Discussion Breast cancer is the most common malignancy in women. Surgery is the mainstay of treatment. Modified radical mastectomy with or without reconstruction or breast preservation in addition to axillary lymph node dissection are common surgical procedures in breast cancer. Surgery of the axilla is associated with numerous complications, including infection, lymphedema of the ipsilateral upper extremity and collection of fluid in surgical site (seroma). Most common complication after breast cancer surgery is wound seroma. The exact etiology of seroma formation remains controversial. Several interventions have been reported with the aim of reducing seroma formation including the use of ultrasound scissors in performing lymphadenectomy [ 13 ], buttress suture [ 14 ], fibrin glue [ 15 ], fibrin sealant [ 16 ], bovine thrombin application [ 17 ], and altering surgical technique to close dead space [ 18 ]. However, it has been suggested that although the use of these interventions might reduce the risk of seroma formation, further studies are needed to verify the real impact on long-term morbidity of such techniques [ 19 ]. Several studies have been performed to investigate factors related to post-surgical seroma. These studies have observed that the early removal of drains might led to increased incidence of seroma [ 12 ], whereas others have shown that drains removal time had no influence on seroma formation [ 3 ]. The findings from our study also indicated that the length of time drains are left did not influence the seroma rate (Table 2 ). Similar observation was reported by a recent study where the use of drains did not prevent seroma formation. On the other hand it was associated with a longer postoperative hospital stay and more pain after surgery for breast cancer [ 16 ]. It has been suggested that the restriction of arm movements may also reduce the incidence of seroma formation [ 8 ]. This observation however was challenged by others who showed that there is no significant disadvantage in early arm motion [ 9 ]. Porter et al reported that the use of electrocautery to create skin flaps in mastectomy reduces blood lose but increased the rate of seroma formation [ 11 ]. In addition, an association of postoperative seroma formation with neoadjuvant chemotherapy was also noted [ 4 ]. Compression dressing to prevent seroma rate is a common method used by many surgeons. A study demonstrated that routine use of a pressure garment to reduce postoperative drainage after axillary lymph node dissection for breast cancer is not warranted [ 12 ]. However, we think that the use of pressure garment and prolonged limitation of arm activity not only reduces seroma formation but also may increase the incidence of seroma formation after removal of drain [ 12 ] and even might cause shoulder dysfunction [ 8 ]. In the present study no relationship was observed between the incidence of seroma and the patients' age, tumor size, and lymph node status. However, the study found that the MRM was associated with higher rate of postoperative seroma formation (OR = 2.83, P = 0.04). Similarly, Gonzalez et al , demonstrated that patients who underwent modified radical mastectomy had a greater incidence of seroma formation than patients who underwent breast preservation surgery [ 10 ]. They also showed that there was a direct correlation between age and the development of seroma [ 10 ]. A recent study by Lumachi et al indicated that the tumor size and total amount of drainage represented the principal factors of seroma formation following axillary dissection in patients underwent surgery for breast cancer [ 19 ]. The results of our study suggest that seroma formation after breast cancer surgery is independent of duration of drainage, compression dressing and other known prognostic factors in breast cancer patients except the type of surgery, i.e there is a 2.5 times higher risk of seroma formation in patients undergoing MRM compared to BP. The small sample size of present study is a limitation and hence the power of the study is low. A number of questions remain unanswered and more research is needed to answer these. Conflict of interest The authors declare that they have no competing interests. Source of Funding None Contributors EH designed the study, collected the data and wrote the first draft of the manuscript. AK, MN, ME and HH all contributed to patient recruitment and the preparation of first draft of the manuscript. AM contributed to the study design, data analysis, and edited the final version. All authors read and approved of the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC543447.xml
555543	Two new plumage mutations in the Japanese quail: "curly" feather and "rusty" plumage	Background The genetics of plumage of Japanese quail is of interest both from a biological standpoint, for comparative studies between avian species, and from a zootechnical standpoint, for identifying commercial selection lines or crosses. There are only few plumage mutations reported in quail, and the present work describes a new color variant "rusty" and a new feather structure "curly", and their heredity from an F1 and F2 segregation experiment. Results Curly feathers result from abnormal early growth caused by transient joining of follicle walls of adjacent feathers around 10 days of age, but the expression of the trait is variable. Rusty plumage color results from the replacement of the wild-type plumage pattern on the tip of the feather by a reddish coloration, but the pigmentation of the bottom part of the feather is not affected. Two lines breeding true for the curly or the rusty phenotype were developed. Both characters are determined by autosomal recessive mutations which are independent. The curly mutation has also a positive effect on body weight at 5 weeks of age. Conclusion The curly line is a new model which may be used for further work on the growth of the feather, and the rusty mutation is a new addition to the panel of plumage mutations available for comparative studies in poultry, and more generally among avian species.	Background Japanese quail is both a model animal in biology and a bird used for meat and egg production under a large variety of settings [ 1 ]. In the recent past, a special attention was given to the study of its plumage, and several major genes have been described [ 2 ]. Since the last compilation of plumage mutations of Japanese quail [ 2 ], new loci were described [e.g. [ 3 ]], linkage and epistasis relationships were explored [ 4 , 5 ], and some genes were recently mapped [ 6 ]. This knowledge has already been put to use for running comparisons between chicken and quail based on plumage genetics [ 7 , 8 ], and for tagging commercial quail lines with a visible plumage trait, like the "fawn" mutation [ 9 ], or with an auto-sexing mutation like the "roux" gene [ 10 ]. Interestingly, some of the mutations described in quail, like the sex-linked "roux" and the lethal "yellow" mutations still have no known homologues in the chicken. Moreover, the fact that some plumage colors, like "lavender", are common to several avian species [ 5 ] is an added incentive to enrich the panel of characterised Japanese quail mutations as potential tools for comparative studies among bird species. In the present work, a new feather structure phenotype (curly) and a new plumage color (rusty) were described, and their mode of inheritance and linkage were studied in two successive generations (F1 and F2) from an F0 made of eight quail with curly feathering and eight birds with rusty plumage. Growth of the F2 quail was also monitored and compared according to their phenotype for the two plumage mutations. Results and discussion Phenotypic description The quail line with the curly feathers was developed in 2001 from group mating 6 founding (G0) curly quail. Starting in G3, only quail for which the curly phenotype observed at 10 days of age was expressed most strongly were kept for breeding. No differential survival was observed in curly quail after hatching, but the hatching rate in the fixed line at G5 under pool mating was only 34%, mainly because some females did not get mated under this type of mating. In curly chicks, the calamus of the growing wing feathers are not independent from one another, but they are connected through the follicle walls which appear to be joined together. This phenomenon is best observed around 10 days of age and is associated with the curly growth (Figure 1 ). The expression of the trait is variable, however, and the penetrance of the curly mutation appears to be incomplete. The difference between normal and curly adults (Figure 2 ) is not as marked as for other feather structure mutations, like porcupine [ 11 ], for example. Following the gene nomenclature proposed for the chicken [ 12 ], the locus for this new mutation was named CU, and the symbols of the allele responsible for the curly mutation and of the normal allele at this locus were CUC and CUN, respectively. In all instances, inheritance of the trait was similar for both sexes, and heterozygotes CUC/CUN had normal plumage structure, indicating that the locus CU was autosomal and the mutation was recessive. Figure 1 Wing of a 10 day Japanese quail with the curly phenotype. In curly quail, the calamus of the adjacent growing wing feathers are not independent from one another, but they are connected through the follicle walls which appear to be joined together. This phenomenon which lasts one to two weeks is best observed around 10 days of age. It hinders the normal growth of the feathers, thereby inducing the permanent curly structure of the feathers. Figure 2 Japanese quail with the curly and the normal feather structures. The adult curly quail is on the left. Its plumage has an overall fluffier look when compared to the normal quail on the right, but the intensity of the curliness varies between curly quail, as the penetrance of this autosomal recessive mutation is not complete. The quail line with the rusty plumage color was started in 2000 from a founding base (G0) made of a single rusty female bred to a wild-type male from another origin. It was followed by sib mating of the all wild-type G1 quail to produce G2 and G3 pedigreed progeny and selection of the few rusty G4 birds. Survival rate to sexual maturity of the rusty descendance of the early generations was poor, and reproductive performance was low (hatched/incubated = 31%) in the fixed line under pool mating at G6, because of the mating system but also of inbreeding derived from having a single rusty ancestor. The plumage of the mutant chicks was rusty, but the down underneath retained the usual wild-type dark-slaty color. A similar color pattern was maintained in rusty adults: their contour and flight feathers had dark barbs on the bottom and rusty colored barbs on the top third of their length which produced the overall rusty look (Figure 3 ). The effect of this mutation appears to be different from the dilution of the pigmentation [ 5 ] associated with the roux mutation which affects the whole feather and produces a paler color (Figure 4 ). The locus for this new mutation was named RU, and the symbols of the allele responsible for the rusty mutation and of the wild-type allele at this locus were RUR and RUN, respectively. In all instances, inheritance of the trait was similar for both sexes, and heterozygotes RUR/RUN had wild-type plumage color, indicating that the locus RU was autosomal and the mutation was recessive. Figure 3 Japanese quail with the rusty plumage color. Quail with the rusty phenotype are homozygous for an autosomal recessive mutation. Figure 4 Comparison of colors of feathers from Japanese quail. On the left, the contour feather of a wild-type Japanese quail shows the usual brownish colored tip with a transversal lighter stripe and slaty colored barbs at the bottom. On the center, the feather of a quail with the rusty phenotype has a different, rusty colored, tip, but keeps the same slaty colored barbs than the wild-type bird at the bottom. On the right, the feather of a roux quail (caused by a sex-linked recessive mutation) shows a diluted color, paler than rusty, over its whole length. Linkage analysis All 68 F1 quail were wild-type birds, confirming that both mutations should be recessive, and not sex-linked. Four phenotypes were obtained in the 531 F2 progeny: 326"wild-type", 95 "rusty", 79 "curly" and 31 "rusty and curly" (Figure 5 ). The high hatching rate (81%) obtained in the F2 confirms that the much lower value obtained in the two lines fixed for the rusty or the curly mutation might have resulted from inbreeding due to the small number of founders and from the pool mating system used rather than from detrimental reproductive effects directly associated with the two mutations. Observed and expected distributions under two different null hypotheses (A: independent segregation given complete penetrance of the curly mutation, and B: independent segregation given incomplete penetrance of the curly mutation) are shown with the corresponding values of χ 2 s (7.5 and 1.9, respectively) in Table 1 . None of the two hypotheses could be rejected because the probabilities for these values of χ 2 s were high enough (p > 0.05 and p > 0.1, respectively). The maximum likelihood estimation of the penetrance parameter (± SE) was 1-λ = 0.83 (± 0.07) which indicates that 17% of F2 curly quail might have been misclassified as wild-type birds. This result is consistent with the observation that the expression of the curly trait was variable, and it would account for the relative deficit of curly birds in the F2. Overall, however, the segregation of the F2 results fits a simple two-locus Mendelian inheritance of two autosomal recessive and independent mutations. Figure 5 Japanese quail with the rusty plumage color and the curly feather structure. This phenotype was the least frequent of the four quail types produced in the F2 between the curly and the rusty lines. These quail are homozygous for the two recessive mutations. Table 1 Segregation of plumage color and feather structure in the F2 from rusty plumage and curly feather quail lines Phenotype Observed (n = 531) Expected under independent segregation And complete penetrance (λ = 0) And incomplete penetrance (0 < λ < 1) Wild-type a = 326 9n/16 = 298.7 9n/16+3λn/16 = 315.6 ¥ Rusty b = 95 3n/16 = 99.6 3n/16+λn/16 = 105.2 Curly c = 79 3n/16 = 99.6 3n(1-λ)/16 = 82.7 Rusty and curly d = 31 n/16 = 33.2 n(1-λ)/16 = 27.6 Goodness of fit χ 2 s = 7.5 £ χ 2 s = 1.9 $ ¥ : calculated using 1-λ = 0.83 £ :0.10 > p > 0.05. $ :0.5 > p > 0.1 Growth At hatching, none of the plumage mutations were associated with differences of body weight (Table 2 ), but after two weeks quail with curly feathers had become significantly heavier than wild-type ones, and the difference reached 3.4% (p < 0.001) of the average body weight at 35 days of age (Table 2 ). On the opposite, quail with rusty plumage were 2.3 % lighter (p < 0.05) than wild-type ones. The mean body weight at 5 weeks of age was 176.8, 170.5, 181.0 and 182.0 g, respectively for the "wild-type", "rusty", "curly" and "rusty and curly" quail. The fact that both "curly" and "rusty and curly" birds had a similar high body weight, whereas "rusty" quail were the smallest ones is an indication that the effect on growth of the "curly" mutation might be epistatic over that of the "rusty" mutation. The association of plumage color mutations with lower growth have been previously reported in albino [ 13 ] and roux [ 10 ] quail, with respectively 9 and 3% lower body weight at a similar age, but no mutation with a favorable effect had been found so far. A large scale experiment focused on the associated effects of the curly mutation on growth is needed, however, to confirm and extend the present results. Table 2 Analyses of variance of body weight of the F2 from rusty plumage and curly feather quail lines Age At hatch 1 week 3 weeks 5 weeks Sample size 1 348 348 346 347 Mean body weight (g) (SD) 8.1 (0.8) 27.6 (4.5) 105.7 (12.5) 176.7 (17.0) R 2 0.82 0.45 0.47 0.50 Significance of main effects Hatch ns * * Family * * * * Sex ns ns * Feather structure ns ns * Plumage color ns * ** * Contrast (g) "curly" – "wild-type" -0.08 ns 0.8 ns 3.8 6.1* "rusty" – "wild-type" 0.04 ns -2.1* -3.6 -4.1* 1 : The quail sample was made up of 215 wild-type, 59 curly, 58 rusty, and 16 curly and rusty quail from 17 full-sib families and born in two successive hatches. : p < 0.05; : p < 0.01; *: p < 0.001; ns : not significant. Conclusion The two new "curly" and "rusty" mutations will enrich the small number of plumage mutations already available in Japanese quail for studying the genetics and the biology of feathers, a field of research with many perspectives [ 14 ]. They may have also some interest from a zootechnical standpoint to tag commercial lines, and, if the positive effect of the "curly" mutation was confirmed, this gene might be worth introgressing in parental meat quail lines. Methods Birds The two mutations originated in the experimental quail population maintained and selected on behavioral traits at the INRA Station de recherches avicoles in Nouzilly, France. After two quail lines were established by fixing separately the curly and the rusty phenotypes, eight reciprocal single pair matings (three "rusty × curly" and five "curly × rusty") were set up with G5 rusty and G4 curly quail from the two pure lines to produce the F1. Then, 531 F2 birds were produced in three consecutive hatches from 17 single pair matings of F1 birds. Sib-mating was avoided, and the hatching rate across all pair matings and hatches was 81%. All chicks were pedigreed, and they were phenotyped for plumage color at hatching and for feather structure at 10 days of age. F2 quail from the first two hatches were raised in two group pens (one per hatch) with free access to ad libitum commercial feed and drinking water, and they were weighed weekly until 5 weeks of age. Genes Gene nomenclature used in this paper followed recommendations published for chicken genes [ 12 ], with a two-part symbol: "locus abbreviation""allele abbreviation/locus", and "locus abbreviation*N" as the symbol for the wild-type allele. Statistical analyses Analysis of the segregation in the F2 to test for linkage was carried out using maximum likelihood methodology and the χ 2 test [ 15 ]. Penetrance and its standard error were estimated as: 1-λ = (3c/(a+c)) + d(b+d) and SE = (9ac/(a+c) 3 ) + bd/(b+d) 3 , derived for misclassification of phenotypes due to "partial manifestation" [ 16 ]. In the formulae, a, b, c and d are the numbers of observations in the four phenotypic classes described in Table 1 . Five-way analyses of variance of individual body weights (BW) between hatching and 5 weeks of age were carried out by the GLM procedure [ 17 ] for the 348 quail hatched alive in the first two hatches, using the following linear model: BW=(overall mean) + hatch + family + sex + (feather structure phenotype) + (plumage color phenotype) + error. The number of classes for the five main effects were respectively, 2, 17, 2, 2 and 2. Contrasts between least-squared means for each mutant phenotype and for the wild-type quail were estimated from the analyses of variance, using data adjusted for systematic effects of hatch, family and sex. Authors' contributions FM coordinated the study and wrote the paper, DG and CM participated in the design of the study and carried out the the data collection.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555543.xml
545778	Genomic transcriptional response to loss of chromosomal supercoiling in Escherichia coli	Microarray analysis shows that transcription of 306 E. Coli genes is affected by changes in the level of chromosome supercoiling, suggesting that supercoiling transmits regulatory signals from the environment to many cellular pathways.	Background The chromosome of Escherichia coli is a circular double-stranded DNA molecule that is maintained in a negatively supercoiled state. Supercoiling induces torsional tension in the DNA, and thus can influence processes that involve the opening of the double helix, such as replication initiation [ 1 ], DNA looping [ 2 ] and transcription [ 3 ]. A number of external stimuli, such as osmotic stress, oxygen tension, nutritional shifts, and temperature change affect supercoiling (for review see [ 4 ]), suggesting that supercoiling is a mechanism by which environmental changes could be communicated to the transcriptional machinery. In E. coli , supercoiling is maintained at a precise range during log phase growth by the topoisomerases DNA gyrase, topoisomerase I (topo I), and topoisomerase IV (topo IV) [ 5 - 7 ]. DNA gyrase and topo IV are ATP-dependent type II topoisomerases that introduce negative supercoils and remove positive supercoils, respectively [ 8 - 10 ], whereas topo I is a type IA topoisomerase that removes negative supercoils [ 11 ]. Together, these activities remove the topological effects of translocating proteins, such as RNA polymerase, that create (+) supercoils in front and (-) supercoils behind the moving protein [ 12 , 13 ]. The balanced activities of these enzymes result in a steady-state level of negative supercoiling. In turn, supercoiling modulates the expression of the genes for gyrase ( gyrA and gyrB ), and for topo I ( topA ). Relaxation of the chromosome upregulates gyrA and gyrB and downregulates topA as a form of feedback control [ 14 - 16 ]. This dual response also indicates that (-) supercoiling can promote, as well as inhibit, gene expression. It is perhaps not surprising that transcription of topoisomerase genes may be sensitive to supercoiling changes. Yet transcription of other genes, such as fis (a nucleoid-associated protein and transcriptional regulator), ilvG (an amino-acid synthase subunit) and cydAB (an oxidase involved in aerobic respiration), has been found to be sensitive to supercoiling [ 17 - 19 ], suggesting that a wider class of genes whose expression is sensitive to supercoiling may exist. Furthermore, a recent search for osmotic shock genes found a cluster of genes with enhanced sensitivity to supercoiling [ 20 ]. If supercoiling is used as a mechanism to sense environmental changes, we predict that genes from many functional classes would be affected by supercoiling, because environmental changes such as temperature and osmotic strength will affect many different reactions in the cell. Determining which genes are supercoiling sensitive may illuminate principles of promoter activation, such as common sequence characteristics in promoters and regulation of transcription initiation [ 14 , 17 , 18 ]. In this study, we used cDNA microarrays [ 21 , 22 ] representing nearly the entire E. coli K-12 genome to systematically identify those genes that respond to relaxation of the chromosome during log-phase growth. We used antibiotics and mutations in the topoisomerase genes to change supercoiling levels by independent mechanisms and thus discerned the general effects of chromosome relaxation. We classify supercoiling-sensitive genes, or SSGs, according to their response to DNA relaxation. Therefore, we call 'relaxation-induced genes' those genes whose expression is increased upon DNA relaxation, and 'relaxation-repressed genes' those whose expression is repressed by DNA relaxation. An extensive statistical analysis of our experimental results revealed 200 relaxation-repressed genes and 106 relaxation-induced genes; in total, around 7% of all genes in the genome were found to be significantly affected by supercoiling changes. Many of these genes are more sensitive to supercoiling than gyrA or topA , and their expression patterns correlated with the supercoiling level of a reporter plasmid in the cells. SSG transcripts have the same rates of RNA decay as non-SSG transcripts, and thus the changes in expression were due to a change in the rate of RNA synthesis, rather than RNA decay. We discovered that the sequences of the relaxation-induced genes are significantly ( p < 0.0001) AT-rich in their upstream sequences, and also have AT-rich coding regions. Relaxation-repressed genes had a corresponding preference for GC-rich sequences. The SSGs are dispersed throughout the chromosome, and fall into many different functional classes. We propose that the large number and functional diversity of the SSGs reflects the role of supercoiling as a second messenger that responds to environmental changes and feeds into different regulatory networks. Results Topoisomerase inhibition We sought to determine the genes that are activated or repressed by relaxation of the (-) supercoils in the chromosome. To isolate the expression changes due to the loss of supercoiling from those due to the experimental approach, we used three different methods to relax the chromosome. In two of the methods we inhibited gyrase and topoIV with either quinolone or coumarin antibiotics, and in the third we used a temperature-sensitive strain in which gyrase is inhibited at 42°C. Because it is technically difficult to quantify the supercoiling state of the bacterial chromosome, we used a plasmid, pBR322, in the strains as a reference. Co-transcriptional translation of the tetA gene in pBR322 anchors this plasmid to the membrane [ 23 ], and thus this plasmid has been used as a model for the chromosome [ 7 ]. The superhelical density, or σ, of plasmids can be readily measured. Plasmid σ values for all of the relaxation experiments are shown in Table 1 . Inhibition of topoisomerases by norfloxacin The quinolone antibiotic norfloxacin selectively and immediately inhibits gyrase and topo IV [ 24 - 26 ]. We used isogenic strains with resistance mutations in the genes for gyrase ( gyrA and gyrB ) or topo IV ( parC and parE ) as controls, to separate expression changes due to undiscovered drug targets from those directly due to changes in supercoiling. When we inhibited gyrase by treating gyrA + parC R cells with 15 μg/ml norfloxacin, the reporter plasmid in the cells was rapidly relaxed (Table 1 ). In a parallel experiment, plasmid DNA in a drug resistant gyrA R parC R strain remained (-) supercoiled. After 30 minutes, there was a 10 3 -fold drop in viability in the sensitive strain, but only a 17% drop in the resistant strain. A norfloxacin concentration of 50 μg/ml fully inhibited both gyrase and topoisomerase IV in the wild-type strain (data not shown), while the resistant strain retained wild type levels of (-) supercoiling and showed only a slight drop (15%) in viability, indicating that we did not overcome the drug resistance. At bacteriocidal concentrations similar to these, quinolones cause a decrease in the sedimentation coefficient of the nucleoid, indicating relaxation of the chromosomal supercoils [ 27 ]. The reference RNA sample was from cells removed immediately before addition of the drug ( t = 0) and was labeled with Cy3 (green). RNA samples taken 2, 5, 10, 20 and 30 minutes after drug addition were labeled with Cy5 (red). The labeled experimental and reference samples were mixed in equal amounts before hybridization to a microarray. Inhibition of topoisomerases by quinolones leads to double-strand breaks in the chromosome [ 28 ]. Thus, norfloxacin not only reduces supercoiling, but also induces the SOS response to DNA damage [ 29 ]. We found that the induction of the SOS response by norfloxacin was significantly slower and less extensive than either the responses of the SSGs (see below) or the SOS induction caused by UV treatment (see Additional data file 1). We conclude that the induction of SOS by norfloxacin is not a significant impediment to our search for SSGs. Inhibition of topoisomerases by a coumarin antibiotic We also relaxed the chromosome using novobiocin, a coumarin antibiotic that inhibits gyrase, and at a higher concentration, topo IV [ 30 , 31 ]. Novobiocin inhibits the ATPase activity of the enzyme [ 32 , 33 ], and the mechanism of inhibition is completely different from that of norfloxacin [ 34 ]. We treated cells with 20, 50 and 200 μg/ml novobiocin for 5 minutes and measured the DNA relaxation by gel electrophoresis (Table 1 ) and the gene-expression changes by microarray. In addition to changes due to a loss of topoisomerase activity, we saw changes in a set of non-overlapping genes between the norfloxacin and novobiocin experiments, indicating that there are also drug-specific transcriptional effects. Since we focused our analysis on the genes that responded to supercoiling independent of the relaxation method used, these drug-specific changes were removed from consideration. Inhibition of gyrase by mutation We also used a mutant that is temperature-sensitive for gyrase activity [ 35 ], which results in relaxation of the chromosome at the restrictive temperature [ 36 ]. We measured expression changes in gyrB234 cells upon shift to the restrictive temperature and subsequent relaxation of the DNA (Table 1 ). To control for the effects of the temperature shift on gene expression, we compared the changes in the gyrB TS mutant to those in an identically treated isogenic wild-type strain. The gyrB TS data were combined with the norfloxacin and novobiocin data to make a body of experiments and controls where the transcriptional effects of relaxation were isolated from effects due to the method used to relax the chromosome. Identification of supercoiling-sensitive genes by statistical analysis We obtained a dataset from a total of 35 arrays. Fourteen of the arrays were controls in which either drug was added to resistant cells or the temperature was shifted for wild-type cells. The supercoiling of the reporter plasmid did not change in these controls (Table 1 ). The remaining 21 arrays represented experiments in which the DNA was relaxed by different methods and over various time courses. This rich dataset allowed us to use statistical methods to determine those genes whose expression significantly varied with supercoiling levels. Using threshold ratio values (for example, requiring a twofold change in expression) to determine which genes change significantly during an experiment can bias expression analysis towards genes with very low or variable expression levels [ 37 ]. We used statistical methods to minimize the bias. To assess the significance of the difference in gene expression between supercoiled and relaxed samples we used the method described by Dudoit et al . [ 38 ]. Briefly, we performed a t -test for each gene and corrected the obtained p -values for multiple testing by a step-down procedure [ 39 ]. The corrected p -value represents the probability that the differences in gene expression between the controls and relaxation experiments could have arisen by chance, after taking the expression of all genes into consideration. We obtained p -values ranging from 0.000125 to 1. As an independent metric of supercoiling sensitivity, we measured how closely gene expression followed the level of DNA supercoiling, by calculating the correlation of the expression of each gene across all of the experiments with the level of supercoiling in the reporter plasmid. Relaxation-induced genes showed a positive correlation with plasmid linking number (that is, as (-) supercoiling is lost, both linking number and gene expression increase), up to a maximum observed Pearson correlation coefficient of 0.91. Relaxation-repressed genes showed a negative correlation with plasmid linking number to a minimum Pearson coefficient of -0.88. The majority of genes (3,190, or 80%) showed no strong correlation with plasmid supercoiling, resulting in Pearson coefficients between 0.5 and -0.5. The p -value represents the robustness of the response to relaxation, whereas correlation with plasmid supercoiling may represent sensitivity to changes in supercoiling levels. For example, a gene that is always completely repressed in response to relaxation will have a low p -value, but may show little correlation with intermediate levels of supercoiling. Similarly, a gene with more variable expression may have a higher p -value, but may also have a higher sensitivity to intermediate supercoiling levels. Taken together, these metrics provide a detailed account of supercoiling sensitivity. The p -values for all of the genes versus their correlation to plasmid supercoiling are plotted in Figure 1a . The great majority of the genes have both high p -values and little correlation with plasmid supercoiling. Those genes with the lowest p -values (and thus, the most significant expression change upon relaxation) tended to be more strongly correlated (or anticorrelated) to plasmid supercoiling. The data for all genes can be found in Additional data file 2. Among all genes there is a continuous variation in both p -value and correlation to plasmid supercoiling. We found a total of 306 genes at p < 0.05, which we define as SSGs. Of these, 106 genes were induced by DNA relaxation and have a positive correlation with plasmid linking number, while 200 genes were repressed by relaxation and these have a negative correlation with plasmid linking number. The correlations of the SSGs with plasmid supercoiling are shown in Figure 1b , which is an expansion of the significant region of the plot in Figure 1a . All the SSGs have a correlation with plasmid supercoiling with an absolute value greater than 0.5, which validated our selection on the basis of p -value. Just over half of the SSGs have high significance, p < 0.005. The high redundancy of our data (21 arrays measuring responses to DNA relaxation, and 14 control arrays with negatively supercoiled DNA) minimized the influence of any single array measurement. Thus we can be confident that the genes we classed as SSGs have a reproducible response to supercoiling changes. Figure 2a shows the expression changes in the 200 relaxation-repressed genes across the 35 conditions tested, with each numbered column representing one array. Each row represents the expression of one gene across all experiments, ranked by p -value (from the top). Each colored entry in the diagram corresponds to one spot on one array (that is, expression of a gene for a point in a given experiment: red if expression increased during the experiment, green if it decreased). Conversely, these relaxation-repressed genes should have low ratios (and black squares) in the control experiments 1 to 14. The significant difference in SSG expression between the controls and relaxation experiments is reflected by the sharp contrast between the mostly black controls and the bright green relaxation experiments. At the top we have shown a model expression profile representing the supercoiling of the reporter plasmid in each experiment (Table 1 ), with black indicating no change in plasmid supercoiling and bright green indicating complete relaxation of the plasmid. These plasmid relaxation data match very well the expression data of the SSGs. The names of the top 10% of genes (those with the lowest p -value) are listed, along with their correlations to plasmid supercoiling levels. The 106 genes that are induced by relaxation are similarly shown in Figure 2b . Red squares indicate expression at a higher level when the DNA is relaxed. Once again there is a striking difference in color between the control and relaxation experiments, and the SSGs show a strong similarity to the model profile at the top (in this model profile, red color indicates relaxation of the reporter plasmid). Several of the relaxation-induced genes are marginally repressed (shown by green color) in some control experiments. This is due to the fact that our statistical selection did not require the SSGs to be unchanged in the controls, but only required a significant difference in expression between the controls and relaxation experiments. However, this trend highlights the large expression change (from repression to induction) caused by chromosomal relaxation. It is striking how many genes respond significantly to a loss of chromosomal supercoiling (7% of the total genes). The full list of SSGs, with their p -values, correlations to supercoiling, and expression levels in each experiment can be found in Additional data file 3. Kinetic analysis of gene expression and supercoiling We expected that changes in SSG expression that are a direct effect of supercoiling changes (rather than mediated through other genes) should respond quickly to relaxation. We used a finer time-course experiment to determine which genes had the fastest response to chromosomal relaxation. When 15 μg/ml norfloxacin was added to gyrA + parC + cells, plasmid supercoiling levels changed dramatically within the first minute (Figure 3 ). Significant changes in gene expression followed by 2 minutes (Figure 4 ). We ranked the SSGs according to their correlation to plasmid supercoiling levels in this experiment. Thus, genes with transcriptional changes that match the kinetics of plasmid relaxation have high correlations. About 90% of the SSGs had a correlation higher in absolute value than 0.5, and more than half had correlations better than 0.75. The expression profiles of all of the SSGs, ranked by their correlation to plasmid supercoiling, are shown in Figure 4 . The correlation of the SSGs to plasmid relaxation kinetics shows the sensitivity of gene expression to changes in supercoiling, while the p -value is a good indicator of the reproducibility of the response to supercoiling across the different experimental conditions we tested. The speed of the transcriptional response to relaxation, combined with the strong correlations to supercoiling of the reporter plasmid in the cells, is strong evidence that the SSGs are directly regulated by supercoiling changes. Furthermore, given that E. coli mRNAs have a mean half-life of 5.2 ± 0.3 minutes in LB media [ 40 ], RNA synthesis of the relaxation-repressed genes must have slowed almost immediately upon DNA relaxation, in order to produce the quick changes we recorded (Figure 4 ). More than half of the relaxation-repressed genes changed by twofold or more in the first 5 minutes of this experiment. We found no correlation of p -value with the published values of RNA half-life [ 40 ] and in general the mRNA half-lives of the SSGs were not significantly different from those from the rest of the genome (data not shown). We conclude that the changes in SSG expression are direct effects on transcription, rather than an effect on RNA degradation. Sequences surrounding the start codon of supercoiling sensitive genes We searched for a basis of supercoiling sensitivity at the nucleotide sequence level by examining the AT content in and around the SSGs. We considered only the first genes in an operon. Whereas relaxation-repressed genes have a slightly depleted AT content both upstream of their promoters and within the coding sequence, relaxation-induced genes have an emphatic elevation of AT content in similar regions. The AT content of relaxation-induced genes from 800 nucleotides before to 200 nucleotides after the start codon is 54.6%, compared with 51.7% for non-SSGs. To illustrate the very low probability of selecting by chance a set of genes with such an elevated AT content, we randomly selected groups of first-in-operon non-SSGs 50,000 times and calculated AT content within the same window. We never found a set with an AT content as high as the relaxation-induced genes (red circle, Figure 5a ). The difference in AT content is highly statistically significant ( p = 3E-4). This is not the only region in which the AT content of SSGs deviates from the rest of the genome. Figure 5b shows the mean AT content in a 100-nucleotide window for relaxation-induced, relaxation-repressed, and non-SSGs from 2 kilobases (kb) upstream to 1.5 kb downstream of the start codon. Nearly all genes, including non-SSGs, have elevated AT content upstream and just downstream of the start codon. The relaxation-induced genes, however, have a higher maximum AT richness and the elevated AT content extends over a wider region. Also, the relaxation-repressed genes showed a highly statistically significant reduction in AT content from -400 to +1,000 relative to the start codon ( p = 1E-6). Striking as these differences in AT content are for SSGs as a group, they are not sufficient to distinguish an individual SSG from a non-SSG. That is, not all genes with high or low AT content were supercoiling sensitive in our experiments. Although such genes are rare in the non-SSG population, the greater size of the pool of non-SSGs results in many genes with wide variations in AT content. Also, supercoiling sensitivity cannot solely be due to differences in AT content, as a few SSGs were highly sensitive to supercoiling changes in spite of having an AT content similar to the rest of the genome. Discussion In this analysis of supercoiling effects on transcription, we identified 306 genes that quickly and reproducibly respond to chromosomal relaxation. The comprehensive nature of our investigation, with responses of 93% of the genome (4,003 protein-coding genes) in 21 different relaxation experiments and 14 control experiments, allowed us to be more stringent than previous studies in our definition of SSGs, and to identify those genes that had statistically significant changes after the chromosome was relaxed by different methods. Genes that are sensitive to relaxation but are also affected by temperature shifts (including topA [ 41 ] and gyrA [ 42 ]) showed changes in our control experiments, and thus had less significant p -values. Accordingly, although the topoisomerase genes topA and gyrA both clearly respond to supercoiling (see Figure 1 and [ 14 - 16 ]), they have p -values of 0.058 and 0.062, respectively (compared to the p -value of 0.001625 for gyrB ). The omission of these topoisomerase genes from our list of SSGs reflects the conservative statistical approach we used to define the list. There are probably other genes that respond to supercoiling changes in different conditions from those we tested (log-phase growth in rich media). Also, we defined SSGs by focusing on the immediate effects of relaxation, and thus considered only primary transcriptional changes, rather than downstream effects mediated by other gene products (though we note that 14 of the SSGs encode known transcriptional regulators). When downstream effects are considered, changes in supercoiling are likely to affect transcription of an even greater proportion of the genome. There have been several previous attempts to measure the effects of supercoiling on gene expression in E. coli . Two early studies used either nylon membranes or two-dimensional protein gels to compare topoisomerase mutants with slightly different homeostatic levels of supercoiling, and neither study found a large number of genes [ 43 , 44 ]. This could be due to the lower sensitivity of these earlier studies and because they measured steady-state gene expression, generations after the initial mutations and subsequent adjustment to the new supercoiling levels. A more recent analysis by Church and colleagues used microarrays to gauge the osmotic stress response of E. coli [ 20 ]. Surveying 2,146 genes that were above their threshold of detection, the authors scored a subset of 30 genes that should be significantly enriched for supercoiling-sensitive transcription. Four of the genes identified are on our list of SSGs ( ynhG , yrbL , otsB and yifE ). Seven others had p < 0.1 in our relaxation experiments, and the rest had still higher p -values in this study. It is possible that these genes are only responsive to supercoiling changes in the context of osmotic stress. Just as supercoiling is affected by many environmental changes, such as osmotic shock, oxygen tension, nutrient upshift and temperature change, so too do changes in supercoiling affect genes in a large number of classes. Not surprisingly, a substantial fraction (6.9%) of the SSGs encode genes involved in DNA replication, recombination, modification and repair. However, the SSGs span many other classes, and thus are well positioned to feed into many different regulatory networks. Thus, supercoiling can act as a second messenger that quickly translates environmental changes to transcriptional programs, inducing and repressing specific genes independently of protein synthesis. Several of the SSGs warrant further inspection. For example, the repression of the smtAmukBEF operon on loss of supercoiling is intriguing, given the importance of mukB , mukE and mukF in chromosome supercoiling and segregation [ 45 , 46 ]. Consistent with this, the XerC site-specific recombinase, which is needed for proper chromosome partitioning, is also repressed by relaxation. As (-) supercoiling promotes chromosome segregation in E. coli [ 47 ], these genes may represent part of a supercoiling 'checkpoint' that senses whether supercoiling levels are sufficient for proper chromosome segregation. Thus, if there is insufficient (-) supercoiling to support chromosome segregation, transcription of these genes may be suppressed until supercoiling is re-established. Another relaxation-repressed gene that may be involved in chromosomal maintenance is yrdD , a 'putative topoisomerase'. yrdD encodes a 19 kilodalton (kDa) protein 30-40% identical to the carboxy-terminal domain of topoisomerase I from Bacillus subtilis , Helicobacter pylori and Methanococcus jannaschii . The function of YrdD is unknown, but the repression by chromosomal relaxation provides an intriguing lead. Chromosomal relaxation leads to the repression of cls (cardiolipin synthase) and ileS (isoleucine tRNA synthetase), which is consistent with the earlier discovery that these genes were involved in sensitivity to gyrase inhibitors [ 48 ]. Also, we noted that the nucleotide salvage genes deoA and deoC are induced on relaxation. For these genes, DNA relaxation may be a signal of DNA damage, and their induction would allow the cell to recycle nucleotides necessary for DNA repair. Finally, the induction of rpoD , the σ 70 subunit of RNA polymerase, may help the cell compensate for the increased difficulty of melting the relaxed DNA template. What is the basis of supercoiling sensitivity? Most of the well controlled analyses of supercoiling-sensitive promoters, notably of the lacp s and ilv G P [ 18 , 49 - 51 ], were done on plasmids in vitro . The more relevant issue is promoter regulation on chromosomes in vivo , where other factors may dominate. The CRP protein increases lac operon transcription at the low to moderate superhelicities found in vivo , and the nucleoid-associated protein IHF is implicated in the supercoiling sensitivity of the ilvGMEDA operon [ 52 ]. Also, the relative levels of the nucleoid-associated proteins IHF, H-NS and, especially, Fis, can influence the local topology of DNA and accordingly affect transcription of nearby promoters [ 53 - 55 ]. We found no significant enrichment of genes regulated by IHF, H-NS, or Fis in our list of SSGs. However, we found that chromosomal relaxation affected different promoters to varying extents, and it is possible that the effect of changes in supercoiling may be amplified or attenuated at specific promoters by the actions of DNA-binding proteins. Finally, the proximity of genes to surrounding promoters and other barriers to supercoil diffusion may affect the response to supercoiling. For example, the modulation of the Salmonella leu-500 promoter by supercoiling requires that the promoter is either on the chromosome or on a plasmid anchored to the cell membrane by transcription and translation of a gene such as tetA [ 23 ]. Further analysis of supercoiling-sensitive promoters will be more straightforward with the set of genes identified in this paper and our finding that relaxation-induced genes have an enriched AT content in the promoter and initially transcribed sequences. It is striking that there are so many relaxation-induced genes that are relatively repressed when the chromosome is (-) supercoiled. This is surprising because (-) supercoiling should favor formation of an open promoter complex. The promoter regions of many of the genes induced by relaxation are AT rich, which will make it easier to form an open promoter complex even when the DNA is relaxed and the energy required is greater. Alternatively, the difference in AT content could reflect structural features such as curvature or flexibility. Curved sequences of DNA can influence the position of plectonemic supercoils, and thus could serve to localize a promoter sequence to the apex of a superhelical loop [ 56 ]. We note that the AT richness for the relaxation-induced genes extends on both sides of the transcriptional start site. It has been previously shown that promoter activity can be regulated by the initial transcribed sequence [ 57 , 58 ]. Moreover, in their analysis of the gyrA and gyrB promoters, Menzel and Gellert [ 14 ] found that base-pairs downstream of the transcriptional start were important for the supercoiling sensitivity of these promoters. These authors proposed that promoter clearance may be the rate-limiting step during relaxation-induced transcription of gyrA and gyrB . Promoter clearance has also been invoked in the mechanism of supercoiling sensitivity of some promoters in vitro [ 51 ]. As our group of relaxation-induced genes is AT rich over this region, we can extend this hypothesis to transcription of many relaxation-induced genes in vivo , and propose that promoter clearance is generally a key regulatory step for supercoiling sensitive transcription. The AT-rich regions of our relaxation-induced genes extend downstream of the translational start site, and thus may involve transcription elongation in addition to promoter clearance. There is growing appreciation of the regulation of transcription elongation [ 59 - 61 ]. The AT-rich regions deep within the coding sequence of relaxation-induced genes may reflect such regulation; easily melted regions of DNA may facilitate the continued movement of RNA polymerase along a relaxed, covalently closed template. At a given level of (-) supercoiling, there is likely to be an optimum AT content that facilitates both unwinding and subsequent closure of the transcription bubble. This hypothesis is strengthened by the fact that the genes with the opposite response, the relaxation-repressed genes, have a significantly depressed AT content over the same region. The SSGs are useful as topological probes of the chromosome in living cells. While the SSGs are scattered throughout the chromosome, they are not evenly spread, but rather have regions of high and low density. The SSGs are plotted on a chromosomal map in Figure 6 . The density of SSGs as a percentage of all genes in a 20-kb region varies from 2% to more than 20%. The regions with high SSG density may reflect spatial covariations in transcription which were recently described in the E. coli chromosome [ 62 ]. The distribution of the SSGs may also be influenced by the organization of the chromosome into topologically separate domains of supercoiling. We have already used the SSGs as local reporters of supercoiling to test the domains hypothesis. In recent work, we monitored expression from the SSGs after cleaving the chromosome with a restriction enzyme, and found that the SSGs accurately reported the resulting relaxation of the chromosome [ 63 ]. Relaxation diminished rapidly with distance from a restriction site, indicating that there are about 450 topologically separate domains in the chromosome. We also monitored transcription from the SSGs during replication in synchronized cells [ 64 ]. Here we found that the relaxation-induced and relaxation-repressed genes reported that supercoiling is re-established very quickly after the passage of the replication fork, again consistent with a large number of topological domains. Thus, the SSGs are not only a useful tool to study promoter regulation and the physiological effects of supercoiling changes, but also can lead to new findings about chromosome structure. Conclusions We have shown that supercoiling acts as a transcription factor, with positive and negative effects on specific genes while leaving the majority of the genome unchanged. Like other transcription factors such as TrpR [ 65 ] and ArcA-P [ 66 ], supercoiling affects transcription from a wider array of genes than at first anticipated. The 306 genes that we identified as robust SSGs are classified into many different functional groups [ 67 ], including transcriptional regulators and genes in the SOS, PhoB and stringent-response regulons [ 68 ]. Transcriptional changes from the SSGs will affect a variety of transcriptional and regulatory networks, and thus supercoiling level is a global regulator that can affect a wide array of processes in the cell. As the topology of the chromosome is affected by anoxia, ionic strength and growth conditions, the cell can use supercoiling levels both to sense the environment and to effect appropriate transcriptional responses. Materials and methods PCR materials and conditions Amino- and carboxy-terminal primers for protein-coding open reading frames (ORFs) of E. coli K-12, strain MG1655 (Sigma-Genosys), were generously supplied by Fred Blattner (University of Wisconsin) and Carol Gross (University of California San Francisco). ORFs were amplified from MG1655 genomic DNA using ExTaq polymerase (PanVera) and failed PCR reactions were attempted again using Platinum Taq (Invitrogen) or previous successful reactions as the DNA template. Ninety-six percent of the ORFs were successfully amplified. PCR conditions were set according to those supplied with the primers. DNA was precipitated with isopropanol and prepared for microarray printing as described in [ 69 ]. We did not include the RNA-coding genes on the arrays because primers for these genes were not initially available, though we note that some genes, such as tyrT , have been shown to respond to changes in supercoiling [ 70 ]. Microarray printing and processing Detailed instructions on slide preparation, microarray printing and processing microarrays can be found online [ 69 ]. 384-well plates were dried down between prints and resuspended in deionized water each time after the first print. RNA preparation and microarray hybridization E. coli cells were grown with shaking in LB media to an OD 600 = 0.45-0.55 at 37°C, or at 30°C for temperature-sensitive strains. Samples of cells were withdrawn at intervals and added to a 1/10 volume of either 95% ethanol plus 5% phenol or 2 M NaN 3 to stop transcription. Cells were then quickly harvested by centrifugation in a microcentrifuge. The supernatant was aspirated and the pellets frozen in liquid N 2 . Total RNA was prepared using the Qiagen RNeasy mini kit, except that 4 mg/ml lysozyme and a 30 sec incubation was used in the first step. For each microarray, 20 μg total RNA was primed with 1-2 μg of random hexamers and labeled by reverse transcription in the presence of Cy3- and Cy5-conjugated dUTP (Amersham Biosciences). For each experiment or condition, a Cy5-labeled experimental sample was combined with a Cy3-labeled reference sample and hybridized to a processed microarray as described [ 69 ]. After 5-7 h hybridization, microarrays were washed and scanned at 10 μm resolution with a GenePix 4000A scanner (Axon Instruments). Image processing Scanned array images were visually inspected, and non-uniform spots were excluded from further analysis. The background was subtracted from the images that were then (median) normalized using the Scanalyze 2.0 program, v. 1.44 (Michael Eisen, Lawrence Berkeley National Laboratory) such that the total fluorescence in each channel was equal. Data analysis We tested several methods of imputation to estimate the values of spots missing due to hybridization defects (described in [ 71 ]), and after error analysis of the different methods we chose the weighted mean of K-nearest neighbors for K = 20. With this method we obtained a total of 4,003 genes, or 93% of the total number of E. coli genes, that could be considered for further study. Because we were interested in changes in expression levels due to variations in supercoiling rather than to drug or genetic effects, we used a two-sample comparison approach (comparing the mean over all relaxation experiments with that of the control experiments) rather than a factorial analysis approach. We tested two commonly used methods to determine differentially expressed genes in the comparison of two samples. We found that the method of Dudoit et al [ 38 ], which controls the family-wise error (that is, the probability of finding at least one false positive) was slightly more stringent for our data than that developed by Tusher et al [ 37 ]. Northern analysis Samples were run on formaldehyde-MOPS 1% agarose gels and blotted onto a nylon membrane [ 72 ]. 32 P-labeled DNA probes for gyrB and asnB (as a loading control) were synthesized from their respective PCR products, and radioactivity was quantified by a phosphorimager. Assays of DNA topology Plasmid DNA was extracted from cells by the alkaline lysis method [ 72 ] or the Qiagen spin miniprep kit. The norfloxacin-resistant mutants and the gyrB234 mutant are in a C600 strain background, but all strains used have been described in greater detail elsewhere [ 26 , 35 , 73 ]. To increase the intracellular concentration of novobiocin, we used a ΔacrA strain that greatly slows drug efflux [ 74 ]. The superhelical density, σ, of pBR322 was determined by band counting [ 75 ] from the mean of the topoisomer distributions to a relaxed, covalently closed reference plasmid (σ = 0) which had been relaxed with calf thymus topoI. σ of pBR322 was calculated with the formula σ =Δ Lk/Lk 0 , where Lk 0 for pBR322 = 4,361 bp/10.5 bp/turn = 415. Samples were run on parallel 20-cm gels containing 0, 2.8 or 10 μg/ml chloroquine for 18-26 h at 48-52 V with constant buffer recirculation, which allowed visualization of the entire distribution of topoisomers. Gels were southern blotted [ 72 ], and hybridized with a 32 P-labeled probe made from random priming of pBR322. Radioactive blots were quantified using a phosphorimager. Microarray validation We tested the validity of our microarrays in three ways. First, we compared gene expression ratios measured with microarrays to values obtained by northern hybridization. We measured induction ratios for gyrB by both methods 5 min after addition of the gyrase inhibitor novobiocin to ΔacrA cells at 5, 20, 50 and 200 μg/ml. The microarray ratios for these concentrations were 2.3, 4.8, 4.9 and 6.3, respectively, while the ratios from northern hybridizations were 2.8, 4.7, 4.7 and 5.1. Second, as an internal control we compared the transcription of genes in 153 known polycistronic operons. We found no operons with genes that changed expression more than 1.5-fold in opposite directions (data not shown). Third, we compared two identically grown cultures with the same microarray (see Additional data file 4). We used two strains that were isogenic, except that one had point mutations conferring norfloxacin resistance on gyrase and topo IV. The correlation coefficient of the gene-expression levels was 0.982, indicating the negligible variation between the cultures. In contrast, when we treated cells with the gyrase inhibitor norfloxacin (see Additional data file 4), the correlation coefficient with respect to the untreated cells was only 0.391 and hundreds of genes showed large differences in expression. We conclude that gene-expression changes resulting from slight genotypic changes or experimental repeats were negligible compared with the changes resulting from topoisomerase inhibition, and that the E. coli microarrays are a reliable method for quantifying these changes. Selection of supercoiling-sensitive genes We limited the list of SSGs to those whose expression difference between treatments and controls was statistically significant ( p -values < 0.05) over a total of 35 experiments, in which DNA gyrase was inhibited with novobiocin, norfloxacin or by a mutation that rendered gyrase temperature-sensitive. Next we determined the correlation of gene expression with the σ of a reference plasmid in the same cells. To calculate the correlation of gene expression to plasmid supercoiling, we created a model profile made up of the ratio of plasmid σ in each experiment to plasmid σ in the (supercoiled) reference for that experiment (Table 1 ). The maximum ratio was scaled to 2.5, representing a σ of 0 (complete relaxation) and the minimum ratio was scaled to 1, representing native supercoiling levels (-0.06). The model repression profile is simply the inverse of the model induction profile. Changes of the arbitrary scaling values did not alter the results. Correlation coefficients in Figure 4 were calculated with respect to those 13 arrays only. Additional data files The following additional data files are available with the online version of this paper: Additional data file 1 contains data on the induction of the SOS response to DNA damage; Additional data file 2 contains gene-expression ratios for all genes across all experiments; Additional data file 3 contains gene-expression ratios for supercoiling-sensitive genes across all experiments; Additional data file 4 contains data on the reproducibility of microarray measurement of RNA levels. Supplementary Material Additional data file 1 Data on the induction of the SOS response to DNA damage Click here for additional data file Additional data file 2 Gene-expression ratios for all genes across all experiments Click here for additional data file Additional data file 3 Gene-expression ratios for supercoiling-sensitive genes across all experiments Click here for additional data file Additional data file 4 Data on the reproducibility of microarray measurement of RNA levels Click here for additional data file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545778.xml
555557	Inflammatory cytokine levels correlate with amyloid load in transgenic mouse models of Alzheimer's disease	Background Inflammation is believed to play an important role in the pathology of Alzheimer's disease (AD) and cytokine production is a key pathologic event in the progression of inflammatory cascades. The current study characterizes the cytokine expression profile in the brain of two transgenic mouse models of AD (TgAPPsw and PS1/APPsw) and explores the correlations between cytokine production and the level of soluble and insoluble forms of Aβ. Methods Organotypic brain slice cultures from 15-month-old mice (TgAPPsw, PS1/APPsw and control littermates) were established and multiple cytokine levels were analyzed using the Bio-plex multiple cytokine assay system. Soluble and insoluble forms of Aβ were quantified and Aβ-cytokine relationships were analyzed. Results Compared to control littermates, transgenic mice showed a significant increase in the following pro-inflammatory cytokines: TNF-α, IL-6, IL-12p40, IL-1β, IL-1α and GM-CSF. TNF-α, IL-6, IL-1α and GM-CSF showed a sequential increase from control to TgAPPsw to PS1/APPsw suggesting that the amplitude of this cytokine response is dependent on brain Aβ levels, since PS1/APPsw mouse brains accumulate more Aβ than TgAPPsw mouse brains. Quantification of Aβ levels in the same slices showed a wide range of Aβ soluble:insoluble ratio values across TgAPPsw and PS1/APPsw brain slices. Aβ-cytokine correlations revealed significant relationships between Aβ1–40, 1–42 (both soluble and insoluble) and all the above cytokines that changed in the brain slices. Conclusion Our data confirm that the brains of transgenic APPsw and PS1/APPsw mice are under an active inflammatory stress, and that the levels of particular cytokines may be directly related to the amount of soluble and insoluble Aβ present in the brain suggesting that pathological accumulation of Aβ is a key driver of the neuroinflammatory response.	Background Alzheimer's disease is a progressive neurodegenerative disorder characterized by intra-cellular abnormally phosphorylated tau protein and extra-cellular beta amyloid plaques. It has been suggested that inflammation may be a key player in the pathophysiology of AD as evidenced by epidemiological studies which have revealed that the long term use of non-steroidal anti-inflammatory drugs reduces the risk of developing AD [ 1 - 3 ]. Transgenic mouse models of Alzheimer's disease that over-express β-amyloid (Aβ) exhibit significant cerebrovascular inflammation and microgliosis around areas of plaque deposition [ 4 - 7 ]. Chronic administration of ibuprofen can reduce plaque pathology and brain Aβ levels in these animal models of AD [ 8 , 9 ]. There are numerous reports of increased levels of cytokines in the brains of Alzheimer's disease patients, and in transgenic mouse models of Alzheimer's disease [ 10 - 12 ]. However, all these reports have focused on a small number of cytokines within the same sample. It is not clear which cytokines are key in promoting and maintaining the inflammatory environment in the AD brain. Furthermore, it is unclear which Aβ species (1–40, 1–42, soluble or insoluble) are most closely related to cytokine levels. Multiplex technology enables the simultaneous quantification of many cytokines within a single sample. By examining different mouse models of AD using multiplex technology, it is possible to more clearly characterize the particular cytokines which maintain the inflammatory environment and to relate them to particular forms of Aβ (1–40, 1–42, soluble or insoluble). There is considerable debate over which length of Aβ and which conformations are most potently toxic. Recently, specific oligomeric forms have been shown to be most toxic to neurons. These soluble species of Aβ differ from the higher-molecular-weight aggregated insoluble forms that are found precipitated in the AD patient and mouse brain. This study sought to determine whether soluble or insoluble Aβ fractions were most closely related to cytokine levels. Materials and methods Organotypic brain slice cultures Mouse brain slice cultures were prepared as previously described [ 29 ]. Briefly, 15-month-old PS1 (M146L), TgAPPsw (K670M / N671L), PS1/APPsw and wildtype littermates were humanely euthanized and the brains extracted under sterile conditions. One-mm-thick brain slices were sectioned from co-ordinates 1 to -4 from bregma using a mouse brain slicer. Sections were cultured in neurobasal medium with 5% B27 supplement (Gibco-Invitrogen, CA) and Penicillin-Streptomycin-Fungizone mixture (Cambrex Corp., NJ). After 40 hours, media was collected for quantification of cytokine levels. Multi-plex cytokine array analysis was performed using the Bio-plex protein multi-array system, which utilizes Luminex-based technology [ 13 ]. For the current experiments, a mouse 12-plex assay was used according to the recommendations of the manufacturer (BioRad, CA). Measurement of Aβ levels in brain slices Brain slices were washed with PBS (BioSource, CA), and 300 μl of lysis buffer was added. Lysis buffer consisted of mammalian protein extraction reagent (Pierce-Endogen, IL) with 1X protease inhibitor cocktail XI (Calbiochem, CA), 100 μM Sodium Orthovanadate, and 1 μM Phenylmethylsulfonyl Fluoride (PMSF) (Sigma-Aldrich, MO). The resulting mixture was sonicated using a sonic dismembrator (Fisher Scientific, PA) Protein content in each slice was determined using the bicinchoninic acid (BCA) protein reagent kit (Pierce-Endogen, IL), as per the manufacturers protocol. Insoluble Aβ was extracted using 70% formic acid as previously published [ 14 ]. Aβ content in brain slices was determined using human Aβ 1–40 and Aβ 1–42 ELISA detection kits (Biosource, CA), as per the manufacturers protocol. Statistical analyses For statistical analyses, ANOVA and t-tests were performed where appropriate using SPSS for Windows release 10.1. Hierarchical cluster analysis of Aβ-cytokine data from brain slices were performed with the R program . A correlation matrix was constructed using the raw data and subsequently converted to a distance matrix by subtracting each element in the correlation matrix from 1. The distance matrix was used as the dissimilarity matrix for building an hierarchical cluster using the averaging method. The resulting dendrogram consists of closely related members under the same node. The farther one needs to traverse across the tree to reach another member, the higher the dissimilarity represented. The distance from the base in the y-axis represents dissimilarity or 1-r, where r is the correlation co-efficient. Results Cytokine production by organotypic brain slice cultures Cytokine production was evaluated by multi-plex cytokine array analysis using the cell culture supernatant of organotypic brain slice cultures from control, PS1 (Presenilin 1 mutant heterozygotes), TgAPPsw, and TgPS1/APPsw mice at 15 months of age. We chose non-transgenic littermates as controls for the TgAPPsw mice and the PS1 animals as controls for the PS1/APPsw mice as the PS1 animals were the littermates of the PS1/APPsw mice. There were no significant differences in cytokine production between control slices and PS1 slices showing that PS1 over-expression does not directly induce inflammatory events. Compared to control slices, production of IL-1α, TNF-α, GM-CSF and IL-6 was increased in TgAPPsw slices (figs. 1 , 2 ). Compared to TgAPPsw slices, PS1/APPsw brain slices produced significantly more IL-12p40, IL-1β, IL-1α, TNF-α, GM-CSF and IL-6. Across control, TgAPPsw, and PS1/APP transgenic brain slices, there was a graduated increase in IL-1α, TNF-α, GM-CSF and IL-6. Figure 1 Cytokine production by brain slices from transgenic mouse models of AD at 15 months of age. Freshly harvested brain slices were incubated in neurobasal medium with B27 supplement. Media was collected after 24 hours, and cytokine levels measured. Mean concentrations (N = 15) +/- standard error are expressed in picograms per milligram of protein. P < 0.05 was considered statistically significant. Figure 2 Cytokine production by brain slices from transgenic mouse models of AD at 15 months of age. Freshly harvested brain slices were incubated in neurobasal medium with B27 supplement. Media was collected after 24 hours, and cytokine levels measured. Mean concentrations (N = 15) +/- standard error are expressed in picograms per milligram of protein. P < 0.05 was considered statistically significant. Correlation between Aβ level and cytokine production by transgenic mouse brain slices Quantification of amyloid levels in brain mouse slices revealed that PS1/APPsw mice produce significantly more total Aβ as compared to TgAPPsw mice at the same age, and levels of insoluble and soluble Aβ (both 1–40 and 1–42) correlated well with each other (Table 1 ). Analysis of the ratio of soluble:insoluble Aβ revealed a wide range of values across the TgAPPsw and PS1/APPsw mouse brain slices, with a 15.3-fold variance for Aβ 1–40 and a 5.4-fold variance for Aβ 1–42 (for Aβ 1–40, comparison of soluble:insoluble ratios revealed an average difference of 3.9 fold, and an average 1.7-fold difference for Aβ 1–42). Table 1 Quantification of Aβ levels in TgAPPsw and PS1/APPsw mouse brain slices. Data expressed as picograms/mg protein, mean ± S.E.M. for 13 determinations. TgAPPsw PS1/APPsw Soluble Aβ1–40 331.15 ± 35.36 4957.79 ± 322.30 Soluble Aβ1–42 68.11 ± 6.82 1644.29 ± 90.30 Insoluble Aβ1–40 67619.38 ± 7089.61 4095442 ± 409212.3 Insoluble Aβ1–42 6837.22 ± 2741.70 286463.3 ± 31395.63 Although all the cytokines that changed in the transgenic brain slices were correlated with increases in Aβ levels, some showed a closer relationship than others to Aβ levels (Figs. 3 , 4 , and 5 ). A table of r-correlation values is given in Additional file 1 . It is important to note that the dendrograms depict the closeness of a correlation between a particular cytokine and Aβ levels, and that all the members in the dendrograms are in fact highly correlated with Aβ levels (1% significance was considered as r >= 0.496, and 5% significance was considered as r >= 0.388). IL-4 and IL-5 were not produced in detectable amounts, were therefore omitted from the dendrograms. Of all the cytokines, IL-12p40 showed the strongest correlation with levels of both Aβ1–40 and 42 (soluble or insoluble). IL-1α and IL-1β were also highly correlated with Aβ1–40 and 42 (soluble or insoluble). Figure 3 Dendrogram correlations of Aβ1–40 and Aβ1–42-cytokine relationships. Closely related members appear under the same node. The farther one needs to travel across the tree to reach another member, the greater the dissimilarity. Figure 4 Dendrogram correlations of Total Aβ (Aβ1–40+Aβ1–42)-cytokine relationships. Closely related members appear under the same node. Total Aβ levels were calculated by adding soluble and formic acid extracted Aβ. The farther one needs to travel across the tree to reach another member, the greater the dissimilarity. Figure 5 Dendrogram correlations of (Aβ1–42:40 ratio)-cytokine relationships. Total Aβ1–42:40 ratio's were calculated for both soluble and formic acid extracted Aβ. Closely related members appear under the same node. The farther one needs to travel across the tree to reach another member, the greater the dissimilarity. Discussion Levels of both peripheral and local CNS cytokines are elevated in AD patients, indicating that there is cellular activation occurring in response to inflammatory stimuli [ 15 - 20 ]. However, there is still considerable debate over exactly what is triggering this inflammation. Studies using mouse models of AD have shown that ibuprofen is effective in reducing plaque pathology and also in improving behavioral deficits characteristic of these transgenic models [ 8 , 21 ]. The transgenic mouse models used to study AD exhibit some of the pathological features seen in the AD patient brain and show an increased production of inflammatory markers such as COX-2, PGE 2 and also increased levels of the pro-inflammatory cytokines IFN-γ and IL-12, TNF-α, IL-1α, IL-1β and IL-6 [ 12 , 22 ]. Pathological analysis of tissue from AD patients and from mouse models of AD shows that there is extensive astrocytic and microglial activation around areas of Aβ plaque deposition [ 6 , 7 ]. In addition, the chronic use of non-steroidal anti-inflammatory drugs (NSAIDs) has been associated with a reduced risk of developing AD [ 23 , 24 ], suggesting that inflammation is an important contributor to the pathophysiology of AD. One aim of this study was to create a cytokine expression profile for organotypic brain slice cultures from transgenic mouse models of Alzheimer's disease, and to further relate this increase to the level of Aβ present in the brain. Another purpose of our study was to determine whether inflammatory events may be correlated with the accumulation of particular forms of Aβ; either soluble or insoluble. In the current study, we used the organotypic brain slice culture model to assess multiple cytokine production in the culture medium surrounding brain slices from transgenic mice that are engineered to over-produce Aβ. Cytokine production from 15-month-old control, PS1, TgAPPsw and PS1/APPsw mouse brain slices was assessed using the Bioplex cytokine multi-array system. Cytokine levels were not significantly elevated in PS1 brain slices compared to control slices, indicating that the PS1 (M146L) mutation does not have a significant impact on cytokine production. No significant change in the production of IL-4 and IL-10 was observed in the brains of these transgenic mice compared to their respective controls, indicating the absence of an anti-inflammatory response. All of the cytokines that were increased in the TgAPPsw brain slices (IL-1α, TNF-α, GM-CSF and IL-6) were further increased in the PS1/APP brain slices. This suggests that the presence of these inflammatory molecules is related to the amount of β-amyloid protein present, in agreement with a pro-inflammatory effect of Aβ [ 25 - 29 ]. A recent report has also shown increases in IL-1β, IL-6 and TNFα in-vivo after intra-cerebral administration of fibrillar Aβ into rat brain [ 30 ]. In order to further understand the correlation between the amount of Aβ and cytokine levels in the brains of transgenic mice, levels of both soluble and insoluble (formic acid-extracted) Aβ1–40 and 1–42 were quantified in the same slices from which cytokine production was measured, allowing a direct correlation of Aβ-cytokine levels. Levels of soluble and insoluble Aβ1–40 correlated well with each other, and the same was observed for Aβ1–42. As expected, quantification of Aβ levels generally revealed significantly higher amyloid levels in the PS1/APPsw mouse brain slices compared to TgAPPsw (for soluble Aβ, approximately 15 fold more Aβ1–40, and 20 fold more 1–42) but there was considerable slice-to-slice variation in soluble and insoluble Aβ levels within and between genotypes. The TgAPPsw and PS1/APPsw mice express equal levels of the APPsw molecule, but the PS1/APPsw model produces greater levels of Aβ and develops plaques at an earlier age (10 weeks) [ 31 - 33 ]. This increased deposition of Aβ in the PS1/APPsw mouse is due to a PS1 mutation, resulting in increased production of Aβ1–42 [ 34 - 36 ]. The Aβ data in the current report found a significant range of values for soluble:insoluble Aβ ratios between brain slices. This broad spread of values allowed correlation with equally wide ranges of cytokine production. This approach of examining Aβ-cytokine correlations within the same slices in the same aged animals eliminated the confounding factor of age related changes in cytokine production. Both Aβ1–40 and 1–42 correlated closely with all the cytokines that changed in the brain slices, but the correlation was particularly striking with IL-12p40. IL-12 is a hetero-dimeric cytokine which can comprise two subunits; IL-12p40 and IL-12p35. It is produced mainly by monocytes and macrophages and is a crucial factor in directing the T-cell response to infection, by inducing a Th1-type cytokine response. Our data agrees with that of previous reports showing that IL-12p40 is strongly up-regulated in-vitro (in response to an inflammatory stimulus) and in-vivo in the cerebral cortex of TgAPPsw mice [ 12 , 37 , 38 ]. IL-1, which was increased in the transgenic brain slices, is a major immune-response molecule functioning in the periphery and brain. The family comprises three related proteins (IL-1α, IL-1β and IL-1 receptor antagonist (IL-1ra)). IL-1α and IL-1β are two different isoforms of IL-1 that have similar affinities for their receptor IL-1R, and therefore have similar activities. Both are capable of inducing inflammatory cascades in-vivo and in-vitro, and it has been shown that they are capable of up-regulating expression of astrocyte-derived S100B and APP [ 39 , 40 ]. It has been shown that IL-1β can promote β-secretase cleavage of APP in human astrocytes and thereby increase production of Aβ1–40 and 1–42 [ 41 , 42 ]. It is also known that accumulation of plaques and the formation of neurofibrillary tangles are correlated with increased IL-1 levels in the AD brain [ 43 - 45 ]. Certain polymorphisms of IL-1A (the gene for IL-1α) are associated with late onset AD, although there is controversy as to whether all IL-1 gene polymorphisms represent risk factors for AD [ 46 - 50 ]. Microglia, in particular, have been shown to locally up regulate IL-1α at both the protein and mRNA level when inflamed, a situation that occurs in chronic disease states such as AD [ 51 ]. Both IL-1α and IL-1β can enhance the translation of APP mRNA in human astrocytes [ 52 ]; an up-regulation of IL-1α/β production in-vivo could therefore increase Aβ production, and an inflammatory cycle with increased Aβ levels may further increase IL-1α/β production. The Aβ 1–42:40 ratio is also of considerable interest in relation to cytokine levels and although there are currently no studies correlating Aβ 1–42:40 ratio with cytokine levels in-vivo, certain reports have suggested that cytokines can modulate Aβ production [ 53 - 55 ]. PS1 mutations are known to cause a shift in the production of Aβ species, favoring the production of Aβ1–42 over 1–40 and causing an increase in the Aβ1–42:40 ratio [ 56 ]. Since TNF-α correlated better with the level of Aβ1–42 than with that of Aβ 1–40, and correlated particularly well with the Aβ1–42:40 ratio in our study, TNF-α levels may be partly determined by this ratio. Higher levels of Aβ1–42 can promote the formation of toxic oligomers [ 57 - 59 ], and it therefore seems possible that the increased level of Aβ oligomers in PS1/APP mice (compared to APPsw) and the level of oligomeric forms present in the brains of our transgenic mice may be related to the amount of TNF-α being produced. It is important to consider the nature of the exact form of Aβ that may be most responsible for the inflammatory events seen in AD brains. Aβ can exist in various forms (monomeric, dimeric, oligomeric and fibrillar), but it is not yet clear which of these forms are most potent in inducing inflammatory cellular responses [ 57 , 60 , 61 ]. This is of interest because the oligomeric forms of Aβ which are thought to be the most toxic are produced more readily by Aβ1–42 (for review see [ 62 ]). Future studies will assess the relative proportions of monomers/dimers, oligomers or fibrils occurring in these mice brains and their relationship with the cytokine increases observed. List of abbreviations AD: Alzheimer's disease APP: Amyloid precursor protein APPsw: Amyloid precursor protein Swedish mutation PS1: Presenilin 1 Aβ: Beta-amyloid Tg: Transgenic TNF: Tumor necrosis factor IL-x: Interleukin-x IL-1ra: Interleukin-1 receptor antagonist GM-CSF: Granulocyte macrophage colony stimulating factor PBS: Phosphate buffered saline COX-2: Cyclo-oxygenase-2 PGE2: Prostaglandin E2 IFN: Interferon NSAID: Non-steroidal anti-inflammatory drug Competing interests The author(s) declare that they have no competing interests. Authors' contributions NP carried out the in-vitro brain slice assays, processed brain tissues, performed the Bio-plex assay, ELISAs and drafted the manuscript. DP conceived the design of the study, carried out Bio-plex assays, performed statistical analyses and aided in manuscript preparation. VM analyzed data and constructed dendrograms. AQ aided in ELISA and Bio-plex assays and collected mouse brain tissues. FC oversees management of the mouse colonies. MM aided in manuscript preparation and gave critical analysis of the manuscript. Supplementary Material Additional File 1 Correlation table of levels of different β-amyloid species with cytokines in transgenic mouse models of Alzheimer's disease. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555557.xml
548512	The effect of IL-13 and IL-13R130Q, a naturally occurring IL-13 polymorphism, on the gene expression of human airway smooth muscle cells	Background Growing evidence shows that interleukin 13 (IL-13) may play an essential role in the development of airway inflammation and bronchial hyper-responsiveness (BHR), two defining features of asthma. Although the underlying mechanisms remain unknown, a number of reports have shown that IL-13 may exert its deleterious effects in asthma by directly acting on airway resident cells, including epithelial cells and airway smooth muscle cells. In this report, we hypothesize that IL-13 may participate in the pathogenesis of asthma by activating a set of "pro-asthmatic" genes in airway smooth muscle (ASM) cells. Methods Microarray technology was used to study the modulation of gene expression of airway smooth muscle by IL-13 and IL-13R130Q. TaqMan™ Real Time PCR and flow cytometry was used to validate the gene array data. Results IL-13 and the IL-13 polymorphism IL-13R130Q (Arg130Gln), recently associated with allergic asthma, seem to modulate the same set of genes, which encode many potentially interesting proteins including vascular cellular adhesion molecule (VCAM)-1, IL-13Rα2, Tenascin C and Histamine Receptor H1, that may be relevant for the pathogenesis of asthma. Conclusions The data supports the hypothesis that gene modulation by IL-13 in ASM may be essential for the events leading to the development of allergic asthma.	Background Recent reports using murine models of allergic asthma have shown that the Th2 type cytokine IL-13 may play a critical role in the pathogenesis of asthma, either by regulating airway inflammation, mucus hyper-secretion or airway hyper-responsiveness [ 1 - 5 ]. Evidence suggests that the potential role of IL-13 in asthma may come from its aptitude to directly interact with airway resident cells, such as epithelial cells or airway smooth muscle (ASM) cells, as shown by the ability of IL-13 to stimulate a set of different pro-asthmatic genes including inflammatory cytokines such as thymus and activation-regulated chemokine (TARC), eotaxin, monocyte chemotactic protein-1 (MCP-1) as well as growth factors such as vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (bFGF) [ 6 - 10 ]. The ability of IL-13 to modulate ASM responsiveness to G-protein coupled receptor (GPCR) agonists, either by increasing contractile agonist-evoked calcium responses [ 11 ], and/or by impairing ASM responsiveness to β2-adrenoceptor stimulation [ 6 ], may also explain, at least in part, the putative role of IL-13 in allergen-associated BHR reported in animal models [ 1 - 4 ]. Previous reports have shown that other cytokines such as tumor necrosis factor alpha (TNFα) or interleukin (IL)-1β, may also participate in airway hyper-responsiveness by modulating ASM responsiveness to contractile GPCR agonists [ 12 - 14 ]. These data strongly support the current concept that cytokine modulation of ASM, an effector cell thought to solely regulate bronchomotor tone [ 12 ], may play an important role in the development of airway inflammation and bronchial hyper-responsiveness, the two main features of asthma. The molecular mechanisms by which IL-13 induces "pro-asthmatic responses" in ASM have not been clearly established. Identifying the expression profile of "pro-asthmatic" genes activated by IL-13 in ASM cells may therefore provide new insight into the design of novel therapeutic approaches for asthma. Using complementary molecular approaches, we investigated the effect of IL-13 on the transcription of "pro-asthmatic" genes in human airway smooth muscle cells (HASMC). The effect of IL-13 was compared to that of IL-13R130Q, a naturally occurring isoform resulting in a change from glutamine to arginine residues in the coding region that is associated with high serum IgE levels [ 15 ]. Interestingly, no report has yet investigated whether both IL-13 and IL-13R130Q share the same or have different biological activities. We found that both IL-13 and IL-13R130Q stimulate the same set of important genes that encode for proteins which may be clinically relevant for regulating airway hyper-responsiveness, airway inflammation and airway remodeling, key characteristics of asthma. Methods Cell Culture Human tracheas were obtained from lung transplant donors, in accordance with procedures approved by the University of Pennsylvania Committee on Studies Involving Human Beings. A segment of trachea just proximal to the carina was removed under sterile conditions and the tracheal muscle was isolated. The muscle was then centrifuged and resuspended in 10 ml of buffer containing 0.2 mM CaCl 2 , 640 U/ml collagenase, 1 mg/ml soybean trypsin inhibitor and 10 U/ml elastase. Enzymatic dissociation of the tissue was performed for 90 min in a shaking water bath at 37°C. The cell suspension was filtered through 105 μm Nytex mesh, and the filtrate was washed with equal volumes of cold Ham's F12 medium (Gibco BRL Life Technologies, Grand Island, NY) supplemented with 10% FBS (HyClone, Logan, UT) 100 U/ml penicillin (Gibco), 0.1 mg/ml streptomycin (Gibco), and 2.5 μg/ml fungizone (Gibco). Aliquots of the cell suspension were plated at a density of 1.0 × 10 4 cells/cm 2 . The cells were cultured in Ham's F12 media supplemented with 10% FBS, 100 U/ml penicillin, 0.1 mg/ml streptomycin and this was replaced every 72 h. Human ASM cells in subculture during the second through to fifth cell passages were used because, during these cell passages, the cells retain native contractile protein expression, as demonstrated by immunocytochemical staining for smooth muscle actin and myosin [ 16 ]. Unless otherwise specified, all chemicals used in this study were purchased from Sigma/Aldrich (St. Louis, MO). RNA isolation Total cellular RNA was isolated from IL-13 (50 ng/ml), IL-13R130Q (50 ng/ml) or control treated HASMC using the RNeasy mini kit (Qiagen, Inc. Valencia, CA) as per manufacturer's instructions. The IL-13 was purchased from R&D Systems (Minneapolis, MN) and the IL-13R130Q was generated in house at Centocor Inc. The quality and quantity of RNA was assessed using the Agilent 2100 Bioanalyzer (South Plainfield, New Jersey). Samples that demonstrated high quality (ratio of 28S rRNA and 18S rRNA is greater than 1.7) were submitted for microarray analysis. Microarray Processing A complimentary DNA (cDNA) microarray, or DNA chip (Target B), containing a total of 8159 human gene cDNA clones from Research Genetics (IMAGE consortium, Huntsville, AL), Incyte Genomics (Santa Clara, CA) and internal sources was used in this study. All clones have been verified by DNA sequencing and are printed as 2 independent spots on a given chip. Duplicate chips were used for each RNA sample. Non-linear normalization between duplicate chips allowed each clone to be averaged to a single intensity value for each RNA sample. RNA amplification, probe synthesis and labeling, cDNA chip hybridization and washing were performed as described previously [ 17 ]. Agilent Image Scanner was used to scan the cDNA chips (Agilent Technologies, Palo Alto, CA). Fluorescence intensity for each feature of the array was obtained by using ImaGene software (BioDiscovery, Los Angeles, CA). Microarray data analysis In this study, fifty one-color cDNA microarrays were used to profile gene expression in human airway smooth muscle cells from 3 donors stimulated with IL-13, or its variant IL-13R130Q at 2 time points (6 hr and 18 hr). Untreated samples from the same group of donors were used as control. The samples being analyzed are listed in Table 1 . Table 1 Summary of number of samples from each donor and treatments Time Donor Untreated IL-13 IL-13R130Q 6 hr Donor 1* - - - Donor 2 1 2 2 Donor 3 1 2 2 18 hr Donor 1 1 2 2 Donor 2 1 2 2 Donor 3 1 2 2 The samples from Donor 1 at the 6 hr time point were not included due to poor quality of RNA. Purified cDNA probes were hybridized to two microarrays, each containing two spots for each cDNA. Raw intensity data from the cDNA arrays were first normalized within each sample. Linear normalization and then nonlinear normalization was performed within each sample. Outlier intensity data points (greater than 1.4 fold away from the median of replicate measurements) were identified and removed from the data sets. The average intensity was generated by calculating the arithmetic mean of nonoutlier intensity values. The averaged intensity for each clone was further normalized across all samples. Chip-to-chip normalization was performed by dividing the averaged intensity of each clone by the 50.0 th percentile of all measurements in that sample. The intensity of each clone was then normalized to the median intensity of that clone in the untreated control group. The normalized intensity was then log transformed. Using Partek Pro™ 5.1, sources of variance, such as batch effects, were identified by Principle Component Analysis (PCA) and appropriate factors were taken into account in the Analysis of Variance (ANOVA). ANOVA was performed to identify the genes that were differentially expressed by cytokine stimulation. Treatment (IL-13 and IL-13R130Q), time (6 hour and 18 hour), and donor (1, 2, and 3) were the three main effects considered in ANOVA. P-value cutoff was 0.05. Benjamini and Hochberg false discovery rate (FDR) was performed for multiple testing correction. After comparing the gene lists from IL-13 and IL-13R130Q treatments, it was clear that these two treatments resulted in the regulation of the same set of genes. Subsequently, samples from these two treatments were combined and regarded as replicates in ANOVA. Furthermore, outliner samples in the data set were detected by PCA and removed to improve the detection power. As an alternative approach, fold change comparisons (cutoff = 1.5 fold) between a treated condition and the control were carried out within each donor by using GeneSpring™ 6.2 [ 18 ]. A gene was considered as reliably detected in a given condition if more than half of the replicates representing the same condition had a raw expression intensity of more than 50, CV smaller than 25%, and raw intensity being generated from 2 or more of the duplicate spots representing the clone. A pair-wise comparison between a treatment and its untreated control was performed only on the genes that were reliably detected in at least one condition of the pair. The genes that showed at least 1.5 fold differential expression in two or more donors were identified for each cytokine treatment at a time. Reverse Transcription and Real Time PCR 1 μg of total RNA from each of the IL-13 (50 ng/ml) or IL-13R130Q (50 ng/ml) or control treated HASMC were used for the reverse transcription (RT) reaction. The RT reaction was performed as per protocol using TaqMan ® RT reagents (Applied Biosystems) at 37°C for 120 min followed by 25°C for 10 min. Forty ng of cDNA per reaction were used in the Real Time PCR using the ABI Prism ® 7900 sequence detection system (Foster City, California). In the presence of AmpliTaq Gold DNA plolymerase (ABI biosystem, Foster City, California), the reaction was incubated for 2 min at 50°C followed by 10 min at 95°C. Then the reaction was run for 40 cycles at 15 sec, 95°C and 1 min, 60°C per cycle. Assays-on-Demand™ primers and probes (Applied Biosystems) were used in the PCR. The Real Time PCR data was analyzed using the standard curve method. Flow Cytometry Flow cytometry was performed as described previously with slight modifications [ 19 ]. Briefly, adherent cells treated with IL-13 for 24 hr were washed with PBS, detached by trypsinization (2 min, 37°C) and then washed with Ham's-F12 (10% FCS) media, centrifuged, and transferred to microfuge tubes (1.5 ml). Cells were incubated with anti-IL-13Rα2 (5 μg/ml, Santa Cruz Biotech) antibody followed by 1 hr incubation with a fluorescein isothiocyanate-conjugated secondary antibody (Jackson ImmunoResearch Laboratories; West Grove, PA). In parallel experiements, cells were incubated with the FITC-conjugated mouse anti-VCAM-1 antibody (2 μg/ml, Santa Cruz Biotech) for 1 h at 4°C. The cells were then centrifuged and resuspended in cold PBS in microfuge tubes. Samples were then analyzed using an EPICS XL flow cytometer (Coulter, Hialeah, FL). VCAM-1 and IL-13Rα2 levels were expressed as the increase in mean fluorescence intensity over un-stimulated cells. Results IL-13 regulates gene expression of HASMCs IL-13 may exert its deleterious effects in asthma by directly altering gene expression in airway resident cells such as epithelial cells or ASM cells [ 5 - 7 , 20 ]. In order to determine which genes are regulated by IL-13 in airway smooth muscle cells, we employed the cDNA microarray technology. We also wanted to ascertain if the effect of IL-13R130Q, a naturally occurring isoform of IL-13 and associated with high serum IgE levels [ 15 ], was any different than IL-13 in terms of modulating gene expression. The concentrations of IL-13 (10–100 ng/ml) used in our study were shown previously to stimulate gene expression in human ASM cells [ 7 , 8 , 10 ], although the in vivo relevance of these particular concentrations remains unknown. Three donors were used and two types of analyses were carried out (Fold change analysis; Statistical Analysis). Both IL-13 and IL-13R130Q generated a similar expression profile i.e., genes regulated by IL-13 were the same as those regulated by IL-13R130Q at the 1.5 fold cutoff. Table 2 lists genes of interest that were identified from analyzing the data and divides them into one of three categories. Genes involved in all three characteristics of asthma (airway inflammation, remodeling and bronchial hyper-responsiveness) were identified. Of particular interest are vascular cellular adhesion molecule (VCAM)-1, Tenascin C, IL-13Rα2 and Histamine Receptor H1. Table 2 Summary of genes up regulated by IL-13 and IL-13R130Q. Category Gene(s) Fold change Airway Inflammation Adhesion Molecules VCAM-1 ↑ 2 fold ALCAM Selectin P ligand Laminin B1 Chemokines Chemokine Ligand 2 Chemokine Ligand 11 Chemokine Ligand 26 Chemokine Ligand 27 Cytokine receptors IL-13 Rα2 ↑ 1.6 fold Interleukin 1 receptor Airway Remodeling Extracellular matrix Tenascin C ↑ 2 fold Tenascin R Collagen Type I Collagen Type VI Collagen Type III Fibulin 1 CD44 Cell proliferation Pim-1 eEF1A Cytokines PDGFC Retinoic acid Receptor Interferon beta 1 Bronchial Hyper-responsiveness Cytoskeletal constituants Vimentin Tropomyosin 1 Tropomyosin 2 Actin Calcium regulators Phospholipase D Calreticulin hGIRK1 TRPC4 TRPC6 Sphingosine kinase 1 Rho GDP dissociation inhibitor FKBP1A Receptor Histamine H1 receptor ↑ 1.3 fold The fold changes correspond to the genes in bold. Real Time PCR validation TaqMan™ Real Time PCR was used to validate VCAM1, IL-13Rα2, Tenascin C and Histamine Receptor H1. As shown in Figure 1A , VCAM1 was upregulated between 2 and 2.5 fold upon IL-13 or IL-13R130Q treatment at the 6 and 18 hour time points in both donors. This is comparable to the microarray data (Table 2 ). In Figure 1B , IL-13Rα2 mRNA is upregulated with IL-13 or IL-13R130Q. However, the upregulation is more pronounced at the 18 hour time point compared to 6 hour. In Figure 2A Tenascin C is upregulated with IL-13 and IL-13R130Q and in Figure 2B , Histamine Receptor H1 shows an upregulation of about 1.5 fold in both donors at both time points and with both treatments. Again, this is comparable to the microarray data (Table 2 ). Figure 1 Real Time PCR (Taqman ® ) analysis showing the level of A) VCAM1 B) IL-13Rα2 upon treatment of ASM from two donors with IL-13 or IL-13R13Q for 6 or 18 hrs. The quantity of each gene is normalized to 18S and relative to the untreated sample. Values shown are mean ± standard deviation from an n = 6. Figure 2 Real Time PCR (Taqman ® ) analysis showing the level of A) Tenascin C and B) Histamine receptor H1 upon treatment of ASM from two donors with IL-13 or IL-13R13Q for 6 or 18 hrs. The quantity of each gene is normalized to 18S and relative to the untreated sample. Values shown are mean ± standard deviation from an n = 6. Validation of VCAM-1 and IL-13Rα2 at the protein level In order to validate the modulatory effect of IL-13 on VCAM-1 and IL-13Rα2 genes at their protein level, flow cytometry was performed to confirm the up regulation of VCAM-1 and IL-13Rα2 in HASMC by IL-13. As shown in Figure 3 and 4 , IL-13 (10–100 ng/ml, 24 hr) differentially stimulates the expression of VCAM-1, with levels increasing in a dose-dependent manner, while IL-13Rα2 levels were identical at 10, 30 and 50 ng/ml. At 100 ng/ml IL-13, VCAM-1 and IL-13Rα2 levels were significantly increased by 20% and 35% over basal, respectively (n = 3, p < 0.05). Increases in VCAM-1 and IL-13Rα2 proteins by IL-13 nicely correlate with changes in mRNA levels (Figure 1A and Figure 1B ), suggesting that the IL-13 regulates the expression of inflammatory proteins via transcriptional mechanisms. Figure 3 Effect of IL-13 on VCAM-1 expression. ASM cells were incubated with the indicated concentrations of IL-13 for 24 hr. VCAM-1 expression was assessed by flow cytometry as described in Methods. Values shown are mean ± SEM and are significantly different from basal, n = 3 different experiments. P < 0.05 significant from untreated cell. Figure 4 Effect of IL-13 on IL-13Rα2 expression. ASM cells were incubated with the indicated concentrations of IL-13 for 24 hr. IL-13Rα2 expression was assessed by flow cytometry as described in Methods. Values shown are mean ± SEM and are significantly different from basal, n = 3 different experiments. *P < 0.05 significant from untreated cell. Discussion Recent evidence using various experimental approaches such as gene-deficient mice, soluble inhibitors or transgene overexpressing IL-13 in the airways have highlighted the critical role of IL-13 in the pathogenesis of allergic asthma, possibly due to its ability to regulate goblet cell metaplasia, mucus hypersecretion and airway hyper-responsiveness [ 1 , 3 , 21 ]. Few reports that used microarray analyses applied to cultured human ASM cells demonstrated that IL-13 differentially regulates a number of important genes that are relevant for the pathogenesis of asthma [ 7 , 10 ]. Although the functional relevance of such gene microarray analyses remains yet uncertain, these studies strongly suggest that IL-13 may be involved in the pathogenesis of asthma by directly modulating physiological responses of the ASM. Compared to the previous gene microarray reports [ 7 , 10 ], we did confirm the physiological relevance of the microarray data using two different experimental approaches. At least for four different genes, VCAM-1 (an adhesion protein), IL-13Rα2, Histamine H1 receptor (a G protein-coupled receptor) and Tenascin C (an extracellular matrix glycoprotein), there was a close correlation between the data obtained from gene microarray with those obtained by real time PCR analyses. In addition, we showed that IL-13 stimulates the expression of VCAM-1 and IL-13Rα2 at the protein level, showing the physiological relevance of the gene array data. It is interesting to note that no difference in gene expression profile were noticeable between cells exposed to IL-13 or IL-13R130Q, an IL-13 polymorphism recently found to be associated with elevated serum and allergen-specific IgE [ 15 , 22 ]. Our report is the first to suggest that this particular IL-13 polymorphism is equally effective as IL-13 in the transcriptional regulation of the genes examined in the present study. Our present study further supports the concept that IL-13 regulates the expression of different "pro-asthmatic" genes that are potentially important in the regulation of all three key features of asthma, i.e., airway inflammation, airway remodeling and bronchial hyper-responsiveness (for details see Table 2 ). Previous reports using cultured ASM cells also demonstrated that IL-13 can stimulate the expression of other pro-inflammatory proteins, such as eotaxin [ 8 , 23 , 24 ], TARC [ 9 ] or VEGF [ 25 ] and tenascin [present report and [ 10 ]]. Upregulation of tenascin C and R, glycoproteins that contribute to extracellular matrix structure [ 26 ], may play an important role in airway remodeling, a characteristic of chronic asthma. The stimulatory role of IL-13 on VCAM-1 may be important in asthma since VCAM-1 has been regarded as a key player in the development of airway inflammation [ 27 ]. The ability of IL-13 to increase the expression of different G-protein coupled receptors (GPCR), such as Histamine H1 [present report and [ 10 ]] or CysLT1 receptor [ 28 ] represents one potential mechanism by which IL-13 promotes airway hyper-responsiveness to GPCR agonists previously described both in vivo [ 1 , 3 , 21 ] or in vitro in isolated airways preparations [ 11 , 29 , 30 ]. Additional studies are needed to determine whether IL-13 also modulates ASM responsiveness to Histamine. The receptor complex by which IL-13 regulates cellular function comprises the IL-13Rα1, which binds IL-13 and forms a complex with the IL-4Rα to initiate signal transduction via the JAK/STAT6 pathway [ 31 ]. IL-13Rα2, the other cell surface protein binds IL-13 with high affinity but the complex is not functionally active. One previous report using transgenic mice showed that overexpressing IL-13 in the airways induced a marked increase in both IL-13Rα1 and IL-13Rα2 mRNA levels, mostly in epithelial cells and macrophages [ 32 ]. Our study is the first to show that the expression of IL-13Rα2 is also transcriptionally increased by IL-13 in ASM cells. Although the functional significance of such regulation remains unknown, it is possible that the newly induced IL-13Rα2 could function as a decoy receptor to limit IL-13 signaling in ASM cells. Additional studies are needed to further support this hypothesis. Conclusions These data further support the hypothesis that gene modulation by IL-13 in ASM may be essential for the events leading to the development of allergic asthma. Additional studies are clearly needed to define the transcriptional regulation of the different "pro-asthmatic" genes by IL-13, which may lead to novel therapeutic approaches for the treatment of allergic asthma. Authors' contributions FS, LL, YA and RAP participated in the conception and design of the study. FS coordinated the study and along with YA drafted the manuscript. OT performed the RNA isolation and flow cytometry experiments. CH and KL performed the microarray data analysis. MB performed the Real Time PCR experiments. DG and BA proof read the manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC548512.xml
554113	Protein domains and architectural innovation in plant-associated Proteobacteria	Background Evolution of new complex biological behaviour tends to arise by novel combinations of existing building blocks. The functional and evolutionary building blocks of the proteome are protein domains, the function of a protein being dependent on its constituent domains. We clustered completely-sequenced proteomes of prokaryotes on the basis of their protein domain content, as defined by Pfam (release 16.0). This revealed that, although there was a correlation between phylogeny and domain content, other factors also have an influence. This observation motivated an investigation of the relationship between an organism's lifestyle and the complement of domains and domain architectures found within its proteome. Results We took a census of all protein domains and domain combinations (architectures) encoded in the completely-sequenced proteobacterial genomes. Nine protein domain families were identified that are found in phylogenetically disparate plant-associated bacteria but are absent from non-plant-associated bacteria. Most of these are known to play a role in the plant-associated lifestyle, but they also included domain of unknown function DUF1427, which is found in plant symbionts and pathogens of the alpha-, beta- and gamma-Proteobacteria, but not known in any other organism. Further, several domains were identified as being restricted to phytobacteria and Eukaryotes. One example is the RolB/RolC glucosidase family, which is found only in Agrobacterium species and in plants. We identified the 0.5% of Pfam protein domain families that were most significantly over-represented in the plant-associated Proteobacteria with respect to the background frequencies in the whole set of available proteobacterial proteomes. These included guanylate cyclase, domains implicated in aromatic catabolism, cellulase and several domains of unknown function. We identified 459 unique domain architectures found in phylogenetically diverse plant pathogens and symbionts that were absent from non-pathogenic and non-symbiotic relatives. The vast majority of these were restricted to a single species or several closely related species and so their distributions could be better explained by phylogeny than by lifestyle. However, several architectures were found in two or more very distantly related phytobacteria but absent from non-plant-associated bacteria. Many of the proteins with these unique architectures are predicted to be secreted. In Pseudomonas syringae pathovar tomato , those genes encoding genes with novel domain architectures tended to have atypical GC contents and were adjacent to insertion sequence elements and phage-like sequences, suggesting acquisition by horizontal transfer. Conclusions By identifying domains and architectures unique to plant pathogens and symbionts, we highlighted candidate proteins for involvement in plant-associated bacterial lifestyles. Given that characterisation of novel gene products in vivo and in vitro is time-consuming and expensive, this computational approach may be useful for reducing experimental search space. Furthermore we discuss the biological significance of novel proteins highlighted by this study in the context of plant-associated lifestyles.	Background The Proteobacteria comprise a phylum of Gram-negative bacteria that includes an extraordinary diversity of lifestyles, ecology and metabolism. At one end of a spectrum are free-living organisms such as Pseudomonas aeruginosa , which has a relatively large genome that encodes enormous regulatory and metabolic flexibility, allowing it to colonise diverse niches. At the other extreme are highly specialised intracellular symbionts ( Buchnera species, Rickettsia species), whose small genomes have undergone reductive evolution and which lack many common metabolic and regulatory features. With the availability of complete genome sequences for many model plant-associated bacteria, we are particularly interested in how genome analyses can be used to gain insights into the mechanisms and evolution of associations between bacteria and plants. There are complete annotated genome sequences available for several phylogenetically diverse proteobacterial plant pathogens and symbionts, along with many of their non-pathogenic and non-symbiotic relatives. For example, among the alpha-Proteobacteria, complete genome sequences are available for the phytopathogen Agrobacterium tumefaciens [ 1 - 3 ], the nitrogen-fixing symbionts Bradyrhizobium japonicum [ 4 ], Mesorhizobium loti [ 5 ] and Sinorhizobium meliloti [ 6 , 7 ], the non-pathogenic free-living Caulobacter crescentus [ 8 ], and the animal pathogenic Rickettsia species [ 9 - 11 ]. Ralstonia solanacearum [ 12 ] is the sole completely sequenced plant pathogen amongst the beta-Proteobacteria, a division that also includes animal pathogens in the genera Neisseria [ 13 , 14 ] and Bordetella [ 15 ] and the free-living chemolithoautotroph Nitrosomonas europaea [ 16 ] whose genomes have been sequenced. Among the available complete genome sequences for the gamma-Proteobacteria are those of the plant pathogens Xylella fastidiosa [ 17 , 18 ], Xanthomonas campestris [ 19 ], Xanthomonas axonopodis [ 19 ] and Pseudomonas syringae pathovar tomato [ 20 ] as well as P. aeruginosa [ 21 ], which is an occasional pathogen of plants as well as animals. Each of these three divisions of the Proteobacteria contains a wide variety of different lifestyles, so it is logical to assume that bacteria-plant interactions have evolved independently in multiple separate Proteobacterial lineages. Ultimately the differences between these lifestyles are determined by the organisms' genes acting through their expressed proteins and RNAs. Given the abundance of complete genome sequence data now available, a high priority is to understand which features of an organism's proteome determine its lifestyle, and the evolutionary processes underlying environmental adaptation and evolution of novel traits. Two main sources have been proposed for the evolution and acquisition of novel traits by bacteria: (i) duplication, mutation and recombination of existing genes within a single lineage, and (ii) lateral gene transfer between lineages. A combination of both bioinformatic and experimental studies are needed to determine the relative importance of these two processes in the evolution of plant-associated lifestyles in bacteria. Evolution of new complex biological behaviours tends to arise (but not exclusively) by novel combinations of existing building blocks. The functional and evolutionary building blocks or units of the proteome are protein domains. Protein domains can be classified into families; examples of widely used classification schemes are those of Pfam [ 23 ] and SMART [ 24 ]. We hypothesised that systematic identification of proteins having domain architectures that are exclusive to plant-associated bacteria would identify good candidates for proteins with specific involvement in plant-microbial interactions, or in a plant-associated lifestyle, and would also generate insight into the distribution and evolution of novel traits in plant-associated bacteria. Results and discussion Hierarchical clustering of completely-sequenced prokaryotic proteomes To gain an overview of the similarities and differences between their protein domain content, we classified representative prokaryotes into hierarchical clusters based on their complement of protein domain families described. For each proteome we generated a 7,677 binary state element vector where each element represented the presence or absence of one of the 7,677 Pfam protein domain families. Pairwise distances were calculated for each pair of proteomes based on the level of similarity between the pair of vectors, and tree was built by neighbour-joining (see Methods for more details). One hundred trees were built, each time leaving out 10 % of the vector elements, selected at random. The tree shown in Figure 1 represents the consensus of these 100 jacknife trials. Figure 1 Clustering of complete prokaryotic proteomes based on their protein domain content . 100 jacknife trials were performed, each leaving out a random 10% of the data. The tree in Figure 1 illustrates the similarities and differences between prokaryotes with respect to their repertoire of recognisable protein domain families. There is clearly a correlation between domain complement and phylogeny; for example, the Archaea form a distinct cluster that is clearly separated from the Bacteria. Furthermore, within the Bacteria, the Cyanobacteria, Gram-positive Bacteria, chlamydias and mycoplasmas each fall into distinct clusters. However, there are some striking discrepancies between the protein domain-based clustering and phylogenetic classification. For example, the oral pathogen Treponema denticola (marked with an asterisk in Figure 1 ) clusters with the dental bacterium Fusobacterium nucleatum rather than with its fellow spirochetes T. pallidum and Borrelia burgdorferi . It is notable that the Proteobacteria do not form a single distinct cluster in the protein-domain based classification in Figure 1 . The cluster that contains the gamma-proteobacterial Pseudomonas and Xanthomonas species also contains the beta-Proteobacteria R. solanacearum and Chromobacterium violaceum . This probably reflects that these organisms have relatively large genomes and therefore share in common some common protein domains that are not encoded in smaller more streamlined genomes. Conversely X . fastidiosa , which has a relatively small genome, falls into a cluster with Neisseria meningitidis . Interestingly, the plant pathogen E. caratovora fell into a cluster with Yersinia pestis , Salmonella species and E. coli , which are animal pathogens and commensals. This indicates that despite differing lifestyles, these species have diverged relatively little with respect to loss and gain of protein domain families. Overall, the results of clustering bacterial proteomes on the basis of their domain content suggested that in addition to phylogeny, an organism's domain repertoire may reflect other factors, possibly including genome size and lifestyle. These preliminary observations led us to investigate whether it is possible to identify any particular domains or domain architectures that may be characteristic of a plant-associated lifestyle. Protein domain families restricted to plant-associated bacteria We queried the Pfam 16.0 database to determine the species distribution of each of the 7,677 domain families. Of these, 85 were found in at least one of the completely sequenced plant associated bacteria but absent from all other completely sequenced bacteria. Most of these domain families are restricted to a single species or group of very closely related organisms. For example, domain of unknown function DUF1484 (Pfam:PF07363) appears to be restricted to Ralstonia solanacearum , whilst DUF1520 (Pfam:PF07480) is restricted to Bradyrhizobium japonicum and Sinorhizobium meliloti . Although it is possible that these species-specific domain families are involved in pathogenesis or symbiosis it is equally likely that they have some unrelated function. However, several domains are potentially interesting from the point of view of plant-microbe interactions either because they are found in phylogenetically disparate species of phytobacteria or because they are also found in eukaryotes. Table 1 lists the domain families that are found in plant-associated members of more than one subdivision of the Proteobacteria, but are not found in any non-plant-associated bacteria. Several of these are already implicated in host-plant interactions. For example, proteins belonging to the NolX family (Pfam:PF05819) include HrpF from the gamma-proteobacterium X. campestris and NolX from the alpha-proteobacterium Rhizobium fredii and Rhizobium species NGR234. In these rhizobia, NolX (also referred to as NopX) has been shown to play a role in nodulation specificity and is exclusively expressed during the early stages of interactions with plants [ 25 , 26 ]. NolX is thought to facilitate protein secretion into the plant host via a type III secretion system [ 27 ], and a similar role has been postulated for X. campestris HrpF [ 28 ]. The importance of members of the NolX family in microbe-plant interactions is reinforced by our observation that they are also found in several other plant-associated alpha- and gamma-Proteobacteria as well as in the phytopathogenic beta-proteobacterium R. solanacearum (see Table 1 ), but are not found in any other completely sequenced genomes. Similarly, the Avirulence domain (Pfam:PF03377) is restricted to the phytopathogens R. solanacearum and Xanthomonas species [ 29 ]. Table 1 Pfam protein domain families found in phylogentically disparate plant-associated bacteria and not found in non-plant associated bacteria. Pfam domain family Species distribution Avirulence PF03377 X. avirulence protein, Avr/PthA R. solanacearum; X. axonopodis (pv. citri); X. campestris (pv. citri); X. campestris (pv. vesicatoria); X. campestris; X. manihotis; X. oryzae (pv. oryzae); X. oryzae; DspF PF06704 DspF/AvrF protein Erwinia amylovora; E. carotovora subsp. atroseptica SCRI1043; Erwinia pyrifoliae; Erwinia stewartii; Pantoea agglomerans (pv. gypsophilae) (Erwinia herbicola); Pectobacterium atrosepticum; P. syringae (pv. tomato); P. syringae; DUF1427 PF07235 Domain of unknown function A. tumefaciens (strain C58 / ATCC 33970); B. japonicum; P. aeruginosa; R. solanacearum; Rhizobium leguminosarum (biovar trifolii); Rhizobium meliloti (Sinorhizobium meliloti); X. campestris (pv. campestris); DUF811 PF05665 Domain of unknown function P. aeruginosa; R. solanacearum; HrpE PF06188 HrpE protein Erwinia amylovora; E. carotovora subsp. atroseptica SCRI1043; Erwinia chrysanthemi; Erwinia pyrifoliae; Erwinia stewartii; Pectobacterium atrosepticum; Pectobacterium carotovorum (subsp. carotovorum) (E. carotovora (subsp. carotovora)); P. fluorescens; P. syringae (pv. glycinea); P. syringae (pv. phaseolicola); P. syringae (pv. savastanoi); P. syringae (pv. syringae); P. syringae (pv. tabaci); P. syringae (pv. tomato); P. syringae; HrpF PF06266 HrpF protein Erwinia amylovora; E. carotovora subsp. atroseptica SCRI1043; Erwinia chrysanthemi; Erwinia pyrifoliae; Erwinia stewartii; Pectobacterium atrosepticum;Pectobacterium carotovorum (subsp. carotovorum) (E. carotovora (subsp. carotovora)); P. syringae (pv. glycinea); P. syringae (pv. phaseolicola); P. syringae (pv. savastanoi); P. syringae (pv. syringae); P. syringae (pv. tabaci); P. syringae (pv. tomato); Ice_nucleation PF00818 Ice nucleation protein repeat Bordetella phage BPP-1; Erwinia herbicola; Pantoea ananas (Erwinia uredovora); P. fluorescens; P. syringae (pv. syringae); P. syringae; X. campestris (pv. campestris); X. campestris (pv. translucens); NolX PF05819 NolX protein R. solanacearum; Rhizobium fredii (Sinorhizobium fredii); Mesorhizobium loti; Rhizobium sp. (strain NGR234); X. axonopodis (pv. citri); X. axonopodis pv. glycines; X. campestris (pv. campestris); X. campestris (pv. vesicatoria); X. oryzae (pv. oryzae); VirK PF06903 VirK protein A. tumefaciens (strain C58 / ATCC 33970); A. tumefaciens; B. japonicum; P. syringae (pv. tomato); R. solanacearum; Rhizobium sp. (strain NGR234); X. axonopodis (pv.citri); X. campestris (pv. campestris); X. fastidiosa (strain Temecula1 / ATCC 700964); X. fastidiosa; A further protein family limited to plant-associated bacteria is characterised by the ice nucleation repeat (Pfam:PF00818)and is found in proteins that may have a role in frost damage to host plants. It remains to be seen whether the remaining two domain families (DUF811 and DUF1427) are involved in the plant-associated lifestyle. DUF1427 (Pfam:PF07235) is restricted to several plant-associated alpha-Proteobacteria, the beta-proteobacterium R. solanacearum and the gamma-Proteobacteria P. aeruginosa and X. campestris (Table 1 ). Although their functions are unknown, proteins containing DUF1427 are thus candidates for involvement in interactions with plants or may at least have a role in plant-associated lifestyles. Several of these proteins have predicted signal peptide sequences and / or predicted transmembrane regions, suggesting an extracytoplasmic location. This may be indicative of a role in extracellular interactions with plants or with other components of the environment. Table 2 lists the 13 protein domain families that appear to be restricted to plant-associated bacteria and to eukaryotes and/or Archaea. Interestingly, this highlights at least one example of a protein domain that has probably been recruited into plant-associated bacteria from a plant host. Proteins containing a RolB/RolC-like domain (Pfam:PF02027) are found to be restricted to plant-associated alpha-Proteobacteria and to plants of the genus Nicotiana (see Table 2 and Figure 2 ). The activity of these proteins in plants may lead to an increase in intracellular auxin activity caused by the release of active auxins from inactive beta-glucosides [ 30 , 31 ]. The presence of many Agrobacterium- like proteins in Rhizobium (Agrobacterium) vitis reflects another key feature of the biology of these plant-associated bacteria, the fact that many of the genes involved directly in Agrobacterium and Rhizobium- plant interactions are encoded on large plasmids that facilitate lateral gene transfer of complex and novel traits between bacteria. Rhizobium (Agrobacterium) vitis is not a symbiont, but rather causes a tumorigenic disease of grapevine through the action of a number of A. tumefaciens- like genes [ 32 ]. Table 2 Pfam protein domain families restricted to plant-associated bacteria and eukaryotes. Pfam domain family Species distribution (not exhaustive) CBM_14 PF01607 Chitin binding Peritrophin-A domain Ralstonia solanacearum; Metazoa; Fungi; Viruses CD225 PF04505 Interferon- induced transmembrane protein Xanthomonas campestris (pv campestris); Metazoa; DUF726 PF05277 Protein of unknown function (DUF726) Pseudomonas syringae (pv tomato ) ; Metazoa; Plants; DUF763 PF05559 Protein of unknown function (DUF763) Mesorhizobium loti; Sinorhizobium meliloti; Xanthomonas axonopodis (pv. citri); Xanthomonas campestris (pv. campestris); Archaea; GDA1_CD39 PF01150 GDA1/CD39 (nucleoside phosphatase) family Pseudomonas syringae (pv. Tomato); Plants; Fungi; Metazoa; Het-C PF07217 Heterokaryon incompatibility protein Het-C Pseudomonas syringae (pv. tomato); Fungi; PAX PF00292 'Paired box' domain Rhizobium etli; Mesorhizobium loti; Metazoa; PPR PF01535 PPR repeat Ralstonia solanacearum; Plants; Metazoa; Fungi; Rhamnogal_lyase PF06045 Rhamnogalacturonate lyase family Erwinia carotovora subsp. atroseptica SCRI1043; Erwinia chrysanthemi; Plants; Ribosomal_60s PF00428 60s Acidic ribosomal protein Ralstonia solanacearum (Pseudomonas solanacearum); Plants; Metazoa; Archaea; RolB_RolC PF02027 RolB/RolC glucosidase family Agrobacterium rhizogenes; Agrobacterium tumefaciens (strain Ach5), and Agrobacterium tumefaciens (strain 15955); Agrobacterium tumefaciens (strain Ach5), and Agrobacterium tumefaciens; Agrobacterium tumefaciens (strain Ach5); Agrobacterium tumefaciens (strain C58 / ATCC 33970); Agrobacterium tumefaciens; Agrobacterium vitis (Rhizobium vitis); Plants; SBP56 PF05694 56 kDa selenium binding protein (SBP56) Bradyrhizobium japonicum; ; Plants; Metazoa; Archaea; ST7 PF04184 ST7 protein Rhizobium loti (Mesorhizobium loti); Metazoa; Figure 2 Examples of proteins containing a RolB/RolC domain. Protein domain families that are over-represented in plant-associated bacteria Bacterial physiology and behaviour is determined not only by the presence or absence of particular proteins but also by numbers of representatives of protein families. For example, gene duplication events may lead to a lineage-specific expansion that results in novel orthologues that can take on novel functions different from that of the parent gene. Therefore we investigated whether any protein domain families were over-represented in the plant-associated proteobacteria with respect to the background distribution of domains in all Proteobacteria for which complete sequences were available. For each of the 7,677 Pfam domain families, we counted the numbers of proteins in which that domain family occurs in the complete proteomes of Erwinia carotovora , Pseudomonas syringae pathovar tomato , Ralstonia solanacearum , Sinorhizobium meliloti , Bradyrhizobium japonicum , Mesorhizobium loti , Agrobacterium tumefaciens (Washington strain and Dupont strain), Xanthomonas campestris pathovar campestris , Xanthomonas axonopodis pathovar citri , Xylella fastidiosa and Xylella fastidiosa (strain Temecula1). We then calculated a P value for the probability of observing at least this number of occurrences given the background frequency in the Proteobacteria and assuming a binomial distribution. The smaller the P value, the less likely that the observed frequency occurred by chance. In other words, the smaller the P value, the more over-represented is the domain family. The most over-represented domains are listed in Table 3 . Table 3 Protein domain families over-represented in plant-associated proteobacteria. Domain family Expected number of proteins Observed number of proteins P Pfam accesion Pfam ID PF00211 Guanylate_cyc 33.39 70 2.17E-008 PF00296 Bac_luciferase 46.36 81 2.56E-006 PF04828 DUF636 36.58 65 1.40E-005 PF04679 DNA_ligase_A_C 17.65 38 1.76E-005 PF01068 DNA_ligase_A_M 24.03 47 2.18E-005 PF02738 Ald_Xan_dh_C2 35.72 63 2.33E-005 PF03758 SMP-30 19.35 40 2.63E-005 PF01638 DUF24 37 64 3.51E-005 PF01757 Acyl_transf_3 54.86 87 3.75E-005 PF00067 p450 24.24 46 5.31E-005 PF02746 MR_MLE_N 50.18 80 6.30E-005 PF02894 GFO_IDH_MocA_C 66.35 100 6.97E-005 PF01799 Fer2_2 31.26 55 7.69E-005 PF06169 DUF982 11.06 26 8.88E-005 PF07536 HWE_HK 23.82 44 1.35E-004 PF01022 HTH_5 68.47 101 1.38E-004 PF03573 OprD 14.03 30 1.41E-004 PF00656 Peptidase_C14 14.89 31 1.73E-004 PF03459 TOBE 83.78 139 2.48E-004 PF02627 CMD 51.89 79 2.75E-004 PF01188 MR_MLE 56.78 85 2.79E-004 PF07506 RepB 10.84 24 3.81E-004 PF01261 AP_endonuc_2 85.48 122 4.91E-004 PF00150 Cellulase 11.06 24 4.97E-004 PF01408 GFO_IDH_MocA 85.7 130 5.36E-004 PF00941 FAD_binding_5 21.05 38 5.58E-004 PF01315 Ald_Xan_dh_C 29.35 49 5.60E-004 PF00353 HemolysinCabind 36.15 57 8.12E-004 PF06823 DUF1236 8.93 20 9.64E-004 The domain with the statistically most significant over-representation in the plant-associated bacteria was the guanylate cyclase domain (Pfam:PF00211). This domain was particularly abundant in B. japonicum (32 proteins) and S. meliloti (24 proteins). No other fully-sequenced proteobacterium encodes more than three, although the spirochaete Leptospira interrogans encodes 17 proteins matching PF00211). Cyclic-diGMP, the product of guanylate cyclase, is a secondary messenger that plays a role in cell-cell and cell-surface contact in several bacteria by regulating cellular adhesion genes [ 33 ]. Such interactions are very important in initiating bacterial infection of eukaryotic organisms and this may account in part for the high numbers of such domains in these plant-associated bacteria. Of particular interest is the observation that one response regulator from C. crescentus has been shown to become sequestered to the cell pole following phosphorylation [ 35 ]. This is coupled to the activation of the guanylate cyclase domain, suggesting that localised synthesis of this secondary message could induce local effects within specific regions of the bacterial cell. Another domain with statistically significant over-representation in the plant-associated bacteria was the bacterial luciferase-like monooxygenase domain (Pfam:PF00296). This domain was particularly abundant in the plant-associated alpha-Proteobacteria with 15 proteins in Agrobacterium tumefaciens , 11 proteins in B. japonicum and 9 proteins in M. loti containing this domain. The related alpha-Proteobacteria C. crescentus , B. melitensis , B. suis and Rhodopseudomonas palustris have 3, 2, 2 and 0 luciferase (PF00296) proteins respectively. Other species containing large numbers of luciferase-like proteins include Mycobacterium bovis (13 proteins) and M. tuberculosis (14 proteins). Several domains of unknown function are amongst those most over-represented in the phytobacteria. For example, DUF636 is unusually abundant in the rhizobia with 16 representative proteins in B. japonicum and 14 and 13 in M. loti and S. meliloti respectively. Other prokaryotes encode between 0 and 5 DUF636 proteins, whilst Arabidopsis thaliana and Homo sapiens each encode one. Domain architectures The functionality of the proteome depends not only on the repertoire of protein domains but also on the interactions and cellular context of those domains. One important aspect of this context is the range of combinations of domains within a protein; that is the domain architecture of proteins. We used the Pfam database to ascertain the domain architecture of every protein sequence from each bacterial species for which a complete annotated genome sequence was available. 3,774 distinct protein domain architectures were found in R. solanacearum , P. aeruginosa , E. carotovora (subspecies atroseptica ), P. syringae (pathovar tomato ), B. japonicum , S. meliloti , M. loti , A. tumefaciens , X. fastidiosa , X. campestris , X. axonopodis . 459 of the 3,774 domain architectures encoded in genomes of plant-associated bacteria were absent in all other bacteria for which complete genome sequences were available. These 459 architectures are listed in the supplementary data. However, many of these architectures were restricted to a single species or several closely related species and so were of limited interest for this study. We were particularly interested to discover whether any domain architectures are related to plant-associated lifestyle rather than simply resulting from phylogeny. The 15 protein architectures illustrated in Table 4 were each found in plant-associated bacteria from at least two different divisions of the Proteobacteria and were not found in any other non-plant-associated organisms. For example, polypeptide sequences consisting of an N-terminal domain of unknown function DUF442 fused to a metallo-beta-lactamase domain are restricted to A. tumefaciens , M. loti , S. meliloti , X. fastidiosa and X. fastidiosa .The metallo-beta-lactamase domain (Pfam:PF00753) is common and widespread, being found in over 2000 different proteins from a wide range of organisms. However, only in these proteins from plant-associated bacteria is the metallo-beta-lactamase domain fused to DUF442. This suggests that the catalytic domain may have been recruited to some new function connected to a plant-associated lifestyle in these bacteria. Table 4 Domain architectures found in phytobacteria of two or more subdivisions of the Proteobacteria and not found in non-plant-associated bacteria. Domain architecture Species distribution Proteins DUF763 Aeropyrum pernix; Archaeoglobus fulgidus; Bradyrhizobium japonicum; Methanobacterium thermoautotrophicum; Methanopyrus kandleri; Picrophilus torridus; Pyrobaculum aerophilum; Pyrococcus abyssi; Pyrococcus furiosus; Pyrococcus horikoshii; M. loti; S. meliloti; Sulfolobus solfataricus; Sulfolobus tokodaii; Thermoplasma acidophilum; Thermoplasma volcanium; X. axonopodis (pv. citri); X. campestris (pv. campestris); Hypothetical protein XCC1094. ( Q8PBM5 ); Hypothetical protein XAC1190. ( Q8PN83 ); Hypothetical protein APE1824. ( Q9YAX1 ); Hypothetical protein ST0586. ( Q974S6 ); Hypothetical protein PF0611. ( Q8U361 ); Hypothetical protein. ( Q97VZ2 ); Hypothetical protein PH0745. ( O58515 ); Hypothetical protein SMb21455. ( Q92U57 ); Hypothetical protein. ( Q9UZ46 ); Mlr6856 protein. ( Q987Y3 ); Bll3834 protein. ( Q89NK4 ); Uncharacterized conserved protein. ( Q8TYA4 ); Hypothetical protein PAE0766. ( Q8ZYH9 ); Hypothetical protein TVG0468151. ( Q97BH6 ); Hypothetical protein Ta1095. ( Q9HJ77 ); Hypothetical protein AF1496. ( O28776 ); Hypothetical protein. ( Q6L1J8 ); Hypothetical protein MTH448. ( O26548 ); Hypothetical protein MTH449. ( O26549 ); VirK A. tumefaciens (strain C58 / ATCC 33970); A. tumefaciens; Bradyrhizobium japonicum; P. syringae (pv. tomato); R. solanacearum; Rhizobium sp. (strain NGR234); X. axonopodis (pv. citri); X. campestris (pv. campestris); X. fastidiosa (strain Temecula1 / ATCC 700964); X. fastidiosa; VirK (Tiorf135 protein). ( O50246* ); VirA/G regulated gene. ( Q7CNV8 ); Hypothetical 15.8 kDa protein in pinF2 3'region (ORF2). ( Q44433* ); Hypothetical 15.6 kDa protein y4WH. ( P55686* ); PUTATIVE SIGNAL PEPTIDE PROTEIN. ( Q8XX33* ); VirK protein. ( Q8PDC2* ); VirK protein. ( Q8PQ93 ); ID299. ( Q9ANE2* ); Blr1847 protein. ( Q79UP9 ); VirK protein. ( Q87D31 ); VirK protein. ( Q9PC40* ); Hypothetical protein. ( Q880Z8 ); DUF1427 A. tumefaciens (strain C58 / ATCC 33970); Bradyrhizobium japonicum; P. aeruginosa; R. solanacearum; Rhizobium leguminosarum (biovar trifolii); S. meliloti; X. campestris (pv. campestris); Hypothetical protein XCC2052. ( Q8P914 ); Bsl6958 protein. ( Q89EW2 ); Hypothetical protein. ( Q93EB2 ); HYPOTHETICAL TRANSMEMBRANE PROTEIN. ( Q8Y2U1* ); AGR_L_1747p. ( Q8U4X9* ); Hypothetical protein. ( Q92Y85 ); Bsr4258 protein. ( Q89MD5 ); Hypothetical protein. ( Q9I0E5* ); DUF1486 A. tumefaciens (strain C58 / ATCC 33970); Neurospora crassa; P. aeruginosa; P. syringae (pv. tomato); R. solanacearum; M. loti; S. meliloti; Hypothetical protein. ( Q7SFH5 ); Hypothetical protein Atu3018. ( Q8UBJ8 ); Hypothetical protein. ( Q92YL1 ); Mlr2224 protein. ( Q98IW1 ); Hypothetical protein. ( Q9I3U3 ); Hypothetical protein. ( Q9JP27 ); AGR_L_3571p. ( Q7CRD4 ); Hypothetical protein RSc0819. ( Q8Y171 ); RepB A. tumefaciens (strain C58 / ATCC 33970); P. syringae (pv. tomato); M. loti; S. meliloti; Msr9757 protein. ( Q98P91 ); Mll8115 protein. ( Q983Y2 ); Hypothetical protein. ( Q88BH6 ); Hypothetical protein Atu5040. ( Q8UKR0 ); AGR_pAT_52p. ( Q7D423 ); Hypothetical protein. ( Q92XS2 ); Hypothetical protein. ( Q930E6 ); Hypothetical protein. ( Q930E5 ); DUF442~Lactamase_B A. tumefaciens (strain C58 / ATCC 33970); M. loti; S. meliloti; X.fastidiosa (strain Temecula1 / ATCC 700964); X. fastidiosa; Metallo-beta-lactamase superfamily protein. ( Q8UAA9 ); Hypothetical protein. ( Q92ZB8 ); AGR_L_2726p. ( Q7CSJ2 ); Hypothetical protein. ( Q87AD6 ); Mlr2158 protein. ( Q98J12 ); Hypothetical protein. ( Q9PFB0 ); GAF~Phytochrome Bradyrhizobium sp. ORS278; X. axonopodis (pv. citri); Phytochrome-like protein. ( Q8PEQ2 ); Bacteriophytochrome. ( Q8VUB6 ); Glyco_hydro_6~CBM_2 Microbispora bispora; Micromonospora cellulolyticum; R. solanacearum; Thermomonospora fusca; X. fastidiosa (strain Temecula1 / ATCC 700964); X. fastidiosa; Cellulose 1,4-beta-cellobiosidase. ( Q87E00 ); 1,4-beta-cellobiosidase. ( Q9PDW2 ); PROBABLE EXOGLUCANASE A (1,4-BETA-CELLOBIOSIDASE) PROTEIN (EC3.2.1.91). ( Q8XS97 ); Endoglucanase A precursor (EC 3.2.1.4) (Endo-1,4-beta-glucanase) (Cellulase). ( P26414* ); Endoglucanase E-2 precursor (EC 3.2.1.4) (Endo-1,4-beta-glucanase E-2)(Cellulase E-2) (Cellulase E2). ( P26222* ); Endo-beta-1,4-glucanase. ( Q53488 ); DUF811 P. aeruginosa; R. solanacearum; Hypothetical protein. ( Q9I6E4* ); Hypothetical protein. ( Q9I6E5* ); Hypothetical protein RSc3082. ( Q8XUV1 ); Condensation~Condensation~AMP-binding~PP-binding~Condensation~AMP-binding~PP- binding~Condensation~AMP-binding~PP- binding~Condensation~AMP-binding~PP- binding~Condensation~AMP-binding~PP- binding~Thioesterase~Thioesterase P. syringae (pv. tomato); R. solanacearum; Probable peptide synthesis protein. ( Q8XS39 ); Non-ribosomal peptide synthetase, terminal component. ( Q881Q3 ); NolX R. solanacearum; Rhizobium fredii (Sinorhizobium fredii); M. loti; Rhizobium sp. (strain NGR234); X. axonopodis (pv. citri); X. axonopodis pv. glycines; X. campestris (pv. campestris); X. campestris (pv. vesicatoria); X. oryzae (pv. oryzae); HrpF protein. ( Q8PBA6 ); HrpF protein. ( Q8PQD2 ); HrpF. ( Q83XD5 ); HrpF. ( O33967 ); HrpF. ( Q6F5A9 ); HrpF. ( Q9KW22 ); Type III secretion system component. ( Q6QJ83 ); SECRETED PROTEIN POPF2. ( Q8XRF4 ); SECRETED PROTEIN POPF1. ( Q8XPT2 ); Nodulation protein; NolX. ( Q989P8 ); Nodulation protein nolX. ( P55711 ); Nodulation protein NolX. ( Q93LZ2 ); Nodulation protein NolX. ( Q9EUG7 ); Nodulation protein nolX. ( P33213 ); DUF802~DUF802 R. solanacearum; X. axonopodis (pv. citri); Hypothetical protein XAC3753. ( Q8PG64* ); Probable transmembrane protein ( Q8XQ05* ); Avirulence~Avirulence R. solanacearum; X. axonopodis (pv. citri); X. campestris (pv. citri); X. campestris (pv. vesicatoria); X. campestris; X. oryzae (pv. oryzae); X. oryzae; Avirulence protein AvrXa7-3M. ( Q6GWX1 ); Avirulence protein AvrXa7-1M. ( Q6GWX7 ); Avirulence protein. ( Q9EZV3 ); Avirulence protein AvrXa7-4M. ( Q6GWX4 ); Avirulence protein. ( Q9F0D0 ); Hypothetical 122 kDa avirulence protein in avrBs3 region. ( P14727 ); AvrBs3-2 protein. ( Q07061 ); PROBABLE AVRBS3-LIKE PROTEIN. ( Q8XYE3 ); Apl3 protein. ( Q9Z3F5 ); Avirulence protein. ( Q8PRG7 ); PthA protein. ( Q56780 ); Apl1 protein. ( Q9R7J3 ); Avirulence protein AvrXa7-2M. ( Q6GWX3 ); Avirulence protein. ( Q8PRN6 ); Avirulence protein AvrXa10. ( Q56830 ); PthB. ( Q7X130 ); Apl2 protein. ( Q9Z3F6 ); Avirulence protein. ( Q8PRM3 ); Avirulence protein. ( Q8PRK7 ); RgpF-RgpF M. loti; Rhizobium sp. (strain NGR234); X. axonopodis (pv. citri); X. campestris (pv. campestris); Mll4799 protein. ( Q98D97 ); Hypothetical protein XAC3576. ( Q8PGP0 ); Hypothetical protein wxcX. ( O34262 ); Hypothetical 45.0 kDa protein y4gN. ( P55470 ); TPR_2~TPR_1~Sulfotransfer_1 M. loti; X. axonopodis (pv. citri); uncultured bacterium 560; TPR domain/sulfotransferase domain protein. ( Q6SGF7 ); Mlr4028 protein. ( Q98EY4 ); Hypothetical protein XAC3051. ( Q8PI47 ); One regulatory domain found in large numbers in Pseudomonas genome is the PAS domain (Pfam PF00989) [ 36 ], which is present in 25 ORFs in P. aeruginosa PAO1 and 30 ORFs in P. syringae pathovar tomato . The average number of PAS-containing ORFs in complete proteobacterial genomes is about 10. Although PAS domains are only found in a limited subset of bacterial regulators, they are at the forefront of molecular innovation with 9 of the novel architectures identified in P. aeruginosa , and 5 of those in P. syringae pathovar tomato containing PAS domains (see supplementary data for more details). Xanthomonas genomes also encode a large number of PAS-containing polypeptides, (18 and 21 in X. axonopodis and X. campestris respectively). However, each X. fastidiosa encodes only one: PhoR, a regulator generally associated with responses to phosphate limitation. Ten novel PAS architectures are present in each Xanthomonas genome, of which 7 are common and 3 are unique to each strain (some of which are illustrated in Figure 3 ). PAS domains, which are involved in sensing light, oxygen and other environmental factors, have particular importance in helping bacteria to adapt to a changing environment, an ability of little value to X. fastidiosa in its restricted and relatively constant niches. Figure 3 Examples of proteins containing phytochrome domains. One intriguing signal transduction domain identified in unique domain architectures from both P. syringae and Xanthomonas was a phytochrome domain (Pfam:PF00360) (Figure 4 ). This domain enables light-mediated signal transduction in plants and bacteria, through binding a light-sensitive chromophore [ 37 , 38 ]. Phytochrome-containing proteins are used to detect light, and to discriminate between different wavelengths of light. Phytochromes are used for shade avoidance by plants, and to detect depth in soil or water or other conditions where light is attenuated. The short list of bacteria that contain phytochromes includes photosynthetic species ( e.g . Rhodospirillum centenum , Anabaena species strain PCC7120 and Synechocystis species strain PCC6803) as well as plant associated bacteria ( e.g . R. leguminosarum , A. tumefaciens ) and soil bacteria ( e.g. P. putida ) [ 38 , 39 ]. An unusual photosynthetic strain, Bradyrhizobium species ORS278 uses phytochrome to regulate the photosynthesis gene cluster and a similar induction was seen with Rhodopseudomonas pallustris but not with several other photosynthetic bacteria [ 40 ]. It is not known why phytochrome proteins are retained in non-photosynthetic bacteria but it has been suggested that the phytochrome-like sensor kinases in Agrobacterium may play a role in detecting depth in soil strata as a means of optimising interactions with roots [ 39 ]. Most of the bacterial phytochrome proteins have a PAS domain and a GAF domain at the N-terminus and a histidine kinase domain at the C-terminus (see Figure 4 ), though a phytochrome from Rhodobacter sphaeroides (UniProt:Q8VRN4; see Figure 4 ) has a more complex domain architecture [ 40 ]. The presence of two phytochromes in P. syringae , one of them with a unique architecture, may reflect the recruitment of phytochrome to a novel regulatory function unique to P. syringae . Protein PSPTO2652 from P. syringae is unique in that it has an additional C-terminal histidine kinase. Another unusual domain architecture is the PAS-GAF-Phytochrome-PAS organisation found in Xanthomonas proteins XAC4293 and XCC4154 (Figure 4 ), which, if shown to be functional, may represent a new phytochrome protein family. Figure 4 Examples of proteins containing phytochrome domains. Further analysis of novel Pseudomonas protein domain architectures The availability of multiple finished and unfinished Pseudomonas genomes allowed us to study in more detail the distribution, genomic context and properties of Pseudomonas gene products highlighted by this analysis. Closer examination of the genomic context of the P. syringae genes encoding proteins with unusual domain architectures showed that most were flanked on either or both sides by genes that have few or no orthologues in other Pseudomonas strains, suggesting that these novel genes have been recruited simultaneously with other genes, possibly of related function, or that they have recombined into the genome at hotspots for recombination and insertion of alien DNA. To further address the hypothesis that at least some of these architectures have been acquired by horizontal gene transfer we examined the GC content and third position GC content of each of these genes, in comparison to the total genome (0.593 GC, 0.716 GC3). Sixteen of the genes deviated from the average GC3 content by more than 0.05. High GC3 content genes include pvsA , PSPTO4084, PSPTO2413 and cfa6 . Low GC3 content genes include hrpZ , PSPTO3210, glf , PSPTO4696, hopPtoS (1,2 & 3), PSPTO2259, PSPTO0400, avrF and PSPTO1070. The GC content of flanking genes frequently reflected that of the novel gene, most strikingly for glf , PSPTO2441, PSPTO4696, hopPtoS (1,2 &3), PSPTO4699, PSPTO1070 & PSPTO2632, which were each associated with low GC regions containing few ORFs with orthologues in other Pseudomonas genomes. One other feature frequently associated with horizontally transferred genes is the presence of IS elements, tRNAs, plasmid and phage genes in flanking regions. PSPTO3229, PSPTO4569, PSPTO2312, PSPTO2829, PSPTO2310, Glf, PSPTO2441, PSPTO4696 and PSPTO2326 are all located in close proximity to IS elements and phage-like sequences, or in defined regions of the genome flanked by IS elements and phage-like sequences (see Figure 5 ). Figure 5 Genetic islands unique to Pseudomonas syringae . Genes encoding transposases are marked with an asterisk (*) and the asparaginyl tRNA gene is marked 'tAsn'. Black diamonds indicate genes encoding unique domain architectures [49]. Overall, this analysis suggests that a large number of the novel architectures present in P. syringae pathovar. tomato are uniquely associated with this species or pathovar of Pseudomonas , and that many of these genes have been acquired by horizontal gene transfer and are located in regions of the genome with a high potential for recombination and rearrangement. Conclusions Our initial observations, from the clustering of complete prokaryotic proteomes on the basis of domain content, motivated us to test whether any protein domains or domain architectures are specifically associated with a plant-associated lifefstyle. We identified nine protein domain families that are found in phylogenetically diverse plant-associated bacteria but not in non-plant-associated Bacteria (Table 1 ). Inevitably, there is an element of random chance in the species distribution of domain families; however, we observed that most of domains whose functions are at least partly known are implicated in the plant associated lifestyle. Therefore it seems possible that the two domains of unknown function (DUF811 and DUF1427) may also turn out to be significant for this lifestyle. Several domain families were also found only in plant pathogenic bacteria and in eukaryotes (Table 2 ). For example the RolB/RolC-like domain family is restricted to plant-associated bacteria and to plants of the genus Nicotiana , and is implicated in modulating auxin activity. Having investigated patterns of presence or absence of domains within bacterial proteomes, we next identified which domains are most over-represented in the plant-pathogenic Proteobacteria as compared with the frequency of occurrence in all the sequenced Proteobacteria (Table 3 ). Amongst the most over-represented domains was the guanylate cyclase domain. This was largely due to the large number of guanylate-cyclase-like proteins encoded by B. japonicum and S. meliloti . Although this approach may have revealed some potential leads for further investigation, it should be remembered that this analysis was rather crude and susceptible to the biased phylogenetic distribution of the organisms for which complete genome sequence data are currently available. However, detailed analysis of the frequency distributions of protein domain families in various organisms may yield rewards. As well as the repertoire of domains, another important aspect of a proteome is the repertoire of domain architectures; that is the combinations of domains found within a single protein. Just as for the repertoire of domains, the species distribution of a domain architecture might be explained by chance. Nevertheless, the proteins listed in Table 4 may be a good starting point for further investigation of bacterium-plant interactions. Many of these protein identified in this study have N-terminal predicted signal peptide motifs, suggesting that they are secreted. Further experiments are required to determine whether proteins of unknown function will also have a role in plant-specific functions. Many proteins involved in bacteria-plant interactions, such as TTSS-secreted effectors have subtle or conditional phenotypes, and would not be identified in conventional mutant-phenotype screens. Assays to detect subtle differences in growth in planta or in disease development are labour-intensive. Bioinformatic analyses such as this one represent useful and informative tools for reducing experimental search space, particularly when combined with other post-genomic techniques such as microarray analyses. We found relatively little evidence of lateral dissemination of niche-specific novel architectures between phylogenetically distinct divisions in the Proteobacteria, with less than 20 phytobacteria-specific domain architectures present in two or more divisions of the Proteobacteria. We did identify a number of domain architectures and domains that were uniquely conserved in both plant-associated prokaryotes and eukaryotes. The methodology used in this study makes no prior assumptions about the nature or cause of "uniqueness". Unique architectures identified using this approach include rare domains, novel domain combinations and architectures that are truncated relative to the majority of similar proteins (which may represent deletions and loss of function mutations). Some proteins will inevitability be included or excluded because of the limitations of current domain prediction technology. However, in addition to identifying protein candidates for further investigation, this type of analysis can be used to challenge and improve current models for domain prediction and expose errors and limitations of genome sequence data and protein prediction. For example, consider a case in which a protein is identified as having the "unique" architecture B~C~D. Additional examination of the protein may reveal that the protein has a similar sequence to proteins with the architecture A~B~C~D. The absence of the A domain may indicate a genuine alteration in structure and potentially in function, or a frameshift in the genome sequence data, or a functional "A" domain that fails to meet current predictive criteria. Each of these hypotheses can be tested by further research and experimentation, both in silico and in the lab. Although our approaches to identifying candidate genes and proteins of significance to lifestyle have led to several potential leads and interesting hypotheses, there are some caveats. Firstly, evolution does not proceed exclusively through loss and gain of domains and domain shuffling; for example, protein innovation can also occur through mutation and divergence within domain families. Also, it is becoming increasingly apparent that an organism's physiology, behaviour and ecology depend as much on higher order 'systems level' phenomena as on the inventory of molecular components. We chose to base our surveys of protein domains on the Pfam because this mature database is relatively comprehensive in its coverage ( e.g . compared with SMART) and its data is of high quality. Furthermore, its data is distributed in a form that is ideally suited for constructing database queries such as those in this study. Another advantage is that in Pfam no two domains ever overlap in their coverage of a protein sequence, which significantly simplifies the analysis. However, it should be noted that Pfam is not absolutely infallible and some of its threshold values are rather stringent, leading to failure to identify some 'outlying' members of a domain family. In summary, this study has described and applied a new approach for identifying architectural innovation and potentially important domains in proteins from genome sequence data. The data generated in this study have highlighted a large number of interesting and largely uncharacterised novel proteins and suggested new insights into the molecular basis of interactions between bacteria and their plant hosts, which will provide inspiration for future experimental research. Methods The Pfam relational database data files were downloaded from the Pfam website [ 46 ]. The census of domains and architectures were taken from Pfam release 16.0 (November 2004) using custom PERL scripts to wrap SQL queries against the Pfam relational database. The complete bacterial genomes included in Pfam 16.0, and hence considered in this study, are listed in the supplementary data. We excluded from the analysis of domain architectures all protein sequences in UniProt [ 47 ] that are designated as fragments. A file listing the presence or absence of each Pfam domain in each proteome can be found in the supplementary data. Each row in this file represented a vector used for the clustering of bacterial proteomes. Neighbour-joining was performed using PHYLIP [ 41 ]. Trees were visualised using ATV [51]. BLAST [ 42 ] searches were performed using the NCBI [ 48 ] and Expasy [ 49 ] web servers. Comparison between Pseudomonas genomes was aided by use of PseudoDB [ 50 ]. Transmembrane and signal peptide predictions were taken from Pfam, which in turn uses TMHMM [ 45 ] and SignalP [ 43 ]. It should be remembered that predictive methods often have difficulty distinguishing between signal peptides and N-terminal transmembrane helices [ 44 ]. Authors' contributions DJS and GMP conceived the original study, carried out the bioinformatics analyses, and drafted the manuscript. JAD proposed extending the study to symbionts as well as pathogens. All the authors contributed to interpretation of the data and to writing the final manuscript. Supplementary Material Additional File 1 This table lists the 459 domain architectures that are found in one or more plant-associated bacteria but are absent from other bacteria for which complete sequence data is available. Click here for file Additional File 2 Prokaryotic genomes included in Pfam16.0 (and hence in this study). Click here for file Additional File 3 "domains.tab.gz" Species distribution of each of the 3,774 Pfam domains. This tab-delimited file has been compressed using gzip. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC554113.xml
534803	Stimulation of allergen-loaded macrophages by TLR9-ligand potentiates IL-10-mediated suppression of allergic airway inflammation in mice	Background Previously, we demonstrated that OVA-loaded macrophages (OVA-Mφ) partially suppress OVA-induced airway manifestations of asthma in BALB/c mice. In vitro studies showed that OVA-Mφ start to produce IL-10 upon interaction with allergen-specific T cells, which might mediate their immunosuppressive effects. Herein, we examined whether IL-10 is essential for the immunosuppressive effects of OVA-Mφ in vivo , and whether ex vivo stimulation of the IL-10 production by OVA-Mφ could enhance these effects. Methods Peritoneal Mφ were loaded with OVA and stimulated with LPS or immunostimulatory sequence oligodeoxynucleotide (ISS-ODN) in vitro . The increase of IL-10 production was examined and, subsequently, ex vivo stimulated OVA-Mφ were used to treat (i.v.) OVA-sensitized mice. To further explore whether Mφ-derived IL-10 mediates the immunosuppressive effects, Mφ isolated from IL-10 -/- mice were used for treatment. Results We found that stimulation with LPS or ISS-ODN highly increased the IL-10 production by OVA-Mφ (2.5-fold and 4.5-fold increase, respectively). ISS-ODN stimulation of OVA-Mφ significantly potentiated the suppressive effects on allergic airway inflammation. Compared to sham-treatment, ISS-ODN-stimulated OVA-Mφ suppressed the airway eosinophilia by 85% (vs. 30% by unstimulated OVA-Mφ), IL-5 levels in bronchoalveolar lavage fluid by 80% (vs. 50%) and serum OVA-specific IgE levels by 60% (vs. 30%). Importantly, IL-10 -/- Mφ that were loaded with OVA and stimulated with ISS-ODN ex vivo , failed to suppress OVA-induced airway inflammation. Conclusions These results demonstrate that Mφ-derived IL-10 mediates anti-inflammatory responses in a mouse model of allergic asthma, which both can be potentiated by stimulation with ISS-ODN.	Background Chronic asthma is driven and maintained by the persistence of a subset of chronically activated memory T lymphocytes. The development of allergen-specific CD4 + T-helper 2 (Th2) immunoresponses is responsible for the cellular and molecular events underlying the initiation and progression of allergic asthma [ 1 , 2 ]. The Th2 lymphocyte, therefore, is potentially an important target cell for therapy in allergic asthma. Dendritic cells (DC) are well defined as antigen presenting cells (APC) able to initiate and regulate T cell responses [ 3 ]. Besides skewing T-cell responses into Th1 or Th2 responses [ 4 ], DC have been shown to mediate the induction of antigen-specific regulatory T (Treg) cells, like CD4 + Th3 cells and CD4 + T regulatory 1 (Tr1) cells [ 5 , 6 ]. Macrophages (Mφ), however, can also serve as APC and play a pivotal role in controlling and directing immune responses [ 7 , 8 ]. To exert these functions, Mφ express MHC-II molecules and secrete a variety of mediators. By secreting pro-inflammatory cytokines, such as IL-1, IL-6 and TNF-α, Mφ can trigger immune responses against microbial pathogens [ 8 , 9 ]. Moreover, by releasing IL-12 Mφ can specifically skew immune responses towards Th1 responses [ 10 - 12 ]. Although Mφ favor the induction of Th1 responses [ 13 , 14 ], it has also been demonstrated that Mφ can induce differentiation of Th2 lymphocytes [ 15 , 16 ]. Similar to DC, Mφ are nowadays thought to be capable of suppressing immune responses by secreting anti-inflammatory mediators, such as PGE 2 , TGF-β and IL-10 [ 7 , 9 , 17 ]. In the lung, alveolar Mφ participate in the maintenance of immunological homeostasis. By secreting pro-inflammatory cytokines and chemokines they direct the recruitment and activation of inflammatory cells, while they also play a key role in dampening immune responses against non-pathogenic antigens [ 9 , 18 ]. Alveolar Mφ have been shown to suppress T-lymphocyte proliferation in vitro [ 19 , 20 ] and APC-function of DC in vitro and in vivo [ 21 ]. Additionally, several studies have demonstrated that Mφ induce tolerance against inhaled allergens, likely at the level of allergen-specific T lymphocytes [ 22 - 24 ]. Interestingly, selective elimination of alveolar Mφ potentiated IgE Ab production in response to inhaled allergen, indicating a key role for alveolar Mφ in tolerance against allergen inhalation [ 25 ]. Moreover, we [ 26 ] and others [ 12 , 27 ] demonstrated that treatment with allergen-loaded Mφ effectively suppresses allergen-induced airway manifestations of asthma. In vitro studies demonstrated that allergen-specific T cells induced IL-10 production by OVA-loaded Mφ (OVA-Mφ), suggesting that the immunosuppressive effects of OVA-Mφ might be mediated by IL-10 [ 26 ]. In this study, we investigated whether stimulation with toll like receptor 4 (TLR4)-ligand LPS [ 28 ] and the TLR9-ligand immunostimulatory sequence oligodeoxynucleotide (ISS-ODN) [ 29 ] increases the IL-10 production by allergen-loaded Mφ and, thereby, can potentiate their immunosuppressive effects. Subsequently, using Mφ isolated from IL-10 -/- mice, we examined whether Mφ-derived IL-10 is crucial in the suppression of allergen-induced allergic airway inflammation. Methods Animals Animal care and use were performed in accordance with the guidelines of the Dutch Committee of Animal Experiments. Specific pathogen-free (according to the Federation of European Laboratory Animal Science Associations [ 30 ]) male BALB/c mice (6 wk old) were purchased from Charles River (Maastricht, The Netherlands). The mice were housed in macrolon cages in a laminar flow cabinet. Breeding pairs of IL10 -/- BALB/c mice were kindly provided by DNAX (Palo Alto, CA) and were housed in macrolon cages with filter top. All mice were provided with food and water ad libitum . Materials OVA (chicken egg albumin, grade V) and purified LPS from Escherichia coli 0111:B4 were purchased from Sigma-Aldrich (St. Louis, MO). CpG-containing phosphorothioate ISS-ODN and control phosphorothioate mutated oligodeoxynucleotide were synthesized by Isogen Bioscience BV (Maarsen, The Netherlands). The ISS-ODN used had the sequence 5'-TGACTGTGAA-CGTTCGAGATGA-3' and the mutated-ODN had the sequence 5'-TGACTGTGAA-GGTTAGAGATGA-3' [ 31 ]. Loading and stimulation of Mφ Peritoneal Mφ were isolated from naïve BALB/c mice as described previously [ 26 ]. For in vitro experiments, Mφ were plated in triplicate wells of a 96-well round-bottomed plate (Greiner Bio-One GmbH, Kremsmuenster, Austria) at 1 × 10 5 Mφ/well in RPMI 1640 enriched with 2% FCS, penicillin/streptomycin (all GIBCO BRL) and 50 μM β-mercaptoethanol (Sigma-Aldrich). Mφ were loaded with 2 mg/mL OVA and stimulated with different concentrations of LPS, ISS-ODN or mutated-ODN, for 20 h at 37°C and 5% CO 2 . Subsequently, supernatants were harvested and the amount of IL-10 was determined using an IL-10-specific sandwich ELISA. Stimulation with 10 μg/mL LPS or 3 μg/mL ISS-ODN triggered the highest IL-10 production by Mφ. For in vivo studies, 1 × 10 7 Mφ/mL were loaded with 2 mg/mL OVA and were stimulated with 10 μg/mL LPS or 3 μg/mL ISS-ODN. After incubation for 3 h at 37°C and 5% CO 2 , the Mφ were extensively washed (3 times with 50 mL saline) to remove all residual soluble OVA, LPS, and ISS-ODN. Sensitization, treatment and challenge Mice were sensitized to OVA by active sensitization with 7 i.p. injections of 10 μg OVA in 0.5 mL pyrogen-free saline on alternate days [ 32 ]. Treatment was performed 17 days after the last sensitization by administration (i.v.) of 3 × 10 5 Mφ in 50 μl saline. As an additional control group, mice were i.v. injected with 50 μL saline (sham treatment). One week after treatment, mice were exposed to OVA (2 mg/mL saline) aerosol challenges for 5 min on 8 consecutive days. Determination of OVA-specific IgE levels in serum Mice were sacrificed and were bled by cardiac puncture. Subsequently, serum was collected and stored at -70°C until analysis. OVA-specific IgE in serum was measured as described [ 33 ]. A reference standard was obtained by i.p. immunization of mice with OVA and arbitrarily assigned a value of 1000 experimental units/mL (EU/mL). The detection level of the IgE ELISA was 0.5 U/mL for IgE. Analysis of the cellular composition in the bronchoalveolar lavage fluid Bronchoalveolar lavage (BAL) was performed immediately after bleeding of the mice by lavage of the airways through a tracheal cannula with 1 mL saline containing 2 μg/mL aprotinine (Roche Diagnostics) and 5% BSA (Sigma-Aldrich). Cytokines in the supernatant of this first mL of the BAL fluid (BALF) were determined by ELISA. Subsequently, mice were lavaged 4 times with 1 mL saline. The cells in the BALF were pooled in cold PBS (including those from the first mL) and subsequently differentiated into mononuclear cells (monocytes, Mφ and lymphocytes), eosinophils, and neutrophils as described previously [ 33 ]. Cytokine ELISAs IL-5, IL-10, IL-12p70 ELISAs (all BD PharMingen) were performed according to the manufacturer's instructions. The detection limit of the IL-5 ELISA was 10 pg/mL, of the IL-10 ELISA 15 pg/mL, and of the IL-12p70 ELISA 62.5 pg/mL. Statistical analysis All data are expressed as mean ± standard error of mean (SEM). Statistical analysis on BALF cell counts was performed using the non-parametric Mann-Whitney U test (2-tailed). For ELISA, results were statistical analyzed using a Student's t test (2-tailed, homoscedastic). Results were considered statistically significant at the P < .05 level. Results IL-10 production by Mφ is increased by LPS and ISS-ODN In vitro studies suggest that the immunosuppressive effects of OVA-Mφ could be mediated by Mφ-derived IL-10 [ 26 ]. To further enhance these immunosuppressive effects we attempted to increase the IL-10 levels produced by Mφ. Since Mφ express TLR4 and TLR9 [ 28 , 29 ], we tested whether activation of these receptors (using LPS and ISS-ODN, respectively) would increase the IL-10 production by peritoneal Mφ. As Figure 1 shows, stimulation with LPS or ISS-ODN highly increased the IL-10 production by OVA-Mφ in vitro , while control mutated oligodeoxynucleotide did not. The IL-10 levels produced by OVA-Mφ increased 2.5-fold upon stimulation with LPS and 4.5-fold upon stimulation with ISS-ODN. IL-12p70 was not detectable in any of these cultures (data not shown). Figure 1 LPS and immunostimulatory sequence oligodeoxynucleotide (ISS-ODN) enhance the IL-10 production by OVA-Mφ. 1 × 10 5 Mφ/well were loaded with 2 mg/ml OVA and stimulated with either 10 μg/ml LPS or 3 μg/ml ISS-ODN for 20 h. As a control, Mφ were stimulated with mutated-ODN (M-ODN, 3 μg/ml). One of four representative experiments is shown. Increased production of IL-10 potentiates the suppressive effects of OVA-Mφ To examine the in vivo effect of the increased production of IL-10 by OVA-Mφ, peritoneal Mφ were isolated and subsequently loaded with OVA and stimulated with LPS (10 μg/mL) or ISS-ODN (3 μg/mL) for 3 h. The stimulated and OVA-loaded Mφ were administered (i.v.) to OVA-sensitized mice. In sham-treated mice, OVA-inhalation challenge strongly increased OVA-specific IgE Ab in serum (Figure 2 ). Treatment with OVA-Mφ that were not stimulated or stimulated with LPS caused no significant suppression in the up-regulation of serum OVA-specific IgE (Figure 2 ). In contrast, ISS-ODN-stimulated OVA-Mφ significantly suppressed (60%, P < .05) the up-regulation of serum OVA-specific IgE (Figure 2 ). OVA-specific IgG2a levels in serum of sham-treated mice were also increased upon OVA-inhalation challenge. However, these levels were not affected upon treatment with OVA-Mφ or stimulated OVA-Mφ (data not shown). Figure 2 OVA-specific IgE levels in serum are significantly suppressed upon treatment with ISS-ODN-stimulated and OVA-loaded Mφ. OVA-sensitized BALB/c mice were treated (i.v.) with saline (sham), OVA-Mφ, ISS-ODN-stimulated OVA-Mφ (ISS-ODN/OVA-Mφ), or LPS-stimulated OVA-Mφ (LPS/OVA-Mφ). Subsequently, these mice were challenged by OVA inhalation. Serum OVA-specific IgE levels were measured prior to and after challenge. Values are expressed as the mean ± SEM (n = 6 to 8). * P < .05 compared with sham-treated and OVA-challenged mice. † P < .05 compared with mice treated with OVA-Mφ and that were OVA-challenged. The BALF of mice, sensitized and challenged with OVA, contained high numbers of eosinophils (Figure 3A ). OVA-Mφ partially suppressed (30%, not significant) the influx of eosinophils into the BALF. Ex vivo stimulation of OVA-Mφ with LPS further enhanced the suppression of airway eosinophilia (60%, P < .05), compared with sham-treated mice (Figure 3A ). OVA-Mφ stimulated with ISS-ODN effectively suppressed the airway eosinophilia. The number of eosinophils in the BALF were significantly ( P < .01) suppressed by 85% compared to sham-treated mice and by 79% compared to mice treated with OVA-Mφ (Figure 3A ). Figure 3 ISS-ODN-stimulated and OVA-loaded Mφ (ISS-ODN/OVA-Mφ) significantly suppress airway eosinophilia and IL-5 levels in the bronchoalveolar lavage fluid. The number of eosinophils (eo), neutrophils (neutro) and mononuclear cells (MNC) in the BALF (A), and IL-5 levels in the BALF (B) after OVA inhalation challenge. Values are expressed as the mean ± SEM (n = 6 to 8). * P < .01 and † P < .05 compared with sham-treated mice. ‡ P < .01 and § P < .05 compared with mice treated with OVA-Mφ. The BALF of sham-treated mice contained high levels of the Th2 cytokine IL-5 (Figure 3B ), that correlates with the numbers of eosinophils. Treatment with OVA-Mφ or LPS-stimulated OVA-Mφ reduced the IL-5 levels in the BALF by 50% (p = 0.07 and p = 0.06, respectively), compared to sham-treated mice (Figure 3B ). ISS-ODN-stimulated OVA-Mφ, significantly reduced ( P < .01) the IL-5 levels in the BALF by 80%, compared to sham-treated mice. IL-10 was not detectable in the BALF of any of the mice (data not shown). Since ISS-ODN-stimulated OVA-Mφ produced the highest levels of IL-10 and most strongly suppressed OVA-induced airway inflammation, we used these Mφ to further analyze the underlying mechanism of immunosuppression by allergen-loaded Mφ. IL-10 produced by OVA-Mφ suppress OVA-induced airway inflammation To prove that IL-10 produced by OVA-Mφ indeed mediates the observed immunosuppressive effects, we isolated peritoneal Mφ from IL-10 -/- BALB/c mice. After loading with OVA and stimulation with ISS-ODN ex vivo , the IL-10 -/- Mφ were administered (i.v.) to OVA-sensitized BALB/c mice. Serum OVA-specific IgE levels of allergic mice that were treated with ISS-ODN-stimulated IL-10 -/- Mφ were as high as that of sham-treated mice (Figure 4 ). However, the up-regulation of serum OVA-specific IgE levels was partially affected by ISS-ODN-stimulated IL-10 -/- OVA-Mφ (Figure 4 ). The serum OVA-specific IgE levels were approximately 50% suppressed compared with unloaded ISS-ODN-stimulated IL-10 -/- Mφ. Still, these IgE levels were 45% higher ( P < .05) than in mice treated with ISS-ODN-stimulated OVA-Mφ. Figure 4 The suppression of OVA-specific IgE in serum by ISS-ODN-stimulated and OVA-loaded Mφ is partially mediated by IL-10. Sensitized mice were treated (i.v.) with saline (sham), ISS-ODN-stimulated OVA-Mφ (ISS-ODN/OVA-Mφ), ISS-ODN-stimulated IL-10 -/- Mφ (ISS-ODN/IL10 -/- Mφ), or ISS-ODN-stimulated IL-10 -/- OVA-Mφ (ISS-ODN/IL10 -/- OVA-Mφ). Serum OVA-specific IgE levels were measured prior to and after OVA challenge. Values are expressed as the mean ± SEM (n = 6 to 8 per group). * P < .05 compared with sham-treated and OVA-challenged mice. † P < .05 compared with mice treated with ISS-ODN/IL10 -/- OVA-Mφ and that were OVA-challenged. Importantly, after OVA-inhalation challenge, treatment with ISS-ODN-stimulated IL-10 -/- OVA-Mφ did not suppress airway eosinophilia (Figure 5A ). Unloaded IL-10 -/- Mφ that were stimulated with ISS-ODN had also no effect on airway eosinophilia, while immunotherapy with ISS-ODN-stimulated OVA-Mφ suppressed the influx of eosinophils by 88% ( P < .05), compared to sham-treated mice (Figure 5A ). Figure 5 IL-10 is crucial in the suppression of airway eosinophilia and IL-5 levels in the bronchoalveolar lavage fluid by ISS-ODN-stimulated and OVA-loaded Mφ (ISS-ODN/OVA-Mφ). (A) Number of eosinophils (eo), neutrophils (neutro) and mononuclear cells (MNC) in the BALF after OVA challenge. (B) Levels of IL-5 in the BALF after OVA challenge. Values are expressed as the mean ± SEM (n = 6 to 8 per group). * P < .05 compared with sham-treated mice. † P < .05 compared with mice treated with ISS-ODN/IL10 -/- OVA-Mφ. The BALF of sham-treated mice, sensitized and challenged with OVA, contained high levels of IL-5 (Figure 5B ). Treatment with ISS-ODN-stimulated IL-10 -/- OVA-Mφ as well as with ISS-ODN-stimulated IL-10 -/- Mφ had no effect on the IL-5 levels in the BALF (Figure 5B ). In contrast, these IL-5 levels were significantly reduced by 78% ( P < .05) upon treatment with ISS-ODN-stimulated OVA-Mφ (Figure 5B ). Together, IL-10 produced by OVA-Mφ mediated the anti-inflammatory effects of allergen-loaded Mφ on allergen-induced airway eosinophilia and IL-5 production. Discussion Previously, we showed that allergen-loaded Mφ partially suppress allergen-induced airway manifestations in mice [ 26 ]. Here, we demonstrated that the anti-inflammatory effects of allergen-loaded Mφ are IL-10 dependent and that both the IL-10 production and the immunosuppressive effects can be potentiated by stimulation of ISS-ODN. Stimulation with ISS-ODN, a ligand for TLR9 [ 29 ], readily increased the IL-10-production by peritoneal Mφ that were loaded with OVA. In contrast, these Mφ produced no detectable IL-12p70. We (data not shown) and others [ 34 ] confirmed these data using the Mφ-like cell line RAW264.7. In our mouse model of allergic asthma, stimulation of OVA-Mφ with ISS-ODN significantly potentiated their immunosuppressive effects. The suppression of the OVA-induced serum OVA-specific IgE levels, airway eosinophilia and IL-5 levels in the BALF was enhanced. Measuring the enhanced pause (Penh) in response to inhalation of different doses of methacholine (data not shown), we confirmed our previous finding that OVA-Mφ significantly suppressed OVA-induced airway hyperresponsiveness to methacholine [ 26 ]. As potentiation of the IL-10 production by Mφ (using LPS or ISS-ODN) did not further suppress the allergen-induced airway hyperresponsiveness (data not shown), it can be speculated that the mechanisms underlying the suppression of airway hyperresponsiveness and of allergic airway inflammation, at least in part, differ. However, we would like to stress that Penh values may not correlate with changes in pulmonary resistance [ 35 ]. Compared to ISS-ODN, stimulation with LPS showed an intermediate capacity to enhance the immunosuppressive effects of OVA-Mφ. OVA-Mφ produced higher levels of IL-10 upon stimulation with ISS-ODN compared to stimulation with LPS, suggesting a correlation between the levels of IL-10 produced by the Mφ and the extent of suppression of allergen-induced airway inflammation. As LPS and ISS-ODN trigger signaling via different intracellular pathways [ 36 ], we hypothesized that stimulation of Mφ with a combination of LPS and ISS-ODN could further increase the production of IL-10. However, the levels of IL-10 produced by OVA-Mφ stimulated with LPS and ISS-ODN were as high as when stimulated with ISS-ODN only (data not shown). Implying that stimulation of Mφ with ISS-ODN results in maximal production of IL-10. Using Mφ isolated from IL-10 -/- BALB/c mice, we demonstrated that Mφ-derived IL-10 is crucial in the suppression of airway eosinophilia and IL-5 levels in the BALF, while the suppression of serum IgE is partially IL-10 dependent. Although a lack of IL-10 production could up-regulate the MHC class II and B7 expression in Mφ of IL-10 -/- mice, as increased IL-10 levels could down-regulate these molecules [ 37 , 38 ], the shift in expression of these molecules will most probably not be the underlying mechanism of suppression of airway eosinophilia and Th2 cytokines, because low levels of MHC class II or B7 itself do not result in suppressive functions. Furthermore, we can not exclude that there are, at present unknown, developmental changes in Mφ derived from IL-10-deficient mice, that may affect their capacity to suppress an allergic inflammatory response. The observation that the suppression of IgE is partially IL-10 independent suggests that the suppression of serum IgE levels is only slightly correlated to Th2-cytokine levels. This is in agreement with the finding that memory IgE responses are inferior mediated by Th2 cytokines [ 39 ]. These data indicate that a second, IL-10 independent, suppressive pathway has to be induced by OVA-Mφ that causes a further suppression of serum IgE levels. Mφ can reside for long period of time in tissue, including the lung [ 40 ]. By secreting the immunosuppressive cytokine IL-10 the Mφ could, allergen-independently, suppress allergen-induced airway inflammation. In the past, IL-10 has been shown to suppress allergen-induced airway manifestations of asthma in the mouse [ 41 - 43 ]. However, we found that, after i.v. treatment, the ISS-ODN-stimulated OVA-Mφ specifically migrate to the spleen of OVA-sensitized mice. Subsequently, at this site, an allergen-specific and long-lasting immunosuppressive response is induced (preliminary results by Vissers JLM et al). These results demonstrate that the IL-10 produced by the Mφ is not directly responsible for the suppression of allergic inflammation in the lungs, but that an allergen-specific suppressive T-cell subset is induced in the spleen. This hypothesis is supported by the finding that IL-10 production by OVA-Mφ, upon recognition of OVA-specific T cells in vitro , is dependent on MHC class II/TCR interaction [ 26 ]. Direct targeting of OVA to alveolar Mφ, for instance by intratracheal treatment with OVA-loaded liposomes, could demonstrate whether Mφ in the lung can directly induce suppression of OVA-induced airway inflammation. However, we (data not shown) and others [ 44 ] found that alveolar Mφ from OVA-sensitized mice do not produce IL-10 upon stimulation with LPS or ISS-ODN. Although we were not able to test the suppressive capacity of OVA-loaded alveolar Mφ in our mouse model, targeting of allergens to alveolar Mφ could still be promising to induce immunosuppressive effects in humans, because alveolar Mφ from patients with allergic asthma produce IL-10 [ 45 , 46 ]. In this study and in our previous study [ 26 ], we observed no indications for an increased Th1 response upon immunotherapy with OVA-Mφ that could counteract the Th2 response. In contrast, others demonstrated that allergen-loaded Mφ, stimulated with IFN-γ ex vivo , promote a switch from Th2 cells to Th1 cells [ 12 , 27 ]. This dissimilarity can mainly be explained by the difference in cytokines which are produced by the Mφ used. IFN-γ-stimulated Mφ produce IL-12 upon allergen-specific interaction with T cells [ 12 ], while our Mφ produce IL-10 upon antigen recognition [ 26 ]. Likely, IL-12 will favor skewing towards Th1 [ 4 , 11 ], whereas IL-10 will act as a suppressive cytokine. In our model, allergen-loaded Mφ will, most probably, induce Treg cells via secretion of IL-10. Antigen-induced Treg cells are typically induced in microenvironments with APCs presenting antigens and local high levels of IL-10 [ 6 ]. This T-cell subset plays a pivotal role in the maintenance of T-cell tolerance against foreign-antigens. They exhibit their suppressive activity by secreting the suppressive cytokine IL-10 (Tr1 cells) or TGF-β (Th3 cells) [ 47 ]. By using CD4 + T lymphocytes, ex vivo transduced to express IL-10, it was shown that allergen-specific lymphocytes can suppress allergen-induced asthma manifestations via production of IL-10 [ 43 ]. Recently, Akbari and colleagues found that pulmonary dendritic cells from mice exposed to respiratory allergen produced IL-10 and, thereby, induced allergen-specific Tr1 cells [ 5 , 48 ]. Furthermore, treatment of mice with killed Mycobacterium vaccae induced allergen-specific Treg cells that produced IL-10 and TGF-β [ 49 ]. In agreement with our study, these studies indicate a pivotal role for IL-10 in limiting allergen-induced asthma manifestations. Conclusions Here, we demonstrated, in a mouse model of allergic airway inflammation, that treatment with allergen-loaded Mφ suppress asthma manifestations in an IL-10-dependent manner. Importantly, the IL-10 production and anti-inflammatory effects of allergen-loaded Mφ can be potentiated by stimulation with ISS-ODN. Further detailed analysis of the mechanisms underlying this Mφ-based immunotherapy may lead to the development of new strategies to induce tolerance against allergen-specific Th2 responses in allergic diseases, including asthma. Authors' contributions JLMV carried out the allergic model, subsequent analysis, writing and preparation of the manuscript. BCAMvE, PVJ and GAH assisted with the allergic model. AJMvO participated in the direction of the study as well as writing and preparing the manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC534803.xml
549630	Genetic variation analysis of the Bali street dog using microsatellites	Background Approximately 800,000 primarily feral dogs live on the small island of Bali. To analyze the genetic diversity in this population, forty samples were collected at random from dogs in the Denpasar, Bali region and tested using 31 polymorphic microsatellites. Australian dingoes and 28 American Kennel Club breeds were compared to the Bali Street Dog (BSD) for allelic diversity, heterozygosities, F-statistics, G ST estimates, Nei's DA distance and phylogenetic relationships. Results The BSD proved to be the most heterogeneous, exhibiting 239 of the 366 total alleles observed across all groups and breeds and had an observed heterozygosity of 0.692. Thirteen private alleles were observed in the BSD with an additional three alleles observed only in the BSD and the Australian dingo. The BSD was related most closely to the Chow Chow with a F ST of 0.088 and also with high bootstrap support to the Australian dingo and Akita in the phylogenetic analysis. Conclusions This preliminary study into the diversity and relationship of the BSD to other domestic and feral dog populations shows the BSD to be highly heterogeneous and related to populations of East Asian origin. These results indicate that a viable and diverse population of dogs existed on the island of Bali prior to its geographic isolation approximately 12,000 years ago and has been little influenced by domesticated European dogs since that time.	Background Bali, a province of the Republic of Indonesia, is an island just 87 km from north to south and 142 km from east to west and home to more than 2.9 million people [ 1 ]. Approximately 800,000 stray dogs (Fig. 1 ) also live on the island based on a survey conducted by the Bali Street Dog Foundation (personal communication). Only a small percentage of these dogs live in homes or are provided routine veterinary care [ 2 ]. Figure 1 Typical Balinese street dogs. Their phenotypic appearance is similar to that described for randomly breeding feral dog subpopulations in other parts of the world. More than 90% of the residents of Bali are Hindu [ 3 ] with myth and ritual playing a vital part of daily life [ 1 ]. The dog is also an important part of Balinese life and mythology. A popular tale from the Mahabharata [ 4 ] describes King Yudisthira's journey to Heaven's Gate, and his love for a dog that befriended him on his arduous and tragic journey (Fig. 2 ). As a direct result of such mythology, BSDs are treated with a degree of reverence and are often provided ceremonial food offerings [ 2 ]. The deliberate killing of street dogs is not typically practiced, because Balinese people believe that all things should be allowed to die naturally [ 2 ]. These cultural mores have contributed to the current overpopulation of dogs on the island. Figure 2 The Story of Yudisthira [4]. As a result of overpopulation, many BSDs suffer from chronic skin diseases, internal parasites, parvo- and distemper-virus infections, and malnutrition. In an effort to reduce the dog population and to care for their medical needs, the Bali Street Dog Foundation (Yayasan Yudisthira Swarga) was founded in 1998 [ 2 ]. They provide emergency care, treatment for skin disease and parasites, sterilization, public education on the plight of feral dogs, and improved veterinarian training. Twenty to 30 dogs are sterilized each day, with more than 9,000 dogs sterilized to date. The BSD population is of interest for both its genetic diversity and historical relationships. It is also a population that has bred more or less randomly for thousands of years with limited genetic influx, due mainly to geographic barriers and a strict rabies control program in effect since 1926. The present study is concerned with the genetic diversity of this unique canine population and its relationship to other canine subpopulations in Asia and throughout the world. Data presented herein was derived from the DNA testing of 40 BSD samples from the Denpasar city region of Bali with 31 polymorphic microsatellite loci. The genetic diversity of the BSD was compared to that of the Australian dingo and 28 American Kennel Club (AKC) breeds. Results Locus diversity Analysis of locus diversity across all 30 subpopulations revealed that the number of observed alleles ranged from six to 20 with a total of 366 for all loci (Table 1 ). Overall heterozygosity of the loci was high, with an average of 0.779, and all but four loci having H T values greater than 0.700. Average H S was 0.577 for the 30 subpopulations, with all but three loci having H S values greater than 0.500. The H S and H T values were closest for C23.123 and farthest for C22.279 and C10.404. HWE analysis revealed that all but one locus had at least one population out of equilibrium for the 30 populations sampled. C01.424, C31.646 and CPH16 had 7 populations out of HWE and AHT130 did not have any populations with p values below 0.05. The level of locus diversity attributable to subpopulation structure was evaluated with two statistics – R ST and F ST . Both statistics gave similar average values at 0.230 and 0.236 respectively. However, R ST ranged from 0.098 to 0.486 while F ST ranged from 0.179 to 0.328. Table 1 Observed number of alleles, average total heterozygosity (H T ), average subpopulation heterozygosity (H S ), number of populations out of HWE, average p values, R ST , F ST , R ST /F ST ratio, G ST and pairwise F ST values for 31 loci. BSD Pairwise F ST by Locus Chr. Num. Observed Alleles H T H S Num. Loci with p value <0.05 Average p value R ST F ST R ST /F ST × Dingo × Chow CPH16 CFA20 11 0.829 0.610 7 0.407 0.098 0.233 0.420 0.005 0.132 C08.618 CFA08 9 0.744 0.553 4 0.481 0.148 0.229 0.646 0.071 0.185 FH2001 CFA23 13 0.791 0.593 4 0.433 0.167 0.225 0.741 0.083 0.100 C20.446 CFA20 10 0.729 0.553 3 0.541 0.173 0.215 0.804 0.123 0.105 C01.424 CFA01 9 0.716 0.476 7 0.475 0.258 0.320 0.807 0.125 0.110 CPH02 CFA32 9 0.693 0.520 4 0.499 0.194 0.223 0.871 0.148 0.066 FH2004 CFA11 18 0.809 0.611 3 0.522 0.187 0.214 0.873 0.057 0.060 AHT137 CFA11 14 0.861 0.672 2 0.427 0.176 0.197 0.893 0.064 0.188 C03.877 CFA03 12 0.730 0.509 1 0.524 0.268 0.275 0.977 0.244 0.114 C06.636 CFA06 12 0.652 0.483 6 0.435 0.233 0.237 0.982 0.069 0.024 AHT121 CFA13 18 0.865 0.632 1 0.511 0.255 0.251 1.016 0.098 0.063 VIASD10 CFA07 9 0.759 0.555 2 0.525 0.249 0.233 1.066 0.268 0.030 C31.646 CFA31 14 0.814 0.566 7 0.410 0.301 0.281 1.075 0.088 0.036 RVC1 CFA15 9 0.774 0.569 4 0.458 0.270 0.242 1.119 0.355 0.216 LEI002 CFA27 11 0.737 0.551 4 0.534 0.259 0.229 1.127 0.107 0.149 LEI004 CFA37 13 0.667 0.509 4 0.510 0.246 0.219 1.128 0.133 0.051 C28.176 CFA28 10 0.735 0.546 6 0.400 0.270 0.234 1.152 0.249 0.013 C22.279 CFA22 11 0.836 0.529 2 0.441 0.262 0.214 1.226 0.201 0.109 PEZ02 Unlinked 12 0.762 0.600 1 0.567 0.232 0.187 1.242 0.134 0.027 FH2054 CFA12 10 0.848 0.654 4 0.524 0.251 0.199 1.261 0.055 0.069 C23.123 CFA23 8 0.766 0.636 4 0.402 0.351 0.278 1.263 0.110 0.002 CPH08 CFA19 11 0.765 0.582 4 0.465 0.284 0.222 1.276 0.082 0.056 C14.866 CFA14 10 0.840 0.604 3 0.473 0.330 0.255 1.293 0.180 0.112 PEZ08 CFA17 17 0.859 0.684 5 0.465 0.235 0.179 1.312 0.121 0.079 AHT130 CFA18 11 0.829 0.614 0 0.539 0.313 0.235 1.331 0.116 0.105 AHT111 CFA02 11 0.785 0.582 4 0.371 0.348 0.246 1.414 0.214 0.006 C10.404 CFA10 13 0.865 0.558 3 0.526 0.486 0.328 1.479 0.177 0.160 C09.250 CFA09 10 0.830 0.582 1 0.491 0.412 0.272 1.513 0.016 0.042 FH2140 CFA05 20 0.795 0.621 2 0.560 0.297 0.192 1.546 0.119 0.072 AHT139 CFA15 6 0.664 0.508 4 0.452 0.330 0.206 1.604 0.085 0.078 CPH03 CFA06 15 0.815 0.612 2 0.526 0.394 0.229 1.724 0.017 0.180 All 366 0.779 0.577 108 0.481 0.230 0.236 1.135 0.126 0.088 Bali street dog diversity Overall, the BSD was the most genetically diverse population surveyed here, displaying 239 total alleles out of the 366 seen in all 30 subpopulations or 65.3% of the total observed alleles (Table 2 ). The Australian dingo displayed 144 alleles and the AKC breeds displayed 138.8 on average. Analysis of expected (H E ) and observed (H O ) heterozygosities (Table 2 ) revealed that the BSD had a 44.0% higher H E than the Australian dingo (0.736 vs. 0.511) and a 28.4% higher H E than the average AKC breed (0.573). H O was also highest in the BSD at 0.692, versus 0.426 in the Australian dingo and 0.563 in the average AKC breed. Table 2 Total number of alleles (N A ) observed, range of the lowest and highest number of observed alleles per locus, expected heterozygosity (H E ), observed heterozygosity (H O ), F IS , number of loci out of HWE and average p values for all 31 loci for the BSD, a bootstrapped sampling of the BSD, the Australian dingo, the American Kennel Club breeds and for all subpopulations. N A Observed N A Range H E H O F IS Num. Loci with p value <0.05 Average p value Bali Street Dog 239 3 – 14 0.736 0.692 0.097 4 0.357 Bali Street Dog 20 214.6 --- 0.727 0.692 --- --- --- Australian dingo 144 2 – 9 0.511 0.426 0.194 12 0.284 AKC Breeds 138.8 2.2 – 7.8 0.573 0.563 0.137 3.29 0.492 All Populations 366 6 – 20 0.577 0.562 0.136 3.60 0.492 In order to evaluate the bias of sampling twice the number of BSDs, twenty samples were taken at random from the total pool of 40 for bootstrap determinations, and the number of observed alleles, H E , and H O were calculated. This process was repeated for 10,000 iterations and the average value for each measurement was determined. The average bootstrap value for the number of observed alleles for 20 BSDs was 214.6. The average bootstrapped H E and H O values for the BSD were 0.727 and 0.692 respectively. To understand the loss of approximately 24 observed alleles after bootstrapping, the allele frequencies for the BSD at each locus was examined. While the BSD had the highest number of alleles, they also had the highest number of alleles with a frequency below 5% (67 out of 239, data not shown). F IS estimates were calculated to assess the level of inbreeding for each subpopulation (Table 2 ). The BSD had the lowest value at 0.097 and the Australian dingo had the highest at 0.194 with the average of AKC breeds at 0.137. Private alleles Allele frequency analysis also revealed that 10 private alleles were observed in the BSD as well as three alleles shared only with the Australian dingo (Table 3 ). The majority of private alleles in the BSD were below 5% in frequency with the exception of AHT121 where the private alleles had a combined frequency of 18.75%. The BSD and the Australian dingo also shared three alleles not seen in any AKC breed at locus C10.404 with a combined frequency of 16.25% in the BSD and 75% in the Australian dingo. Table 3 Private alleles for the BSD and Australian dingo subpopulations relative to the 28 comparison AKC breeds. Locus Allele Pop Freq. Pop Freq. AHT111 92 BSD 0.013 AHT121 82 BSD 0.075 AHT121 90 BSD 0.113 C06.636 158 BSD 0.013 C10.404 168 BSD 0.038 Dingo 0.075 C10.404 170 BSD 0.063 Dingo 0.450 C10.404 172 BSD 0.063 Dingo 0.225 C10.404 174 BSD 0.038 C23.123 154 BSD 0.013 FH2140 160 BSD 0.013 FH2140 171 BSD 0.025 PEZ02 144 BSD 0.013 VIASD10 94 BSD 0.013 Asian alleles Several additional unique alleles were found only in the BSD, Australian dingo, Chow Chow and Akita; demonstrating a closer relationship of the BSD to Asian versus non-Asian dogs (data not shown). Appearing at the highest frequency was allele 201 of locus CPH08 in the BSD, Australian dingo and Chow Chow at frequencies of 21.3%, 10% and 20%, respectively. Further, the BSD, Australian dingo, Chow Chow and Akita share allele 113 of locus C22.279 at frequencies of 23.8%, 67.5%, 15% and 5%, respectively. Results for locus PEZ08 demonstrate a lack of influence of European alleles where a high frequency of deviations from n+4 alleles were observed in the AKC breeds sampled yet no deviation from n+4 alleles was observed in the BSD or the Australian dingo. A pairwise F ST analysis was performed for each locus between the BSD and the two closest subpopulations: the Australian dingo and Chow Chow (Table 1 ). The BSD was most similar to the Australian dingo at locus CPH16 with a F ST of 0.005 and to the Chow Chow at locus C23.123 with a F ST of 0.002. The BSD was most dissimilar to the Australian dingo and Chow Chow at locus RVC1 with a F ST of 0.355 and 0.216 respectively. Several loci had similar distances for both population pairs, such as locus C01.424 or locus C20.446 and may indicate areas of the genome that are neutral to either environmental or human selection. Genetic distance relationships Further distance analysis was performed for all 31 loci between all 30 subpopulations using both Nei's DA distance and pairwise F ST estimates (Table 4 ). Across all loci, the BSD shared allele frequencies most closely with the Chow Chow (DA = 0.242, F ST = 0.088) and the Australian dingo (DA = 0.242, F ST = 0.126), and least closely with the Airedale Terrier (DA = 0.454, F ST = 0.258). Table 4 Nei's DA distance (lower triangle) and mean F ST estimates (upper triangle) between each pair of 9 dog subpopulations represented graphically in Figure 3. BSD Dingo Chow Akita AES AS AT BCO BLT MBT BMD BS BT BU BX BZ DP GH GR JRT KE LR NE PG PM PPN PWC RR WM YT BSD 0.13 0.09 0.13 0.14 0.14 0.26 0.19 0.26 0.25 0.21 0.15 0.17 0.18 0.33 0.18 0.26 0.17 0.14 0.11 0.17 0.17 0.18 0.23 0.13 0.11 0.15 0.15 0.16 0.13 Dingo 0.24 0.23 0.26 0.27 0.27 0.41 0.32 0.42 0.40 0.37 0.30 0.29 0.29 0.50 0.31 0.41 0.30 0.26 0.25 0.31 0.30 0.31 0.39 0.27 0.25 0.29 0.29 0.29 0.25 Chow 0.24 0.40 0.19 0.22 0.21 0.34 0.26 0.33 0.32 0.29 0.24 0.23 0.24 0.41 0.26 0.34 0.26 0.21 0.17 0.23 0.25 0.25 0.31 0.20 0.17 0.22 0.22 0.23 0.17 Akita 0.29 0.41 0.36 0.19 0.19 0.33 0.26 0.35 0.32 0.31 0.20 0.22 0.25 0.40 0.23 0.30 0.24 0.18 0.17 0.23 0.22 0.26 0.30 0.20 0.17 0.20 0.20 0.22 0.19 AES 0.28 0.43 0.43 0.37 0.10 0.29 0.16 0.33 0.30 0.23 0.14 0.16 0.19 0.32 0.24 0.24 0.13 0.12 0.10 0.14 0.16 0.18 0.25 0.12 0.10 0.17 0.13 0.16 0.13 AS 0.27 0.42 0.38 0.36 0.20 0.27 0.15 0.32 0.29 0.24 0.16 0.17 0.18 0.35 0.21 0.23 0.16 0.13 0.10 0.17 0.16 0.19 0.23 0.13 0.10 0.16 0.13 0.17 0.12 AT 0.45 0.55 0.50 0.45 0.40 0.36 0.35 0.46 0.44 0.37 0.34 0.32 0.34 0.50 0.31 0.46 0.32 0.29 0.26 0.34 0.32 0.32 0.42 0.27 0.28 0.34 0.32 0.31 0.28 BCO 0.35 0.49 0.48 0.44 0.26 0.25 0.45 0.39 0.37 0.24 0.21 0.21 0.25 0.41 0.29 0.30 0.22 0.21 0.15 0.24 0.23 0.25 0.30 0.17 0.16 0.20 0.22 0.21 0.16 BLT 0.43 0.55 0.51 0.52 0.46 0.42 0.51 0.54 0.10 0.43 0.35 0.40 0.35 0.49 0.40 0.44 0.36 0.30 0.25 0.34 0.32 0.35 0.46 0.33 0.27 0.36 0.36 0.36 0.30 MBT 0.41 0.52 0.48 0.48 0.41 0.38 0.48 0.50 0.09 0.54 0.52 0.54 0.38 0.46 0.50 0.52 0.48 0.43 0.36 0.42 0.41 0.34 0.44 0.30 0.26 0.35 0.34 0.33 0.29 BMD 0.39 0.51 0.47 0.49 0.36 0.35 0.45 0.36 0.54 0.42 0.23 0.28 0.30 0.42 0.35 0.36 0.28 0.23 0.18 0.29 0.25 0.28 0.37 0.21 0.20 0.26 0.24 0.26 0.21 BS 0.32 0.46 0.45 0.38 0.28 0.28 0.48 0.35 0.52 0.34 0.36 0.20 0.24 0.35 0.23 0.28 0.17 0.15 0.13 0.19 0.15 0.22 0.30 0.15 0.11 0.18 0.15 0.19 0.17 BT 0.36 0.48 0.44 0.41 0.29 0.30 0.43 0.35 0.59 0.37 0.42 0.35 0.23 0.40 0.24 0.28 0.19 0.20 0.16 0.22 0.24 0.23 0.28 0.17 0.18 0.22 0.18 0.24 0.17 BU 0.31 0.41 0.42 0.42 0.29 0.27 0.41 0.39 0.42 0.31 0.41 0.40 0.36 0.36 0.27 0.30 0.22 0.22 0.17 0.26 0.24 0.28 0.32 0.22 0.16 0.24 0.19 0.26 0.19 BX 0.47 0.58 0.55 0.54 0.37 0.40 0.51 0.48 0.49 0.46 0.46 0.45 0.50 0.36 0.44 0.44 0.40 0.33 0.31 0.39 0.39 0.40 0.46 0.35 0.31 0.40 0.36 0.40 0.31 BZ 0.35 0.46 0.45 0.41 0.40 0.33 0.39 0.44 0.52 0.37 0.47 0.36 0.39 0.36 0.50 0.32 0.21 0.23 0.17 0.27 0.23 0.28 0.34 0.20 0.19 0.24 0.25 0.27 0.20 DP 0.45 0.57 0.55 0.48 0.39 0.37 0.60 0.44 0.57 0.41 0.47 0.44 0.43 0.41 0.48 0.46 0.29 0.28 0.23 0.29 0.30 0.34 0.37 0.24 0.23 0.29 0.26 0.32 0.26 GH 0.35 0.46 0.46 0.44 0.27 0.27 0.41 0.34 0.50 0.35 0.40 0.30 0.32 0.35 0.49 0.33 0.43 0.19 0.16 0.21 0.21 0.23 0.28 0.17 0.15 0.22 0.15 0.23 0.16 GR 0.33 0.43 0.42 0.39 0.25 0.26 0.37 0.33 0.43 0.29 0.37 0.31 0.35 0.35 0.40 0.37 0.44 0.34 0.11 0.18 0.13 0.18 0.31 0.15 0.10 0.17 0.16 0.16 0.14 JRT 0.25 0.43 0.32 0.34 0.21 0.20 0.37 0.27 0.38 0.24 0.30 0.28 0.30 0.27 0.39 0.31 0.38 0.29 0.25 0.15 0.11 0.15 0.26 0.11 0.08 0.14 0.13 0.16 0.07 KE 0.33 0.50 0.42 0.41 0.27 0.28 0.44 0.35 0.47 0.32 0.42 0.36 0.36 0.38 0.46 0.42 0.42 0.36 0.34 0.28 0.22 0.25 0.32 0.15 0.15 0.22 0.17 0.23 0.16 LR 0.33 0.45 0.46 0.39 0.29 0.28 0.42 0.36 0.42 0.31 0.36 0.26 0.38 0.37 0.46 0.35 0.44 0.32 0.26 0.24 0.37 0.20 0.33 0.15 0.12 0.23 0.19 0.17 0.16 NE 0.35 0.45 0.44 0.45 0.28 0.30 0.42 0.38 0.45 0.45 0.37 0.37 0.38 0.39 0.46 0.43 0.46 0.34 0.33 0.28 0.40 0.34 0.36 0.19 0.17 0.21 0.21 0.22 0.20 PG 0.39 0.54 0.48 0.45 0.38 0.35 0.51 0.44 0.60 0.56 0.50 0.45 0.43 0.44 0.52 0.51 0.52 0.43 0.49 0.41 0.48 0.46 0.51 0.27 0.25 0.30 0.28 0.30 0.27 PM 0.26 0.46 0.39 0.38 0.23 0.25 0.41 0.29 0.49 0.45 0.35 0.29 0.29 0.34 0.45 0.34 0.41 0.30 0.30 0.22 0.27 0.28 0.34 0.43 0.11 0.19 0.16 0.14 0.13 PPN 0.26 0.42 0.37 0.35 0.22 0.21 0.39 0.28 0.42 0.40 0.32 0.26 0.34 0.28 0.38 0.32 0.37 0.30 0.24 0.20 0.27 0.24 0.31 0.39 0.24 0.15 0.13 0.13 0.11 PWC 0.31 0.45 0.41 0.39 0.30 0.27 0.47 0.32 0.50 0.48 0.41 0.33 0.37 0.39 0.49 0.37 0.48 0.37 0.31 0.27 0.35 0.34 0.35 0.45 0.32 0.28 0.17 0.21 0.17 RR 0.28 0.43 0.40 0.37 0.25 0.22 0.39 0.33 0.49 0.45 0.36 0.29 0.27 0.29 0.45 0.38 0.41 0.27 0.29 0.26 0.27 0.30 0.32 0.38 0.28 0.24 0.31 0.21 0.16 WM 0.32 0.45 0.41 0.41 0.29 0.28 0.39 0.34 0.47 0.44 0.37 0.31 0.36 0.40 0.47 0.40 0.45 0.37 0.29 0.29 0.36 0.26 0.34 0.41 0.28 0.27 0.35 0.28 0.19 YT 0.30 0.44 0.37 0.41 0.29 0.22 0.40 0.31 0.44 0.42 0.35 0.31 0.32 0.31 0.40 0.32 0.40 0.29 0.30 0.18 0.30 0.29 0.34 0.43 0.26 0.25 0.31 0.29 0.32 Genetic distance relationships amongst the five Asian subpopulations were further explored using neighbor-joining dendograms with four non-Asian subpopulations for comparison (Fig. 3 ). The BSD, Chow Chow, Australian dingo and Akita clustered together in 90% of the trees. The BSD, Chow Chow and Australian dingo further clustered in 87% of the trees. The BSD and Australian dingo maintained their relationship within the larger cluster in 84% of the trees. In the remainder of the tree, the Rhodesian Ridgeback, Greyhound, Airedale Terrier and Borzoi maintained a relationship in 51% of the trees and the Airedale Terrier/Borzoi cluster was seen in 63% of the trees. The Pug did not maintain a relationship with any other breed in this analysis, but was intermediate to the Asian and non-Asian subpopulations. Figure 3 a. Unrooted neighbor-joining dendogram showing the genetic relationships among 9 dog subpopulations based on DA genetic distance. b. Rooted neighbor-joining dendogram showing the genetic relationships among 9 dog subpopulations based on DA genetic distance. In both versions of the dendogram the Pug did not cluster with any population but is placed intermediate between the Asian and non-Asian subpopulations. Discussion Population diversity Microsatellites have been previously used to assess genetic diversity and relationships in feral dog subpopulations [ 6 , 7 ]. Kim et al. [ 6 ] found that H O was high in three feral dog subpopulations of Korea, Sakhalin and Taiwan, ranging from 0.539 in the Taiwanese to 0.717 in the Korean dogs. Given that the loci used in that study had an average allele number of 7.75, these values are similar to the H O of 0.692 observed in the BSD. Wilton et al. [ 7 ] surveyed a population of Australian dingoes and found an average H O of 0.387 using microsatellites with an average allele number of 6.93, similar to the H O of 0.426 for the Australian dingoes reported herein with an average allele number of 11.8. Given the size of the island of Bali, it is extraordinary that 800,000 feral dogs can thrive and maintain such high levels of genetic diversity. Of all the subpopulations surveyed here, the BSD has the highest number of observed alleles, the highest heterozygosity, the fewest number of loci out of Hardy-Weinberg equilibrium and the lowest F IS . Even after adjusting for sample size, the BSD maintains their status as the most heterogeneous population in the study. Unlike the Australian dingo which exhibits a much lower level of diversity, the BSD findings suggest either a large founding population on Bali and/or a consistent genetic influx since the geographic isolation of ~12,000 years ago. This data also supports that the BSD appears to approximate a randomly breeding population with little selection pressure. When comparing the heterogeneity of the BSD to that observed within the AKC breeds some caveats should be addressed. One may initially expect long established, well-defined dog breeds to be much less heterogeneous than reported here. While some breeds do have a low H E , such as the Boxer with a H E of 0.320, breeds like the Jack Russell Terrier have a high H E of 0.713 and overall their H E is higher than that of the dingo. Of first note, the selection of the dogs that contribute to a breed composition mostly occurs prior to official breed recognition primarily by genetic drift due to geographic isolation and selection for particular working or physical characteristics. After official breed recognition future breeding choices are based primarily on the availability of sires and dams that approximate the breed standard. As a result, there is a founding population that proceeds to breed mostly by convenience. Also, many breed standards have changed considerably over the years resulting in retention of a certain level of diversity within each breed, some breeds retaining much more than others. Finally, dogs comprising the comparison AKC breeds were sampled from across the United States, removing any geographical bias of the genotypes observed and slightly elevating the heterozygosities. Locus diversity The average allelic diversity of the loci used in the present study was 11.8 alleles per locus, versus 7.75 in the Kim et al work [ 6 ]. However, the average number of alleles observed is 4.6 among the subpopulations in the present study and the average H T is 0.577. The average values for the 11 subpopulations surveyed in the Kim et al [ 6 ] work were 4.34 and 0.547, respectively. The higher total allelic diversity in the present study is likely due to the fact that nearly three times more subpopulations were studied. R ST and F ST values were nearly identical across all subpopulations and all loci, indicating that approximately 23% of the differences observed in allele frequencies can be attributed to differences between subpopulations. F ST provides an unbiased estimate of genetic drift between subpopulations by comparing alleles identical by state. R ST takes advantage of the stepwise mutation model, which assumes that mutations most often occur as whole repeat unit losses or gains from the original allele size. As a result, the number of mutations provides an estimate of time from divergence. It is interesting, therefore, to compare R ST and F ST values by locus. Eighteen of the 31 loci studied have an R ST to F ST ratio greater than 1.1 (Table 1 ) indicating that the populations have been separated for a sufficient amount of time for mutations to impact genetic structure. An interesting exception is observed at CPH16 where the ratio is 0.420. CPH16 may have a mutation pattern where both stepwise additions and subtractions occur at equal and high frequency. Of note, the average pairwise R ST value between the BSD and each of the 29 comparison subpopulations is 0.056 at locus CPH16. The highest R ST to F ST ratio occurs at locus CPH03 with a value of 1.724. Interestingly, the BSD and the Australian dingo have a pairwise R ST value of 0.017 at CPH03, whereas the average value of the BSD compared to the other 28 subpopulations has a value of 0.254. The distance between the BSD and the Australian dingo at CPH03 may support that those two populations were isolated most recently from each other relative to the other 28 subpopulations. Bali street dog origin The origin of the people of Bali is clouded by myth and a scarcity of archeological findings. Therefore, the origin of the dog on Bali is also speculative. Nonetheless, a hypothesis can be formed based on known human and dog histories. Current evidence points to an early migration of humans from Africa through Indonesia and into Australia approximately 60,000 to 70,000 years ago [ 8 , 9 ]. Recent excavations have also revealed that there was a great expansion into Indonesia from China between 4,000 and 5,000 years ago that could have contributed to a population pre-existing on Bali [ 1 ]. Supportive evidence that Indonesia was populated prior to 5,000 years ago is a higher degree of heterogeneity in the Indonesian population than seen in the North Asian population, suggesting that the Indonesia was populated earlier than regions to the North [ 10 ]. The "Slow Boat Model" for the peopling of Polynesia also suggests a prolonged mixing of Southeast Asians with Indonesians, which predated migration to the East [ 11 ]. In short, Indonesia appears to be a human genetic melting pot with genetic influences over tens of thousands of years. The dog on the island of Bali may also be a parallel "canine genetic melting pot." While the domestication date of the dog is in much dispute [ 12 ], approximately 14,000 years ago is accepted as a late date. During the earliest human migrations through Indonesia however, it is highly possible that wolf packs or feral dogs traveled the same routes, establishing a feral population on Bali in the process. Even if humans were not capable of taming the dog at that time, dogs could still have benefited from close proximity to humans. Figure 4 shows a superimposition of the proposed geographic origin for five Asian and four non-Asian dog subpopulations presented herein and the major theorized human migration routes. It is noteworthy that the BSD, Chow Chow and Australian dingo, related breeds by genetic analysis, all share one proposed human migration route. Figure 4 Human migration patterns proposed in "Tracing the road down under" [8], a summary of the Modern human origins: Australian perspectives conference at the University of New South Wales, September 2003 with locations of origin for 5 Asian and 4 non-Asian dog subpopulations. If a feral dog population was established on the island of Bali more than 14,000 years ago, then that population became isolated approximately 10,000 years ago when the sea levels drastically rose, submerging the land bridges of the Indonesia archipelago [ 13 ]. Geographic isolation was unlikely to have been absolute; genetic diversity of the BSD was invariably enhanced at various times by the influx of new dogs. At the time humans migrated to Indonesia from China, dogs were known to be domesticated and undoubtedly accompanied people as companions [ 17 ]. Mitochondrial DNA sequencing evidence suggests that the dingo was introduced into Australia about that time from the Indonesian archipelago [ 15 , 8 , 9 ]. Bali's documented history of repeated war and trade spanning the last 2,000 years [ 1 , 16 , 17 ] represents actions that are often associated with the introduction of new animals. Indeed, a somewhat free movement of dogs probably occurred up to 1926, when the import of dogs to Bali was greatly curtailed as a means to prevent the introduction of rabies [ 5 ]. This policy greatly reduced, though not eliminated, new outside introductions of new dogs to the island. In contrast to the Australian dingo population, which appears to have undergone a severe population bottleneck or founder effect based on microsatellite alleles and mtDNA [ 18 ], the BSD population maintains a high level of genetic variation. There is no evidence for a genetic bottleneck or small founding population for the BSD. The relatedness of the BSD to the Australian dingo and the Chow Chow is evidenced by common unique alleles and allele frequencies despite the very different levels of genetic diversity between the subpopulations. According to the hypothesis presented herein, one could imagine that feral dog subpopulations were established throughout Indonesia with much mixing until ~12,000 years ago. At that time, each population became closed with little influx of new genetic material until humans migrated south from Asia between 4,000 and 5,000 years ago. The degree of influx since that period would have been influenced by the frequency of trade and conflict, factors determined by accessibility, available natural resources, and political structure. The island of Bali is historically a less visited island than it's neighbor Java and therefore the indigenous dog population would have been subjected to less influence. Conclusions This study into the diversity and relationship of the BSD to other domestic and feral dog populations shows the BSD to be highly diverse and related to populations of East Asian origin. These results indicate that a viable and diverse population of dogs existed on the island of Bali prior to its geographic isolation approximately 12,000 years ago and has been little influenced by domesticated European dogs since that time. It would be of interest to study feral subpopulations on other islands in the archipelago to determine if the same level of diversity is observed elsewhere, or if the situation on Bali is truly unique. Y-chromosome, mitochondrial and MHC marker typing on the BSD, as well as feral dogs from other regions, would help to determine if indeed dogs followed the same migration routes as their likely human companions. Methods Animal selection BSDs were randomly captured and taken to a BSD Foundation field clinic for treatment or sterilization and simultaneously sampled for DNA collection with buccal swabs. Familial relationships of the BSDs sampled could not be easily determined; therefore the sample population was doubled (40 vs. 19–20 samples) over that of other study groups. Blood samples from the Australian dingo were taken from captive animals in Australia. Australian dingoes were known to be unrelated by at least one generation. Dogs from 28 American Kennel Club (AKC) breeds, equally representing the AKC group designations, were sampled with buccal swabs for a previous study [ 19 ]. Twenty dogs were tested for each breed, with the exception of two breeds (Doberman Pinscher and the Border Collie) that comprised 19 individuals. The 28 breeds included were: Airedale Terrier, Akita, American Eskimo, Australian Shepherd, Belgian Tervuren, Bernese Mountain Dog, Border Collie, Borzoi, Boxer, Brittany, Bull Terrier, Bulldog, Chow Chow, Doberman Pinscher, Golden Retriever, Greyhound, Jack Russell Terrier, Keeshond, Labrador Retriever, Miniature Bull Terrier, Norwegian Elkhound, Papillon, Pembroke Welsh Corgi, Pomeranian, Pug, Rhodesian Ridgeback, Weimaraner, and Yorkshire Terrier. Dogs within each breed were unrelated by at least one generation. Marker selection Thirty-one of the 100 microsatellites multiplexed into 12 PCRs by the Veterinary Genetics Laboratory [ 20 ] had been previously used to evaluate the Australian dingo samples (unpublished data). For comparison purposes, those same 31 microsatellites were selected for use in the present study. All markers but one (PEZ02) were mapped on either the 1999 canine genetic linkage map [ 21 ] or the Radiation hybrid map [ 22 ]. Loci selected for study represented 25 of the 38 autosomes of the dog, with five autosomes represented by two loci. The average distance for the markers on chromosomes CFA06, CFA11, CFA20 and CFA23 is 23.5 cM and 23.4 Mb between AHT139 and RVC1 on CFA15. As a result, only 25 loci are known to be unlinked. PEZ02 has not been mapped and may be linked to a marker in the study. Forward primers were synthesized and dye labeled with either Fam, Hex or Vic, or Tamra or Ned (Applied Biosystems, Inc. (ABI), Foster City, CA). Reverse primers were synthesized by Operon (Alameda, CA). Primer sequences and concentrations for all markers are available as Additional file 1 . Sample preparation and PCR conditions BSD and AKC breed DNA was derived from buccal cells harvested from the inside of the cheek with nylon bristle cytology brushes (Medical Packaging Corp., Camarillo, CA). Samples were collected by owners or field volunteers and submitted directly to the laboratory. DNA was extracted by heating a single swab for 10 min at 95°C in 400 μl 50 mM NaOH and then neutralized with 140 μl 1 M Tris-HCl, pH 8.0. Australian dingo DNA was extracted from blood using a standard sodium hydroxide digest. A 2 μl aliquot of extract was used in each PCR which equates to approximately 50 ng DNA. All markers and DNAs were amplified with a PCR reagent mix of 1X PCR buffer (ABI), 4.17 mM MgCl 2 , 200 μM of each dNTP (Hoffmann-La Roche Inc, Nutley, NJ), 0.6 unit AmpliTaq (ABI), and 2% DMSO (Sigma) then covered with 15 ul Chill-out™ Liquid Wax (MJ Research, Inc., Waltham, MA) to prevent evaporation. One of five thermal cycler programs was used for each primer mix ranging from 56° to 64° degrees for the annealing temperature. All PCR work was done in polycarbonate 96-well v-bottom microtiter plates (USA Scientific, Ocala, FL) on MJ Research PTC-100 thermal cyclers (MJ Research, Inc., Waltham, MA). Protocols are also available in Additional 1 . Gel electrophoresis conditions and DNA fragment analysis One μl aliquots of PCR product were mixed with 2 μl Fluorescent Ladder (CXR) 60–400 (Promega 400) or Internal Lane Standard 600 (Promega 600) (Promega, Madison, WI) fluorescent size standard, denatured on MJ Research PTC-100 thermal cyclers for three minutes at 95°C, then held at 5°C or placed on ice for at least one minute before gel loading. Two μl aliquots were then loaded onto a 6% denaturing polyacrylamide gel and run on an ABI 377 Automated Sequencer using ABI 10" × 7 1/8" short plates (12 cm). Gels were run at 1.10 kV (constant) voltage, 60.0 mA current, 200 W power, 51°C and 40.0 mW (constant) laser power for up to 2 hours when using Promega 400, and up to 3 hours using Promega 600. DNA fragment analysis was performed with in-house designed STRand software [ 23 ], which replaces ABI Genotyper and Genescan software. This data was then transferred to an in-house database compatible with the STRand software. Statistical analysis Allelic diversity and observed heterozygosities (H O ) were determined by direct counting for each of the 30 subpopulations. Hardy-Weinberg equilibrium (HWE) tests were performed using Genepop version 3.4 [ 24 ]. Pairwise F ST estimates and subpopulation expected heterozygosities (H E ) for the 30 breeds or dog groups were performed using Genepop version 3.4 [ 24 ]. F IS estimates (inbreeding coefficient of each subpopulation) for each allele following Weir and Cockerham [ 25 ] were calculated using Genepop version 3.4 and are presented as averages across all loci. Gene diversity or total population heterozygosity (H T ) and its associated parameters, H S (average heterozygosity among subpopulations) and G ST (coefficient of genetic differentiation), were calculated across all loci using the public domain software, DISPAN [ 26 ]. Two additional measures of variance, F ST [ 25 ] and R ST [ 27 , 28 ] were calculated using Genepop version 3.4. A pairwise genetic distance matrix using Nei's DA distance was also created using DISPAN with bootstrapping. Genotype data for all populations is available in Additional file 2 . Phylogenetic tree construction Allele frequencies were used to compute a matrix of genetic distances [ 29 ], which were then used to construct a phylogenetic tree of relationships among 5 Asian and 4 non-Asian dog subpopulations. Takezaki's [ 30 ] POPTREE program was used to create a neighbor joining tree using DA distances with 1000 bootstrap replications. The output of POPTREE was then converted to the New Hampshire format for editing in the stand alone program TREEVIEW version 1.6.6 [ 31 ] and bootstrap values were added. Abbreviations BSD: Bali Street Dog F IS , F ST R ST , G ST : F-statistics indices H S , H T , H E , H O : Heterozygosity indices HWE: Hardy-Weinberg equilibrium N A : Number of alleles AES: American Eskimo Dog AS: Australian Shepherd AT: Airedale Terrier BCO: Border Collie BLT: Bull Terrier BMD: Bernese Mountain Dog BS: Brittany Spaniel BT: Belgian Tervuren BU: Bulldog BX: Boxer BZ: Borzoi Chow: Chow Chow Dingo: Australian dingo DP: Doberman Pinscher GH: Greyhound MBT: Miniature Bull Terrier PG: Pug RR: Rhodesian Ridgeback GR: Golden Retriever JRT: Jack Russell Terrier KE: Keeshond LR: Labrador Retriever NE: Norwegian Elkhound PG: Pug PM: Pomeranian PPN: Papillon PWC: Pembroke Welsh Corgi RR: Rhodesian Ridgeback WM: Weimaraner YT: Yorkshire Terrier Authors' contributions DNI performed the majority of data acquisition and analysis, wrote first draft of the manuscript and prepared the final draft for submission. ALS performed the majority of sample processing, assisted in data acquisition and the writing of the subsequent drafts of the manuscript as well as final draft preparation. SG sampled the dogs tested, provided background for the manuscript and assisted in the final draft preparation. ANW provided the Australian dingo data for comparison and assisted in the subsequent drafts of the manuscript. NCP directed the research and assisted in the writing of the manuscript. All authors read and approved the final manuscript. Supplementary Material Additional File 1 "MS 2784776144318916 Supplement1.xls" and contains the primer sequences, expected size range, primer concentration used and annealing temperatures used. In a separate sheet within the same file the protocols for each PCR reaction are listed. Click here for file Additional File 2 "MS 2784776144318916 Supplement2.xls" and contains the individual genotype data for each animal used to derive the statistics and phylogenetic results presented herein. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC549630.xml
526778	Grouping of tooth surfaces by susceptibility to caries: a study in 5–16 year-old children	Background The decline in caries has slowed and this may be indicative of variation in the susceptibility of differing teeth to caries. This study tests the hypothesis that in children, there are groups of tooth sites that exhibit differences in caries susceptibility. Methods Probit analysis of caries data collected from a 4-year longitudinal study of 20,000 schoolchildren aged between 5 and 16 years in 10 differing locations in the United States. Results The development of dental caries within the mouth followed a fixed hierarchy indicating that tooth surfaces show variation in caries susceptibility. Certain teeth and tooth sites have similar susceptibilities and can be grouped, the sizes of the groups vary. The most susceptible group consists of six tooth surfaces: the buccal pits and occlusal fissured surfaces of the first molar teeth. The second group consisted of 12 sites on the second molar and premolar teeth. The group formed by the least susceptible sites included the largest number of tooth surfaces and consists of the majority of the lower anterior teeth and canines. Conclusion Variation in the caries susceptibility of tooth surfaces exists. Surfaces can be grouped according to caries susceptibility. An effect that reduces the cariogenic challenge of one of the sites within a group is likely to affect all the other sites within the particular group.	Background The decline in caries that has occurred in industrialized countries over the past 30 years has been accompanied by major changes in the pattern of caries within the mouth. While the absolute levels of disease have declined, a relatively higher proportion of pit and fissured surfaces and lower proportion of approximal and smooth surfaces are involved. An additional feature in the pattern of dental caries is the existence of a surface hierarchy in susceptibility to caries [ 1 - 4 ]. These authors have reported that the most susceptible surfaces are pit and fissured followed by approximal surfaces on posterior teeth, and the least susceptible, approximal surfaces on anterior teeth. There is also a reported degree of symmetry both between the upper and lower jaws in the posterior sextants [ 5 ] and left and right side of the risk of caries. The concept is so well accepted that some survey systems for recording dental caries will only examine one side of the mouth and then double the score to give the total DMFS [ 6 ]. The fact that caries occurs bilaterally in the same type of tooth suggests that when a decline or increase in caries occurs, it presents in increments of 2. For a given DMF-T or S, there is a specific pattern of caries within a population. A working rule is that 'As caries in the population declines, caries in the least susceptible surfaces (approximal and smooth surfaces) decreases considerably more than in the most susceptible surfaces (pits and fissures)'. This pattern is independent of the presence of fluoride [ 7 ]. Furthermore, changes in mean DMF scores are not linear, but 'stepped' [ 4 ]. This stepped model of the changing patterns of caries suggests certain groups of teeth and tooth sites may have a similar 'resistance' to caries. When the resistance of one site in a group is increased, for example by fluoride, then all sites with similar resistance levels will also benefit and not become carious. Indeed, the existence of groups by resistance may explain the rapid stepped rates of decline of caries in the past 20 years [ 8 ]. A major reason for this rapid decline can be explained by the fact that groups of teeth or tooth sites with similar propensity to decay respond as a group to increase of resistance or reduction in the challenge. This would lead to a marked decline in overall caries levels rather than a gradual reduction. The present study tests the hypothesis that in children, there are groups of tooth sites by caries susceptibility. Methods Study population This study used the data from the National Preventive Dentistry Demonstration Programme (NPDDP) in the United States [ 9 ]. The NPDDP data set was used as it contains the most extensive and comprehensive longitudinal data available on caries preventive regimes using standardised DMF criteria that have been shown to be reliable through extensive critical analyses of the project. Perhaps most importantly for this project, the caries data range was very wide: both within the individual ethnic groups and according to water fluoridation status. The NPDDP included 20,052 children aged from 5 to 16 years of age from 10 locations in the USA: five fluoridated and 5 non-fluoridated communities. The mean DMF-S at the commencement of the programme was 2.43, and was 4.51 four years later. Preventive interventions were introduced that allowed an analysis of the changes in caries patterns with the decline in caries to be examined [ 9 - 11 ]. The caries surface data were selected for each child at the beginning of the study. Statistical methods To test the hypothesis probit analysis was used. Probit analysis is a method for examining any dose-response relationship where the dependent variable, i.e. caries, is dichotomous (caries/no caries). Since all tooth surfaces may not 'respond' in a similar manner, the problem must be formulated in terms of the proportion responding (diagnosed as caries) at each level of challenge. With probit analysis, any changes are constant in proportion so that changes on a log scale will also be constant. In a probit transformation each of the observed proportions were translated into the value of the standard normal curve below which the observed proportion of the area was found. For example, if half the subjects in a caries trial had one particular site carious, the probit value would be 0, since half the area in a standard normal distribution falls below a z score of 0. When using standard normal values, negative scores can occur. To overcome this, the constant 5 was added. For example, if a particular surface scored 1 and half the subjects studied had caries, the probit value would be 0, since half of the area in a standard normal curve falls below a z score of 0. When the constant is added, the transformed value for the proportion becomes 5. If the observed proportion of individuals in whom the site was carious was 0.84, the probit value would be 0.34, i.e. a z score of 1, which would give a transformed value of 6. The actual proportion of each of the tooth sites recorded carious at each DMF-S level was calculated and replaced with the value of the standard normal curve below which the observed proportion of the area below the curve was found. The logarithmic transformation of the data provides a linear relationship between the probability of an event occurring for any given value. The proportion of each surface within a population diagnosed as carious at a particular DMF-S score can be calculated. For example, take the occlusal surface of the first right lower molar. At a DMF of 1, a certain proportion of this site within a population would be carious. This proportion would increase for a DMF of 2, and so on. For each DMF-S score, the percentage of tooth surfaces exhibiting caries was calculated to give a probability score of between 0 and 1. Subsequently, the log transformation of the probabilities for each DMF-S against the actual DMF-S score was plotted for every surface. The common reference value, of 0.5 outlined above was used to establish the susceptibility of each surface to a given caries challenge. As some random variation can be expected, the susceptibilities are grouped within bands rather than as individual sites. To establish whether a hierarchy for caries susceptibility exists, the probability of finding each site carious for a given DMF-S score was calculated. The probability was calculated by adopting a common reference value for the proportion of the tooth sites that become carious. In the present study 0.5 was used to provide the most accurate value, although any value of proportion can be used. The question can then be phrased in terms of the value of the DMF-S at which 50% of the sites or teeth would be expected to have become carious. The probability scores derived are then aggregated to produce an overall picture of tooth and site susceptibilities. The probability of an event occurring ranges from zero to 1. As the overall DMF score rises, the probabilities of any site being carious will change. For example, if a group of 128 individuals, each having a DMF-S score of 1 with all sites exhibiting the same propensity for caries. The distribution of sites affected within the mouth should be random, the probability for a particular site being carious would then be 1/128. If, however, the group all had a DMF-S of 128, the relative probability of finding a particular site carious remains the same, but this time the absolute probability changes from 1/128 to 1. As the DMF score increases the probabilities alter. However, once a probability of 1 has been reached no further increase is possible. Thus, when examining changes in the distribution of caries at different DMF-S scores, the ratio between individual probability scores is unimportant. The crucial factor is the overall ranking exhibited by the probabilities for each site. The order of susceptibly will be determined by the relative values of the probabilities. Whether an individual site is twice as likely to become carious as another cannot be determined using this approach. However, certain sites may exhibit similar probabilities. For example, a particular site on the left hand side of the mouth may have a similar mean probability as the corresponding site on the right hand side. Other factors may influence the distribution of probabilities. For example, it has been suggested that fluoride has a more beneficial effect on approximal sites when compared to occlusal. The probabilities derived are for each site for each individual and are then aggregated for the sample population. The aggregation of probabilities gives rise to a distribution, approximately normal in character, and the mean of this distribution is subsequently reported and used in the analyses. Data analyses were performed using SPSS. The data files of the NPDDP were supplied by the Rand Corporation in ASCII format and subsequently read onto the mainframe system. Two of the five data files were utilised in this project: the master file containing the demographic information of each individual and the clinical file containing the status of each tooth site. Results Figure 1 shows the distribution of probabilities of caries grouped into 5 categories. The categories were formed by grouping together sites with similar probabilities of having experienced caries. The most susceptible groups of sites were defined as having a probability of being carious within the range 0.34 to 0.23, the next group 0.18 to 0.04, then 0.03 to 0.01, then 0.008 to 0.002 and, finally, the least susceptible sites which formed the remaining group (Figure 1 ). Figure 1 Distribution of probabilities of site susceptibilities. The emerging pattern indicates a left:right side symmetry with the propensity of attack similar for the two sides of the mouth. Furthermore, there is a degree of symmetry between the upper and lower jaws in the posterior sextants. For the anterior sextants, teeth in the upper jaw are more prone to attack than those in the lower jaw. An important finding is the relative sizes of the groups by the probability of susceptible sites having experienced caries. The group with most susceptible sites consisted of six tooth surfaces: the pit and fissured surfaces of the first molar teeth. The second group consists of 12 sites on the second molar and premolar teeth. The fifth group, and least susceptible group, is the largest and consists of the majority of the lower anterior teeth and canines. The groups, in order of susceptibility, allowing for some left:right asymmetry, were: 1. occlusal surfaces of 1st molars and buccal pits of lower 1st molars; 2. occlusal surfaces of 2nd molars and buccal surfaces of lower 2nd molars and occlusal surfaces of all 2nd premolars; 3. occlusal surfaces of 1st premolars, palatal surfaces of upper lateral incisors, approximal surfaces of 1st molars, lingual surfaces of lower 1st molars and buccal surfaces of upper 1st molars and palatal surfaces of upper 2nd molars; 4. all approximal surfaces of 2nd premolars, all approximal surfaces of upper 1st premolars, mesial and lingual surfaces of lower 2nd molars and distal and buccal surfaces of upper 2nd molars, approximal surfaces of upper central incisors, some approximal surfaces of upper and lower lateral incisors, all approximal surfaces of lower central incisors and distal approximals of upper canines, and approximal surfaces of 2nd molars; and 5. all surfaces of lower canines, buccal and mesial and labial aspects of upper canines, all smooth and approximal surfaces of lower 1st premolars, smooth surfaces of lower central incisors, approximal surfaces of lateral incisors (Figure 1 ). The next analysis was to establish the probability of finding a particular surface type carious for each DMF score. The probability, by surface, was converted into a ratio as in Figure 1 . Figure 2 combines the findings shown in Figure 1 with the patterns by tooth surface. To facilitate comparison by surface, the probability of each surface type was calculated, the total for each DMF-S score being equal to one. As the DMF score increased, the ratio of smooth to approximal to pit and fissured surfaces changes, although for the lower DMF scores, any major change in the ratios did not occur until a DMF-S of 9. Only then did the contribution of lesions on approximal or smooth surfaces make a significant contribution to the overall DMF-S. Figure 2 Proportion of each tooth surface type affected by caries at each DMF-S score. Discussion This study shows that that a number of tooth sites exhibit similar susceptibilities to caries. Susceptibility of tooth sites is not only similar for homologous pairs but also for grouped sites. For example, at a DMF score of 1, six sites have very similar probabilities of being carious. These involve the pit and fissured surfaces of all first molar teeth. Most authors, whilst accepting a degree of left:right symmetry in caries attack, have assumed that it is either the same site affected on the left hand side of the mouth as on the right [ 12 - 16 ]. This study reported that precise symmetry of caries did not occur by site but rather that symmetry of caries existed by groups of sites. For example, in the anterior incisor region, where several sites have similar propensities, the left:right symmetry may have been lost because the mirror image site had not undergone cavitation. However, due to the similarities in susceptibilities, another site on the opposite or even the same side has cavitated. In both cases symmetry will be lost. What exists is groups of teeth by susceptibility to caries. The concept of groups of susceptible sites has a number of important implications. If the application of a caries preventive strategy leads to a reduction in either the attack intensity or an increase in the resistance of the sites within a group to a value at which a particular site was protected then all sites in the group would also be protected. Depending upon the size of the group, several sites may well be protected. The existence of groups of sites goes some way to explain the stepped nature exhibited by changing patterns of caries. This may explain the apparent rapidity with which DMF levels decreased initially with the decline in caries reported in many industrialised countries since the 1970s. The adoption of a preventive strategy such as the use of fluoride toothpaste to reduce the attack intensity or increase in resistance for the least susceptible group, would lead to substantial savings in the total number of cavitated sites. There is a change in the changing ratio of smooth and approximal sites on the one hand, and pit and fissured surfaces on the other, for different levels of DMF-S, confirming findings by Burt [ 17 ], Dummer et al., [ 18 ], and Vehkalahti et al., [ 19 ]. For example, at a DMF-S of 1, the ratio of pit and fissured to smooth or approximal surfaces is 99:1. At a DMF-S of 10, the ratio had changed to 3:1. However, the changing ratios should be considered against the overall decline in caries. When considering the hierarchy of susceptibility by groups of sites the types of sites must be considered. With a change in the surface type ratio in which pit and fissured surfaces predominate to one in which approximal surface involvement becomes important, the consequences of the latter are smaller. The sites with a similar high propensity for caries attack at a DMF of 10, which includes a higher proportion of approximal surfaces, are not necessarily ten times as great as for an individual with a DMF of 1. This has important implications for planning preventive strategies. More effort would be required to reduce caries from low to very low levels than from high to low caries. Finally, the mathematical relationships identified by Batchelor and Sheiham [ 20 ] describing the distribution of caries at the population level could be combined with the findings of this study, to develop a model to assess the impact of caries preventive strategies. The model would help provide a scientific basis on which to formulate caries preventive strategies. Conclusions The findings show that there is a hierarchy by tooth types and sites in the pattern of dental caries attack in children. While it was not possible to identify precisely the order that each tooth or site succumbed to caries, groups of sites or teeth can be placed in a hierarchy of risk of caries. This concept expands the proposals of a hierarchy of 'within mouth' zones [ 21 - 23 ]. More importantly, there are groupings of tooth sites by susceptibility to caries. The size of the groupings varies. The impact of preventive agents that increased the resistance or reduced the intensity of the challenge affects most sites within a particular group. The larger sized groups occur at high caries levels. Increasing their resistance or lowering the intensity of the caries challenge would lead to a substantial drop in cavitated sites providing the agent or agents offered sufficient protection for any single site in the larger groups. Competing interests The author(s) declare that they have no competing interests. Authors' contributions PAB conceived, undertook the design of the study, performed the statistical analysis and drafted the manuscript. AS participated in the study design and coordination and helped to draft the manuscript. Both authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526778.xml
544857	From bench to clinic and back: Perspective on the 1st IQPC Translational Research conference	Translational Research (TR) provides a set of tools and communication context for scientists and clinicians to optimize the drug discovery and development process. In the proceedings of a Princeton conference on this timely topic, the strengths and needs of this developing field were debated. Outcomes and key points from these discussions are summarized in this article which covers the topics of defining what we mean by translational research (both theoretically and in operational terms), ways in which to engender the TR mindset and embed it in organizations such as the pharmaceutical industry in order to optimize the impact of available technologies (including imaging methods), the scientific basis and under-pinnings of TR including genomics knowledge, information sharing, as well as examples of application to drug discovery and development. Importantly, it should be noted that collaborations and communications between the stakeholders in this field, namely academia, industry and regulatory authorities, must be strengthened in order for the promise of TR to be delivered as better therapies to patients.	Introduction There are many challenges facing pharmaceutical companies in the post-genome era not least of which is declining productivity and innovation. Not surprisingly, there is agreement between Industry, Academia and Regulatory communities that the drug discovery and development process needs to change in order to meet the future needs of patients with effective and desirable drugs. A key part of the strategic solution is to leverage the application of TR principles and practices, which if implemented will go a long way towards addressing the challenge posed by FDA's Critical Path Initiative [ 1 ] (for more detail on this initiative see section on "Optimizing the Impact of TR" and ). Successful drug development requires satisfying a matrix of domains from relevance to the disease and the drug-ability of the target through feasibility and convenience of drug delivery, demonstration of favorable benefit-risk profile in order to achieve to a drug label that reflects physician and patient acceptance. Herein lies a key role for TR in helping to navigate this journey. In order to promote discussion on this topic, the International Quality and Productivity Center (IQPC) organized a Translational Research Conference (20–22 September 2004 Princeton, NJ) that hosted a small group of clinical and basic science researchers and individuals from the pharmaceutical industry. Among the topics discussed were how to define "Translational Research", how to expedite the transfer of pre-clinical findings to influence development plans, how to select biomarkers to ensure support for decisions, how new strategies can be effectively translated to practical tactics, and what team players and collaborations are necessary to conduct successful TR. Success factors identified include: Identification and validation of novel drug targets, development of robust and validated assays to screen drug leads for safety and potential efficacy in humans, and the identification of suitable patients for expedited but informative trials. Defining Translational Research While the goal of TR is to implement in vivo measurements and leverage preclinical models that more accurately predict drug effects in humans, TR itself can be defined in many ways. At its core however, is the thesis that information gathered in animal studies can be translated into clinical relevance and vice versa, thus providing a conceptual basis for developing better drugs. It could in fact be argued that the designation of a special term or definition for TR might be unnecessary or even misleading. Historically the term was assigned to create awareness and advocacy for the general public, clinicians and scientific communities, and especially for the government and other private sponsors [ 2 ] in this evolving discipline. Nonetheless, the basis for TR lies in sound scientific and clinical research principles. Whatever the precise definition TR, it should serve as a forum to find a "common language" for clinicians and scientist in navigating the complexities of basic scientific approaches, data analysis and information processing. It clearly implies the need for an intensive training for scientists and clinicians in multiple disciplines to acquire expertise and experience to conduct TR. For the purposes of this symposium, the scope of translational research was defined as the application of scientific tools and methods to drug discovery and development. This can be achieved by integrating information concerning a) exposure (pharmacokinetics), b) biological activity (pharmacodynamics including safety profiles) delineating differences between species and leading to the validation of target and mechanism biomarkers, and c) outcomes leading to an understanding of efficacy and safety between species and ultimately to the qualification or linkage of biomarkers to clinical outcome (for a fuller discussion on Biomarkers and Surrogate Endpoints see definitions in [ 3 ]). Thus TR can be used to mitigate risk and enhance drug development opportunities. Therefore, in taking a pragmatic or operational rather than a definitional approach, a key to a successful translation of non-human research to human clinical trials lies in the choice of biomarkers. While biological pathways tend to be homologous across species and more so than pharmacokinetic parameters such as absorption and clearance, animal models themselves have a poor record of predicting human disease outcome. Nonetheless, biomarkers are the key for prediction of biological activity if not drug efficacy in humans. At least three types of biomarkers can be identified: (1) target biomarkers measuring the interactions between a drug and its target; (2) mechanism biomarkers measuring their downstream biological effects and (3) outcome biomarkers that reflect efficacy and safety. A second dimension can also be ascribed to biomarkers to help drug developers assign risk assessment to such approaches. This sub-classification links desired utility to points on a risk continuum; e.g. low, medium and high, in which 'low' describes a biomarker applied solely to animal models for example for selecting compounds for progression into humans, whereas 'medium' is association with utility for some aspects of early clinical profiling of efficacy and safety including across species correlation, and 'high' is associated with reproducibility and qualification as an outcome or even regulatory tool in humans. Additionally, TR itself undergoes an evolution from pathfinding (hypothesis generating) to discovery research, to development, and finally to application. Each of these operational phases is amenable to being evaluated or supported by biomarkers, either for the definition of objectives, proof of principle or in assessing risk and feasibility. Consequently the right choice of biomarkers can help drive decision-making and lower the costs and cycle-time for progression of a new drug from the bench into the clinic. In summary, whatever the definition or classification ultimately used, in practical terms translational tools should be developed and applied on a "fit for purpose" basis with prior assessment and agreement of attendant risks. Optimizing the Impact of Translational Research Traditional Research and Development (R&D, also referred to as Discovery and Development) paradigms have accentuated the boundaries between the territories of discovery and development worlds and have not been conducive to bridging key transition points. This is unfortunate since the development world tends to lag behind advances made in discovery, a point recognized by FDA in launching the Critical Path Initiative [ 1 ]. In brief, this initiative challenges Industry and others to develop and implement better tools, such as biomarkers, trial modeling and simulation and other solutions, in order to optimize the development and regulatory stages of a product's life. While advances have been made on streamlining forward progression of R&D through organizational linkages, what has not happened to the same degree is a bi-directional flow of information, namely flow of information from the clinic (e.g. clinical validation or lack thereof) back into the hands of the discovery scientist. The consequence of this is that the biological models used to qualify drug candidates may fail to be predictive of subsequent drug responses in the clinical setting. Thus a practical outcome of TR is to improve the overall probability for technical success (POS) in drug development. Consequently the next paradigm for R&D optimization depends not only on leveraging emerging technologies such as pathway mapping and in silico modeling, but also the need to empower key scientists and clinicians with the task of enhancing the prediction and iteration learning cycle. Since there are different organizational solutions for embedding the TR mindset within an organization, a key element is to provide TR expertise to drug development teams. Furthermore, innovation and productivity values are critically linked through information exchange. Rapid iteration (e.g., learn-confirm cycles) and transfer of knowledge gained from prototype development experience will enable more rapid compound redesign against the highly desired target and be reflected as enhanced innovation. On the productivity side, the tools outlined in the Critical Path Initiative [ 1 ], once effectively implemented, will lead to enhanced development productivity but only if information exchange occurs efficiently across different functions. Hence a backbone for TR is support by user-friendly informatics systems. The journey however starts at understanding the scientific foundations of physiology and pathophysiology, thus providing a rational linkage between the gene, its expressed product, disease expression and ultimately outcome. The discipline of biomarker identification and development as mentioned previously encompasses these principles and is a core tool in the TR scientist's armamentarium. Biomarkers (which are not necessarily Surrogate endpoints and few are in fact) are key tools for escorting the drug candidate from the bench to the bedside and back. That is they can be both animal "diagnostic" as well as human "diagnostic" tools. A key implementation tool is therefore to identify early on which biomarkers may be of value and to study these in the relevant animal models, that is, specifically include them in preclinical screening paradigms, as well as identify their role (e.g., go / no go decision factors) in the clinical development plan. Biomarkers, which include imaging techniques as well as protein and genetic markers, may fulfill several roles in R&D from compound screening and selection through dose justification, decision-making and risk mitigation, however the key is to overtly link them to the discovery and development plans with a priori agreed performance characteristics, such that there is agreement on the utility of the marker. There are many good examples of the value or non-value of preclinical models in predicting subsequent human response and safety. The journey from preclinical experience to the clinic is a well-worn one (e.g., Xenograft testing for oncology), albeit without the degree of overall predictiveness we would desire. On the other hand there is a marked paucity of examples in which clinical experience or observation was translated back into a legitimate drug target and discovery effort (e.g. Viagra). Thus, a major opportunity lies in both developing more sensitive and specific animal models of disease (e.g. knock in/out) as well as fully leveraging novel clinical observations. At the same time it is the ultimate validation in the clinic that counts, and rapid feedback of that information will allow the conditional probabilities and learning cycle to be enhanced. By enabling these principles through organizational and cultural change, the impact of TR will be determined by direct impact on high-quality mid-phase transitions as well as reduced cycle-times and resource burdens. Basic science, genomics and Translational Research The era of genome-scale biology has seen an increase in, and production of, vast amounts of biological data together with an extensive increase in biology-oriented databases. To make the best use of biological databases and the knowledge they contain, different kinds of information from different sources must be integrated in ways that make sense to biologists. A major component of the integration effort is the development and use of annotation standards such as ontologies. Ontologies offer a conceptualization of domains of knowledge and facilitate both communication between researchers and the use of domain knowledge by computers for multiple purposes. Therefore, the Gene Ontology (GO) project was founded in 1998, in an attempt to provide consistent descriptors for gene products, in different databases; and to standardize classifications for sequences and sequence features. Since then, the GO Consortium has grown to include many databases, including several of the world's major repositories for plant, animal and microbial genomes [ 4 ]. Despite vast differences in genome size among various species, genes can be highly conserved at the level of protein sequence allowing searching for an unknown human protein function in yeast. As new genome sequences are being rapidly generated, and where comparative genome analysis requires the integration of data from multiple sources, it is especially relevant to provide rigorous ontologies that can be shared by the scientific community at large. In the past, biological processes and the underlying genes, proteins, other molecules and environmental factors, have been studied separately more than on an integrated basis. The challenge, however, for future research on human disease is to understand not only the mechanistic basis, but also the underlying dynamics of gene product expression. Thus, biological research should emphasize the analysis of pattern of gene expression over individual measurements. GO has been developed to predict behavior of entire biological systems, being assigned to three aspects: (1) Molecular Function describes activities, such as catalytic or binding activities, at the molecular level, e.g. kinase activity. (2) Biological Process describes biological goals accomplished by one or more ordered assemblies of molecular functions, e.g. 'cell death' can have both subtypes, such as 'apoptosis', and subprocesses, such as 'apoptotic chromosome condensation'. (3) Cellular Component describes locations, at the levels of subcellular structures and macromolecular complexes, e.g. 'nuclear inner membrane' with the synonym 'inner envelope' [ 4 ]. The powerful use of comparative gene expression analysis in human disease was exemplified with a recent study on gene expression profiles of gastric cancer patients and their correlation to survival. Leung et al. [ 5 ] have shown that Phospholipase A2 group IIA (PLA2G2A) expression is associated with prolonged survival and less frequent metastasis by studying gene expression patterns in human gastric cancers. This observation was confirmed in an independent set of patient samples by using quantitative RT-PCR. Beyond its potential diagnostic and prognostic significance, this result suggested that the activity of PLA2G2A may suppress progression or metastasis of human gastric cancer. In summary, the application of mathematical models and computer simulations to analyze gene expression profiles and to compare complex data sets of various origins may provide new insight into the pathogenesis of cancer progression and metastasis. The Gene Ontology (GO) project provides structured, controlled vocabularies and classifications that cover several domains of molecular and cellular biology and are freely available for community use in the annotation of genes, gene products and sequences. Translational Research in Drug Discovery: Strategies for Complex Systems Cancer vaccines are promising therapeutics designed to elicit immune responses against antigens expressed by tumor cells. However, vaccines that have worked well in preclinical models have not translated into consistent responses in the clinic. Since vaccines are comprised of multiple components, multiple immunological endpoints are used to identify the least effective vaccine components in cancer patients. Post-clinical research strategies are subsequently designed with a focus on improving the least effective vaccine components. To improve the performance of cancer vaccines in the clinics, which are traditionally judged by clinical endpoints, novel endpoints and biomarkers are needed to assist in understanding why cancer vaccines are not working. From clinic to bench, a systematic strategy is needed for pre-clinical optimization that addresses vaccine limitations identified in the clinics; and from bench to clinic, performance criteria need to be established for a follow-up clinical study. After gathering the therapeutic options, testing has to be prioritized on the basis of: a) already available data; b) availability of the therapeutic modality; c) models and assays available internally; d) turnaround time; and e) on the patent landscape. Prioritization and rapid evaluation of novel therapeutics will decrease the turnaround time and facilitate decision-making. However, several tools are needed to make this a reality. For example, complex therapeutic strategies require biomarker or even surrogate endpoints from clinical trials to direct development of second-generation therapeutics. The rapid qualification and choice of surrogate endpoints should be based on knowledge gathered by an "early-stage therapeutic opportunities database". This comprehensive database should include data on therapeutic targets, models, assays and published results and indeed the plethora of new therapeutic strategies in preclinical stages can only be managed by accessing informative databases. Moreover, pre-clinical compound optimization can be facilitated by establishing quantitative endpoints of short duration and lastly go / no go decision points must be established for surrogate endpoints and clinical responses in animal models. However, several current issues of scientific basis also have to be addressed, such as the importance of clinical surrogate endpoints, the relevance of animal models, lack of concordances between assays, and the lack of concordance between surrogate endpoints and the clinical response, in order to improve cancer vaccine development strategies. Applying Translational Research to Drug Development A core principle of TR revolves around validation of targets, biomarkers and treatment modalities in humans. These activities and drug development itself cannot be undertaken without patients or clinical data. How TR can be integrated in a multi-center, multi-cultural organization involving patient accrual from more than 38 different countries worldwide, for the research and treatment of cancer can be exemplified by EORTC , a non-profit organization conducting more than 100 clinical trials and treating 7000 cancer patients yearly. Advancement in basic science and immunology and an overwhelming revolution of biotechnology have changed the targets and endpoints in cancer trials from the mere assessment of cytotoxicity to defined mechanisms for potential anti-tumor effect. That is in the era of "targeted therapies" molecular therapeutics are now being designed to target "strategic" checkpoints that underlie the malignant phenotype. The challenges to be met are: 1) dealing with new compounds affecting novel molecular targets, 2) innovation in design and analysis of clinical trials, 3) cooperation between translational researchers and network of clinical investigators and 4) informed patients. The major concerns in conducting clinical trials are rising costs coupled with efficacy rates as low as 5% in cancer patients, making signal to noise detection not only difficult but expensive. The need for research on tumor tissue requires the set-up of tumor banks and the associated administrative burden often discourages young oncologists. EORTC established a tumor bank comprising real tissue samples but including a "virtual review" by pathologists. This ensures the availability of a well-categorized and prognostically evaluated collection of primary tumors and allows an online-searchable bank for researchers to access. Indeed the tumor bank harbors paraffin-embedded tumors, as well as frozen tumor tissues and storage of tissue is de-centralized at the institute where it is collected. To assure equal quality of tissue, which in outcome of scientific experiments can be compared, standardization of the collection and storage methods is fundamental. Therefore, protocols for storage, retrieval and tracking of tissues will be standardized and implemented in all participating laboratories. Access to the tumor bank allows screening of many available tumor samples for the expression of molecular targets and will help to unravel novel biomarkers for diagnosis and treatment. Such access will allow us to overcome the missed opportunities due to lack of tissue collection in clinical trials, which could have allowed better pre-screening of potentially responsive patients based on expression of certain biomarkers e.g., expression of bcl-2, and the treatment of target positive patients may have ensured a better clinical outcome in this target class. The challenge of testing promising new modalities for the cure of disease that had shown efficacy in experimental models lies in a lack of understanding of the underlying mechanism, heterogeneity of human genetic backgrounds and a lack of suitable controls in human studies. Strategies have been developed at the NIH for the global monitoring of patients by studying, with high-throughput technology, the systemic effects of treatments as well as their effect within the target organ. For this bedside to bench effort, a systematic sampling of human tissues of local (site of immunogen application), systemic (circulation) and peripheral (tumor site) origin needs to be standardized to ensure high quality of samples avoiding degradation of protein, RNA and DNA. This TR approach allows experimental studies in human samples during or after therapy through amplification of transcripts for analysis of minimal sample tissue, and the application of monitoring techniques for genetic profiling. Further, proteomic-based approaches allow following the kinetics of the mechanism of actions of therapeutics. Studying the effects of treatment in a bedside to bench approach provides markers for the characterization of disease process and/or testing hypotheses generated by experimental models. Therefore, the nature of research in the clinical setting can realistically be described as 'hypothesis generating", rather than 'hypothesis driven', through a discovery-driven approach. Analysis of the genetic background can reveal polymorphism of genes involved in immune reactions, such as cytokines and their receptors, which might influence the outcome of immunological interventions in different patient populations [ 6 ]. Analysis of disease heterogeneity can be approached by transcriptional analysis, through linear amplification of RNA and subsequent analysis by cDNA array and transcriptome array, and/or functional protein analysis, through protein characterization by proteomics [ 2 , 7 ]. Numerous tumor-antigen based cancer vaccine studies have shown that there is a functional dissociation between systemic circulating cytotoxic T cells and tumor infiltrating T cells (TIL). Tumor antigen-specific T cells have been demonstrated to have a quiescent phenotype and consequently cell cycle activation requires antigen-specific stimulation, as well as non-specific stimulation by IL-2 [ 8 ]. In addition, the local release of immune inhibitory factors by tumor cells is influencing the T cell phenotype and cytotoxicity leading either to tumor regression or recurrence [ 2 ]. To understand these complex mechanisms, it is important to study the tumor microenvironment by collection of large libraries of relevant clinical specimen, such as excisional biopsies or fine needle aspirates (FNA). FNA have the advantage to allow serial sampling of the same tumor site over time and treatment and to allow a prospective follow up of a given lesion. Studying of the tumor microenvironment will provide invaluable insights into mechanisms involved in disease progression and/or changes affected by therapy, in terms of genes whose expression changed due to (1) genetic instability, (2) immune selection or (3) immune regulation. Despite the many obstacles in monitoring therapeutic effect in early phase clinical trials and the lack of hypothesis, the scientific significance of these trials should be reviewed assuming that the new treatment will not be beneficial. Desirable outcomes include learning about the disease process, the primary goal of the therapy and the reasons for its failure. Another concern should be if we have taken advantage of the patient population accrued at least to learn something, although independent of treatment, about the disease process itself. Clinical trials should therefore be designed, within ethical constructs, to look at questions beyond the ones related solely to treatment. This can be achieved through (1) establishment of libraries of relevant clinical samples for immediate or future studies, (2) prospective collection of data into a consistent format, and (3) tight link between clinical and scientific data. Developing better therapies for chronic inflammatory diseases also exemplifies the use of the latest technological advances in TR such as proteomics, transcriptomics and cellomics, for identification or application of biomarkers. Chronic inflammation frequently precedes the development of cancer in adults, such as lung [ 9 ], esophageal, gastric and pancreatic cancers. This may be due to a switch from apoptotic (scheduled) to necrotic (unscheduled) tumor cell death induced by mechanisms related to the chronicity of the inflammatory response. Acute inflammatory processes caused by viral or bacterial infections are in most cases effectively cleared by the immune system of a competent host. Some infections and other causes of inflammation such as solar exposure to the skin, prolonged tobacco smoke or chemicals, can also lead to prolonged inflammatory processes. In these chronic up-regulated situations, products of cyclooxygenase activity, or nitric oxide accumulating at the local inflammatory site lead to augmented cell proliferation and death. This is often be linked to hypermethylation of promoter regions in tumor-suppressor and/or pro-apoptotic genes. Persistence of defects in the apoptotic machinery provokes necrotic cell death and the release of cellular contents, which in turn enhances cell growth, cancer progression and infiltration of leukocytes including tumor-associated mast cells and macrophages. Several factors, such as: 1) the nuclear protein HMGB1, 2) the S100 family of molecules; 3) purine metabolites, ATP, AMP and uric acid, and 4) heat shock proteins have emerged as relevant mediators or "endogenous damage or danger signals" to recruit inflammatory cells, to promote wound healing and associated stromagenesis, angiogenesis; and ultimately to modulate immune functions [ 10 ]. Until recently, methods to measure necrotic death in patients were not available. The application of proteomics to identify factors, such as HMGB1 in serum of cancer patients, has revealed elevated serum levels in patients with metastatic melanoma, pancreatic cancer and others [ 10 ]. The correlation of these serological markers of necrotic cell death with histological patterns, genetic resistance to apoptotic death in animal models could lead to novel targets for immune therapy, such as antibodies to HMGB1, in order to interrupt the "circolo vizioso" of this "addiction to death" which promotes tumor growth [ 9 ]. Current attempts for cancer therapy focused on vaccination to antigenic targets or application of cytokines have resulted in measurable anti-tumor reactivity in the blood; however, these therapies have mostly failed to show a correlation with tumor outcome or progression. Therefore, to more completely understand and identify factors assessing tumor death could inform and drive the development of more effective biological therapies for cancer patients. Sample acquisition in the blood includes serum/protein collection for Seldi-Tof mass spectrometry; and the collection of cells for microarray, proteomics, and high contents screening via cellomics. Protein chip Seldi-Tof MS has been already successfully used to discriminate serum expression profiles in various cancer types [ 11 - 13 ]. The complexity of these advanced, high-throughput technologies will exponentially increase the amount of data, with the consequence that the main activities of future biological and medical laboratories will be in data analysis and integration rather than in data collection. Therefore, specialized teams are required for collaboration efforts in order to manage data warehousing, mining and analysis, and thus establishing networks for the identification and application of biomarkers. Beside proper study design, the models chosen to perform data classification and to estimate classification errors are highly critical for the complex data analysis. The identification of diagnostic markers for cancer, or markers to identify responders vs. non-responders to therapy requires systematical analysis of healthy vs. diseased, then of benign inflammatory disease vs. malignant cancer. Thus, methods to perform statistical analysis (e.g. permutation, randomization) are powerful, intuitive and provide an objective position from which to assess results. To handle these complex data analysis problems, the University of Pittsburgh has formed the Pittsburgh Supercomputing Center (PSC) headed by Dr. Arthur W. Wetzel, in a joint effort with Carnegie Mellon University and Westinghouse Electric Company, and is to date the most powerful open-resource computer available. Imaging tools and Technologies for Translational Research There are many examples of the value of weaving molecular imaging into Investigational New Drug Development. At the same time, the scale of the initial investments required vs. perceived benefits may not gain the necessary support of decision makers for application into development programs. There is a clear need to educate on the power and limitations of nuclear imaging techniques within the context of enhancing new drug development. Within this context, a primary goal for TR is to emphasize the cultural and operational shifts required of various stakeholders including academia, in order to better partner with industry. The term imaging covers a range of available techniques, including discovery autoradiography, small animal imaging (PET and MRI), traditional anatomical imaging (Ultrasound, MRI, CT), functional imaging (MRI, PET, SPECT) and many new tracers are available as are techniques with increased sensitivity to enable micro-doing studies (AMS) [ 14 ]. Nuclear imaging techniques are powerful tools and can be used for a number of development objectives. These include a number of goals described below. Firstly demonstrating drug penetration into the tissue of interest and co-localization or binding with the intended target through receptor occupancy (e.g., labeled ligand displacement), including describing dose vs. target occupancy curves remains a key objective an done used frequently in early clinical research. A second objective involves the quantification of a compound's pharmacokinetic (PK) profile using radio-labeled compound, an analysis that can be performed on a region of interest basis e.g. to assess time on target as well as potential therapeutic benefits vs. side effects. Additionally, imaging can be used to quantify pharmacodynamic (PD) effects of drug action and their relationship to administered dose. In combination, PK/PD information thus derived can be used to select a dose with which to test the clinical hypothesis or help quantify the therapeutic index. From a TR perspective all these techniques can be applied in the discovery and preclinical phases to facilitate compound selection and optimization as well as in the clinical phases. A key question emerges in applying these technologies: "How best to get it done" and the debate of internal imaging centers vs. external networks and academic relationships quickly emerges. On balance, it is clear that there is not one ideal solution here rather in general a collaborative approach between industry and academia is recommended. As a consumer of medical imaging, industry is a critical player in driving innovation and the paradigm shift towards more frequent yet appropriate utilization. However, a partnership approach ultimately generates better value and cost-effectiveness for the Imaging discipline as a whole. Conclusions and path forwards TR is an approach to foster communication between the scientific community and clinical practitioners. To maximize the value this can bring requires that public and governmental education has to be improved in order to leverage understanding and advocacy. There are many benefits to be accrued from this, not least of which being for the patient that is waiting for meaningful therapeutic advances. New drugs have to be developed fast and show effect on the right target at the earliest possible stage of development in order for industry to become more innovative and productive and medicines to be less expensive. Amongst other specific aspects required, are the strengthening of educational opportunities for physician scientists to help prepare them to conduct effective TR. At the same time, discovery science should be conducted by scientists who have been trained in relevant disciplines including cell biology and pharmacology as well as molecular biology. This in turn requires grant support for TR-related projects. Specifically, young scientific investigators should have more access to grants from governmental bodies and foundations in order to conduct research on clinical samples. This funding is largely in the hands of government leadership. Other points for disseminated education include the availability of a plethora of tools available to conduct and advance TR and development opportunities that include high quality clinical sample collection. Lastly, since TR is information intensive, considerable efforts are required to provide accessible databases and share knowledge. To help ameliorate this gap and provide access to information derived from human experimentation and to optimize the communication between clinicians and scientist, Dr. Marincola founded the Journal of Translational Medicine , an Open Access, peer-reviewed online journal, so that more therapeutic insights may be derived from new scientific ideas – and vice versa . In conclusion, TR represents a team effort, since no single constituency can be fluent in all aspects, and thus a concerted effort is needed amongst translational researchers to convince stakeholders and legislators of the need to support TR efforts, and thus maximize its potential.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544857.xml
529463	The prevalence and severity of oral impacts on daily performances in Thai primary school children	Background Traditional methods of measuring oral health mainly use clinical dental indices and have been complemented by oral health related quality of life (OHRQoL) measures. Most OHRQoL studies have been on adults and elderly populations. There are no systematic OHRQoL studies of a population-based sample of children. The objective of this study was to assess the prevalence, characteristics and severity of oral impacts in primary school children. Methods Cross-sectional study of all 1126 children aged 11–12 years in a municipal area of Suphanburi province, Thailand. An OHRQoL measure, Child-Oral Impacts on Daily Performances index (Child-OIDP) was used to assess oral impacts. Children were also clinically examined and completed a self-administered questionnaire about demographic information and oral behaviours. Results 89.8% of children had one or more oral impacts. The median impact score was 7.6 and mean score was 8.8. Nearly half (47.0%) of the children with impacts had impacts at very little or little levels of intensity. Most (84.8%) of those with impacts had 1–4 daily performances affected (out of 8 performances). Eating was the most common performance affected (72.9%). The severity of impacts was high for eating and smiling and low for study and social contact performances. The main clinical causes of impacts were sensitive tooth (27.9%), oral ulcers (25.8%), toothache (25.1%) and an exfoliating primary tooth (23.4%). Conclusions The study reveals that oral health impacts on quality of life in Thai primary school children. Oral impacts were prevalent, but not severe. The impacts mainly related to difficulty eating and smiling. Toothache, oral ulcers and natural processes contributed largely to the incidence of oral impacts.	Background Contemporary concepts of health suggest that dental health should be defined in physical, psychological and social well-being terms in relation to dental status [ 1 , 2 ]. That is why Cohen and Jago considered that the greatest contribution of dentistry is to the improvement of quality of life because most oral diseases and their consequences interfere with, or have impacts on, daily life performances [ 3 ]. Therefore, disruptions in normal physical, psychological and social functioning are important considerations in assessing oral health. Despite these suggestions, traditional methods of measuring oral health use mainly clinical dental indices and focus on the absence or presence of oral diseases. They do not inform us about the oral well-being of people in terms of feelings about their mouths, or, for example, their ability to chew and enjoy their food. The inadequacy of the normative approach in measuring oral health has been recognised and lead to the development of measures of oral health-related quality of life (OHRQoL) [ 4 ]. A number of socio-dental or OHRQoL measures have been developed and used for assessing oral well-being and to describe oral impacts on people's quality of life [ 5 ]. Generally, they measure the extent to which oral conditions disrupt normal social role functioning and lead to major changes in behaviours, such as changes in ability to work or attend school, or undertake parental or household duties [ 6 , 7 ]. In addition to describing oral impacts on quality of life, some OHRQoL measures were designed specially to assist dental service planning by incorporating them with traditional normative measures in the process of dental needs assessment [ 8 , 9 ]. Most studies using OHRQoL to assess oral impacts of the mouth and teeth have been on adults and elderly populations. Few studies have been done on children possibly because no OHRQoL measures designed for use with children existed until recently. A single measure, dental pain, has been used on children in Malaysia [ 10 ] and in South Africa [ 11 ]. They found a high prevalence of pain that affected daily living. Similarly, a study in New Zealand found that most school children complained of at least one dental symptom [ 12 ]. To date, there are no systematic OHRQoL studies of a large population-based sample of children. In particular, the OHRQoL of primary school children, who are frequently the main target group for dental health services, has not been assessed. Therefore the objective of this study was to use an OHRQoL measure, the Child-OIDP, to assess the prevalence, characteristics and severity of oral impacts in primary school children. Methods A cross-sectional survey was carried out in a municipal area of Muang district, Suphanburi province, Thailand. The sample was all 1,126 students aged 11–12 years, in the final year class of all primary schools (grade 6) in the area. Data were collected through: a) an interview for oral impacts using the Child-OIDP [ 9 ], by one interviewer b) a self-administered questionnaire for demographic information such as age, sex and occupation of the father and mother, or male and female guardians [ 13 ] and oral health behaviours and c) an oral examination by four calibrated community dentists, mainly based on the WHO guidelines [ 14 ]. Orthodontic normative treatment needs were assessed by the Index of Orthodontic Treatment Need (IOTN) [ 15 ]. Oral hygiene was also assessed using the Simplified-Oral Hygiene Index (OHI-S) [ 16 ]. All documents were translated from English to Thai and the validity was checked by a back-translation method, involving blind re-translation into English. The validity of the translation was verified by experts in the use of questionnaires in both languages. This was also checked after wording modifications, in order to ensure the conceptual and functional equivalences of the questionnaires. A pilot study was carried out to validate all questionnaires before using them in the main data collection. The psychometric properties of the Child-OIDP in terms of face, content and concurrent validity as well as internal and test-retest reliability were excellent. The index was also practical to use with this age group. Full description of the validation process of the Child-OIDP can be found elsewhere [ 9 ]. For the main data collection, test-retest reliability of data was tested by ten percent random duplication. Weighted kappa score for the Child-OIDP was 0.91, kappa scores of self-administered questionnaires were 0.7–1.0, and those of intra- and inter-examiner for oral examinations were 0.7–1.0 and 0.6–1.0 respectively indicating good to excellent agreement. The SPSS and Stata programmes were used for statistical analysis. The protocol of the study was approved by the Ethical Committee of the Ministry of Public Health of Thailand. Primary education and local health authorities as well as all primary schools in the study areas gave permission. Positive consent forms and letters informing parents were sent to parents. Measuring oral impacts and calculating their severity Two comprehensive OHRQoL measures specifically for use with pre-adolescent children have recently been developed; the Child Perceptions Questionnaire (CPQ11-14) [ 17 ] and the Child Oral Impacts on Daily Performances (Child-OIDP) [ 9 ]. Both were validated on a cross-sectional study using a proxy, because no gold standard is available; therefore, at this stage, they should be considered discriminative and not yet evaluative OHRQoL measures. However, they differ mainly in their aims and theoretical frameworks. The Child-OIDP index was developed on a large population-based sample with the aim of being used for dental health service planning. Its theoretical framework is the same as for the original OIDP, namely oral health consequences are categorised into different levels and the index measures only oral impacts on daily performances at the ultimate level [ 8 , 9 ]. The Child-OIDP has the advantage over the CPQ11-14 in that it specifies the different clinical causes of each oral impact and therefore the treatments needed. Although the objective of current study is to assess oral impacts of children, a broader aim of the project was to assess the implications of using measures of oral impacts to estimate dental needs of children. Therefore the Child-OIDP was selected for this study as it is specifically designed to be incorporated into a needs system. The procedure for using the Child-OIDP began with a self-administered questionnaire carried out with all children as a group in their classroom. The questionnaire contains a list of all oral problems that children are likely to perceive and also include an open answer for any unexpected perceived problem. It was developed during a pilot study, as a modification from the one used in the original OIDP. Children were asked to identify oral problems that they perceived in the last three months. This step aimed to focus children's attention to their oral health problems and to lead to the oral impacts assessment later. Their answers here were used only as a guide to investigate oral impacts on daily performances in the next step and were referred to when they were asked about the causes of oral impacts in individual interviews. Thereafter, children were individually interviewed, irrespective of their answers at the first step, to assess oral impacts on daily life in relation to 8 daily performances. The 8 performances were: a) eating, b) speaking, c) cleaning teeth, d) relaxing, including sleeping, e) smiling, laughing and showing teeth without embarrassment, f) maintaining emotional state, g) study, including going to school and doing homework and h) contact with other people. The individual interviews were aided by 16 pictures (negative and positive pictures for each performance). If children reported an impact on any performance, the frequency of the impact and the severity of its effect on their daily life were scored. Children were also asked to identify oral problems that in their opinion, caused the impact. The oral problems were identified from the list complied in the first step of the assessment. The oral impact score of each performance is obtained by multiplying severity and frequency scores, 0, 1, 2 or 3 each, in relation to that performance. Therefore scores can range from 0 to 9 per performance. The overall impacts score is the sum of all 8 performances (ranging from 0 to 72) divided by 72 and multiplied by 100. An alternative method of reporting the severity of oral impacts, from the same data set, is to use the 'intensity' and 'extent' of impacts. The intensity refers to the most severe impacts on any of the 8 performances or the highest performance score. It is classified into 6 levels; none, very little, little, moderate, severe and very severe (Table 1 ). The idea behind this is to differentiate between for example, a child with minor impacts (score of 1) on 6 performances and another child with severe impacts (score of 6) on only 1 performance. In the former case, the child will be in the 'very little', and in the latter one, in the 'severe' category. The extent refers to the number of performances with impacts (PWI) affecting a child's quality of life over the past three months. It ranges from 0 to 8 PWI. The relationships between the impact score and intensity as well as between score and extent were statistically significant (p < 0.001) [ 18 ]. Intensity and extent of impacts represent an alternative method of describing or comparing oral impacts on children. They are more straightforward and could give a simpler and clearer picture of impacts than using a single score. Therefore, they provide a more practical aspect to the OHRQoL assessment making it more easily applicable to dental service planning. Table 1 Classification of the intensity of oral impacts on a performance The intensity of impacts Severity score Frequency score Performance score Very severe Severe (3) × Severe (3) 9 Severe Severe (3) × Moderate (2) 6 Moderate (2) Severe (3) Moderate Moderate (2) × Moderate (2) 4 Severe (3) × Little (1) 3 Little (1) Severe (3) Little Moderate (2) × Little (1) 2 Little (1) Moderate (2) Very little Little (1) × Little (1) 1 No impact None (0) × None (0) 0 Results 1101 of the 1126 children returned positive consent forms approved by their parents. 1034 children (91.8% of the total) completed all stages of the survey. 52.4% were male and 47.6% were female. Their mean age was 11.3 years (sd = 0.6). The highest percentage of their fathers were agricultural workers or labourers (34.5%), 30.5% worked in business/private, 27.5% in governmental sectors, 2.1% did not work and 5.4% of children did not have a male guardian. The highest percentage of mothers worked in business/private (38.6%), 24.5% in agriculture, 21.1% in governmental organisations. 14.9% did not work and 1.0% did not have a female guardian. This population had a low level of dental caries: 43.1% were caries free and the DMFT scores ranged from 0 to 12 with a median score of 1.0 and a mean of 1.5 (±1.8). Almost all children (97.0%) had a Community Periodontal Index (CPI) score of 1 or more; 84.2% had calculus. In terms of oral hygiene status, 5.4% had good, 69.1% had moderate and 25.5% had poor oral hygiene. OHI-S scores ranged from 0.5–5.5 with a median of 2.5 and mean score of 2.5 (±0.9), indicating a moderate level of oral hygiene. The prevalence of oral impacts was high; 89.8% of children had experienced some kind of oral impact on their daily life during the past three months. There was no difference between the prevalence of impacts in girls and boys (Chi-square test). Impacts on Eating were the most prevalent (72.9%). The prevalence of impacts on Emotion (58.1%), Cleaning teeth (48.5%) and Smiling (40.1%) were also relatively high. The remaining prevalences of impacts were lower, namely Study (15.4%), Relaxing (14.7%), Contact with people (12.2%) and Speaking (9.9%) (Table 2 ). Table 2 Prevalence, intensity and score of oral impacts in Thai school children Performances Oral impacts on daily performances Overall impacts Eating Speaking Cleaning teeth Relaxing Emotion Smiling Study Contact Prevalence (%) 89.8 72.9 9.9 48.5 14.7 58.1 40.1 15.4 12.2 Impact intensity (% of children with impacts) - Very little 15.9 27.9 37.4 33.2 37.4 43.7 25.5 57.8 49.2 - Little 31.1 39.0 33.3 38.8 44.2 37.2 28.2 31.2 38.5 - Moderate 31.7 21.8 19.2 20.8 14.3 13.9 27.4 9.7 10.7 - Severe 18.7 10.8 9.1 6.6 3.4 4.7 15.7 1.3 1.6 - Very severe 2.6 5.5 1.0 0.6 0.7 0.5 3.2 0.0 0.0 Impact score - Range 0–59.7 0–9 0–9 0–9 0–9 0–9 0–9 0–6 0–6 - Mean (sd) 8.85 (7.4) 1.87 (1.8) 0.23 (0.9) 1.13 (1.6) 0.30 (0.7) 1.17 (1.4) 1.21 (2.0) 0.25 (0.7) 0.21 (0.7) - Percentiles (25,50,75) 2.8,7.6,12.5 0,2,2 0,0,2 0,0,0 0,0,0 0,1,2 0,0,2 0,0,0 0,0,0 Extent and Severity of impacts Among the children with impacts, the extent of impacts varied from 1 to 8 performances with impacts (PWI); 16.2% had 1 PWI, 23.3% had 2, 26.9% had 3 and 18.4% had 4 PWIs. Few children had 5 or more PWIs. About 1 in 5 children had severe or very severe intensity of impacts; 18.7% had severe and 2.6% had very severe intensity of impacts.15.9% had very little, 31.1% had little and 31.7% had moderate intensity of impacts (Table 2 ). The intensity of impacts on each performance showed that Eating and Smiling were the most severely affected while Study and Contact were the least. 16.3% of children with impacts on Eating and 18.9% of those on Smiling had severe or very severe impacts, while the same intensity was reported by 1.3–10.1% of children having impacts on other performances. 57.8% of children with impacts on Study and 49.2% of those on Contact had a very little or little level of impact intensity, whereas none had a very severe intensity of impacts on those two performances. The distribution of overall impact scores was skewed (Table 2 ). They ranged from 0.0 to 59.7 with a median of 7.6 and a mean score of 8.8 (sd = 7.4). No difference in overall impact scores were identified between different sexes (Mann-Whitney U test). Mean scores of impacts on each of the 8 performances ranged from 0.21 to 1.87 (maximum possible score is 9). Mean impact score for Eating (1.87) and Smiling (1.21) were the highest while those for Study (0.25) and Contact (0.21) were the lowest (Table 2 ). 'Causes' of the impacts There were various oral and dental problems that children perceived as the causes of their overall oral impacts (Table 3 ). The more prevalent problems leading to impacts were a sensitive tooth (27.9%), oral ulcers (25.8%), toothache (25.1%) and an exfoliating primary tooth (23.4%). Furthermore, oral conditions that related to appearance frequently affected children; position of teeth (20.0%) and colour of teeth (16.2%) were quite frequently cited. In addition, swollen or inflamed gums were related to overall impacts in 13.8% of children. Table 3 Frequency of oral conditions perceived as causing overall oral impacts Oral conditions causing overall impacts Frequency (%) Toothache (t-ache) 25.1 Sensitive tooth (t-sensitive) 27.9 Tooth decay, hole in tooth 5.0 Fractured permanent tooth 4.6 Colour of teeth (colour) 16.2 Shape or size of teeth 2.7 Position of teeth (position) 20.0 Bleeding gum (bleed) 7.4 Swollen or inflamed gum (swollen) 13.8 Calculus 0.9 Bad breath 7.2 Oral ulcer (ulcer) 25.8 Exfoliating primary tooth (exfoliat) 23.4 Tooth space (due to unerupted permanent tooth) (space) 5.3 Erupting permanent tooth 4.9 Deformity of mouth or face 0.4 Missing permanent tooth 0.7 The main perceived causes of impacts on each of the 8 performances are shown in Figure 1 . Toothache and oral ulcers were among the main perceived causes of impacts on 6 performances. The majority of impacts on Eating were caused by toothache (64.5%) and on Speaking by oral ulcers (57.8%). An exfoliating primary tooth was one of the main perceived causes of impacts on the following 5 performances; Eating (17.9%), Cleaning (29.5%), Relaxing (11.2%), Emotion (17.5%) and Study (17.6%). Position of teeth was among the main perceived causes of impacts on 3 performances; Smiling (40.7%), Contact (19.8%) and Emotion (10.0%). Space due to a non-erupted permanent tooth (after exfoliation) was one of the main reasons for impacts on Smiling (11.1%). Bad breath was the most frequent perceived cause of impacts on social Contact (27.0%). Figure 1 Main oral conditions causing impacts on each of the eight performances. Abbreviations refer to Table 3. Discussion The prevalence of oral impacts experienced during the past three months by the study population was very high (89.8%). This is surprising in that this was a low caries population in an area with a free accessible school dental service. Although there is no study using OHRQoL index with a population-based sample of 12 year olds to compare with, findings of previous OHRQoL studies suggest that oral impacts are very common in children of this age. In Brazilian adolescent populations, the prevalence of impacts was 32% [ 19 , 20 ] and 62% in Uganda [ 21 ]. In child populations, a 88% prevalence of dental pain was reported in South African 8–10-year-old school children [ 11 ] and 73% of New Zealand children with good oral status had at least one dental symptom in the past year [ 12 ]. That was higher than the 60.1% reported in Malaysian children who also had good oral status and received successful school dental services [ 10 ]. A study using the CPQ11-14 index with paedodontic patients found that all the children had oral impacts in the past three months [ 17 ]. These findings indicate that oral impacts may be higher in children than in adults. For example, compared to studies using the original OIDP index [ 8 ] with other older age groups, the prevalence of oral impacts in a Thai adult population was 73.6% [ 22 ] and 52.8% for a Thai elderly population [ 23 ]. In a UK national survey of elderly people the prevalence of OIDP impacts was 17% for edentate and 14% for dentate participants [ 24 ]. Despite the fact that oral impacts were prevalent in this Thai child population, they were not severe. For example, half of this population had Child-OIDP score less than 7.6 and half of those with impacts had very little or little intensity of impacts (Table 2 ). Moreover, many clinical causes that contributed to the prevalent impacts do not last long; that is oral ulcers, exfoliating teeth and spaces due to a non-erupted permanent tooth. This study found that eating was the most important aspect of OHRQoL of children. Difficulty with eating due to oral problems was the most common impact (72.9%), and led to more severe oral impacts on children's quality of life than impacts on other performances. Oral ulcers and exfoliating primary teeth contributed to eating difficulties in nearly half of those with impacts. The finding that eating was the most common performance affected is similar to all studies using the OIDP in all age groups [ 19 , 21 - 24 ]. They are also similar to a study using the CPQ11-14 with paedodontic patients where impacts on functional limitations were more common than impacts on emotional and social well-being [ 17 ]. Difficulty with smiling was another important aspect of children's OHRQoL. It affected 40% of children. The most prevalent cause was position of teeth. Dissatisfaction with position of teeth, moreover, accounted for oral impacts in 1 in 5 of all children (Table 3 ). Although there is no study documenting the extent of pre-adolescent children's concern about their oral appearance, it is evident that the concern is important when they reach adolescence [ 25 ]. For example, de Oliveira and Sheiham found that adolescents with untreated malocclusions were significantly more likely to report oral impacts on their daily lives than those who had completed orthodontic treatment [ 26 ]. Chen and Hunter found that psychological impacts of oral health, such as avoiding laughing and being teased about teeth, were more prevalent in children than in adults and elderly [ 12 ]. Gum problems were the other important oral conditions affecting children's OHRQoL. More than one fifth of children perceived that bleeding and swollen gums caused oral impacts on their life, particularly in relation to difficulty cleaning, a problem experienced by nearly half of all children (Table 3 , Figure 1 ). Children with difficulty cleaning their teeth because of gum inflammation are unlikely to achieve good levels of oral hygiene because brushing may lead to bleeding, and their gum problems would undoubtedly remain or even get worse. This problem would not be solved by the traditional dental treatment without understanding the affects of oral impacts on behaviour. An interesting finding was that impacts relating to social dimensions, such as study being affected and contact with people, were less common and least severe. Schor suggested that children's social performances rely more on their physical and psychological performances than adults [ 27 ]. It is apparent that an important reason for the high prevalence of oral impacts in children is natural processes such as exfoliating primary teeth or spaces due to a non-erupted permanent tooth. They contributed largely to the high incidence of impacts in these pre-adolescent children. On the other hand, these conditions were not reported as important causes of oral impacts in other age groups [ 19 , 22 ]. The findings on the other clinical causes of oral impacts in this study was consistent with what Jaafar found in Malaysian children, namely, toothache and oral ulcers [ 10 ]. Moreover, it is noteworthy that despite the fact that this was a low caries population having access to free dental service, sensitive teeth and toothache were frequently reported causes across the various impacts, particularly so with respect to the more common impact of difficulties with eating. Although children could often not specify precisely which impairments led to impacts, the question of perceived clinical causes should exclude impacts from some conditions which are definitely not related to actual impairments as well as to treatment needs. For example, toothache, ulcers and conditions relating to appearance definitely require different treatment and could be easily differentiated. However, the accuracy of detecting perceived impairment is limited in a population-based study, while it can be improved at the individual level of investigation. The specific age group under investigation, particularly in relation to their stage of development, may have influenced the high prevalence of oral impacts. Developmental changes unavoidably affect HRQoL between childhood and adolescence [ 28 ]. Maturity and an increase in age generate a more sophisticated understanding and perceptions about health and illness [ 29 ]. Therefore, perceptions about health and quality of life of children change as they mature [ 28 , 30 ]. This might make younger children more sensitive to oral symptoms than older age groups. Because of those considerations the modification of the Child-OIDP addressed the main possible problems that might arise when employing adult measures with children [ 30 , 31 ]. They include the adjustment of the 8 items of daily performances, simplification of rating scales, decrease of the time frame and rearrangement and clarification of the complex questions that were beyond the capability of children under 12 years according to Piaget's cognitive development theory [ 32 ]. Moreover, the use of pictures as aids is considered of value when interviewing children [ 33 , 34 ]. In addition to the modification, another advantage of the Child-OIDP lies in its conceptual framework where oral health consequences are divided into three levels; the first level represents oral problems (such as tooth decay), the second or intermediate level represents symptoms (such as pain) and the third or "ultimate level" represents difficulty in daily performances. The index measures impact at the ultimate level only, which could reduce double scoring, by not measuring twice the same impacts experienced at different levels. For example, pain is not scored whereas difficulty with eating due to pain is scored. In addition, this approach could reduce the uncertainty of children's perception and interpretation and therefore make the index more applicable for children [ 35 ]. Fink explained that HRQoL can be measured through different types of information. Measuring impacts on daily functioning is more objective and reliable than measuring reported health problems or symptoms which are more influenced by individuals' perception and interpretation [ 36 ]. Thus, HRQoL measures for children that involve subjective reported problems or symptoms such as pain are frequently problematic, because children's interpretation and perception about health differ from adults [ 30 ]. On the other hand, HRQoL measures that focus on information about functioning, such as the Sickness Impact Profile, may readily be applied to children as well as adults [ 37 ]. Therefore, to reduce a problem with children's interpretation about their health or symptoms, the technique of assessing HRQoL based on activities of daily living is appropriate [ 35 ]. Conclusions The prevalence of oral impacts on daily performances in this child population was very high. Oral impacts affected children's quality of life mainly through difficulty eating and smiling. There are various oral conditions that contributed significantly to the incidence of impacts, namely, sensitive teeth, toothache, oral ulcers and exfoliating primary teeth. Although the prevalence of impacts was high, the severity was not; many children had their quality of life affected at low levels. This reveals a need for further longitudinal studies to better understand and interpret OHRQoL measures in children. Authors' contributions SG carried out all work including data collection, data analysis and writing the paper. GT supervised the project and assisted writing. AS initiated the idea of, and supervised the project and edited writing.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC529463.xml
548274	Treatment of trigeminal ganglion neurons in vitro with NGF, GDNF or BDNF: effects on neuronal survival, neurochemical properties and TRPV1-mediated neuropeptide secretion	Background Nerve growth factor (NGF), glial cell line-derived neurotrophic factor (GDNF) and brain-derived neurotrophic factor (BDNF) all play important roles in the development of the peripheral sensory nervous system. Additionally, these growth factors are proposed to modulate the properties of the sensory system in the adult under pathological conditions brought about by nerve injury or inflammation. We have examined the effects of NGF, GDNF and BDNF on adult rat trigeminal ganglion (TG) neurons in culture to gain a better understanding of how these growth factors alter the cytochemical and functional phenotype of these neurons, with special attention to properties associated with nociception. Results Compared with no growth factor controls, GDNF, at 1 and 100 ng/ml, significantly increased by nearly 100% the number of neurons in culture at 5 days post-plating. A significant, positive, linear trend of increasing neuron number as a function of BDNF concentration was observed, also peaking at nearly 100%. NGF treatment was without effect. Chronic treatment with NGF and GDNF significantly and concentration-dependently increased 100 nM capsaicin (CAP)-evoked calcitonin gene-related peptide (CGRP) release, reaching approximately 300% at the highest concentration tested (100 ng/ml). Also, NGF and GDNF each augmented anandamide (AEA)- and arachidonyl-2-chloroethylamide (ACEA)-evoked CGRP release, while BDNF was without effect. Utilizing immunohistochemistry to account for the proportions of TRPV1- or CGRP-positive neurons under each growth factor treatment condition and then standardizing evoked CGRP release to these proportions, we observed that NGF was much more effective in enhancing CAP- and 50 mM K + -evoked CGRP release than was GDNF. Furthermore, NGF and GDNF each altered the concentration-response function for CAP- and AEA-evoked CGRP release, increasing the E max without altering the EC 50 for either compound. Conclusions Taken together, our results illustrate that NGF, GDNF and BDNF differentially alter TG sensory neuron survival, neurochemical properties and TRPV1-mediated neuropeptide release in culture. In particular, our findings suggest that GDNF and NGF differentially modulate TRPV1-mediated neuropeptide secretion sensitivity, with NGF having a much greater effect on a per neuron basis than GDNF. These findings are discussed in relation to possible therapeutic roles for growth factors or their modulators in pathological pain states, especially as these relate to the trigeminal system.	Background Among many trophic factors that act on sensory neurons, three have been studied extensively: nerve growth factor (NGF), glial cell line-derived neurotrophic factor (GDNF) and brain-derived neurotrophic factor (BDNF). During development, NGF, GDNF and BDNF, along with neurotrophin-3 (NT3), support the survival of subpopulations of sensory neurons through their cognate trk receptors [ 1 - 3 ]. In the adult, uninjured rat, receptors for NGF, GDNF and BDNF are found in partially distinct subpopulations of sensory neurons. NGF-responsive, trkA-containing neurons are mostly small diameter, contain the neuropeptide calcitonin gene-related peptide (CGRP), and are generally thought to have nociceptive properties [ 4 , 5 ]. The BDNF-responsive, trkB-containing population of sensory neurons is predominantly larger diameter and mechanosensitive [ 6 ], although there is considerable overlap with the trkA population [ 7 ]. GDNF receptors are more widespread, with as many as 60% of dorsal root ganglion (DRG) neurons containing c-ret, while GFR components are found throughout the c-ret population, as well as in c-ret negative neurons [ 8 , 9 ]. Additionally, sensory neurons that are postnatally dependent on GDNF for survival bind the isolectin B 4 (IB 4 , [ 2 ]). NGF is thought to play a primary role in the development and maintenance of several pro-algesic states. NGF produces sensitization of nociceptive responses in vivo and increases responsiveness to chemical stimuli associated with nociceptive neurotransmission in vitro [ 10 - 12 ]. NGF increases the expression of the pro-inflammatory neuropeptide CGRP [ 13 ] and increases substance P (SP) and CGRP content in sensory neurons [ 14 - 17 ]. NGF also promotes the development and maintenance of hyperalgesia following chronic constriction injury [ 18 , 19 ]. Additionally, NGF increases capsaicin (CAP) sensitivity both in vivo and in vitro [ 20 - 24 ]. In accordance with this increased CAP sensitivity, NGF increases the expression of the CAP- and noxious heat-sensitive ion channel vanilloid receptor type 1 (TRPV1) through the ras/p38 MAP kinase signaling pathway, thereby promoting thermal hyperalgesia [ 25 , 26 ]. Additionally, NGF has recently been shown to modulate TRPV1 by releasing the receptor from phosphotidylinositol-4,5-bisphosphate-mediated inhibition [ 27 ]. GDNF also plays a role in modulating nociceptive processing, but with contrasting in vitro and in vivo effects. In cultured DRG neurons, GDNF increases CAP sensitivity and TRPV1 expression [ 16 , 26 ] and increases neuropeptide content [ 16 ]. On the other hand, intrathecal injection of GDNF has no effect on C-fiber evoked outflow of SP and does not induce thermal hyperalgesia, while NGF does both [ 15 ]. This, coupled with the observation that GDNF-overexpressing mice do not develop thermal or mechanical hypersensitivity [ 28 ], suggests that GDNF might not lead to nociceptor sensitization in vivo. In fact, in nerve injured rats, GDNF has antihyperalgesic effects [ 29 ] and promotes the functional regeneration of DRG-spinal cord connections [ 30 ]. It is not understood how this dissociation of the in vitro and in vivo effects of GDNF is manifested. Inflammation and nerve injury both increase BDNF content in DRG neurons and in the spinal cord, and this increase in BDNF is associated with the maintenance of a hyperalgesic state [ 31 - 36 ]. Furthermore, BDNF is released in the spinal cord upon noxious afferent stimulation [ 37 , 38 ], and peripheral CAP application increases BDNF release at central terminals of sensory neurons [ 39 ]. Hence, BDNF might act as a neurotransmitter in the pain pathway in adult animals [ 40 ]. The trophic properties of BDNF on adult sensory neurons, particularly nociceptors, are poorly understood. In addition to CAP, a number of other TRPV1 agonists have been described. Among these are compounds that are also cannabinoid receptor agonists, including the endogenous cannabinoid anandamide (AEA, [ 41 , 42 ] and the synthetic AEA analogue arachidonyl-2-chloroethylamide (ACEA, [ 43 ]). AEA causes antinociception in vivo through central [ 44 ] and peripheral mechanisms [ 45 , 46 ]. It is not known, though, whether the endogenous production of AEA can reach sufficient concentrations under pathophysiological states to act as an endogenous activator of TRPV1. One potential mechanism whereby growth factors could contribute to the development of nociceptor sensitization is through the modulation of the efficacy or potency of AEA at TRPV1.Thus, growth factors might unmask conditions under which AEA could be capable of functioning as an endogenously produced TRPV1 agonist, potentially leading to neurogenic inflammation and thermal hyperalgesia. The aim of the present work was to gain a more precise understanding of NGF-, GDNF- and BDNF-dependent alterations of cultured TG sensory neuron survival and phenotype, with attention to markers and functions that are related to nociception. Furthermore, by relating these findings to neurosecretion associated with pharmacological stimulation of TRPV1, we hope to present a rationale for differences in how these neurotrophins might contribute to altered nociception at the level of the TG sensory neuron. Moreover, there is a gap in knowledge concerning neurotrophic factor influence over TG neurons, as the vast majority of our understanding of how neurotrophins alter sensory neurons stems from studies of the DRG. The recent advances of CGRP receptor antagonists for the treatment of migraine [ 47 , 48 ] raises the possibility that manipulations which influence CGRP expression and/or secretion might be beneficial in the treatment of migraine or conditions related to the cerebral vasculature. A fuller understanding of the impact of NGF, GDNF and BDNF on TG neurochemical properties has the potential to lead to novel therapeutic strategies concerning disease states involving neuropeptide secretion in the trigeminal system, such as migraine. Results Effects of NGF, GDNF or BDNF treatment on neuronal survival in vitro TG neuronal cultures were plated at equal densities (~5000 neurons / well) on 8-well culture slides (experimental design shown in Figure 1 ). After five days of growth factor treatment, immunocytochemistry (ICC) for 200 kD neurofilament (NF-H) present in all sensory neurons [ 49 ] and not other ganglion cells (i.e., glia), was performed to determine the number of neurons in each well. Representative photomicrographs are shown in Figure 2 . NGF did not significantly influence the number of neurons in culture following five days of treatment (Fig 3 ). On the other hand, GDNF significantly enhanced the number of neurons per well versus no growth factor treated cultures at both 1 and 100 ng/ml (Fig. 3 ). The enhanced neuronal survival seen with GDNF was biphasic, as 10 ng/ml GDNF did not significantly enhance neuronal survival. It should be noted that in GDNF-treated cultures we observed a robust sprouting of TG neurons, appearing far more robust than in NGF- or BDNF-treated cultures (although sprouting was still evident in these cultures). Effects of NGF, GDNF or BDNF treatment on TRPV1 mRNA and protein levels To directly assess changes in TRPV1 mRNA levels with growth factor treatment, quantitative, realtime PCR was conducted. NGF, GDNF or BDNF (100 ng/ml) did not significantly alter TRPV1 mRNA levels after 5 days of treatment (Fig. 4 ). To evaluate whether growth factor treatment might exert post-transcriptional regulation of TRPV1 gene expression, we next examined their effects on TRPV1 protein levels by Western blot. As shown in Fig. 5 , a major band was detected with the TRPV1 antibody at ~130 kD, consistent with the glycosylated form of TRPV1 [ 25 ]. NGF and GDNF (100 ng/ml) each was demonstrated to have increased TRPV1 protein levels by approximately 50% compared with no growth factor-treated cultures (Fig 5 ). BDNF effects on TRPV1 protein levels were not assessed, because preliminary studies indicated that BDNF did not influence CAP-evoked CGRP release. Effects of NGF, GDNF or BDNF treatment on K + -evoked CGRP release and CGRP content We next examined the effect of growth factor treatment on K + -evoked CGRP release and total CGRP content in TG neuronal cultures. NGF treatment significantly and concentration-dependently increased 50 mM K + -evoked CGRP release at every concentration, peaking (10.5-fold increase vs. control) at 100 ng/ml (Fig. 6A ). Similarly, GDNF significantly and concentration-dependently increased K + - evoked release at every concentration, again peaking (9.7-fold increase vs. control) at 100 ng/ml. On the other hand, BDNF did not alter K -evoked CGRP release. Because there were differences in neuron numbers between growth factor conditions, we examined under each condition the proportion of neurons that were positive for CGRP, TRPV1 or IB 4 . Representative photomicrographs for each of these conditions are shown in Figure 7 . The proportion of neurons positive for CGRP, TRPV1 or IB 4 are shown in Table 1 . In no cases were CGRP- or TRPV1-immunoreactive or IB 4 -binding neurons not likewise immunoreactive for NF-H. We then utilized the proportion of CGRP-expressing neurons under NGF, GDNF or BDNF treatment conditions to normalize K + - evoked CGRP release to the number of CGRP-immunoreactive neurons under each condition, respectively (see Methods for normalization calculation). Following this transformation, NGF treatment still augmented K + -evoked CGRP release at every concentration tested, with the magnitude of the effect reaching 18 times that of control (Fig 6B ). While GDNF treatment again enhanced K + -evoked CGRP release at all concentrations following normalization, its peak effect was relatively lower. BDNF treatment continued not to have an effect following normalization to the number of CGRP-immunoreactive neurons. Much like K + -evoked CGRP release, total CGRP content was increased significantly by NGF and GDNF treatment, while BDNF had no effect (Fig. 6C ). Both NGF and GDNF increased CGRP content concentration-dependently at every concentration (p < 0.001), with peaks at 10 ng/ml (3.2-fold increase) for NGF and 100 ng/ml (3.0-fold increase) for GDNF (Fig. 6C ). Again, we normalized these data to CGRP-immunoreactive neurons to assess changes in relation to alterations in CGRP-immunoreactive neurons between the growth factor conditions. Interestingly, the GDNF augmentation of CGRP release appeared to be due primarily to the large increases in neuron survival, because only the 10 ng/ml GDNF condition was significantly greater than control (p < 0.001), but still substantial lower compared with the untransformed lysis figures (Fig. 6D ). On the other hand, normalization led to a further augmention in the effect of NGF on total CGRP content, up to a 5.8-fold increase. BDNF treatment did not alter the normalized CGRP content. Effects of NGF, GDNF or BDNF treatment on CAP-evoked CGRP release To assess the effect of growth factor supplementation on the neurosecretory function of nociceptors in culture, we evaluated CAP-evoked CGRP release. Treatment with 100 nM CAP for 10 min led to a significant enhancement of CGRP release over baseline under all conditions. Treatment of cultures with NGF or GDNF resulted in a concentration-dependent enhancement of this effect, with significant increases in CAP-evoked CGRP release at 1, 10 and 100 ng/ml. BDNF also enhanced the ability of CAP to evoke CGRP release; however, the effect was only seen at the 100 ng/ml concentration. In every case, GDNF and NGF (except 1 ng/ml NGF) supplementation significantly enhanced CAP-evoked CGRP release over the same concentration of BDNF (Fig 8A ). To determine to what extent the enhancement of CAP-evoked CGRP release in response to growth factor treatment might derive from increased neuronal survival and/or up-regulation in the proportion of neurons expressing TRPV1, release data were normalized to the number of TRPV1-immunoreactive neurons (Fig. 8B ; TRPV1 neuron proportions shown in Table 1 ). In this case, NGF treatment still significantly increased CAP-evoked CGRP release over no growth factor treatment at 1, 10 and 100 ng/ml. However, this normalization procedure led to a relative reduction in the GDNF effect on CAP-evoked CGRP release, such that only the 10 ng/ml condition reached a significant increase in CGRP outflow. After normalization, BDNF treatment had no significant effect on CAP-evoked CGRP release at 1 or 100 ng/ml and actually reduced the amount of release at 10 ng/ml. Moreover, following normalization, NGF-treated neurons displayed a significantly higher CGRP release than either GDNF- or BDNF-treated neurons at every concentration of growth factor tested (Fig 8B ). Because treatment with either GDNF or NGF induced a large increase in 100 nM CAP-evoked CGRP release, we assessed the effect of supplementation with these growth factors on the receptor pharmacodynamics of this response compared with no growth factor supplementation. GDNF or NGF treatment significantly increased the Hill slope of the concentration-response function from 1.12 ± 0.47 in the no growth factor condition to 3.760 ± 0.057 for 100 ng/ml NGF and to 4.48 ± 0.90 for 100 ng/ml GDNF (Fig 9A ) but had no effect on the EC 50 (41 nM for GDNF or NGF and 47 nM for no growth factor). GDNF (100 ng/ml) or NGF (100 ng/ml) treatment caused a significant, five-fold increase in the E max of the CAP concentration-response curve (Fig. 9B ). A common feature of CAP concentration-response functions is that they frequently exhibit an inverted U-shape, likely attributable to TRPV1 desensitization at higher concentrations. Thus, whereas the concentration-response function in the absence of growth factor displayed desensitization only at 1 uM, that for either GDNF- and NGF-treated TG cultures showed desensitization of the CAP-evoked CGRP response at concentrations above 100 nM (Fig. 9B ). Effects of NGF, GDNF or BDNF treatment on AEA- and ACEA-evoked CGRP release To test whether growth factor-mediated enhancement of neuropeptide release might be generalizable to other TRPV1 secretagogues, we evaluated the dual cannabinoid-vanilloid agonists AEA and ACEA. As in the case with CAP, the effects of AEA and ACEA on evoked CGRP release were augmented by supplementation of TG cultures with NGF (Fig. 10A ) or GDNF (Fig. 10B ). While TG neurons not treated with growth factors were largely unresponsive to AEA or ACEA, in terms of CGRP release, treatment of cultures with either NGF or GDNF for 5 days at 1, 10 or 100 ng/ml caused a concentration-dependent increase in CGRP release evoked by 30 μM AEA or ACEA. Furthermore, the estimated EC 50 values for each of the individual growth factors to increase evoked CGRP release were equivalent for CAP, AEA and ACEA (NGF = 3.5 ng/ml; GDNF = 1.3 ng/ml). In contrast, BDNF supplementation did not augment AEA- or ACEA-evoked CGRP release, while a small, yet significant, increase in CAP-evoked CGRP release was observed with increasing BDNF concentration (Fig. 10C ). When TG neurons were supplemented with either 100 ng/ml NGF or GDNF, the E max and EC 50 values for AEA did not change (F = 2.082 (4,85) ) ; however, when compared with non-growth factor-treated TG neurons, the E max was significantly augmented (Fig. 10D ). Discussion This study demonstrates that NGF, GDNF and BDNF differentially influence neuronal survival, neuropeptide content and stimulated secretion of TG neurons in culture. GDNF significantly augments TG neuronal survival, while both NGF and GDNF modulate CAP sensitivity, and alter the pharmacodynamics of the concentration-response function for CAP-evoked CGRP release. Furthermore, NGF and GDNF increased the releasable pool and total content of CGRP while increasing TRPV1 protein, without increasing its mRNA. These data support the hypothesis that GDNF and NGF, but not BDNF, alter CAP sensitivity in cultured TG neurons, and taken together, suggest that the differential effects of NGF and GDNF in vitro may reflect their differential effects in vivo, particularly with regard to TRPV1-mediated nociception. GDNF, but not NGF or BDNF significantly increased the number of neurons present in TG culture 5 days post-plating. The finding that this effect of GDNF was biphasic, as the 10 ng/ml concentration did not enhance the number of neurons in TG culture, suggests the possible involvement of different receptors with different concentrations of GDNF. Multiple receptors exist in the GDNF receptor family, and they function in concert with the protein c-ret. GDNF binds most readily to GFR alpha-1, but also binds to GFR alpha-2 / c-ret heterodimers [ 50 ]. GFR alpha-1 and -2 are expressed in sensory ganglia and are found mostly in IB 4 -binding neurons [ 51 ] that also contain c-ret [ 8 ]. Notably, it has been suggested that GFR alpha-1 and GFR alpha-2-containing neurons in the adult rat make up two distinct populations in the DRG within the IB 4 -binding class [ 8 ]; hence, the effects observed here may be due to GDNF acting through its high affinity GFR alpha-1 site at the 1 ng/ml concentration and through either or both GFR alpha-1 and GFR alpha-2 / c-ret, for which it has a lower affinity, at the 100 ng/ml concentration. While the observation here that GDNF promotes survival is a novel finding for rat TG sensory neurons in primary culture, the neuroprotective effects of GDNF are well documented. GDNF is protective against the loss of dopaminergic neurons in animal models of Parkinson's disease [ 52 , 53 ], and GDNF reduces the number of apoptotic bodies in DRG explants from adult mice [ 54 ]. GDNF also rescues the reduction of P2X3 expression that occurs in the DRG following axotomy [ 55 ]. Although BDNF and NGF did not significantly increase the number of TG neurons in culture, we did observe a significant, linear trend for increased neuron survival as a function of increasing BDNF concentrations, an observation not present in NGF-treated TG cultures. Hence, BDNF and GDNF both promoted the survival of TG neurons in vitro. Primary cultures generated here were grown in the presence of mitotic inhibitors, which greatly reduce the supportive glial cells and the trophic factors they normally provide TG neurons in the native ganglia (although some of these cells are still present). Removal of astroglial support promotes apoptosis in cerebral neuronal cultures, an effect which is reversed by addition of GDNF or BDNF, but not NGF, to the culture medium [ 56 ]. The addition of GDNF or BDNF to the culture medium here may have re-supplied, at least partially, the withdrawn glial-supplied trophic factors that maintain sensory neurons in vivo, thereby increasing the number of neurons in culture at 5 days. It should be noted, however, that despite the significant effects of the growth factors observed in this study on neuronal survival, from an original density of nearly 5,000 neurons / well in the original culture homogenate, only about 10% of neurons could be counted at 5 days post-platting, in the control condition. A maximum of just over 20% were counted in the 100 ng/ml GDNF treated TG cultures. Therefore, it stands to reason that GDNF and BDNF are supporting the survival of certain classes of sensory neurons that are not supported either without growth factors or with NGF alone. The proportions of TG neurons in culture that expressed CGRP- or TRPV1-immunoreactivity or IB 4 -binding sites were assessed with attention to how these proportions changed in the presence of NGF, GDNF and BDNF. Notably, the proportion of sensory neurons in vitro that expressed CGRP (~65%) was much larger than the known proportion of CGRP-containing neurons in native TG (~35%, [ 57 ]). This likely indicates that the culturing process conditions are either selectively preserving peptidergic neurons or that normally non-peptidergic neurons novelly express CGRP in culture. On the other hand, discrepancies on reports of the percentage of IB 4 -binding neurons in native TG, ranging from ~35–60% [ 58 , 59 ] make it difficult to assess whether there is an increased proportion of IB 4 -binding neurons in culture; however, our data indicate that this proportion is at least in the upper range, if not greater than native TG. Furthermore, a probable significant overlap exists between CGRP-immunoreactive and IB 4 -binding neuronal populations observed here, as both were found in the majority of TG neurons in culture, in agreement with the demonstration that these populations overlap significantly in native, adult rat TG and DRG (Price and Flores, unpublished observations). We observed that a greater proportion of TRPV1-immunoreactive neurons were present with GDNF supplementation (nearly 30% increase), again indicating either that GDNF preferentially supports the survival of TRPV1-expressing sensory neurons or that neurons that do not normally express TRPV1 begin expressing TRPV1 when GDNF is included in the culture medium. The latter proposition is supported by the finding that peripheral treatment with anti-GDNF antibodies suppresses the novel expression of TRPV1 in IB 4 -binding neurons following peripheral inflammation [ 60 ]. The percentage of TRPV1-expressing neurons in TG cultures not treated with growth factors was essentially equivalent to the percentage in native TG [ 57 ]. This finding indicates that, not only does GDNF promote survival of TG neurons in culture, but also enriches for TRPV1-expressing neurons, suggesting that GDNF might be preferentially neuroprotective for sensory neurons in adult animals that express TRPV1. We have also illustrated that chronic application of NGF or GDNF, but not BDNF (except at high concentrations, and to a much lesser degree), increases CAP-evoked CGRP release from TG neurons in vitro. NGF and GDNF each increased TRPV1 protein, and both upregulated the CGRP content of TG neurons. Both the NGF- and GDNF-induced upregulation of TRPV1 appears to be translationally regulated, as neither of these growth factors altered TRPV1 mRNA levels. NGF, in the setting of inflammation, is known to increase TRPV1 protein, but not mRNA, through activation of the p38/MAP kinase pathway [ 25 , 26 ], and NGF-mediated upregulation of TRPV1 is blocked by over-expression of dominant-negative ras [ 26 ]. Interestingly, the study by Bron et al. (2003) indicated that GDNF is also able to upregulate TRPV1 expression; although this conclusion was based on immunofluorescence and cobalt uptake assays, it is consistent with our direct demonstration of GDNF-induced upregulation of TRPV1 protein by Western blot. After normalization to the number of neurons that express TRPV1 and CGRP, we found that NGF had a much greater effect on CAP-evoked CGRP release and neuropeptide content, on a per cell basis, than did GDNF. On the other hand, GDNF increased the proportion of neurons in culture that express TRPV1, such that GDNF-maintained cultures contained a higher number of TRPV1-expressing neurons compared with NGF-maintained cultures. Hence, our findings suggest that NGF increases CAP responsiveness in individual cultured TG neurons. On the other hand, GDNF appears to increase the responsiveness of the in vitro population by altering the proportion of TRPV1 neurons and upregulating the releasable pool of CGRP, as evidenced by the persistent increase in 50 mM K + -evoked CGRP release after normalization. This difference in the ability of GDNF and NGF to enhance individual neuronal neuropeptide content and CAP responsiveness could partially explain why NGF induces hyperalgesia [ 10 , 12 , 19 , 61 ] while GDNF does not [ 15 , 28 , 29 ]. Our findings suggest that while GDNF might play a role in maintaining CAP sensitivity in vitro, its effects in vivo might be of a preservative nature that prevents the development of pain exacerbation following experimental manipulation. Furthermore, NGF and GDNF might differentially/predominantly subserve two of the main physiologic processes following injury: hyperalgesia to protect the organism from further injury (NGF) and regeneration/repair to restore function (GDNF). While it has been shown, using a variety of dependent measures, that both NGF and GDNF increase CAP responsiveness [ 16 , 17 ], this is the first demonstration that these growth factors qualitatively alter the pharmacodynamics of the neuronal response to CAP in TG neurons. The increase in the Hill slope of the CAP response following exposure to NGF or GDNF indicates that these growth factors induce positive cooperativity, possible at the level of TRPV1. While the mechanism underlying this effect is not known, it could involve an alteration in post-translational modifications and/or protein interactions of TRPV1 in response to NGF or GDNF. NGF or GDNF also decreased the concentration of CAP necessary to induce tachyphylaxsis compared with control cultures. These findings indicate that TRPV1-mediated sensory neuron desensitization might be a more efficacious therapeutic strategy in pathologies known to be associated with increased levels of NGF and/or GDNF. While we are unaware of any data linking GDNF with migraine, increased cerebrospinal fluid levels of NGF are associated with chronic headache [ 62 ]. Interestingly, a TRPV1 targeted approach has been utilized in a clinical trial for migraine treatment in which intranasal civamide (a TRPV1 agonist) was shown to be effective for acute treatment of migraine [63], indicating that TRPV1 agonists might be employed in this condition, possibly to desensitize CGRP-containing TG nerve endings. Similarly to CAP-evoked CGRP release, the ability of AEA and ACEA to evoke release was concentration-dependently augmented by NGF or GDNF supplementation but was unaltered by BDNF. Moreover, the respective potencies of NGF and GDNF to enhance the CGRP release in response to CAP, AEA and ACEA were equivalent. This suggests that the pharmacology of CAP, AEA and ACEA at TRPV1, at least with respect to neuropeptide secretion, is similarly regulated in the presence of either of these growth factors. Insofar as NGF and GDNF have been implicated in the development of nociceptor sensitization in a number of pathological states, it should be considered that endocanninoids are apt to have enhanced TRPV1-mediated peripheral neuromodulatory effects under these conditions. Conclusions Taken together, our results illustrate that NGF, GDNF and BDNF differentially alter sensory neuron survival, neurochemical properties and TRPV1-mediated neuropeptide release of TG neurons in culture. GDNF and, to a lesser extent, BDNF promote survival of TG neurons and GDNF enhances the proportion of neurons that exhibit TRPV1-immunoreactivity. GDNF or NGF enhanced CAP-, AEA- and ACEA-evoked CGRP release from TG neurons in vitro and increased TRPV1 protein, likely through translational regulation, and overall CGRP content. On the other hand, our findings suggest that GDNF and NGF differentially modulate TRPV1-mediated neuropeptide secretion sensitivity, with NGF having a much greater effect on a per neuron basis, providing a possible explanation for why NGF promotes thermal hypersensitivity in vivo while GDNF apparently does not. Although the present studies were conducted on cultured neurons under artificially controlled conditions, these findings contribute to a growing body of work concerning neurotrophin modulation of the properties of nociceptors. Thus the results described here have implications for advancing our understanding of how this system might be advantageously manipulated in a therapeutic setting especially as it concerns the TG system. Methods Experimental chemicals Capsaicin (CAP, 8-methyl-N-vanillyl-trans-6-nonenamide) was from Fluka-Aldrich (St Louis, MO). AEA ( N -(2-hydroxyethyl)-5Z,8Z,11Z,14Z-eicosatetraenamide, in water soluble emulsion), and ACEA ( N -(2-chloroethyl)-5Z,11Z,14Z)-eicosatetraenamide were from Tocris (Ellisville, MO). Recombinant rat GDNF and recombinant human (100% homologous to rat) BDNF were from Sigma (St Louis, MO). Rat NGF was from Harlan (Indianapolis, IN). All other chemicals were from Sigma, unless otherwise stated. TG culture Adult, male Sprague-Dawley rats weighing 250–300 g were used in this study. All procedures utilizing animals were approved by the Institutional Animal Care and Use Committee of The University of Texas Health Science Center at San Antonio and were conducted in accordance with policies for the ethical treatment of animals established by the National Institutes of Health. Animals were euthanized by decapitation and their TGs were rapidly dissected (~30 s) and placed in ice-cold Ca ++ - and Mg ++ - free Hank's balanced salt solution (HBSS, Gibco, Carlsbad, CA). TGs were enzymaticaly digested for 30 min with 5 mg/ml collagenase followed by 25 min with 0.1% trypsin type IX supplemented for the last 10 min with 10 units of DNase I (Roche, Indianapolis, IN). TG cell suspensions were then centrifuged at 2000 RPM for 2 min, vortexed briefly and centrifuged again. They were then resuspended in basal culture medium containing high glucose Dulbeco's Modified Eagle's Medium (DMEM, Gibco), 1X pen-strep (Gibco), 1X glutamine (Gibco) and 3 μg/mL 5-FDU and 7 μg/mL uridine as mitotic inhibitors. TG cell suspensions were gently triturated with a Pasteur pipette followed by successive triturations through 19- and 23-gauge needles. TG cell suspensions were then transferred to a separate container and adjusted to the total volume needed for plating at a density of ~5000 neurons / well. The appropriate growth factors were added to this suspension prior to plating. Immunocytochemistry (ICC) and Image Aquisition TG cultures were first washed in PBS and fixed for 1 hr in 3.7% formaldehyde in PBS. Next, TG cultures were washed 3 times in PBS and permeabilized in PBS containing 10% normal goat serum (NGS, Gibco) and 0.2% Triton X-100 (Sigma) for 1 hr. Finally, TG cultures were blocked 3 × 10 min in PBS containing 10% NGS and then exposed to NF-H mouse-monoclonal antibody (1:300; Sigma) overnight at 4°C. Primary antisera were then washed off and goat anti-mouse Alexa-Fluor-488 (1:300, Molecular Probes, Eugene, OR), or goat anti-mouse Alexa-Fluor-594 (1:300) was applied for 1 hr at room temperature. For double-labeling either CGRP rabbit polyclonal (1:750, Peninsula Labs) or TRPV1 guinea pig polyclonal (1:3000, Neuromics) or IB 4 -conjugated to Alexa-Fluor-488 (1:1000, Molecular Probes) was then added overnight at 4°C. CGRP or TRPV1 antisera were followed by goat anti-rabbit Alexa-Fluor-594 (1:300) for 1 hr at room temperature. All images were acquired using a Nikon E600 microscope (Melville, NY) equipped with a Photometrics SenSys digital CCD camera (Roper Scientific, Tucson, AZ) connected to a computer equipped with Metamorph V4.1 image analysis software (Universal Image Corporation, Downingtown, PA). Twenty 20X images were taken of each well to capture the complete cellular area for neuron counts. For these experiments, all neurons displaying fluorescent signal above background were counted as positive for the specific marker. This was defined by scaling the image, using Metamorph's built-in scaling feature, to an average pixel value for negative neurons in the case of TRPV1, CGRP and IB 4 (no scaling was performed for NF-H ICC) and establishing all neurons as positive that were above that threshold. For double labeling experiments, the thresholded images were then overlaid and analyzed for the presence of both signals to assess colocalization. Neuronal Counting (Survival) TG cell suspensions (~5000 neurons/well) were added to 8-well poly-D-lysine Lab-Tek II chamber slides (Nalge/Nunc, Naperville, IL). Neuron density for plating was determined by counting neurons with a hemacytometer with Neubauer rulings and the cultures for each slide were generated independently. Neurons were easily distinguished from other cell types for plating density measurements due to their large size and opaqueness. Medium was changed after 24 hr and 72 hr, and on day 5, ICC was performed (as described above). In total, 12 chamber slides were utilized. The experimental design is shown in figure 1 with 4 slides each for NGF, GDNF and BDNF. Neurons on the NF-H alone slides were not counted as these slides were used only for images shown in figure 2 . To assess neuronal survival, all NF-H- immunoreactive neurons were counted for each growth factor concentration and are presented as mean ± SEM. The 2 matching wells for each slide were averaged and this average was used for 1 observation (as described in Fig 1 ). For colocalization studies, the previously generated NF-H-immunoreactive neuron counts for each well were followed by counting of CGRP- or TRPV1-immunoreactive or IB 4 -binding neurons in the same wells to generate the proportion of neurons expressing these markers. The total for the 2 matching wells were summed to yield the final proportion. At least 500 NF-H-immunoreactive neurons were counted under every growth factor condition to calculate proportions of neurons expressing the three markers examined. In no cases were CGRP-, TRPV1- or IB 4 -binding neurons not likewise positive for NF-H. Colocalization is presented only as an overall percentage. Realtime PCR TG cultures were prepared on 48-well poly-D-lysine pre-coated plates (Becton Dickinson, Franklin Lakes, NJ) at an initial density of ~5000 neurons/well. For each 48 well plate 12 wells received no growth factor, 12 received 100 ng/ml NGF, 12 received 100 ng/ml GDNF and the remaining 12 received 100 ng/ml BDNF. A total of 3 plates were utilized and each culture plate was generated independently. Following 5 days of culture, RNA was extracted from TG cultures using a ToTALLY RNA (Ambion, Austin, TX) total RNA extraction kit and each of the 12 wells per condition were pooled together to yield sufficient RNA. RNA samples were subsequently treated with DNA-free DNAse (Ambion) for removal of trace amounts of DNA. RNA concentrations were determined by UV absorbance, and RNA samples were then diluted to equal concentrations in TE buffer. For realtime PCR assessment of TRPV1 mRNA levels, samples were loaded in triplicate in 96-well reaction plates with each sample containing 150 ng RNA, 2X RT-PCR Taqman Master Mix, 40X Multiscribe and RNase Inhibitor solution, forward primer (tcc agt caa gcc cca cat c), reverse primer (tcc gag tca ccc ttc cca) and Taqman probe (6FAM tca cta cca gga gtc gta ccc ggc ttt TAMRA) (all 300 nM, all reagents Applied Biosystems, Foster City, CA) in a final reaction volume of 50 μL. Controls were run concomitantly using the same reaction recipe with a rodent GAPDH control kit (Applied Biosystems), primers and probe at 50 nM. Reactions were run on an ABI Prism 7700 (Applied Biosystems) with an initial RT step of 48°C for 30 min followed by a 95°C 10 min denaturation step and then a repeating denaturation, extension cycle of 95°C for 15 s and 60°C for 1 min for 55 cycles in order to reach a full plateau for all samples. All data were normalized to GAPDH mRNA levels to account for any variation in RNA concentrations between samples. Western blotting TG cultures were prepared on 48-well poly-D-lysine pre-coated plates at an initial density of ~5000 neurons/well. For each 48 well plate 12 wells received no growth factor, 12 received 100 ng/ml NGF, 12 received 100 ng/ml GDNF and the remaining 12 received 100 ng/ml BDNF. A total of 3 plates were utilized and each culture was generated independently. After five days, total protein was extracted by first lysing cells with lysis buffer (1 mM Na pyrophosphate, 50 mM HEPES, 1% Triton X-100, 50 mM NaCl, 50 mM NaF, 5 mM EDTA and 1 mM Na orthovanadate) supplemented with 1% protease inhibitor cocktail and then homogenizing the combined lysate from the 12 wells per condition by pumping it through a 25 gauge needle 20 times. Extracted proteins were cleared of nuclei and cellular debris by spinning the homogenate at 1000 × g for 5 min at 4°C. Protein levels were then measured by the Bradford method. Proteins were run at a concentration of 20 μg/lane on a 12.5% SDS-PAGE gel and transferred to Immobilon – P membranes (Millipore). Membranes were blocked in 5% dry milk for 1 hr and then exposed to rabbit anti-TRPV1 antibody (Neuromics, Minneapolis MN), at a concentration of 1:1000, overnight at 4°C. Membranes were then incubated with donkey anti-rabbit horseradish peroxidase-linked secondary antibody (Amersham, 1:5000) for 1 hr followed by ECL Western blotting detection for 1 min (Amersham). Blots were next exposed to film and subsequently scanned on a flatbed scanner. To control for protein loading, membranes were then stripped and reblotted for β-actin. Images were assessed for changes in TRPV1 protein using NIH image and normalized to β-actin protein levels. CGRP release and CGRP content All experiments were performed in 48-well poly-D-lysine pre-coated plates. Data shown are representative of at least 3 independently conducted CGRP release experiments with consistent results and neurons were plated at the same density as indicated for survival, ICC, realtime PCR and Western blotting experiments (~5000 neurons/well). Culture medium was changed at 24 and 72 hr, and all CGRP assays were performed on day 5. TG cultures were washed free of culture medium by 2 successive washes with release buffer (Hank's balanced salt solution (Gibco) supplemented with 10.9 mM HEPES, 4.2 mM sodium bicarbonate, 10 mM dextrose and 0.1% bovine serum albumin (BSA), pH 7.4). Growth factors were not included in the release buffer. Following washing, TG cultures were exposed for 10 min to the indicated concentrations of CAP, AEA or ACEA or to 50 mM K + buffer (containing 2.5 mM CaCl 2 , 50 mM KCl, 1.2 mM MgCl 2 , 90 mM NaCl, 25 mM NaHCO 3 , 1 mM NaH 2 PO 4 , 10 mM dextrose, 15 mM Hepes, 16 uM thiophan and 0.1% BSA at pH = 7.4), after which the CGRP-containing supernatant was removed and transferred to glass culture tubes (Fisher). Content was assessed by hypotonic lysis with deionized H 2 O supplemented with 1% protease inhibitor cocktail (Sigma) for 30 min. CGRP release or content for each well was subsequently measured by radioimmunoassay. CGRP radioimmunoassay Following culture release assays, individual aliquots of the superfusate (0.5 ml) were incubated with a C-terminally directed anti-CGRP antiserum (kindly donated by Dr Michael Iadarola, NIDCR, NIH, Bethesda, MD, USA). After 24 h, 100 μL of [125 I ]-CGRP 28–37 (approximately 20000–25000 cpm) and 50 μL of goat anti-rabbit antibody conjugated to ferric beads were added. Following another 24 h, bound peptide was separated from free peptide via immunomagnetic separation (PerSeptive Biosystems, Framingham, MA, USA). All incubations were carried out at 4°C. The minimum detection limit for this assay is approximately 1–2 fmol per tube, with 50% displacement occurring at 20–40 fmol per tube. To account for the possibility of any nonspecific effects on the RIA, all drugs used in the release experiments were included in separate standard curves for the purposes of data analysis. We did not observe any alterations in the standard curve for any of the compounds utilized in these studies Data analysis and statistics All data are presented as mean ± SEM unless otherwise stated. When normalizations for CGRP release or content to neuron numbers for either CGRP or TRPV1-immunoreactive neurons were made the equation shown in figure 11 was used. All data were analyzed using GraphPad Prism for Mac OSX (GraphPad, San Diego, CA). To assess statistical differences, data were analyzed by one-way ANOVA followed by Tukey's post-test, for multiple comparisons, or Dunnett's post-test to compare all groups to the control group. For differences between growth factors at the same concentrations differences were assessed by two-way ANOVA with Bonferroni post-test for between group comparisons. All nonlinear regressions were fit to a sigmoidal curve with variable slope. Abbreviations ACEA: arachidonyl-2-chloroethylamide, AEA: anandamide, BDNF: brain-dervied neurotrphic factor, CAP: capsaicin, CGRP: calcitonin gene-related peptide, DRG: dorsal root ganglion, GDNF: glial cell-line derived neurotrphic factor, IB 4 : isolectin B 4 , ICC: immunocytochemistry, NGF: nerve growth factor, SP: substance P, TG: trigeminal ganglion, TRPV1, transient receptor potential receptor vanilloid type 1 Author's Contributions TJP performed and conceived (or participated in their conception) experiments in each section, analyzed all data and authored the manuscript. MDL performed immunocytochemistry and neuron counts. DCS conducted the intitial CAP-evoked CGRP release study. GOD assisted in conceiving the experiments and performing the pilot studies. NAJ conducted the Western blots. AP assisted in conducting the CGRP release experiments. AD assisted in real-time PCR experiments. AAT assisted in generating cultures and in conducting pilot studies for CAP-evoked CGRP release. KMH assisted in conceiving the experiments and gave critical readings of the manuscript. CMF supervised all studies and conceived the original experimental designs as well as critically editing the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC548274.xml
545208	Painful Horner Syndrome as a Harbinger of Silent Carotid Dissection	A painful Horner's syndrome should alert clinicians to the possibilty of a silent carotid dissection	PRESENTATION of CASE A 43-y-old white female presented to the hospital in July 2004 with pain in the left eye and left upper lid ptosis. She did not perceive any difference in perspiration between the two halves of her face. She was a nonsmoker and denied any history of head or neck trauma, or ocular, cardiac, vascular, or neurologic disease. Neuro-ophthalmological examination was normal except for 1 mm of left upper eyelid ptosis (drooping of the eyelid), miosis (constriction of the pupil), and mild enophthalmos (recession of the eyeball into the orbit) consistent with classic left-sided Horner syndrome ( Figure 1 ). There was no carotidynia (a neck pain syndrome associated with tenderness to palpation over the carotid bifurcation) or carotid bruit. A chest radiograph obtained to rule out an underlying left apical superior sulcus tumor was normal. Magnetic resonance imaging/magnetic resonance angiography of the brain with cross-sectional imaging of the neck was obtained, which revealed extracranial left internal carotid artery dissection ( Figures 2 and 3 ). The patient was treated with unfractionated heparin and coumadin and made an uneventful recovery. The patient was seen in the clinic a few months later and did not have any complications at follow-up. Figure 1 Photograph of Patient Showing Left-Sided Horner Syndrome Figure 2 Magnetic Resonance Imaging/Magnetic Resonance Angiography of the Neck Showing Left Internal Carotid Artery Dissection Figure 3 T2-Weighted Magnetic Resonance Imaging Showing Blood in the Arterial Wall and Narrowing of the Lumen of the Left Internal Carotid Artery This is also known as the “crescent sign,” a hallmark of internal carotid artery dissection. DISCUSSION Horner syndrome—characterized by the constellation of miosis, ptosis, anhidrosis (lack of sweating), enophthalmos, and anisocoria (unequal pupil size)—is present in up to 58% of internal carotid artery dissections [ 1 ]. Most patients experience neck, facial, and head pain ipsilateral to the lesion because of ischemia or stretching of the trigeminal pain fibers surrounding the carotid arteries [ 2 ]. Ophthalmic manifestations have been reported to occur in up to 62% of patients with internal carotid artery dissection [ 2 ]. Common findings in descending order of frequency are painful partial Horner syndrome (due to disruption of the third-order neuron oculosympathetic fibers) as seen in our patient, transient monocular vision loss, and permanent visual loss [ 2 ]. De Bray et al. studied the prognosis of 90 cases of isolated Horner syndrome due to internal carotid artery dissection [ 3 ]. They found that 91% of cases of Horner syndrome due to internal carotid artery dissection were painful. The risk of an early ischemic stroke within the first 2 wk was high (around 17%) without initial antithrombotic treatment [ 3 ]. Internal carotid artery dissection is a potentially life-threatening condition and carries a substantial risk of disabling stroke [ 4 ]. Carotid dissection is under-recognized as a cause of Horner syndrome and can be missed [ 5 ]. It is important to diagnose dissection because anticoagulation can prevent carotid thrombosis and embolism [ 5 ]. The investigation of choice is magnetic resonance imaging and angiography scan of the head and neck [ 5 ].The treatment advocated for dissection is anticoagulation for 3–6 mo [ 5 ]. Learning Points Painful Horner syndrome should alert clinicians to the possibility of a silent carotid dissection until proven otherwise [ 6 ]. Magnetic resonance imaging and angiography scan of the head and neck is the imaging modality of choice to look for dissection [ 5 ]. For patients with carotid dissection, anticoagulation with warfarin and coumadin is recommended for 3–6 mo to prevent carotid thrombosis and embolism [ 5 ].	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545208.xml
529305	Detecting spatiotemporal clusters of accidental poisoning mortality among Texas counties, U.S., 1980 – 2001	Background Accidental poisoning is one of the leading causes of injury in the United States, second only to motor vehicle accidents. According to the Centers for Disease Control and Prevention, the rates of accidental poisoning mortality have been increasing in the past fourteen years nationally. In Texas, mortality rates from accidental poisoning have mirrored national trends, increasing linearly from 1981 to 2001. The purpose of this study was to determine if there are spatiotemporal clusters of accidental poisoning mortality among Texas counties, and if so, whether there are variations in clustering and risk according to gender and race/ethnicity. The Spatial Scan Statistic in combination with GIS software was used to identify potential clusters between 1980 and 2001 among Texas counties, and Poisson regression was used to evaluate risk differences. Results Several significant (p < 0.05) accidental poisoning mortality clusters were identified in different regions of Texas. The geographic and temporal persistence of clusters was found to vary by racial group, gender, and race/gender combinations, and most of the clusters persisted into the present decade. Poisson regression revealed significant differences in risk according to race and gender. The Black population was found to be at greatest risk of accidental poisoning mortality relative to other race/ethnic groups (Relative Risk (RR) = 1.25, 95% Confidence Interval (CI) = 1.24 – 1.27), and the male population was found to be at elevated risk (RR = 2.47, 95% CI = 2.45 – 2.50) when the female population was used as a reference. Conclusion The findings of the present study provide evidence for the existence of accidental poisoning mortality clusters in Texas, demonstrate the persistence of these clusters into the present decade, and show the spatiotemporal variations in risk and clustering of accidental poisoning deaths by gender and race/ethnicity. By quantifying disparities in accidental poisoning mortality by place, time and person, this study demonstrates the utility of the spatial scan statistic combined with GIS and regression methods in identifying priority areas for public health planning and resource allocation.	Background The accidental poisoning mortality rate has been increasing in the United States. In the 21-year period spanning 1981 to 2001, mortality rates due to accidental poisoning more than doubled from 2.0 per 100,000 in 1981 to 4.9 per 100,000 in 2001 [ 1 ]. The burden of accidental poisoning mortality in this period was more than four million years of potential life lost (YPLL) before age 65 [ 1 ]. Accidental poisoning mortality trends in Texas have mirrored national trends with a linear increase in rates from 1.5 per 100,000 in 1981 to 5.2 per 100,000 in 2001. During this period, 10,406 total deaths attributable to accidental poisoning have contributed to more than 250,000 YPLL before age 65 [ 1 ]. Accidental poisoning refers to the physiologic damage caused by inhalation, ingestion, or any other mode of exposure to various licit and illicit drugs, or chemicals such as those found in pesticides, household cleaning products and gases/vapors [ 2 ]. According to the US CDC classification standards, this category excludes poisonings that are associated with suicidal or homicidal intent. Several studies have examined geographic and temporal trends in accidental poisoning morbidity and mortality. Patterns of accidental poisoning have been shown to vary by gender, ethnicity, socioeconomic status, and region thus exhibiting non-random spatial or temporal distributions [ 3 - 5 ]. The study conducted by Hanson and Wieczorek [ 6 ] used the spatial scan statistic to identify spatial clusters of alcohol-related mortality. Altmayer et al. [ 7 ] reported significant disparities in premature mortality due to poisoning among certain geographic areas. An early study conducted in Ontario, Canada found spatial and temporal variation in accidental poisoning mortality rates, which also reported higher poisoning mortality in males than in females [ 8 ]. Kaufmann, Staes & Matte in their study on lead poisoning mortality reported elevated rates in less populated areas and different time trends in mortality by race and gender [ 9 ]. Many of the studies regarding poisoning mortality have examined exclusively excess mortality due to long-term exposure to specific agents and have not explicitly studied the possible persistence of mortality due to accidental poisoning across space and time. They often examined either spatial or temporal attributes of the burden of the disease. A statistic for examining spatial and temporal data concurrently is available through the use of SaTScan software. The objectives of this study are to detect the existence of spatial and temporal clusters of accidental poisoning mortality in Texas and to determine if there are variations with respect to race and gender among these clusters. Results From 1980 to 2001, there were 10,774 deaths attributable to accidental poisoning, resulting in an average annual age-adjusted mortality rate of 2.8 deaths per 100 000. From 1981 to 2001, a linear increase was observed in unintentional poisoning mortality rates among Texas counties (annual percent increase = 1.06, 95% CI: 1.05 – 1.06, p < 0.0001, see Figure 1 ). Age-adjusted mortality rates were observed to be higher in men than in women, and to be highest among the Black population (see Table 1 ). Figure 1 Accidental poisoning mortality trends in Texas, 1981 – 2001. Table 1 Characteristics of study population, Texas, 1980 – 2001 Population Average Total Population (%) Cumulative Deaths (%) Annual age-adjusted rate (per 100,000) All 17,577,292 (100) 10,774 (100) 2.8 Male 8,677,605 (49.37) 7,593 (70.48) 4.0 Female 8,899,687 (50.63) 3,181 (29.60) 1.6 White 10,553,321 (60.04) 6,913 (64.16) 3.0 Male 5,187,210 (29.51) 4,661(43.26) 4.1 Female 5,366,111 (30.53) 2,252 (20.90) 1.9 Black 2,052,901 (11.68) 1,584 (14.70) 3.5 Male 991,066 (5.638) 1,063 (9.866) 4.9 Female 1,061,835 (6.041) 521 (4.836) 2.2 Hispanic 4,582,804 (26.07) 2,210 (20.51) 2.2 Male 2,305,930 (13.12) 1,820 (16.89) 3.6 Female 2,276,873 (12.95) 390 (3.620) 0.8 Other 388,266 (2.21) 67(0.62) 0.8 Male 193397 (1.10) 49 (0.45) 1.2 Female 194869 (1.11) 18 (0.17) 0.4 Poisson regression results After adjusting for age, regression analysis revealed significant associations between gender and race/ethnicity with accidental poisoning mortality. Males were found to be at higher risk of death from accidental poisoning than females (see Table 2 ). Among the race/ethnicity categories, the highest effect estimate (rate ratio) was for the Black population. The White population exhibited the second highest risk estimate, and the populations in the "other" category exhibited the lowest risk of accidental poisoning mortality (see Table 2 ). Table 2 Rate ratios (RR) for accidental poisoning mortality using Poisson regression adjusted for age, gender, and race/ethnicity. Factor Rate Ratio 95% Confidence Interval p-value Gender Male 2.47 2.45 – 2.50 <0.0001 Female 1.00 (reference) --- --- Race/Ethnicity White 1.00 (reference) --- --- Black 1.25 1.24 – 1.27 <0.0001 Hispanic 0.79 0.78 – 0.80 <0.0001 Other 0.24 0.22 – 0.25 <0.0001 Space-Time scan results For the total population adjusting for age, gender and race/ethnicity, spatiotemporal scan analysis for accidental poisoning deaths among Texas counties revealed four significant clusters of high mortality rates (see Table 3 ). The most likely cluster was a "hotspot" cluster (i.e. relative risk > 2.0) [ 10 ] consisting of 16 counties located in the Gulf Coast and Southeast regions of Texas (see Figure 2 ). Another hotspot cluster was detected in the eastern portion of Texas and consisted of five counties. The primary cluster (i.e. most likely cluster) persisted for a shorter time period (four years) than the other clusters (see Table 3 ). Figure 2 Statistically significant accidental poisoning mortality clusters among Texas counties, 1980 – 2001. See Table 3 for descriptive and effect information for each statistically significant cluster. Table 3 Statistically significant spatiotemporal clusters of accidental poisoning mortality in Texas by gender and race/ethnicity Cluster Time Period No. of Observed Cases Annual age-adjusted rate (per 100,000) Relative Risk p-value All populations Primary cluster 1998–2001 1095 5.7 2.056 0.001 Secondary cluster 1 1985–2001 637 6.8 2.448 0.001 Secondary cluster 2 1995–2001 1419 4.5 1.598 0.001 Secondary cluster 3 1995–2001 886 4.6 1.661 0.001 All Males Primary cluster 1985–2001 575 10.9 2.744 0.001 Secondary cluster 1 1995–2001 960 6.8 1.711 0.001 Secondary cluster 2 1998–2001 685 7.4 1.863 0.001 Secondary cluster 3 1994–2001 704 6.4 1.620 0.001 All Females Primary cluster 1997–2001 456 3.9 2.371 0.001 Secondary cluster 1 1996–2001 387 2.8 1.751 0.001 Secondary cluster 2 1997–2001 207 2.9 1.786 0.001 White Population Primary cluster 1998–2001 785 7.4 2.480 0.001 Secondary cluster 1 1996–2001 886 5.2 1.743 0.001 Secondary cluster 2 1995–2001 598 4.8 1.601 0.001 White Males Primary cluster 1998–2001 482 9.3 2.273 0.001 Secondary cluster 1 1995–2001 697 7.2 1.768 0.001 Secondary cluster 2 1994–2001 621 6.2 1.509 0.001 White Females Primary cluster 1998–2001 300 5.6 2.940 0.001 Black Population Primary cluster 1993–2001 345 6.3 1.800 0.001 Black Males Primary cluster 1993–2001 236 9.1 1.867 0.001 Black Females Primary cluster 1996–1998 2 2338.7 1048.700 0.033 Hispanic Population Primary cluster 1985–2001 467 6.0 2.730 0.001 Secondary cluster 1 1994–2001 342 3.8 1.730 0.001 Hispanic Males Primary cluster 1985–2001 410 11.2 3.130 0.001 Secondary cluster 1 1995–2001 239 6.0 1.663 0.001 Hispanic Females Primary cluster 1994–2001 84 1.8 2.264 0.001 The spatiotemporal scan statistic also revealed four significant clusters for the male population (see Table 3 ). The primary cluster for the male population was a hotspot cluster consisting of 10 counties located in East Texas. The second most likely cluster consisted of nine counties in North Central Texas (see Figure 2 ). For the male population, the primary cluster persists from 1985 into the present decade. The primary cluster for the female population closely mirrors that of the total population in location and size and persists from 1997 into the present decade. The second most likely cluster for the female population encompasses a large area of frontier (i.e. sparsely populated) counties in the eastern half of Texas (see Figure 2 ). There were no real hotspot clusters observed for the Black population (see Table 3 ). An extreme effect estimate or fluctuation was observed for the Black female population due to the small number of observed deaths in the numerator. The number of cases was two and the period of persistence extended over two years. The significant cluster for the Black population was observed in three highly populated counties in North Texas (see Figure 2 ). This cluster persists for eight years and into the present decade. A hotspot cluster was observed for the total Hispanic population consisting of ten counties in eastern Texas and it persisted for 16 years and entered into the present decade (see Figure 2 and Table 3 ). Five of the counties in the total Hispanic population cluster included a hotspot cluster for Hispanic males persisting for the same period of time. For Hispanic females, a hotspot cluster was observed in Central Texas consisting of 45 counties. This cluster persisted over a seven-year period beginning in 1994 and entering into the present decade (see Figure 2 ). The non-Hispanic White population exhibited a hotspot cluster in the Gulf Coast region of Texas, consisting of seventeen counties. The cluster persisted for three years and into the present decade (see Table 3 ). The same spatiotemporal cluster was observed for the non-Hispanic White male population, albeit with a slightly different effect estimate. The most likely cluster for the non-Hispanic White female population was also a hotspot cluster. This cluster is located in the same region as the total non-Hispanic White population and persists for the same time period but only includes fifteen counties (see Figure 2 ). No significant clusters were detected for populations in the "Other" category. There were very few accidental poisoning deaths for this population and the population count in many of the counties was zero for this subgroup. This resulted in instability in rates and other estimates for this population group (see Table 1 ). Spatial only and spatiotemporal scan adjusting for time non-parametrically results To verify whether the detected clusters using "spatial-temporal" methods are truly geographic clusters or space-time clusters not explained by the time trend, we further conducted two analyses using 1) the purely spatial scan statistic and 2) the space-time scan statistic non-parametrically adjusting for time. The results of these additional analyses indicated that most clusters detected using the default method were in mostly identical regions detected in the earlier analysis, only with slightly lower relative risks (i.e., observed deaths/expected deaths). For instance for the "all population group", the detected primary cluster in West Texas with a relative risk of 2.4 remained a cluster, but reduced to relative risks of 2.2 and 2.1 using additional methods that adjusted for time trend. In addition, most of the detected clusters still persisted to the present decade after the adjustments of time trend. Similar results were observed for most clusters detected in 'Male", "Female", "Black", "Non-Hispanic Whites" and "Hispanic" subpopulations using additional analyses options (see Tables 4 and 5 ). The summary maps of these additional analyses are presented in Figures 3 and 4 . Table 4 Statistically significant spatial clusters of accidental poisoning mortality in Texas by gender and race/ethnicity Cluster No. of Observed Cases Annual age-adjusted rate (per 100,000) Relative Risk p-value All populations Primary cluster 731 6 2.152 0.001 Secondary cluster 1 2408 3.6 1.285 0.001 Secondary cluster 2 520 3.7 1.326 0.001 Secondary cluster 3 148 4.1 1.468 0.011 Secondary Cluster 4 52 5 1.813 0.049 All Males Primary cluster 624 9.7 2.433 0.001 Secondary cluster 1 1685 5 1.269 0.001 Secondary cluster 2 404 5.7 1.434 0.001 All Females Primary cluster 962 2.1 1.278 0.001 White Population Primary cluster 1725 4 1.351 0.001 Secondary cluster 1 290 4.8 1.622 0.001 Secondary cluster 2 114 4.6 1.539 0.015 White Males Primary cluster 1030 5.7 1.399 0.001 Secondary cluster 1 217 7.3 1.782 0.001 Secondary cluster 2 269 5.8 1.414 0.001 White Females Primary cluster 704 2.6 1.337 0.001 Black Population Primary cluster 482 4.3 1.224 0.001 Black Males Primary cluster 108 7.2 1.476 0.039 Black Females Primary cluster 2 373.8 167.634 0.022 Hispanic Population Primary cluster 500 5.3 2.459 0.001 Secondary cluster 1 120 3.9 1.788 0.001 Hispanic Males Primary cluster 438 10 2.822 0.001 Secondary cluster 1 95 6 1.696 0.001 Hispanic Females Primary cluster 128 1.2 1.524 0.001 Table 5 Statistically significant spatiotemporal clusters of accidental poisoning mortality in Texas by gender and race/ethnicity, adjusting for time nonparametrically Cluster Time Period No. of Observed Cases Annual age-adjusted rate (per 100,000) Relative Risk p-value All populations Primary cluster 1984–2001 654 6.3 2.261 0.001 Secondary cluster 1 1983–2001 2257 3.6 1.306 0.001 Secondary cluster 2 1990–1999 331 4.4 1.590 0.001 Secondary cluster 3 1993–1999 1088 3.5 1.264 0.001 Secondary cluster 4 1992–1994 38 8.9 3.207 0.002 All Males Primary cluster 1984–2001 588 10.0 2.521 0.001 Secondary cluster 1 1982–1993 797 5.9 1.475 0.001 Secondary cluster 2 1990–2001 337 6.4 1.605 0.001 Secondary cluster 3 1993–2001 1084 5.0 1.259 0.001 Secondary cluster 4 1992–1994 27 13.0 3.284 0.027 All Females Primary cluster 1983–2001 916 2.1 1.320 0.001 White Population Primary cluster 1983–2001 1390 4.3 1.439 0.001 Secondary cluster 1 1980–1997 220 5.5 1.832 0.001 Secondary cluster 2 1992–1994 32 11.0 3.679 0.001 Secondary cluster 3 1991–1999 207 4.6 1.536 0.003 Secondary cluster 4 1993–2001 544 3.7 1.246 0.019 White Males Primary cluster 1983–2001 950 5.9 1.439 0.001 Secondary cluster 1 1980–1998 179 8.1 1.987 0.001 Secondary cluster 2 1993–2001 416 5.8 1.421 0.001 Secondary cluster 3 1990–1999 171 6.8 1.653 0.001 Secondary cluster 4 1992–1994 24 16.5 1.037 0.004 White Females Primary cluster 1983–2001 638 2.7 1.409 0.001 Black Population Primary cluster 1993–2001 122 6.2 1.758 0.002 Black Males Primary cluster 1993–2001 236 7.2 1.477 0.001 Black Females Primary cluster 1996–1998 2 2031.6 911.05 0.026 Hispanic Population Primary cluster 1984–2001 477 5.6 2.592 0.001 Secondary cluster 1 1993–2001 368 3.2 1.483 0.001 Hispanic Males Primary cluster 1983–2001 424 10.4 2.923 0.001 Secondary cluster 1 1989–2001 88 7.1 2.005 0.001 Hispanic Females Primary cluster 1994–2001 84 1.5 1.920 0.002 Figure 3 Spatial accidental poisoning mortality clusters among Texas counties, 1980 – 2001. Figure includes statistically significant and non statistically significant clusters to facilitate comparison with Figure 1. See Table 4 for descriptive and effect information for each statistically significant cluster. Figure 4 Accidental poisoning mortality clusters among Texas counties adjusting for time nonparametrically, 1980 – 2001. Figure includes statistically significant and nonstatistically significant clusters to facilitate comparison with Figure 1. See Table 5 for descriptive and effect information for each statistically significant cluster. Discussion The results from this study support the existence of spatiotemporal clusters of accidental poisoning mortality among Texas counties and show variations in these clusters by gender and race. This study also identifies several hotspot clusters. Moreover, the additional analyses (spatial only and adjusting for time non-parametrically) identify most clusters in almost identical regions but with slightly lower relative risks. This observation, along with the fact that both the results of the space-time scans with and without adjusting for time non-parametrically both demonstrate the temporal persistence of accidental poisoning mortality clusters into the present decade (at least 60% of the clusters persisted to the present decade using time-adjusting spatial scan), has provided supporting evidence that the clusters detected using the spatial-temporal method are geographic in nature, rather than an artefact of temporal trend. The burden of excess accidental poisoning mortality was found to be highest in the Black population, followed by the non-Hispanic White population, and the least among the Hispanic population and populations of other race/ethnicity categories. Consistent with the literature, the male population also exhibited an elevated risk of accidental poisoning mortality when compared to the female population. The results of this research presented both agreement and disagreement when compared other the results in which conventional techniques were used to identify geographic disparities by race/ethnicity, and gender. For instance, the results of this study were consistent with the literature in that the Black population was found to be at greatest risk of death from accidental poisoning [ 9 , 11 ]. In addition, the increased risk of accidental poisoning mortality observed in males when compared to females is consistent with what has been reported in other studies conducted in the United States [ 5 , 8 , 9 , 12 ]. On the other hand, the finding that the Hispanic population exhibited more than 20% less risk than the White population is unexpected. Other studies have found the Hispanic population to be either at increased risk or equal risk to the White population [ 5 , 11 , 13 ]. Although there are few studies in the literature that explicitly record cluster detection for accidental poisoning mortality, there are a variety of studies that have examined trends of accidental poisoning mortality in space and time [ 3 - 5 , 11 , 14 ]. Many of these studies have been conducted for specific toxic agents, e.g., alcohol, lead and pesticides but have not recorded the existence of spatial and temporal clustering of poisoning events and poisoning deaths. There is also evidence in the literature to support the existence of temporal clustering of accidental poisoning mortality. In a study on organophosphate poisoning, Sahin, Sahin and Arabaci reported that deaths from accidental poisoning due to organophosphates were most frequent during certain months [ 4 ]. In a review of childhood poisoning, McGuigan reported temporal variations in symptomatic poisoning events over many years [ 14 ]. Sudakin, Horowitz, and Griffin used the space-time scan statistic to identify spatial and temporal clustering of accidental poisoning events due to pesticide exposures [ 15 ]. Therefore, the temporal clustering observed in the present study may be reflective of different types of poisoning agent exposures for each cluster. The results of this study should be interpreted with several considerations. One, the process of cluster detection is necessarily ecological. The objective was to determine if there was an association between certain areas of Texas with excess accidental poisoning mortality. These results cannot be extrapolated to the level of the individual, i.e., one cannot interpret a relative risk above one as increased risk of accidental poisoning mortality for residents living in a given county. However, the risk estimate does provide valuable information about geographic disparity of accidental poisoning mortality. Another consideration concerns the utilization of county-level data. The scan statistic may not have been sensitive to small areas of excess mortality that may have been detected at a higher geographic resolution. However, in Texas the county is the smallest geographic area for which there is routinely reported or available deaths and population estimate information. Also, the county is the political level at which health policy actions are instituted, since many of the local public health departments in Texas are county health departments. A third consideration is that this study relied on county-level mortality data; and there may be variations in coding or reporting practices from county to county. Furthermore, in 1998 the International Classification of Diseases underwent its 10 th revision, and there may be differences in mortality estimates based on the codes in the 9 th and 10 th revisions. This introduces the possibility of misclassification bias in the results. However, there is no evidence to indicate that the misclassification would be differential [ 2 ]. Hence, any misclassification would bias the effect estimates towards the null, thus underestimating the actual effect. Thus, the results of this study still provide useful information concerning health disparities at the county, regional and state level. Conclusions This study has demonstrated the existence of spatiotemporal clusters of accidental poisoning mortality in Texas. Additionally, this study has also shown variations in risk and clustering by gender and race/ethnicity. The results of this study have numerous implications. By quantifying accidental poisoning mortality disparities by geographic area, race/ethnicity, and gender, the results provide an evidence-base for health planning. The clusters identified in this study, specifically those that persist into the present decade, represent areas where health services, e.g., poison control centers and emergency rooms, may be deficient. Knowing the specific areas of excess mortality from accidental poisoning would help health policymakers to focus the scope of prevention programs and health care delivery, thus providing for efficient allocation of public health resources. The findings from this study illustrate the use of the scan statistic and GIS in public health surveillance. Longitudinal data for accidental poisoning mortality was analyzed to identify hotspot clusters not only for the total population but also for subpopulations. Identifying geographic areas and specific populations, which are at an elevated risk of accidental poisoning mortality helps in the process of hypothesis generation for possible etiological mechanisms. For example, the most likely cluster for the total Texas population is in a heavily industrial area. The population in this area may have experienced excess mortality from other occupational exposure-related health outcomes (e.g. different types of cancers). One may further examine into the excess accidental poisoning mortality observed in this area in order to determine if it was due to occupational exposures. The clusters identified in this study warrant further attention. Additional studies should be conducted to determine the causes behind the observed clustering, especially in the Gulf Coast region and the Central North Texas region where clustering was observed for the total Texas population and the two populations at greatest risk of accidental poisoning mortality, the Black population and the White population. More research is needed to assess different types of toxic exposures and their contribution to accidental poisoning mortality. Furthermore, the availability, locality and accessibility of health services such as poison control centers and health care facilities (hospitals and clinics) should be studied, giving special attention to those in non-urban counties. Methods Data sources and data processing Three types of files were downloaded for the analysis. First, death files including the unintentional poisoning mortality data was obtained from Expert Health Data Programming Incorporated's Vitalweb (URL ). The data was extracted as counts by race, gender and age-group for ICD-9 codes 850 – 858 (unintentional poisoning by drugs) and 860–869 (accidental poisoning by solid, liquid, gas) for the years 1980 – 1998 and ICD-10 codes X40 – X49 (accidental poisoning by and exposure to noxious substances). The ICD-10 classification codes include exposure to drugs, gases and vapors, and other unspecified chemicals. Second, population (and estimates) files by year were obtained from the Texas State Data Center. Both deaths and population data included stratified information by age groups (16, representing the ages of 0–4 to 75 and above), gender (2, by male and female) and race (4, representing Blacks, Hispanics, Non-Hispanic Whites and Others). Third, county centroids information of latitude and longitude was downloaded from the 2000 U.S. Gazetteer portion of the US Census Bureau (Available online at ). We performed the data processing by using an automated Microsoft Visual Basic Application (VBA) program to organize the downloaded county information of death counts, population at risk and geographic files (i.e. county centroids), before exporting them to SaTScan for analysis and for GIS mapping. The VBA program first imported, aligned, and then exported data tables to a SatScan compatible format. The program then invoked the SaTScan batch executable file (SaTScan version 4.0, freeware available from: URL: ) to perform scan analysis, and extracted cluster information from the SaTScan output file. Following this, the program linked the cluster dbase files with the county name file, and created a new file for mapping purposes. Using ArcGIS software (ESRI, ), we spatially joined the dbase file to a Texas county layer and produced maps displaying the poison mortality clusters of all populations and those in the populations stratified by gender, race/ethnicity, and gender-race/ethnicity combinations. Poisson regression The association between gender, race/ethnicity, or the combination of both and accidental poisoning mortality was explored using Poisson regression. Poisson regression was conducted on accidental poisoning deaths using age, gender, year and race as predictor variables and the natural logarithm of population size as the offset variable. Regression analysis was conducted using the GENMOD procedure from the SAS system for Windows version 8.2 (Cary, North Carolina: SAS Institute, Inc.1999–2001). To control for overdispersion, the Poisson model was used with the scaled deviance option. Estimates of the annual percent increase, and rate ratios for race and gender were calculated from the results of the analysis. Spatial scan statistic To detect potential spatiotemporal variations of poisoning mortality among Texas counties, this study utilized the Spatial Scan Statistic developed by Kulldorff [ 16 , 17 ] to detect clusters of unintentional poisoning mortality among Texas counties. This statistic has been used previously to detect clusters of breast cancer [ 18 ], and those specifically among Texas counties [ 19 ], prostate cancer [ 20 ], alcohol-related mortality [ 6 ], and diabetes prevalence [ 21 ] among others. Many studies conducted on rare events such as unintentional poisoning deaths have aggregated the events to a higher geographic resolution, and have thus been unable to detect the existence of true clusters. The spatial scan statistic, however, is ideal for cluster detection of rare events since it includes the option of examining Poisson distributed data. The space-time parameter of the spatial scan statistic places a cylindrical window on the coordinates grid for the locations studied and moves the center of the cylinder base over the grid so that the sets of geographic units covered by the window are constantly changing. The statistic assumes no predetermined cluster size and utilizes scan windows of various sizes (of the population at risk) with a maximum window size specified by the user. In the space-time scan, this spatial scan window serves as the base of a cylinder, with time acting as the height of the cylinder. The cylinder is continuously expanded and extended up to the maximum specifications set by the user. Whenever the cylindrical window finds a new death, SatScan calculates a likelihood function to test for elevated risk within the cylinder as compared with outside the cylinder. The likelihood function for any given cylindrical window (under the Poisson assumption) is proportional to: ( d/n ) d ([ D - d ] / [ D - n ]) ( D - d ) I ( ) (1) where D is the total number of unintentional poisoning mortality deaths, d is the number of deaths within the space-time cylindrical window, and n is the expected number of cases after adjusting for any specified covariates. When SatScan is scanning for high rates, the indicator function I ( ) is equal to 1 when the window has more cases than expected, and 0 if the observed cases are equal to or less than expected. The present study utilized the retrospective space-time analysis for high rates using a Poisson model to calculate expected deaths due to unintentional poisoning in each county. The spatial scan window setting was set to a maximum cluster size of 25% of the study population, and the temporal scan window was set to a maximum cluster size of 90% of the study period. The covariates specified for this study were age group, gender, and race depending on the population being studied [ 16 ]. The "clusters" detected using the space-time scan statistic could be purely spatial, purely temporal or truly space-time clusters. To test whether these clusters are truly geographical in nature or are confounded by the temporal trend, i.e., whether there are any statistically significant geographical clusters or space-time clusters not explained by the time trend, we performed further analyses using (1) the purely spatial scan statistic and (2) the space-time scan statistic (non-parametrically) adjusting for time. Authors' contributions ETN was responsible for the design and implementation of the project and the final paper. ETN also contributed to data analysis. CEH contributed to database design, visual basic application programming and data analysis. CEH also assisted in the preparation and editing of the final paper. VIH contributed to the preparation and editing of the final paper, including reviewing the relevant literature. AMH was responsible for the layout and design of the illustrative maps and also assisted in preparing and editing the final paper. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC529305.xml
212692	Genome-Wide RNAi of C. elegans Using the Hypersensitive rrf-3 Strain Reveals Novel Gene Functions	RNA-mediated interference (RNAi) is a method to inhibit gene function by introduction of double-stranded RNA (dsRNA). Recently, an RNAi library was constructed that consists of bacterial clones expressing dsRNA, corresponding to nearly 90% of the 19,427 predicted genes of C. elegans . Feeding of this RNAi library to the standard wild-type laboratory strain Bristol N2 detected phenotypes for approximately 10% of the corresponding genes. To increase the number of genes for which a loss-of-function phenotype can be detected, we undertook a genome-wide RNAi screen using the rrf-3 mutant strain, which we found to be hypersensitive to RNAi. Feeding of the RNAi library to rrf-3 mutants resulted in additional loss-of-function phenotypes for 393 genes, increasing the number of genes with a phenotype by 23%. These additional phenotypes are distributed over different phenotypic classes. We also studied interexperimental variability in RNAi results and found persistent levels of false negatives. In addition, we used the RNAi phenotypes obtained with the genome-wide screens to systematically clone seven existing genetic mutants with visible phenotypes. The genome-wide RNAi screen using rrf-3 significantly increased the functional data on the C. elegans genome. The resulting dataset will be valuable in conjunction with other functional genomics approaches, as well as in other model organisms.	Introduction RNA interference (RNAi) is targeted gene silencing via double-stranded RNA (dsRNA); a gene is inactivated by specific breakdown of the mRNA ( Fire et al. 1998 ; Montgomery et al. 1998 ). It is an ideal method for rapid identification of in vivo gene function. Initial studies on RNAi used microinjection to deliver dsRNA ( Fire et al. 1998 ), but it was subsequently shown that dsRNA can be introduced very easily by feeding worms with bacteria that express dsRNA ( Timmons and Fire 1998 ). Using this technique on a global scale, an RNAi feeding library consisting of 16,757 bacterial clones that correspond to 87% of the predicted genes in Caenorhabditis elegans was constructed ( Fraser et al. 2000 ; Kamath et al. 2003 ). Upon feeding to worms, these clones will give transient loss-of-function phenotypes for many genes by inactivating the target genes via RNAi. By feeding the clones in this library to wild-type Bristol N2 worms, loss-of-function phenotypes were assigned to about 10% of genes. However, RNAi phenotypes were missed for about 30% of essential genes and 60% of genes required for postembryonic development, probably because RNAi is not completely effective ( Kamath et al. 2003 ). Other global RNAi screens have been recently performed in C. elegans using this RNAi library or other techniques ( Gönczy et al. 2000 ; Maeda et al. 2001 ; Dillin et al. 2002 ; Piano et al. 2002 ; Ashrafi et al. 2003 ; Lee et al. 2003 ; Pothof et al. 2003 ). These screens were done using wild-type worms. We have already shown that mutation of rrf-3 , a putative RNA-directed RNA polymerase (RdRP), resulted in increased sensitivity to RNAi ( Sijen et al. 2001 ; Simmer et al. 2002 ). There are four RdRP-like genes in C. elegans. Two of these, ego-1 and rrf-1 , are required for efficient RNAi, as apparent from the fact that these mutants are resistant to RNAi against germline or somatically expressed genes, respectively ( Smardon et al. 2000 ; Sijen et al. 2001 ). A third gene, rrf-2, appears to have no role in RNAi. The rrf-3 strain, mutated in the fourth RdRP homolog, shows an opposite response to dsRNA; this mutant has increased sensitivity to RNAi ( Sijen et al. 2001 ). A more detailed study of RNAi sensitivity of rrf-3 mutants using a set of 80 genes showed that rrf-3 is generally more sensitive to RNAi than wild-type worms ( Simmer et al. 2002 ). RNAi phenotypes in rrf-3 animals are often stronger, and they more closely approximate a null phenotype, when compared to wild-type. In addition, loss-of-function RNAi phenotypes were detected for a number of genes using rrf-3 that were missed in a wild-type background. For example, known phenotypes were detected for many more neuronally expressed genes in the rrf-3 background. These features suggest that the rrf-3 strain could be used to improve and extend functional information associated with C. elegans genes. We have conducted a genome-wide RNAi screen using the rrf-3 strain. In total, we found reproducible RNAi phenotypes for 423 clones that previously did not induce a phenotype (corresponding to 393 additional genes). To explore the variability of global RNAi screens, we performed the rrf-3 screen twice for Chromosome I and carried out a Chromosome I screen with wild-type. These were cross-compared and also compared to the results of the wild-type screen of Fraser et al. (2000 ). From this, we find that rrf-3 consistently allowed detection of more phenotypes than wild-type. In addition, we found that there is a significant screen-to-screen variability (10%–30%). Results Comparative Analysis of RNAi for Chromosome I with Wild-Type and rrf-3 We first conducted a pilot screen of Chromosome I using rrf-3 and found RNAi phenotypes for 456 bacterial clones. We compared these data to those obtained by Fraser et al. (2000 ) for a screen in the wild-type Bristol N2 strain. For 153 of these 456 clones, no phenotypes were reported by Fraser et al. (2000 ) and phenotypes were observed for 303 clones in both screens. The N2 screen done by Fraser et al. (2000 ) resulted in RNAi phenotypes for 40 clones for which no phenotypes were found using rrf-3 ( Figure 1 A). These results indicate that rrf-3 can be used in a global screen to identify loss-of-function phenotypes for additional genes. However, some phenotypes were missed in the rrf-3 screen. To explore the reproducibility and variability of RNAi screens, we next screened the clones of Chromosome I using N2 and rrf-3 side by side. We detected phenotypes for 447 clones: 140 were found only in rrf-3 , 11 only in N2, and 296 in both strains ( Figure 1 B). These data confirm that rrf-3 is more sensitive to RNAi and, in addition, these data indicate that global RNAi screens with rrf-3 will result in more clones with a detectable phenotype. Figure 1 Comparison of Different RNAi Experiments of Chromosome I Using Wild-Type Bristol N2 and rrf-3 Differences between different laboratories or investigators and between experiments done within the same laboratory and by the same investigators are observed. Ovals represent the amount of bacterial clones that gave an RNAi phenotype in an experiment. Areas that overlap represent clones for which in both experiments an RNAi phenotype was detected. Differences and overlap between an RNAi experiment done with the rrf-3 mutant strain and the data obtained by Fraser et al. (2000 ) done with the standard laboratory strain, Bristol N2 (A); N2 and rrf-3 tested at the same time within our laboratory (B); experiments done with N2 in two different laboratories: this study (‘NL') and Fraser et al. (2000 ) (C); two experiments done with the same strain, rrf-3 , within our laboratory (D). Variability of the RNAi Effect When we compared the RNAi results that we obtained using N2 with the Fraser et al. (2000 ) data, we were surprised to find significant differences: we only detected phenotypes for 75% of the clones that gave a phenotype in Fraser et al. (2000 ), and these researchers reported phenotypes for 84% of clones for which we found a phenotype ( Figure 1 C). The differences do not appear to be due to false positives. For example, Fraser et al. (2000 ) detected the predicted phenotype for goa-1 and unc-73 , whereas we did not detect a mutant phenotype. Similarly, we detected the known mutant phenotype for egl-30 and cdc-25.1 , which were not detected by Fraser et al. (2000 ). In addition, we found that the false-positive rate is negligible (see below). It is possible that different laboratories or investigators have slightly different results. However, when we compare the results that we obtained with two independent screens of Chromosome I using rrf-3 in our laboratory, we also see differences. For 394 clones we detected a phenotype in both experiments, 54 are specific for the first experiment, and 34 for the second ( Figure 1 D). Among the clones that only gave an RNAi phenotype in one of the experiments are again clones that induced the predicted phenotype based on the phenotypes of genetic mutants ( unc-40 , gpc-2 , and sur-2 ). These data show that large-scale RNAi screens done within the same laboratory and by the same investigators also give variable results. A few examples of variable RNAi results are shown in Table 1 . Table 1 Variable RNAi Effects Selection of clones that induced variable RNAi results in this study (‘NL') and or in the study by Fraser et al. (2000 ). In this subset of bacterial clones, each corresponds to a gene for which a mutant phenotype is known. The expected phenotypes are detected with RNAi, but not in each experiment, indicating false-negative results. The bacterial clones are indicated by ‘GenePairs Name' (name of genepair used to PCR-amplify a genomic fragment) and ‘Predicted Gene' (predicted gene targeted by the named genepair). ‘Locus' gives the genetic locus; ‘Known Mutant Phenotype' gives the mutant phenotype for the indicated gene described in the literature. The RNAi phenotypes are defined in the Materials and Methods section In conclusion, we find that RNAi results from different laboratories and from experiments done in the same laboratory vary from 10% to 30%. This appears to be due to a high frequency of false negatives in each RNAi screen, even when the same method is used in the same laboratory. The Genome-Wide RNAi Screen Based on the positive results of the Chromosome I screen using the rrf-3 strain, we next screened the complete RNAi library with rrf-3 mutant animals. We obtained results for 16,401 clones and detected phenotypes for 2,079 (12.7%). Of these, we identified phenotypes for 625 clones for which no phenotype was reported in the Fraser et al. (2000 ) or Kamath et al. (2003 ) screens using N2, with the remaining 1,454 generating phenotypes in both screens ( Table S1 ). In addition, there are 287 clones for which only Fraser et al. (2000 ) or Kamath et al. (2003 ) found phenotypes (23 of these were not done in our screen). The clones for which we only detected an RNAi phenotype once and that were specific for the rrf-3 screen were retested. Subsequently, the phenotypes of the clones corresponding to Chromosomes II to X that were not confirmed by this repetition were tested once more. In this way, the clones specific for the rrf-3 screen had two chances to be confirmed. Of the 625 clones for which no phenotype was found in the Fraser et al. (2000 ) and Kamath et al. (2003 ) N2 screens, the phenotypes of 423 clones were confirmed and 202 remained unconfirmed ( Table 2 ; see Table S1 ). Combining the N2 screens and these 423 clones, the percentage of clones with a phenotype increases from 10.3% to 12.8%. Table 2 Genome-Wide RNAi Summary of the bacterial clones that induced detectable RNAi phenotypes (‘Positive Clones'). For 423 clones, RNAi phenotypes were reproducibly detected in our laboratory using rrf-3 , but no RNAi phenotypes were reported in the N2 screens; 1,454 clones induced phenotypes in both laboratories; 264 were specifically detected by Fraser et al. (2000 ) or Kamath et al. (2003 ). For 202 clones, RNAi phenotypes were detected with rrf-3 and no RNAi phenotypes were reported in the N2 screens, but this result could not be repeated. In addition, there are 23 clones for which we did not obtain results that gave a phenotype with N2. In the column with the overlapping clones, the rrf-3 data are mainly from one experiment, whereas the N2 data reported by Fraser et al. (2000 ) and Kamath et al. (2003 ) are from repeated experiments. The phenotypes that were scored are described in the Materials and Methods section Some of the RNAi phenotypes only found with rrf-3 that remained unconfirmed could be confirmed by RNAi phenotypes detected with other clones of the RNAi library corresponding to the same gene or by other laboratories using different RNAi methods. For example, for the clones corresponding to the predicted genes F56D1.1 (a member of the zinc finger C2H2-type protein family) and F27C8.6 (a member of the esterase-like protein family), we detected sterile progeny (Stp) and embryonic lethality (Emb), respectively; these were also found by Piano et al. (2002 ). In addition, some unconfirmed RNAi phenotypes are confirmed by comparing to phenotypes of genetic mutants such as gpc-2 , hlh-8 , and unc-84 . This suggests that many of the unconfirmed phenotypes reflect true gene functions. Analysis of the rrf-3 Results To validate the results obtained using rrf-3 , we first assayed the rate of false positives in the total dataset (all RNAi results obtained with rrf-3 for the 16,401 clones tested). In the assay used by Kamath et al. (2003 ), a set of genes for which it is known that genetic mutants display no lethality was selected. A false positive in the RNAi data is then defined as detecting a lethal RNAi phenotype for any of these genes. In the N2 screen, the false-positive rate was 0.4%. We find that the false-positive rate in the rrf-3 data is similarly low (0 of 152 genes). To further determine the effectiveness of the screen, we compared the RNAi phenotypes with loss-of-function phenotypes of genetic mutants. For all chromosomes except for Chromosome I, the rrf-3 data were confirmed by refeeding only if there was no phenotype detected in the N2 screens by Fraser et al. (2000 ) or Kamath et al. (2003 ). Therefore, to compare the difference in detection of known phenotypes between the rrf-3 and the N2 screens, we used the Chromosome I datasets, where phenotypes were confirmed independently for the two strains. Of 75 genetic loci on Chromosome I, Fraser et al. (2000 ) detected 48% of published phenotypes, compared to 59% for rrf-3 ( Table S2 ). Using the genome-wide rrf-3 dataset (excluding the 202 unconfirmed phenotypes), we detected the published phenotype for 54% of 397 selected loci, compared to 52% for N2 ( Table 3 ; see Table S2 ). Table 3 Effectiveness of the rrf-3 Screen RNAi phenotypes obtained with rrf-3 (confirmed using N2 data or rrf-3 refeeding), and the N2 screens by Fraser et al. (2000 ) or Kamath et al. (2003 ) were compared with those of genes that have known loss-of-function phenotypes. ‘Total Genetic Loci Scored' denotes the number of genes that were analysed by RNAi. All loci have a loss-of-function phenotype that was detectable in our screen. ‘RNAi Phenotype Detected' gives the number of genes for which a phenotype was identified. ‘Published Phenotype Detected' gives the number of genes for which the RNAi phenotype matched the phenotype described in the literature We next asked whether using the rrf-3 strain improved general phenotype detection or whether certain types of phenotypes were particularly increased compared to the N2 screens by Fraser et al. (2000 ) and Kamath et al. (2003 ). To do this, we analysed the detection rate of different types of Chromosome I loci. First, we looked at a set of 23 loci with nonlethal postembryonic mutant phenotypes. Using rrf-3 , we reproducibly detected the published phenotype for 11 of these compared to only two for N2. Of 50 loci required for viability (essential genes), we detected 31 using rrf-3 , compared to 33 for N2. Thus, detection of essential genes was similar in the two strains, but detection of postembryonic phenotypes was improved with rrf-3 . Finally, for the whole genome using rrf-3 , we reproducibly detected the published phenotypes for 34 genetic mutants for which no RNAi phenotype was reported in the N2 screens (nine essential genes, 21 with postembryonic mutant phenotypes, and four with a slow-growth mutant phenotype). By comparison, published phenotypes were detected for 23 loci only with N2 (16 essential genes and seven with postembryonic mutant phenotypes) (see Table S2 ). We conclude that rrf-3 particularly improves detection of genes with postembryonic mutant phenotypes, a class that is poorly detected using wild-type N2. A striking feature of the rrf-3 dataset is the high number of clones where a slow or arrested growth (Gro/Lva) defect was induced, without associated embryonic lethality or sterility. Overall, 619 clones induced a Gro/Lva defect using rrf-3, compared to 276 for N2, whereas the number of essential genes detected was similar (1,040 versus 1,170, respectively). In addition, in the confirmed set of 423 clones with rrf-3 -specific phenotypes, Gro/Lva defects are the largest category (42%), whereas this is only 18% for N2, with the largest category being essential genes (49%). These data suggest that rrf-3 might particularly enhance detection of genes that mutate to a slow-growth phenotype; we cannot easily test this hypothesis, as there are currently few known loci with this mutant phenotype. In some cases, a Gro/Lva phenotype was seen in rrf-3 , whereas a different phenotype was seen in N2 (e.g., either lethality or a weak postembryonic phenotype). This suggests that some of the Gro/Lva phenotypes detected are due to incomplete RNAi of an essential gene (where lethality was seen in N2) or by a stronger RNAi effect (where no growth defect was seen in N2). In addition, it is possible that some of the Gro/Lva phenotypes detected are synthetic effects of using the rrf-3 mutant strain. To summarise, using the rrf-3 RNAi supersensitive strain in large-scale screens increases the percentage of clones for which it is possible to detect a phenotype. Detection of postembryonic phenotypes is particularly increased, whereas detection of essential genes is similar in rrf-3 and N2. In addition, using rrf-3 , there is a high rate of induction of Gro/Lva defects. Positional Cloning of Genetic Mutants with Visible Phenotypes Despite the advantages of RNAi, genetic mutants remain indispensable for many experiments. In the past decades, forward genetic screens identified a large number of genetic mutants, many of which are not yet linked to the physical map. We used the RNAi phenotypes obtained with the genome-wide screens to test whether we could systematically clone genes that are mutated in existing genetic mutants. First, the genetic map positions of all uncloned genetic mutants with visible phenotypes were checked using WormBase ( http://www.wormbase.org , the Internet site for the genetics, genomics, and biology of C. elegans ). Second, we searched for clones near the defined map positions that, when fed to N2, rrf-3, or both, gave phenotypes corresponding to the phenotypes of the genetic mutants. For most genetic mutants, more than ten clones with a similar phenotype were found in the interval to which the genetic mutant was mapped. However, for 21 genetic mutants, only one or a few candidate clones were found. The genes corresponding to these clones were subsequently sequenced in the genetic mutant to determine whether a mutation was present. In total, we sequenced 42 predicted genes for the 21 genetic mutants ( Table S3 ). For seven of these— bli-3 , bli-5 , dpy-4 , dpy-6 , dpy-9 , rol-3 , and unc-108 —we found a mutation in one of the sequenced genes ( Table 4 ). The mutated gene was confirmed by sequencing the same gene in a second or third allele (or both) of these genetic mutants ( Table 4 ). Table 4 Properties of the Genetic Mutants Cloned Using the RNAi Phenotypic Data Genetic mutants were linked to the physical map using RNAi phenotypes. The ‘Genetic Map Position’ is based on WormBase annotation. ‘Mutated Gene’ denotes the predicted gene, which is mutated in the genetic mutant. ‘RNAi Phenotype’ gives the loss-of-function phenotype either using rrf-3 or N2 (the latter is based on findings of Kamath et al. [2003]). The phenotypes that were scored are described in the Materials and Methods section a dpy-6(e2762) has a deletion that removes the first six amino acid residues (aa) of the eighth exon and part of the seventh intron b Multiple mutations in dpy-6(f11) (5′-tcgAaaa[G/T]tt[C/A]aaccccacgccaact[G/T]cc); the AAA→AAAA mutation at position 2792 bp of the F16F9.2 coding sequence causes a frameshift that results in a premature stop in the fifth exon The identification of mutations in unc-108 encoding the homolog of the small GTPase Rab2 is of particular interest. The RNAi phenotype of this gene gives a clue about the genetic property of the mutations in the mutants of unc-108 . With rrf-3 , we find that inactivation of Rab2 (F53F10.4) by RNAi causes uncoordinated movement ( Table 4 ). Mutations in unc-108 were isolated in a screen for dominant effects on behaviour; heterozygous unc-108 mutants display dominant movement defects and are indistinguishable from homozygous mutants ( Park and Horvitz 1986 ). RNAi phenocopies a loss-of-function phenotype, suggesting that the dominant movement defects of unc-108 mutants may be due to haplo-insufficiency. In eukaryotes, Rab2 is involved in regulating vesicular trafficking between the endoplasmic reticulum and Golgi. Based on the movement defects of unc-108 mutants, UNC-108 might be involved in vesicle transport in neurons that regulate locomotion. Thus, the RNAi data are a powerful tool to facilitate rapid cloning of the genes identified by genetic mutants and will provide important starting points for further studies of their function. Discussion With this genome-wide RNAi screen using the hypersensitive strain rrf-3 , we have significantly increased the functional information on the C. elegans genome, and we confirmed many RNAi phenotypes observed previously. We have assigned RNAi phenotypes for 406 genes (corresponding to the 423 extra clones) using rrf-3 . For 13 genes, Kamath et al. (2003 ) or Fraser et al. (2000 ) had already found a phenotype using a different clone from the RNAi library that targeted the same gene, and for at least 44 genes a genetic mutant exists (see Table S2 ). Other investigators have also found RNAi phenotypes for some of the genes using different methods. However, for most genes our result is to our knowledge the first hint about their biological function. Although we have identified new RNAi phenotypes for a substantial number of genes, others will have been missed in our screen for the following reasons. First, besides its increased sensitivity to RNAi, the rrf-3 strain has an increased incidence of males (Him) and displays slightly increased embryonic lethality and a reduced brood size ( Simmer et al. 2002 ). In our rrf-3 experiments, we therefore made some minor adaptations to the original RNAi protocol described by Fraser et al. (2000 ). We did not score for the Him phenotype and had more stringent criteria for embryonic lethality and sterility. This may have reduced the number of extra clones identified with a phenotype. Moreover, the changes in the protocol can also account for some differences in the detection of RNAi phenotypes between rrf-3 and N2. Second, when an RNAi phenotype is detected with N2 and not with rrf-3 , the lack of a detectable phenotype may be the result of variability in the efficiency of RNAi. This is consistent with the fact that we observe differences between experiments done with the same strain. When an RNAi phenotype is detected with rrf-3 and not with N2, this can be due to the increased sensitivity to RNAi of rrf-3 . However, besides the higher sensitivity, we may also be observing synthetic effects with rrf-3 (e.g., embryonic lethality, sterility, or developmental delay). In particular, a large number of clones induced a developmental delay phenotype using rrf-3 . Synthetic effects cannot be excluded without investigating genetic mutants. Again, variability in the efficiency of RNAi will also contribute to these differences, and a small portion may be false positives. In general, the few false positives that occur in the screen are most likely due to experimental errors, whereas the false negatives are due to reduced efficiency of the RNAi. Finally, differences between rrf-3 and N2 do not only involve the absence and presence of an RNAi phenotype, but also differences in the phenotypes for clones that did induce phenotypes in both screens (e.g., embryonic lethal in one screen and a postembryonic phenotype in the other). For example, we detected for unc-112 a 100% embryonic lethal (Emb) phenotype with rrf-3 , whereas Kamath et al. (2003 ) detected an adult lethal (Adl), uncoordinated (Unc), and paralyzed (Prz) phenotype with N2. Conversely, Kamath et al. (2003 ) detected for gon-1 a 100% Emb phenotype and other phenotypes with N2, while we did not detect an Emb phenotype with rrf-3 . What could be the source of the interexperimental variation of RNAi? Different phenotypes for the same gene can possibly occur owing to slight differences in the developmental stage at which the animals are exposed to dsRNA and owing to changes in temperature during the experiment. However, this probably does not account for the differences we see, as we always used animals of the same larval stage (L3/L4) and used incubators for constant temperature. It was shown previously that the level of induction of dsRNA production by isopropylthio-β-D-galactoside (IPTG) can modify the penetrance of the RNAi phenotype ( Kamath et al. 2000 ). Therefore, differences in the induction of the dsRNA either by changes in the concentration of IPTG, temperature, timing, or the bacteria may be an important source of the variation in the outcome of RNAi. RNAi is starting to be used extensively in other systems experimentally, as well as therapeutically and agriculturally. The relative variability of the RNAi effect is an important fact to take in account also for the use of RNAi in other systems. The RNAi data can be a useful starting point for many new experiments, such as positional cloning of genetic mutants. By sequencing candidate genes based on the RNAi phenotypes, we identified the causal mutation in seven genetic mutants. Identification of these mutated genes gives insight into the biological process in which they are involved. In addition, cloning of these genes increases the resolution of the genetic map of C. elegans , since these mutants have been extensively used as visible markers in linkage studies. The complete set of RNAi phenotypes detected for the 2,079 clones using rrf-3 will be submitted to WormBase, annotated as confirmed or unconfirmed. There the data can be evaluated in the context of information on gene structure, expression profiles, and other RNAi results. Materials and Methods Nematode strains. We used the following C. elegans strains: Bristol N2, NL4256 rrf-3(pk1426) , CB767 bli-3(e767) , MT1141 bli-3(n259) , CB518 bli-5(e518) , BC649 bli-5(s277) , CB1158 dpy-4(e1158) , CB1166 dpy-4(e1166) , CB14 dpy-6(e14) , CB4452, dpy-6(e2762) , F11 dpy-6(f11) , CB12 dpy-9(e12) , CB1164 dpy-9(e1164) , BC119 dpy-24(s71) , CB3497 dpy-25(e817) , MT1222 egl-6(n592) , MT1179 egl-14(n549) , MT1067 egl-31(n472) , MT151 egl-33(n151) , MT171 egl-34(n171) , egl-34(e1452) , MQ210 mau-4(qm45) , CB754 rol-3(e754) , BC3134 srl-2(s2507dpy-18(e364) ; unc-46(e177)rol-3(s1040) , CB713 unc-67(e713) , CB950 unc-75 (e950) , HE177 unc-94(su177) , HE33 unc-95(su33) , HE151 unc-96(su151) , unc-96(r291) , HE115 unc-100(su115) , MT1093 unc-108(n501) , and MT1656 unc-108(n777) . RNAi by feeding. RNAi was performed as described elsewhere ( Fraser et al. 2000 ; Kamath et al. 2000 ) with minor adaptations when the rrf-3 strain was used: after transferring L3- to L4-staged hermaphrodites onto the first plate, we left them for 48 h at 15°C instead of 72 h and then plated single adults onto other plates seeded with the same bacteria. Furthermore, we did not remove the mothers from the second plates. The phenotypes assayed are these: Emb (embryonic lethal), Ste (sterile), Stp (sterile progeny), Brd (low broodsize), Gro (slow postembryonic growth), Lva (larval arrest), Lvl (larval lethality), Adl (adult lethal), Bli (blistering of cuticle), Bmd (body morphological defects), Clr (clear), Dpy (dumpy), Egl (egg-laying defective), Lon (long), Mlt (molt defects), Muv (multivulva), Prz (paralyzed), Pvl (protruding vulva), Rol (roller), Rup (ruptured), Sck (sick), Unc (uncoordinated) Thin and Pale. Emb was defined as greater than 10% dead embryos for N2 and greater than 30% dead embryos for rrf-3 . Ste required a brood size of fewer than ten among fed N2 worms and fewer than five among rrf-3 . Each postembryonic phenotype was required to be present among at least 10% of the analysed worms. Sequencing of genetic mutants. The coding sequence and the 5′- and 3′-untranslated region (about 500 bp upstream and downstream of the coding sequence) of the predicted genes, as annotated in WormBase, was analysed for mutations by sequencing amplified genomic DNA of the genetic mutants (see Table S3 ). Nested primers were designed using a modification of the Primer3 program available on our website ( http://primers.niob.knaw.nl/ ). Sequence reactions were done using the ABI PRISM Big Dye terminator sequencing kit (Applied Biosystems, Foster City, California, United States) and were analysed on the ABI 3700 DNA analyser. Sequences were compared to the genomic sequence of C. elegans using the BLAST program ( http://www.sanger.ac.uk/Projects/C_elegans/blast_server.shtml ) or analysed using the PolyPhred program (available from http://droog.mbt.washington.edu/PolyPhred.html ). Supporting Information Table S1 RNAi Phenotypes for Bacterial Clones Using rrf-3 (482 KB PDF). Click here for additional data file. Table S2 Detailed Comparison of RNAi Phenotypes with Those of Known Loci (188 KB PDF). Click here for additional data file. Table S3 Summary of Genes Sequenced in Several Genetic Mutants (25 KB DOC). Click here for additional data file. Accession Numbers RNAi data from this study will be submitted to WormBase ( http://www.wormbase.org ).	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC212692.xml
544843	Human lung cancer cells express functionally active Toll-like receptor 9	Background CpG-oligonucleotides (CpG-ODN), which induce signaling through Toll-like receptor 9 (TLR9), are currently under investigation as adjuvants in therapy against infections and cancer. CpG-ODN function as Th-1 adjuvants and are able to activate dendritic cells. In humans TLR9 has been described to be strongly expressed in B-lymphocytes, monocytes, plasmacytoid dendritic cells and at low levels in human respiratory cells. We determined whether a direct interaction of bacterial DNA with the tumor cells themselves is possible and investigated the expression and function of TLR9 in human malignant solid tumors and cell lines. TLR9 expression by malignant tumor cells, would affect treatment approaches using CpG-ODN on the one hand, and, on the other hand, provide additional novel information about the role of tumor cells in tumor-immunology. Methods The expression of TLR9 in HOPE-fixed non-small lung cancer, non-malignant tissue and tumor cell lines was assessed using immunohistochemistry, confocal microscopy, in situ hybridization, RT-PCR and DNA-sequencing. Apoptosis and chemokine expression was detected by FACS analysis and the Bio-Plex system. Results We found high TLR9 signal intensities in the cytoplasm of tumor cells in the majority of lung cancer specimens as well as in all tested tumor cell lines. In contrast to this non-malignant lung tissues showed only sporadically weak expression. Stimulation of HeLa and A549 cells with CpG-ODN induced secretion of monocyte chemoattractant protein-1 and reduction of spontaneous and tumor necrosis factor-alpha induced apoptosis. Conclusions Here we show that TLR9 is expressed in a selection of human lung cancer tissues and various tumor cell lines. The expression of functionally active TLR9 in human malignant tumors might affect treatment approaches using CpG-ODN and shows that malignant cells can be regarded as active players in tumor-immunology.	Background The Toll gene, the expression of one of it's relatives we are reporting here concerning human malignant tumors, originally was characterized for its role in specifying dorsoventral polarity of the Drosophila embryo[ 1 ]. Since homologues of Toll are also present in plants, mammalian toll-like genes are products of an ancient evolutionary process beginning before the separation of animals and plants [ 2 ]. Within the genome of Drosophila thus far nine toll-like genes were identified, ten different human toll-like genes are currently described. In contrast to Drosophila , the mechanisms taking place in mammalian embryogenesis concerning TLR are widely unknown. The discovery of immune function for Toll in Drosophila led to a new understanding of innate immunity mechanisms. Human TLR recognize pathogen-derived products, also termed pathogen-associated molecular patterns (PAMP) [ 3 ]. These are bacterial lipoproteins (sBLP) [ 4 ], viral double stranded RNA/poly (I:C) [ 5 ], lipopolysaccharides (LPS) [ 6 ], flagellin [ 7 ] and bacterial DNA [ 8 ], which engage TLR2, TLR3, TLR4, TLR5 and TLR9, respectively. All functionally characterized TLR signal via the cytoplasmic Toll/interleukin-1 receptor domain (TIR) leading to activation of transcription factors like activator protein-1 (AP-1) and nuclear factor-κB (NF-κB) [ 9 ]. TLR9, in contrast to the other TLR, is not located at the cell surface, but intracellularily and, therefore, inhibition of endocytosis or endosome formation completely ablates the effects of CpG-ODN [ 10 ]. Different studies show an immunostimulatory capacity of bacterial components which can mediate anti-tumor activity. The first reported use of such a therapy for a nonbacterial disease took place 1890, evaluating the anti-tumor activity of living streptococci directly injected into tumor masses [ 11 ]. Shimada demonstrated that bacterial DNA itself can stimulate the immune system [ 12 ]. Over the past years there has been an enormous increase in the understanding of the molecular and cellular effects of CpG-ODN [ 13 ], which have been found to function as Th-1 adjuvants [ 14 ], and are able to activate dendritic cells [ 15 ]. This led to the idea to utilize CpG-ODN for induction of anti-tumor immune response as an adjuvant therapeutic strategy [ 16 - 18 ]. In order to characterize possible interactions between malignant cells and CpG-ODN, we investigated whether TLR9 is present in malignant tumors. A variety of malignant solid tumors and cell lines were tested for TLR9 expression; in addition, we examined direct effects of CpG-ODN upon apoptosis and chemokine production of tumor cells. Methods Tissues Samples of human tumors and tumor-free tissues were obtained from lobectomies because of lung cancer. Tumor-free tissues were taken at least 5 cm away from the tumor-border. The specimens were fixed and paraffin-embedded using the HOPE-technique [ 19 ]. Sections were cut, mounted, and deparaffinized as described elsewhere [ 20 ]. For increased comparability of the staining intensities in malignant and non malignant cells we additionally performed IHC on tumor-bearing and tumor free lung tissues which have been assembled on one slide by use of a mechanical tissue arrayer device (MTA1, Alphametrix, Germany). Cell culture A549 cells and HeLa cells were grown in 25 cm 2 polystyrene flasks with Dulbecco's modified Eagle's medium DMEM (Sigma) with 10 % heat-inactivated fetal calf serum (PAA Laboratories), 100 μg/ml penicillin G, 100 μg/ml streptomycin and 2 mM L-glutamine (Sigma), maintained under 5 % CO 2 by routine passage every 3 days. Cells were seeded in 35-mm dishes (Nunc). For IHC cells were cytocentrifuged and treated by the HOPE-technique [ 21 ], the cell lines used were: A549, HeLa, NCI-H727, Jurkat, L428, CPC-N, Raji, H23, U937, H157, H125, L428, and DV90. Preparation of the probes Total RNA was extracted from lung tissues according to the manufacturer's recommendations (RNeasy, Qiagen). After destroying residual DNA with DNase (Invitrogen), cDNA was synthesized by reverse transcription [ 22 ]. PCR was performed targeting a 393 bp fragment of human TLR9-mRNA (TLR9 forward: AAC TGG CTG TTC CTG AAG TC; TLR9 reverse: TGC CGT CCA TGA ATA GGA AG). PCR-products were separated on 2 % agarose gels stained by ethidiumbromide. Cycle sequencing confirmed 100 % identity with the human TLR9 wild-type-sequence. Probes were labeled with digoxigenin using High-Prime (Roche) according to the manufacturer's recommendations [ 23 ]. ISH Hybridization, detection of signals and controls were carried out as previously described (concentration of probe 2 ng/μl, hybridization temperature 46°C) [ 20 , 22 ]. IHC Primary antibody (mouse anti-human TLR9, clone 26C593, Imgenex) was applied in a dilution of 1/100 in PBS for 16 h at 4°C. Negative controls comprised omission of the primary antibody. Detection was performed by horseradish-peroxidase labeled streptavidine-biotin technique (LSAB2, Dako) [ 24 ]. RT-PCR/Cell lines A549, HeLa, BEAS 2b, U937, and NCI-H727 cell lines were used. RT-PCR was performed like described above using TLR9 specific primers (forward: 5'CATGCCCTGCGCTTCCTATTCA; reverse: 5'TGGGCCAGCACAAACAGCGTCTT) spanning an amplicon of 260 bp. Mononuclear cells were included as positive control as well as RT-PCR targeting glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (forward: GTCATCATCTCCGCCCCTTCTGC; reverse: GATGCCTGCTTCACCACCTTCTTG) (not shown). PCR-products were separated along with a molecular weight marker (MW8, Roche) using 2 % agarose gels (Fig. 1 ). Figure 1 Immunohistochemistry (IHC) (A-C) for TLR9 detected by a mouse monoclonal antibody. Adenocarcinoma of the lung (A) . Squamous cell carcinoma of the lung (B) . A549 cells (all 600 ×) (C) . In situ hybridization (ISH) targeting mRNA of human TLR9 with a digoxigenin-labeled DNA-probe in a squamous cell carcinoma of the lung (600 ×) (D) . Immunohistochemical staining of TLR9-expression-levels in nonmalignant (E) and malignant tissues (F) derived from the same lungs an stained by the use of tissue arrays. Results of RT-PCR targeting TLR9 in cell lines (G) . M: molecular-weight marker (MW8, Roche). 1: negative control; 2: A549; 3: NCI-H727; 4: BEAS 2b; 5: Mononuclear cells from a healthy human donor. Confocal laser microscopy of A549 cells transiently transfected with a GFP-TLR9 plasmid: Cytoplasmic expression of TLR9 is observable in all cells, while successful transfection led to overexpression of TLR9 resulting in bright GFP signals completely superimposed by the TLR9 antibody signal (H) . Nuclear counterstain was performed with TOTO3. Transfection A549-cells were seeded in 35-mm glass bottom dishes (MatTek Corp.) overnight. Cells were transfected with GFP-huTLR9 using Polyfect (Qiagen) according to the manufacturer's instructions or incubated in medium. Confocal Microscopy Cells were washed in tris-buffered-saline, containing 0.2 % Tween 20 (TTBS), fixed with 4 % paraformaldehyde in phosphate-buffered-saline (PBS) for 10 min on ice, and permeabilized with 0.25 % Triton-X100 (Roche) in PBS for 10 min. Cells were washed with TTBS, blocked with 10 % bovine-serum-albumine (BSA) in TBS for 20 min, and incubated with primary antibody (clone 26C593, Imgenex) or isotype (Mouse IgG1, Jackson ImmunoResearch Laboratories) 1:150 in TBS 10 % BSA for 30 min. Cells were washed with TTBS, incubated for 30 min with Alexa-568/goat-anti-MouseIgG1 (Molecular Probes Inc.) 1:500 in TBS containing 10 % BSA, and washed with TTBS. Counterstaining was achieved using TOTO-3 1:500 in TBS containing 10 % BSA. Cells were washed with TTBS, fixed again as above, mounted and analyzed using a confocal laser microscope. The GFP-TLR9 plasmid was kindly provided by Terje Espevik, Trondheim, Norway. Treatment Protocols For CpG-ODN stimulation the M362 sequence was used in a concentration of 1 μM; as control M383 was used as described by Hartmann et al. [ 25 ] (MWG-Biotech). Human tumor necrosis factor-alpha (TNF-α, Roche) in PBS containing 0.5 % bovine serum albumin was added to the cultures in a concentration of 10 ng/ml. CHX (Sigma) was dissolved in PBS and added in a concentration of 10 μM. Flow cytometry Annexin-V FITC apoptosis kit I and PE-conjugated active caspase-3 apoptosis kit I were used according to the manufacturer's instructions (BD Pharmingen). TLR9 antibody and isotype control (eBioscience, clone: eB72-1665) were stained after fixation and permeabilization using Intraprep (Beckmann Coulter) according to the manufacturer's instructions. Flowcytometric data (FACS Calibur) collected from 10,000 cells are reported as percentages of positive cells (Becton Dickinson). Cytokine assays Cell culture supernatant (50 μl per sample) was analyzed using the Bio-Plex system and a Luminex 100TM analyzer (BioRad) according to manufacturer's instructions. Stimulation of tumor-tissues and RT-PCR Tissue blocks from lung cancer specimens (edge length approximately 0.5 cm) were cultivated in RPMI 1640 at 37°C and 5 % CO 2 for 24 h, and either stimulated or not stimulated with 1 μM of CpG-ODN (M362 sequence). These blocks from adjacent locations of the same lung-tumors were fixed using the HOPE-technique and paraffin embedded. RT-PCR was carried out like described above using primers targeting human MCP-1 (forward: AAAGCACCAGTCAACTGGAC; reverse: AGCGCTTGGTGATGTGCTTT) resulting in a 149 bp PCR-product and GAPDH (forward: AGAACGGGAAGCTTGTCATC; reverse: TGCTGATGATCTTGAGGCTG) resulting in a 257 bp PCR-product. PCR products were separated on 2 % agarose gels along with a molecular weight marker (pBR322- Msp1 ) and the results displayed in figure 4b . Figure 4 MCP-1 secretion in response to CpG-ODN-stimulation in the presence or absence of TNF-α by HeLa and A549 cells (A) . Data are expressed as the mean ± SD (n = 6). Student's t test was used for statistical analysis. RT-PCR targeting mRNA of MCP-1 in human non-small cell lung cancer tissue stimulated with CpG-ODN for 24 h (B) (M = pBR322- Msp1 ). Lanes 2 and 3, as well as lanes 4 and 5 respectively show results of tissue samples from the same tumors either in the absence or presence of CpG-ODN. Results Expression of TLR9 in malignant tumors To investigate the expression of TLR9 in human lung tumors and lung tumor cell lines we used the recently described HOPE-fixation method. HOPE-fixed [ 19 ] specimens showed superior preservation of morphology after in situ hybridization (ISH). The generation of TLR9-signals was achieved within 10 minutes, whereas unspecific signals were not detected in the control preparations. We found high signal intensities for TLR9 transcripts in the cytoplasm of tumor cells in the majority of lung cancer specimens. Immunohistochemistry (IHC) revealed strong TLR9 protein expression within tumor cells of tissues and cell lines. In contrast normal lung tissues sporadically showed weak expression of TLR9 mainly in cells revealing morphological characteristics of alveolar macrophages and alveolar epithelial cells as displayed in figure 1 . Negative control specimens did not display signals. The results are summarized in table 1 ; some representative results of ISH and IHC are displayed in figure 1 . To confirm the results obtained by ISH we analyzed TLR9-transcripts in tumor cell lines by RT-PCR. As shown in figure 1 , we found that all tumor cell lines indeed express TLR9. Table 1 Summarized results of immunohistochemistry (IHC) targeting TLR9 in tumor tissues and cell lines. Entity N * No expression Weak expression Strong expression Adenocarcinoma of the lung 21 1 7 13 Squamous cell carcinoma of the lung 23 1 14 8 Large cell carcinoma of the lung 3 0 2 1 Cell lines ** 13 0 1 12 Total 60 2 24 34 * Number of analyzed specimens ** See methods A cytoplasmic localization of TLR9 was confirmed by confocal microscopy (fig. 1 ). This finding is in agreement with previous studies on the distribution of TLR9 in RAW264.7 cells [ 10 ]. Furthermore, immunostaining of GFP-TLR9 transfected A549 cells verified the specificity of the TLR9 antibody: Only those cells which were successfully transfected as demonstrated by the GFP-dependent fluorescence also stained brightly with the TLR9 antibody. CpG-ODN stimulation reduces spontaneous and tumor necrosis factor-alpha (TNF-α)/Cycloheximide (CHX)-induced apoptosis The expression of TLR9 in tumor cells and cell lines rises up the question, whether this receptor is functional active in these cells. As shown in figure 2a , CpG-ODN decrease the rate of spontaneous and induced apoptosis in HeLa and A549 cells after treatment with TNF-α and CHX. Representative histograms demonstrate the detection of annexin in the presence or absence of CpG-ODN and TNF-α/CHX (Fig. 2b and 2c ). The induction of apoptosis after stimulation with TNF-α/CHX was further verified by the expression of active caspase 3 as shown in figure 2d . In the presence of CpG-ODN the expression was reduced analogous to the reduction of annexin-staining (Fig. 2e ). Figure 2 CpG-ODN-stimulation decreases apoptosis in HeLa and A549 cells. Cells were stained with Annexin-V after CpG-ODN-stimulation in the presence or absence of TNF-α and CHX after 24 h (A) . Data are expressed as the mean ± SD (n = 6). Student's t test was used for statistical analysis. Representative histograms are shown from experiments with HeLa cells after CpG-ODN-stimulation in the absence (B) or presence (C) of TNF-α and CHX. Caspase 3 expression in HeLa cells is shown after incubation with TNF-α and CHX (D) . In the presence of CpG-ODN the expression is decreased (E) . The percentage of positive cells in each sample is indicated. Influence of induced apoptosis on TLR9 expression Here we investigated, whether CpG-ODN can modulate their own receptor. We found no differences in TLR9 expression with and without CpG-ODN stimulation. However, in the presence of TNF-α/CHX the expression of TLR9 was strongly reduced, whereas CpG-ODN stimulation counteracted this downregulation (Fig. 3a and 3b ). Figure 3 TLR9 expression after CpG-ODN-stimulation in HeLa cells: There is no difference in TLR9 expression with and without CpG-ODN-stimulation after 24 h (A) . CpG-ODN partially inhibit downregulation of TLR9 which is induced by TNF-α and CHX (B) . FI = fluorescence intensity. Secretion of MCP-1 in response to CpG-ODN and TNF-α In order to obtain further information about the functional activity of TLR9 in tumors we studied cytokine release upon CpG-ODN stimulation. The measurement of cytokines from stimulated HeLa and A549 cells revealed a significantly enhanced release of monocyte chemoattractant protein-1 (MCP-1) after 24 h of stimulation in response to CpG-ODN or TNF-α (Fig. 4a ). The production was further enhanced when stimulated with a combination of CpG-ODN and TNF-α (Fig. 4a ). There was no effect of CpG-ODN on TNF-α production (data not shown). To verify the induction of MCP-1 by CpG-ODN in cell lines we additionally analyzed human tumor tissues by RT-PCR; the results are shown in figure 4b . The relative amounts of RT-PCR-signals for MCP-1 in relation to GAPDH were higher in the specimens treated with CpG-ODN if compared with the controls confirming the results obtained in cell culture experiments on the tissue level. Discussion By application of a novel fixation technique we specify for the first time the expression of TLR9 protein and mRNA in a selection of human non small cell lung cancer tissues as well as cell lines. Stimulation of the TLR-9 expressing cell lines A549 and HeLa with CpG-ODN showed a marked antiapoptotic effect. In addition, there was substantially enhanced release of MCP-1 from the cell lines upon CpG-ODN stimulation which was also shown in ex vivo experiments. We conclude the expression of a functionally active TLR9 in human malignant tumors. The presence of molecules involved in ontogenesis e.g. the carcinoembryonic antigen (CEA) is frequently observed in malignant tumors suggesting a kind of "shift-back" towards earlier developmental stages [ 26 ]. The significance and underlying mechanisms of this phenomenon are poorly understood; nevertheless, the detection of such molecules is used for diagnostic purposes in cancer [ 27 ]. The role of TLR in mammalian embryogenesis is unknown, and thus far there is no evidence for an endogenous TLR9 ligand homologous to Spaetzle . Such a ligand could play a role for the activation of human TLR9. Whether the expression of TLR9 in human malignant cells takes advantage of TLR9-function in embryogenesis therefore remains unclear. On the other hand TLR9 in malignant cells could have similar functions as in cells of the innate and adaptive immune system. In humans TLR9 has been described to be mainly expressed in B-lymphocytes, monocytes and plasmacytoid dendritic cells [ 28 ]. In addition Platz et al. reported a weak expression in respiratory epithelial cell lines and primary epithelial cells [ 29 ]. The CpG-ODN sequence M362 used in our study is known to potently activate TLR9-expressing immune cells in humans including plasmacytoid dendritic cells and B cells as shown by Hartmann et al. [ 25 ] B cells are induced to proliferate and secrete immunoglobulin in response to CpG-ODN, dendritic cells produce a wide array of cytokines and apoptosis is inhibited [ 30 , 31 ]. These mechanisms are both reflected in the results we obtained in our study after CpG-ODN stimulation of malignant cells: Firstly, stimulation of the A549 and HeLa cells with CpG-ODN showed an antiapoptotic effect. This was demonstrated for spontaneous as well as induced apoptosis with TNF-α and CHX after 24 h. Our observation is consistent with previous evidence in other cell lines. Yi et al. demonstrated antiapoptotic effects of CpG-ODN in a mouse B lymphoma cell line [ 32 ], and similar changes were described in chronic lymphocytic leukemia cells [ 33 , 34 ]. Previous data of systemic administration of bacterial DNA as a single agent in vivo showed anti-tumor effects. However, this anti-tumor effect appears to be effective indirectly and is related to enhanced NK cell activity. In a murine model of lymphoma the immunostimulatory effect of CpG-ODN was demonstrated to be responsible for the observed anti-tumor effects [ 35 ]. Carpentier et al. have shown that CpG-ODN in vivo induced rejection of neuroblastoma xenografts [ 36 ]. In contrast CpG-ODN had no effect on survival in mice inoculated with the 38C13 murine B cell lymphoma. However, a single injection of CpG-ODN enhanced the response to anti-tumor antibody therapy [ 37 ]. To what extent the antiapoptotic effects of CpG-ODN on tumor cells demonstrated in our study affect the tumorbiology in vivo requires further investigation. Secondly, tumor cell lines (A549 and HeLa) stimulated with CpG-ODN showed strong secretion of the CC chemokine MCP-1. Furthermore a similar effect was observed in the investigated tumor tissues. Immunostimulatory properties together with anti-tumor activity of bacterial DNA were initially reported for a DNA fraction derived from mycobacteria by Tokunaga and coworkers [ 38 ]. It is known that such DNA induces enhanced production of various cytokines with anti-tumoral activity in NK cells, B cells, monocytes, macrophages and dendritic cells, such as TNF-α, IL-12, and IFN-γ [ 39 ]. In our study a substantial costimulatory effect in addition to CpG-ODN was achieved using TNF-α. MCP-1 has various biological activities including the induction of increased cytotoxic activity of monocytes and NK cells. Transfection of MCP-1 into a human malignant glioma cell line tested on nude mice did not reduce the tumor mass but was associated with the infiltration of large numbers of NK cells and monocytes at the tumor site [ 40 ]. A further study by Nokihara et al. performed with transfection of the MCP-1 gene into human lung adenocarcinoma cells showed reduced systemic spread of transfected cells inoculated i.v. in NK cell-intact severe combined immunodeficient (SCID) mice. These findings suggest that locally produced MCP-1 suppresses tumor progression by a NK cell-mediated mechanism [ 41 ]. Thus, apart from the direct activation of immune cells, the effect of CpG-ODN stimulation on the secretion of MCP1 by TLR9 expressing tumor cells could possibly lead to anti-tumoral effects due to an increase of local MCP1 production which then might lead to attraction of immune cells. The costimulatory effect of TNF-α as demonstrated in vitro in this study could further enhance this scenario. Regarding TLR9 expression in nonmalignant lung tissue our data confirm the findings of low TLR9 expression in respiratory cells of Platz et al. [ 29 ], who have been working on single cell preparations. However TLR9 expression was only seen sporadically weak in nonmalignant lung tissue. Biological explanations for the TLR9 expression in malignant cells require further investigations. Three possibilities are conceivable: Either this could represent a bystander phenomenon, a side effect of a pathway functional to a different purpose. Secondly the upregulation of TLR9 could be beneficial to the tumor, promoting tumor cell survival. Thirdly, it even might help immune control strategies of the organisms an element of a pathway directing defense mechanisms against malignantly transforming cells. While the first possibility seems unlikely in the light of our findings of a functionality of the receptor in various in vitro and ex vivo experiments, our data provide evidence for the second as well as the third possibility; the sum effect of these two counteracting mechanisms in an in vivo setting can not be estimated from these experiments and could even differ from tumor entity to tumor entity. Conclusions In conclusion, we showed in a selection of samples that human malignant tumors express functionally active TLR9 and respond to CpG treatment with prolonged survival and chemokine release. This might influence the effects of CpG-ODN based anti-tumor therapies. Broad screening approaches will be worthwhile to further substantiate these initial results. While recent strategies in tumor-immunology mainly target a strengthening of the host-defense, we provide evidence that the malignant cells themselves can be regarded active players in the complex struggle between tumor and host. In any case CpG-ODN based anti-tumor therapies should be reconsidered in the light of our findings since CpG-ODN products are currently in Phase I/II clinical trials both as a monotherapy and as part of multi-drug regimens. Author's contributions DD carried out the flow cytometry and cytokine assays and was involved in the design and coordination of the study and drafting the manuscript. DA and AJU carried out the confocal microscopy, RT-PCR with cell lines and were involved in drafting the manuscript. JG was involved in immunohistochemistry of cell lines and the design of the study. DB conducted the surgical part of the study. EV conducted the pathological part of the study and was involved in the design of the study. KD and PZ conducted the clinical part of the study and were involved in the design and coordination of the study. TG performed the immunohistochemistry, in situ hybridization and RT-PCR with tissues and conceived of the study. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544843.xml
523827	Should Health Professionals Screen All Women for Domestic Violence?	Background to the debate: The US and Canadian task forces on preventive health recently declared that there is not enough evidence to recommend for or against routine universal screening of women for domestic violence. Yet some experts argue that routine enquiry is justified.	Ann Taket's Viewpoint: Routinely Asking about Domestic Violence Is Worthwhile Domestic violence is a misunderstood topic. The context of a trusted health professional talking to a woman is one that provides an important opportunity for providing information to counter misconceptions. I deliberately talk about this in terms of asking all women about domestic violence and not in terms of screening women for domestic violence. It is not appropriate or helpful to regard enquiry about being abused as a form of screening. Domestic violence is not a disease present in the body of the person who experiences it—rather it is a health-related risk factor. As such, knowledge of abuse puts health professionals in a position to respond better to the needs of women affected by it. Professionals can respond by providing information on specialist services—usually provided outside the health service—that women may access if they wish. By giving information to affected women, health professionals can also help to reduce women's sense of isolation and stigmatisation. Asking about experience of domestic violence can be seen as a routine part of history taking, just as health professionals regularly and repeatedly ask patients about their smoking behaviour, alcohol use, weight, and exercise. The prevalence of domestic violence among women is such that, even if it is not a personal issue for the woman concerned, it most likely will be for one or more of her relatives, friends, and neighbours [ 1 ]. Since many women experiencing abuse feel alone and ashamed, and their abusers often encourage them to believe that the abuse is their fault, presenting information to counter women's negative feelings is an important preventive strategy. Most women experiencing domestic violence report that the specialised services that exist to respond to their needs were difficult to find out about [ 2 ]. The provision of simple information on the existence of specialised services and how to contact them is relevant to all women. Studies have examined women's views on being asked about domestic violence. These studies have shown that once they have experienced being asked, they are usually in favour of being asked. This is true both for those who have experienced or are experiencing abuse, and those who have not [ 3 ]. It is only a small minority of women who object to being asked, or who find the question uncomfortable. Women who have experienced abuse particularly value being asked directly. Women experiencing violence often feel alone and ashamed (Illustration: Margaret Shear, Public Library of Science) Asking about abuse should be done in a flexible fashion—the particular questions used should respond to the circumstances of the consultation. For example, it is appropriate to ask women about domestic violence as part of a health check in a Well Woman Clinic, but it would be completely inappropriate in a consultation where another adult or a child was present. By being flexible, health professionals can integrate their questioning within a variety of different encounters. Integrating questions about abuse into routine encounters provides for the maintenance of confidentiality and safety. In order to do this, health professionals require training on raising the issue and knowledge about local advice and support services. Committees on both sides of the Atlantic have rejected the notion of screening women for domestic violence, arguing that there is insufficient evidence of the effectiveness of interventions [ 4 , 5 ]. Part of the reason for this lack of evidence is that the systematic reviews on which these committees based their recommendations often excluded the most important types of evidence that do exist [ 3 , 6 ]. For example, these reviews excluded studies done outside the health service setting—they excluded those based in social services, or in the voluntary or community sector. Some excluded studies show the effectiveness of specialised service provision for women experiencing abuse. In one example of an excluded study, researchers used a randomised design to evaluate an advocacy service for women experiencing domestic violence [ 7 , 8 ]. Women were interviewed six times over two years, and women in the intervention group reported a higher quality of life, decreased difficulty in obtaining community resources, and less violence over time than women in the control group. Other studies showing the value of specialised support services provided outside of the health system provide evidence of the potential benefits of asking about abuse [ 2 ]. Systematic reviews have also excluded, or devalued, evidence from qualitative studies. For example, a study of 200 women who had used domestic violence outreach services found that about half were living in situations of domestic violence when they first contacted the service. All of these women reported that the outreach services had helped them to leave the abusive relationship—a valued outcome for them [ 9 ]. Given the health impacts on women who experience domestic violence (not to mention their children) and the prevalence of the problem, routinely asking women about abuse should be seen as an important form of primary and secondary prevention for a wide range of health problems. Nadine Wathen and Harriet MacMillan's Viewpoint: The Decision to Screen Should Be Based on Evidence Screening tools for domestic violence are abundant, and many are effective at identifying women experiencing abuse [ 3 , 10 ]. However, merely identifying a woman as abused has not been shown to actually improve her quality of life or reduce the violence she is experiencing [ 6 , 11 ]. Furthermore, with one exception [ 7 ], we do not know whether interventions for women exposed to violence are effective in reducing violence or improving other health-related outcomes. Interventions for abusive men have shown little effectiveness [ 11 , 12 ]. Given the morbidity and mortality associated with domestic violence, it is tempting to suggest that universal screening for abuse should be integrated into routine clinical care, such that all women, regardless of their reason for presenting to a clinical setting, should be “asked the question.” Some argue that this approach is justified by the need to increase awareness of domestic violence as a significant problem with serious health and social consequences, and to make abused women aware that they are not alone in their experience. These are important considerations. Certainly all women who disclose that they have been exposed to violence should be provided with options regarding seeking help [ 13 ]. Good diagnostic assessment requires that clinicians be able to identify and respond to signs and symptoms of abuse, from patterns of physical injury to mental health concerns, including unexplained pain and depression. Not asking women about exposure to violence during certain diagnostic assessments (such as investigation of chronic pain) may lead to misdiagnosis and a path of inappropriate investigations or treatments that will miss the underlying problem [ 14 ]. It is also imperative that clinicians know about the hospital- or community-based services that exist and ensure that there is a system in place to provide appropriate referral [ 15 ]. However, what about women presenting without obvious signs and symptoms of domestic violence—such as a woman who comes to the clinic for assessment of an upper respiratory tract infection? Should such women be prompted to disclose whether they are being abused? The woman who is not being abused will answer to that effect, and the appointment can carry on. But for the woman who is experiencing violence, who has not volunteered this information, several factors must be considered. An important issue is whether she is ready—both psychologically and in terms of taking specific actions—to confront the issue. A number of excellent qualitative studies have examined the process that women undertake in acknowledging that they are “victims” of “abuse” and embarking on the often long and difficult journey to avoid, reduce, and ultimately stop the violence in their lives [ 16 , 17 ]. Given the enormousness of that task, the key question becomes the extent to which prompting disclosures of abuse through universal screening will actually help women in this process, and help them in a way that they find meaningful. Any potential benefits of screening must then be weighed against its potential harms, including labelling women, prompting potentially premature disclosure, and triggering possible reprisal violence from the abuser if he discovers she has sought help. The last of these might be particularly exacerbated for the woman with the respiratory tract infection who was unprepared to disclose and did not take necessary precautions. Other potential harms include exposure to the ramifications of laws on mandatory child protection reporting, whereby health providers must report such disclosures to child protection authorities. This can lead to an investigation that potentially increases a woman's risk of exposure to violence, and in some cases of having her children placed in foster care. Research has shown that many of these potential harms are of concern to women when mandatory universal screening and/or reporting protocols are in place [ 18 ]. Finally, from a health system perspective, the opportunity cost of not having used this time with the woman to conduct screening or prevention activities for which there is proven benefit, such as counselling about Pap smears or mammograms, should not be discounted. There are potential harms from “asking the question” (Illustration: Margaret Shear, Public Library of Science) Given the lack of clear data on the benefits of screening and of the interventions to which women are referred, and the lack of data on potential harms, we and others have concluded the following [ 3 , 19 , 20 ]. Until these questions are answered, the most appropriate health care system approach is the more targeted case-finding or diagnostic method, which focuses health care resources on those in immediate need of care. Our hope is that studies currently underway (for example, those supported by the Ontario Women's Health Council and the US Centers for Disease Control) will provide information about the effectiveness of domestic violence screening. Let's base the decision about implementation of screening on evaluations of whether such screening does more good than harm in the lives of women. Taket's Response to Wathen and MacMillan's Viewpoint I agree entirely with Nadine Wathen and Harriet MacMillan that practice should be based on evidence. There are further areas of agreement. We agree that there is a lack of knowledge on effective interventions for abusers and on harm occurring as a result of enquiry, and that targeted case finding is important. The key difference that exists between my viewpoint and theirs is the conclusion about whether health professionals should aim to ask all women about domestic violence. Underlying this difference is the issue about how much evidence we need, and of what type. My position is that the evidence that already exists is sufficient to justify the promotion of routine enquiry, aiming to ask all women about their experience of abuse. There is evidence of actual benefits to women—and their children—from interventions provided by specialised services for domestic violence and from brief discussions with health professionals [ 21 ]. Aiming to ask all women has several advantages over targeted case finding [ 22 ]. It contributes to changing social attitudes to domestic abuse, it is less likely to make women experiencing abuse feel stigmatised, and it is less likely to compromise the safety of women experiencing abuse. Furthermore, health professionals report that their perceptions about which women are being abused, and which are not, are often wrong. The twin issues of women's safety and harm minimisation are extremely important, for both routine enquiry and targeted case finding. These issues are important reasons why training and protocols for enquiry are necessary. Standard principles of confidentiality should be reinforced in training and protocols, which need to be tailored to relevant legal requirements, such as when child protection issues are involved. Training and protocols also need to emphasise that the role of routine enquiry is to facilitate, and not force, disclosure. It must remain the woman's choice as to if, when, and to whom, she discloses. Wathen and MacMillan's Response to Taket's Viewpoint We agree with Ann Taket that domestic violence is not a disease, and that the paradigm of “screening for disease” is problematic in this context. At issue, however, is the question of whether domestic violence should be “talked about” with all women or only in situations where asking about it is part of a specific diagnostic assessment. As with screening for a disease, universal screening for domestic violence should not be implemented unless we are sure that interventions are available to help those identified via screening and that screening plus appropriate treatment will do more good than harm. Professor Taket outlines the importance of integrating discussions about abuse in consultations to raise community awareness. Unfortunately, there is no evidence that this type of consciousness-raising occurs, or if it does, what benefit it might have. Given the lack of effectiveness of educational campaigns in general, it is difficult to be optimistic about this approach. We disagree with her conclusion that existing systematic reviews have “excluded studies done outside the health service setting….” Our review included interventions such as the post-shelter advocacy counselling approach to which Professor Taket refers [ 11 ]. This intervention has been recommended by the Canadian Task Force on Preventive Health Care as one to which, where available, clinicians might refer women in these circumstances [ 19 ]. However, since shelters themselves have not been adequately evaluated, the value of linking screening to a post-shelter intervention is unclear. Finally, we concur that qualitative studies are invaluable in understanding domestic violence. Such research has provided insight into the complex process that women undertake to address the violence in their lives. Until there is evidence that universal screening actually helps with this process, the focus should be on developing evidence-based approaches to assist women when they do disclose abuse and on training health professionals to respond appropriately to such disclosures.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC523827.xml
535923	Progression of coronary calcification in healthy postmenopausal women	Background Coronary artery calcium score incrementally improves coronary risk prediction beyond that provided by conventional risk factors. Limited information is available regarding rates of progression of coronary calcification in women, particularly those with baseline scores above zero. Further, determinants of progression of coronary artery calcification in women are not well understood. This study prospectively evaluated rates and determinants of progression of coronary artery calcium score in a group of healthy postmenopausal women. Methods We determined coronary calcium score by computed tomography and recorded demographic, lifestyle and health characteristics of 914 postmenopausal women, a subset of those enrolled in the Women's Health Initiative Observational Study. The 305 women with calcium score ≥10 Agatston units at baseline were invited for repeat scan. This analysis includes the 94 women who underwent second scans. Results Mean age of study participants was 65 ± 9 years (mean ± SD), body mass index was 26.1 ± 6.1 kg/m 2 , and baseline calcium score was 162 ± 220 Agatston units. Mean interval between scans was 3.3 ± 0.7 years. A wide range of changes in coronary calcium score was observed, from -53 to +452 Agatston units/year. Women with lower scores at baseline had smaller annual increases in absolute calcium score. Coronary calcium scores increased 11, 31 and 79 Agatston units/year among women with baseline calcium score in the lowest, middle and highest tertiles. In multivariate analysis, age was not an independent predictor of absolute change in coronary calcium score. Hydroxymethylglutaryl coenzyme A reductase inhibitor (statin) use at baseline was a negative predictor (p = 0.015), whereas baseline calcium score was a strong, positive predictor (p < 0.0001) of progression of coronary calcification. Conclusion Among postmenopausal women with coronary calcium score ≥ 10 Agatston units, rates of change of coronary calcium score varied widely. In multivariate analysis, statin use was a negative independent determinant, whereas baseline calcium score was a strong positive predictor of annual change in coronary calcium score.	Background Coronary calcium, assessed by computed tomography, strongly and independently predicts coronary risk [ 1 - 3 ]. Age is by far the most potent determinant of calcium score [ 4 ], although conventional risk factors also been associated with the extent of coronary calcification [ 4 , 5 ]. The rate of progression of coronary calcification appears to further stratify risk [ 6 , 7 ], but reports have been limited by sample size [ 8 ], retrospective design [ 6 , 8 , 9 ], inclusion of individuals with baseline calcium scores of zero [ 9 , 10 ] and limited interval between tomographic scans [ 8 , 10 , 11 ]. Further, not all studies adjusted for use of hydroxymethylglutaryl coenzyme A reductase inhibitors (statins), which have been reported to attenuate progression [ 9 , 12 , 13 ]. Calcium scores differ in men and women [ 4 ], but progression of coronary calcification has not been reported by gender, except for the Healthy Women Study, which only included women [ 10 ]. Of the 80 women in that cohort, 52 (65%) had calcium scores of zero at baseline. After mean follow up of 18 months, 47 of the 52 (90%) had no coronary calcium on repeat scan. Mean change for the 52 women was 0.4 Agatston units and median change was 0. Among the 28 women with measurable coronary calcium at baseline, mean change was 11 Agatston units for women with baseline calcium score 1–99, and 72 Agatston units for the 9 women with baseline calcium score ≥100. In this study, we prospectively assessed the rate of progression of coronary calcification in an ethnically diverse group of healthy women with coronary calcium scores of at least 10 Agatston units at baseline, and identified independent predictors of progression. Methods Patient population Study participants were a subset of women enrolled in the Women's Health Initiative Observational Study [ 14 ] at the George Washington University and Howard University/Medstar clinical sites between February 1995, and December, 1998. Women who joined this ancillary study provided informed consent in a form approved by the respective institutional review boards. The entire Observational Study cohort comprises 93,676 women at 40 clinical sites. For this ancillary study, participants at the George Washington and Howard/Medstar clinics (n = 4435) were invited for computed tomography. Baseline scans were performed on the 914 women who responded to the invitation. Of these, 528 had no coronary calcium detected and 81 had calcium scores of 1 – 9 Agatston units. The remaining 305 women with calcium score ≥10 Agatston units were mailed a letter inviting them to have a second scan; African-American women received two mailings because of a historically lower response rate. This analysis includes the 94 women with serial scans, which were performed a mean of 3.3 ± 0.7 years after the baseline study. Variables Participants provided data on a wide range of health variables including dietary habits, medical history and anthropometric measures. Questionnaire measures assessed self-reported hypertension, diabetes mellitus (excluding gestational diabetes), current smoking, high cholesterol requiring pills, postmenopausal hormone therapy, and family history of premature coronary disease (father with myocardial infarction at age 55 years or younger, or mother with myocardial infarction at age 65 or younger). Statin use at baseline was assessed by medication inventory. Dietary fat consumption was assessed using a food frequency questionnaire based on instruments used in the Women's Health Trial [ 15 ]. Nutrient estimates from the food frequency questionnaire were similar to those from short-term dietary recall and from four-day food records [ 16 ]. Physical activity was assessed by questions on a frequency and duration scale of four walking speeds and three other types of activity classified by intensity (strenuous, moderate or light)[ 17 ]. For this analysis, we categorized women by the number of weekly episodes, at least 20 minutes in duration, of moderate or strenuous activity. Plasma lipids were only measured in a 1% random subsample of Observational Study participants, so were not included as variables in these analyses. CT image acquisition and analysis Images were acquired using an Imatron C – 150 scanner. Thirty contiguous 3-mm slides (100 ms/slice) were acquired during a single breathhold beginning 1 cm caudad to the carina. Each level was triggered by ECG in end-diastole (80% of R-R interval). Images were obtained with a 30-cm 2 field of view (pixel size, 0.586 mm). Images were analyzed by the Agatson method [ 18 ]. Analysis Descriptive statistics such as frequencies, percentages, means and standard deviations (SD) were used to describe the study population and to explore the relationships between coronary calcium score and several explanatory variables. Group comparisons were made by t test, chi square and, where appropriate, the Mantel-Hanzel chi square test. In Table 2 , the p value is based on ranked scores because of variance heterogeneity. Annual change in coronary calcium score in Table 2 is adjusted for age using ANCOVA's least squares means. Table 2 Annual change in calcium score Baseline calcium score Annual change in calcium score Unadjusted Age-adjusted % change (Agatston units) (Agatston units/y) mean (SD) range mean (SD) range 1 (n = 32) 32 (11) 13 to 51 11 (16) -7 to 38 11 33% 2 (n = 30) 94 (39) 52 to 189 31 (31) 3 to 148 34 33% 3 (n = 32) 559 (292) 194 to 1236 79 (102) -53 to 452 77 14% p across tertiles 0.0001 Determinants of annual change in calcium score were evaluated in a multiple linear regression model which included age, baseline calcium score, and statin use at baseline as independent variables (Table 4 ). In a separate model, hypertension was added. Age and statin use were selected as independent variables because they have consistently been identified as determinants of coronary calcification [ 12 , 13 , 19 ]. Baseline calcium score and hypertension were included because of their relationship to change in calcium score in Table 2 and 3 , respectively. Analyses were carried out using SAS System for Windows v8.02. Table 3 Risk factors by tertile of annual change in calcium score Tertile p value 1 2 3 n 32 30 32 mean (SD) Age, y 64 (7) 65 (7) 67 (12) 0.55 Body mass index 25.1 (4.7) 26 (7.3) 27.1 (6.04) 0.45 % calories from fat 27 (6) 27 (7) 27 (10) 1.00 # days/week moderate-vigorous exercise 1.8 (2.2) 2.7 (2.6) 2.2 (2) 0.33 % Hypertension 33 39 59 0.08 Diabetes 7 0 9 0.60* High cholesterol 16 26 22 0.65 Current smoking 10 10 6 0.63* Family history premature CHD 13 6 19 0.48* Hormone use at baseline 48 42 44 0.87 Statin use at baseline 13 10 9 0.65* * Mantel-Hanzel Chi Square used because of small cell sizes Table 4 Multivariate analysis: Determinants of annual change in calcium score Parameter estimate 95% CI p value Age 0.4 -0.83, 1.63 0.52 Baseline calcium score 0.15 0.11, 0.20 <0.0001 Statin use at baseline -43.95 -79.00, -8.88 0.015 Results Baseline demographic and health characteristics of the 94 women with serial scans are shown (Table 1 ). Among the 305 women with calcium score ≥10 Agatson units, characteristics of the 94 women with a second scan were similar to the 211 women without a second scan (data not shown), except that the former group was enriched for African-American women, 31/94 (31%) vs 11/211 (5%)(p < 0.0001 across ethnic groups). Table 1 Comparison of women with serial scans vs women with calcium score <10 Group A Group B mean (SD) p value N 94 609 Age, y 65 (9) 61 (8) <0.0001 Body mass index, kg/m2 26.1 (6.1) 25.9 (5.6) NS Baseline calcium score, Agatston units 162 (220) 0.5 (1.6) <0.0001 % dietary calories from fat 27 (8) 26.5 (7.4) NS BP, mm Hg Systolic 126 (19) 119 (17) 0.001 Diastolic 75 (10) 73 (13) NS No. (%) Ethnicity <0.0001 White 58 (62%) 490 (80%) Black 31 (33%) 88 (14%) Asian/Pacific islander 3 (3%) 12 (2%) Hispanic 1 (1%) 11 (2%) Unknown 1 (1%) 8 (1%) Hypertension* 41 (44%) 146 (24%) <0.0001 Diabetes mellitus 5 (5%) 7 (1%) 0.004 Current smoking 8 (9%) 24 (4%) <0.05 Self-reported high cholesterol requiring pills 20 (21%) 76 (12%) 0.02 Hormone use at baseline 42 (45%) 366 (60%) 0.005 Statin use at baseline 9 (10%) 51 (8%) <0.05 # days/week moderate/vigorous exercise 2.2 (2.3) 2.5 (2.4) NS Group A, women with baseline calcium score ≥10 Agatston units and serial scans; Group B, women with baseline calcium score <10 * Self-reported hypertension or measured BP systolic >140 or diastolic >90 Table 1 compares the 94 women with serial scans, all with baseline calcium scores ≥10 Agatston units (Group A), with women whose baseline calcium scores were <10 Agatston units (Group B). Women in Group A were older, and more likely to report hypertension, diabetes, current smoking, and high cholesterol requiring pharmacologic therapy. Statin use was slightly more prevalent and postmenopausal estrogen use slightly less prevalent among women in Group A. Baseline calcium scores for white and African-American women were 278 ± 330 and 163 ± 214 Agatston units, respectively (p = 0.05). Annual change in calcium score and age-adjusted change in calcium score are shown by tertile of calcium score at baseline (Table 2 ). Women with higher baseline calcium scores had greater absolute annual increase in both unadjusted (79 vs 11 Agatston units/year for women in the highest vs lowest tertile, p < 0.0001 across tertiles) and age-adjusted calcium score. Changes in calcium scores ranged from -53 to +452 Agatston units/year. Annual change in calcium score did not differ between white and African-American women (data not shown). Coronary risk factors are shown by tertile of annual change in calcium score (Table 3 ). Body mass index, dietary fat consumption and physical activity were not different in women with more or less rapid progression of coronary calcification. Conventional risk factors, including age, cigarette smoking and diabetes, also demonstrated no significant trend across tertiles of progression in calcium score. Hypertension was reported somewhat more frequently by women with greater progression of coronary calcification (33% vs 59% in lowest vs highest tertile, p = 0.08). In multiple linear regression analysis, age was not independently associated with annual change in calcium score (Table 4 ). Statin use was a weak negative predictor of progression (p = 0.015), whereas calcium score at baseline was a strong positive predictor of annual change in calcium score (p < 0.0001). Results were similar when hypertension, which had shown a non-significant trend across tertiles of change in coronary calcium score, was added to the model. Hypertension itself was not an independent determinant of progression (data not shown). Discussion In this ethnically diverse group of Women's Health Initiative observational study participants, the rate of progression of coronary calcification was 33%/year among women with calcium scores between 10 and 190 Agatston units at baseline, and 14%/year for women with higher calcium scores. The rate of change in calcium score ranged widely in individual women, from -53 to +452 Agatston units/year. In multivariate analysis, statin use was negatively associated with progression of coronary calcification, whereas baseline calcium score was a strong and independent positive predictor of progression. One strength of this analysis is the inclusion of a wide range of variables affecting atherosclerotic risk, including body mass index, physical activity, dietary fat consumption, postmenopausal hormone and statin use in addition to conventional risk factors. Another strength is the relatively long interval between scans, 3.3 years (mean), enhancing accuracy of the estimated rate of progression. This study includes only women, a group traditionally underrepresented in studies of coronary computed tomography [ 6 , 8 , 9 , 11 ], only those with measurable coronary calcium at baseline, and 37% non-white participants. Limitations include the sample size and absence of laboratory measures, such as lipids and glucose. These were performed only in a random 1% subsample of Observational Study participants, and consequently are not available for inclusion in this analysis. The rate of progression of coronary calcification observed in this analysis is within the range reported by others [ 6 , 9 , 11 ]. Similarly, the lack of relationship between postmenopausal hormone use and progression of coronary calcification is consistent with prior reports [ 10 , 20 ], as is the observed inverse association of statin use with progression of coronary calcification [ 12 , 13 ]. Efforts to identify independent predictors of progression of coronary calcification have been limited, particularly in women. A retrospective study of 55 high-risk men identified baseline calcium score and Lp(a) as independent predictors of progression in a multivariate model which did not include statin use [ 21 ]. A prospective study of 87 men and 24 women identified baseline calcium score as a weak independent predictor (p < 0.05) of change in calcium score [ 11 ]. Smoking, hypertension, diabetes, age, plasma lipid levels, body mass index, prevalent coronary heart disease and use of lipid-lowering medications, aspirin, beta-blockers, angiotensin converting enzyme inhibitors, calcium antagonists or nitrates were not independent predictors of progression, although the ability to detect such relationships was limited in view of the sample size, which presents a similar limitation in this analysis. In contrast, smoking was found to be an independent determinant of progression in a larger study of 311 men and 184 women, which also identified baseline calcium score as a potent predictor of progression [ 7 ]. Another limitation of this study was use of Agatston score, rather than calcium volume score. The latter provides better reproducibility [ 22 ], but was not in widespread use at the time baseline scans were acquired and was not available for the baseline scans in this study. The difference between Agatston and volumetric scores increases with the coronary artery calcium score. For example, among women aged 60–64 years, the 75 th percentile has been reported as 59 Agatston units or 42 volumetric units, and the 90 th percentile as 202 Agatston units or 163 volumetric units [ 23 ]. We cannot exclude the possibility that areas of coronary calcification may have consolidated through retraction during follow up, a phenomenon which may not have been accurately assessed using the Agatston score. Calcium scores differ in men and women [ 24 ]; for example the 75 th percentile scores for 50–54 year old men and women are 99 and 3, respectively [ 4 ]. For 60–64 year old men and women, the 75 th percentile scores are 247 and 49, respectively. If baseline score is the major determinant of progression, a gender difference in rates of progression would be expected on this basis alone. In fact, male gender has been reported as an independent determinant of progression [ 7 ]. Whether rates of progression differ between men and women after adjustment for baseline score is uncertain. Our observations raise several issues with regard to the potential use of coronary calcium score progression as either a clinical tool or an outcome for atherosclerotic intervention trials. First, the variance for this variable is high, both in our study and in others [ 7 , 9 , 12 ], limiting, at least to some extent, its value as a clinical predictor in individual patients and its appeal as an intermediate outcome. Second, stratification by statin use should be considered. Third progression should be adjusted for baseline calcium score. Conclusions In an ethnically diverse cohort of postmenopausal women with coronary calcium score ≥10 Agatston units, rates of progression or coronary calcification vary widely. In multivariate analysis, statin use was inversely associated with progression, whereas baseline calcium score was a strong positive predictor of progression. Competing interests The author(s) declare that they have no competing interests. Authors' contributions JH study concept, design, data analysis, initial and final manuscript preparation. AK data analysis. AP and JB data acquisition and management, manuscript revision. LLA-C and BVH manuscript revision. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535923.xml
535937	Tumor taxonomy for the developmental lineage classification of neoplasms	Background The new "Developmental lineage classification of neoplasms" was described in a prior publication. The classification is simple (the entire hierarchy is described with just 39 classifiers), comprehensive (providing a place for every tumor of man), and consistent with recent attempts to characterize tumors by cytogenetic and molecular features. A taxonomy is a list of the instances that populate a classification. The taxonomy of neoplasia attempts to list every known term for every known tumor of man. Methods The taxonomy provides each concept with a unique code and groups synonymous terms under the same concept. A Perl script validated successive drafts of the taxonomy ensuring that: 1) each term occurs only once in the taxonomy; 2) each term occurs in only one tumor class; 3) each concept code occurs in one and only one hierarchical position in the classification; and 4) the file containing the classification and taxonomy is a well-formed XML (eXtensible Markup Language) document. Results The taxonomy currently contains 122,632 different terms encompassing 5,376 neoplasm concepts. Each concept has, on average, 23 synonyms. The taxonomy populates "The developmental lineage classification of neoplasms," and is available as an XML file, currently 9+ Megabytes in length. A representation of the classification/taxonomy listing each term followed by its code, followed by its full ancestry, is available as a flat-file, 19+ Megabytes in length. The taxonomy is the largest nomenclature of neoplasms, with more than twice the number of neoplasm names found in other medical nomenclatures, including the 2004 version of the Unified Medical Language System, the Systematized Nomenclature of Medicine Clinical Terminology, the National Cancer Institute's Thesaurus, and the International Classification of Diseases Oncolology version. Conclusions This manuscript describes a comprehensive taxonomy of neoplasia that collects synonymous terms under a unique code number and assigns each tumor to a single class within the tumor hierarchy. The entire classification and taxonomy are available as open access files (in XML and flat-file formats) with this article.	Background On March 19, 2004, the author published a new classification for neoplasms, now called "The developmental lineage classification of neoplasms"[ 1 ]. The classification was described as a schema providing a single class location for every tumor in man. The classification contains 39 class descriptors appearing as XML tags. Good classifications encapsulate all information relating to a knowledge domain. Modern classifications allow us to understand complex entities by grouping them by their shared or inherited properties [ 2 ]. Modern classifications also allow us to retrieve and integrate many different kinds of data under a common conceptual framework [ 3 - 8 ]. Basic to every classification is a taxonomy, the complete listing of the instances of the knowledge domain [ 2 ]. Because computers make it easy to store, organize and retrieve any number of listed items, there is no reason to limit taxonomies to a small number of preferred terms. One of the the best examples of a large taxonomy is taxonomy.dat, which attempts to list every living organism on earth [ 9 ]. So thorough is taxonomy.dat that it not only lists all known variations of an organism's name, it also lists commonly used misspellings of an organism. An example of an entry in taxonomy.dat: ID : 50 PARENT ID : 49 RANK: genus GC ID : 11 SCIENTIFIC NAME : Chondromyces SYNONYM : Polycephalum SYNONYM : Myxobotrys SYNONYM : Chondromyces Berkeley and Curtis 1874 SYNONYM : "Polycephalum" Kalchbrenner and Cooke 1880 SYNONYM : "Myxobotrys" Zukal 1896 MISSPELLING : Chrondromyces A recent version of taxonomy.dat is dated June 20, 2004 and is 55,233,858 bytes in length. It has 246,800 entries The taxonomy.dat file is available for public download through anonymous ftp [ 9 ]. The creation and organization of biological information is one of the most active areas of biomedical research [ 10 , 11 ]. The taxonomy for the developmental lineage classification of neoplasms was loosely modeled on taxonomy.dat. An attempt was made to include the name of every tumor, including every known variant of tumor name, and to assign a unique numeric code to all synonyms for a given tumor. The purposes of this paper are: 1) to publicly release the neoplasm taxonomy database file; 2) to explain its role as the source of concept instances for the developmental lineage classification of neoplasms; 3) to describe the methods used to organize the taxonomy; and 4) to compare the taxonomy with the neoplasm nomenclature contained in the Unified Medical Language System Metathesaurus (UMLS), the largest medical nomenclature in existence, and with the Systematized Nomenclature of Medicine Clinical Terminology (SNOMED-CT) [see Additional file 1 ] [see Additional file 2 ][ 12 , 13 ]. Methods The developmental lineage classification of neoplasms was described in a prior publication [ 1 ]. The classification was intended to be populated by a comprehensive taxonomy. The original publication contained a relatively short first-draft taxonomy, and the current taxonomy was built on the early draft [ 1 ]. To ensure compatibility between the classification and other biomedical databases that include neoplasm terms, terms and classes were formatted in XML [ 14 , 15 ]. Terms were grouped by concept and examined for completeness, with software written in the Perl programming language. Perl is a free, open source language, available for virtually every computer operating system, and widely used in the bioinformaitcs community [ 16 , 17 ]. Nomeclature terms often have alternate forms that can be discovered and accrued with the help of software [ 18 ]. Short Perl scripts were prepared to systematically add variant names for patterned term constructs. For instance, it was noticed that some terms appeared in the form of "adenocarcinoma of [organ]" and other terms appeared as "adenocarcinoma of the [organ]." A Perl script ensured that both forms were included for this and other examples. As the taxonomy enlarged, inadvertent duplications of terms and codes were unavoidable. In addition, duplicate terms were occasionally placed into different classes within the hierarchy. Much of the value of a classification comes from the parsimonious deployment of taxons (i.e. no instance can appear in more than one class). A Perl script was prepared that was executed after each modification to the classification/taxonomy. The Perl script parsed through the updated XML classification/taxonomy file, validating that: 1) each term occurs only once in the taxonomy; 2) each term occurs in only one tumor class; 3) each concept code occurs in only one hierarchical position in the classification; and 4) the file containing the classification and taxonomy is a well-formed XML document. The script finds anomalous records, permitting facile repair of the taxonomy file. The validating Perl script, xmlvocab.pl is included as a supplemental file with this manuscript [see Additional file 3 ] A perl script transformed the taxonomy XML file into a flat-file consisting of line-records for each term in the taxonomy (neoself.txt, 19+ Megabytes in length). The transforming Perl script is neoself.pl and is distributed with this article [see Additional file 4 ]. The taxonomy was compared with the neoplasm terms and concepts included in the UMLS [ 12 ]. UMLS is produced and curated by the U.S. National Library of Medicine. UMLS concepts and terms are drawn from over 100 different medical source vocabularies. It is the largest medical nomenclature in existence. The 2004 version of the UMLS Metathesaurus used in this study includes over 2,697,491 medical terms. This version of the UMLS is the first UMLS version to contain the Systematized Nomenclature of Medicine – Clinical terminology (SNOMED-CT) [ 13 ]. As in prior versions, the 2004 UMLS contains the National Cancer Institute Thesaurus, another rich source of neoplasm terminology [ 6 ]. The International Classification of Diseases – Oncology (ICD-O), is a nomenclature prepared by the World Health Organization [ 19 ]. Although the UMLS does not list the ICD-O as a contributing thesaurus, it can be noted that ICD-O terminology is incorporated into the SNOMED-CT nomenclature included in UMLS [ 20 ]. The UMLS is curated and distributed by the U.S. National Library of Medicine [ 12 ]. Although the UMLS is publicly available, there are numerous restrictions on its use, and those wishing to download the UMLS must enter into a license agreement with the National Library of Medicine before obtaining the Metathesaurus. All terms from the UMLS Metathesaurus that have a neoplasm relationship were obtained through the use of a Perl script [see Additional file 5 ]. The 2004 UMLS files used for the extraction were: MRCON (UMLS metathesaurus file, 2004 version, 198,586,537 bytes in length), containing the terms for each UMLS concept (CUI). MRCXT (UMLS metathesaurus file, 2004 version, 8,347,732,946 bytes in length), containing the relationships for every UMLS code. All MRCXT records with a SNOMED-CT derivation were extracted using a Perl script [see Additional file 6 ]. All MRCON records containing a neoplasm term and having a SNOMED-CT origin were extracted using another Perl script [see Additional file 7 ]. Results and discussion Features of the taxonomy The taxonomy currently contains 122,632 different terms encompassing 5,376 neoplasm concepts (neocl.xml). Each concept has, on average, 23 synonymous terms. The second-largest source of tumor names is contained in the licensed 2004 version of the Unified Medical Language System Metathesaurus (UMLS), which draws neoplasm terms from the National Cancer Institute Thesaurus and SNOMED-CT. SNOMED-CT incorporates the International Classification of Disease – Oncology) [ 6 , 12 , 13 , 19 , 20 ]. The UMLS contains 24,593 unique English neoplasm terms with a specific "neoplasms" relationship, about one fifth the number of terms contained in the taxonomy. However, when one counts the terms in UMLS that have ANY type of relationship to neoplasia, the number UMLS-derived neoplasia terms expands to 64,601, about half of the number of terms contained in the taxonomy. It is difficult, if not impossible, to determine a correct number of neoplasm names contained in UMLS. The reason is that in UMLS, a concept may have many different relationships. For instance, the UMLS concept for "abdominal pain" has 805 relationships. Among these are: colic, constipation, diarrhea, influenza-like symptoms, malaise, multiple organ failure AND GI neoplasm benign, GI neoplasm malignant. The last two items are relationships to neoplasms. These relationships are valid because abdominal pain can be associated with benign or malignant neoplasms. Although the Perl script ca_mrrec.pl [see Additional file 5 ] outputs over 64,000 terms with neoplasm relationships, many of these terms are not names of neoplasms. If all 64.601 terms were reviewed to determine which were valid tumor names, a subjective number would be obtained that would certainly differ with each reviewer. It is probably fair to say that the number of UMLS neoplasm terms is somewhere between 24,593 (terms with a specific "neoplasms" relationship) and 64,601 (terms with any kind of relationship to neoplasms). Similarly, the number of SNOMED-CT terms with a neoplasm relationship is 35,920. This is the number of UMLS records with a SNOMED-CT derivation and with any type of neoplasm relationship [see Additional file 7 ]. This should be considered an upper limit estimate and is less than a third of the number of neoplasm terms included in the taxonomy. In general, taxonomy concepts that were highly generic (such as adenocarcinoma of lung) and the concepts that were the most highly pre-coordinated (i.e. multi-word terms with modifiers) had the greatest numbers of synonymous representations. For example, consider these 48 synonyms for adenocarcinoma of the lung: adenoca arising from lung, adenoca arising in lung, adenoca of lung, adenocarcinoma arising from lung, adenocarcinoma arising from pulmonary, adenocarcinoma arising from the lung, adenocarcinoma arising from the lungs, adenocarcinoma arising in lung, adenocarcinoma arising in pulmonary, adenocarcinoma arising in the lung, adenocarcinoma arising in the lungs, adenocarcinoma of lung, adenocarcinoma of pulmonary, ca arising from lung, ca arising from lungs, ca arising in lung, ca arising in lungs, ca of lung, ca of lungs, cancer arising from lung, cancer arising from lungs, cancer arising in lung, cancer arising in lungs, cancer of lung, cancer of lungs, carcinoma arising from lung, carcinoma arising from lungs, carcinoma arising in lung, carcinoma arising in lungs, carcinoma of lung, carcinoma of lungs, lung adenoca, lung adenocarcinoma, lung ca, lung cancer, lung with adenoca, lung with adenocarcinoma, lung with ca, lung with cancer, lung with carcinoma, lungs with ca, lungs with cancer, lungs with carcinoma, pulmonary adenoca, pulmonary adenocarcinoma, pulmonary ca, pulmonary cancer, pulmonary carcinoma Commonly occurring lesions have many different representations. The taxonomy contains closely-related but non-synonymous concepts as separate entries (e.g. there are 29 synonyms for bronchogenic carcinoma, and 15 synonyms for bronchioloalveolar adenocarcinoma) Other items in the taxonomy that have multiple term-variants are the so-called pre-coordinated terms characterized by modifying phrases. These terms are the hardest to capture in a taxonomy, and seem to return the smallest value on the effort. For instance, the taxonomy contains 118 synonyms for "testis with mixed embryonal carcinoma and endodermal sinus neoplasm with seminoma." The large number of synonyms are the direct result of the many different ways that modifying phrases can be ordered and combined to create the same terms. A few examples of the 118 synonyms for this term are: mixed embryonal cancer and endodermal sinus neoplasm with seminoma arising in testis mixed embryonal cancer and endodermal sinus neoplasm with seminoma of testis testis with mixed embryonal cancer and endodermal sinus tumor with seminoma mixed embryonal cancer and endodermal sinus tumor with seminoma arising in testis testis with mixed embryonal cancer and yolk sac neoplasm with seminoma mixed embryonal cancer and yolk sac neoplasm with seminoma arising in testis mixed embryonal cancer and yolk sac neoplasm with seminoma arising from testis mixed embryonal cancer and yolk sac neoplasm with seminoma of testis testis with mixed embryonal cancer and yolk sac tumor with seminoma The taxonomy database is distributed as an XML or as a flat-file. A short excerpt from the XML file is shown: <mesoderm> <name nci-code = "C3731000">mesoblastic nephroma</name> <name nci-code = "C3731100">cellular mesoblastic nephroma</name> <mesoderm> indicates a class tag. Beneath it are two different terms and concepts. Each is given a unique concept number. The second term is similar to the first term, but not identical. Both code numbers share the first 4 digits. The flat-file version of the taxonomy lists these two terms as line-records. Each line record contains the term name, the term code, and the ancestry of the term within the developmental lineage classification. Each ancestor is separated by its predecessor by an arrow character. mesoblastic nephroma\|C3731000\|mesoderm>non_primitive> embryonic>neoplasms>tumor_classification> cellular mesoblastic nephroma\|C3731100\|mesoderm>non_primitive> embryonic>neoplasms>tumor_classification> Medical informaticians dream of the day when all medical data will be captured by computers in a highly structured format that ensures data uniformity. In this utopian vision, only canonical forms of medical terms will be used. Medical reports will have a uniform format, and will be computer parsable and human readable. Taxonomies will be small. Unfortunately, the current trend in medical reporting seems to favor unstructured narrative data entry. Personally, I can remember the early days of computers when data storage and memory constraints were at a premium. Years were entered as two-digit values (nobody worried about Y2K back in the 60s), and entry-words were selected from lists and typically represented by a single digit. Today, the storage and transmission of textual data are non-issues. Large vocabularies of millions of terms can reside in active memory. Physicians prefer narrative text over structured text [ 21 ], and most of the medical data entered by physicians appears in the form of free-text emails, memoranda, progress notes, hospital reports of every type, research publications, etc. Free expression results in a seemingly unlimited way of describing a single thought, and large taxonomies are sometimes useful tools for organizing and retrieving the many terms found in narrative free-text. The neoplasm taxonomy was created to collect all ways of expressing the names of all human tumors. The purpose of the taxonomy is to make it possible for computer algorithms to index textual information about all tumors regardless of the terms used to describe particular tumors. At first inspection, this may seem like a hopelessly complex and ultimately futile endeavor. Can anyone seriously hope to make sense of narrative text? Won't the taxonomy become larger and more complex as additional clinical, genomic, and proteomic modifiers split tumors into incomprehensible subcategories? Actually, the purpose of a taxonomy is to reduce the complexity of the knowledge domain. By focusing efforts on a relatively small area of medicine, it is feasible to create a product that does not exceed the limits of a single expert's mental capacity. The large number of terms contained in the taxonomy (122,632), is encompassed by just 5,376 concepts. We can parse through text, replacing the many different variants of a term with either a single concept code or with a preferred (so-called canonical) synonym. This means that the taxonomy gives us a method of reducing any document index of neoplastic terms down to a maximum of 5,376 canonical terms. In addition, the 5,376 concepts in the taxonomy are represented by 39 different ancestral classes. The developmental lineage classification of neoplasms is constructed using strict rules: no multi-class inheritance; each subclass endowed by the properties of its ancestor. This means that once we have identified a term, we can easily determine its place among just a few dozen classes, and its ancestry should tell us basic information about the biology of the tumor [ 1 ]. The taxonomy for the developmental lineage classification of neoplams is the largest neoplasm taxonomy. It should come as no surprise that large nomenclatures provide better coverage of textual terms than smaller nomenclatures. A 1997 study by Humphreys et al., showed that by combining controlled vocabularies, the UMLS provided substantially more exact matches to free-text terms than any individual vocabulary in the nomenclature [ 22 ]. Finding new ways of expanding terminologies is an active research area [ 18 ]. The current taxonomy is large and has benefited from the use of Perl scripts that validate the database for internal sense and consistency. Conclusions This manuscript describes a comprehensive taxonomy of neoplasia that collects synonymous terms under a unique code number and assigns each tumor to a single class within a tumor hierarchy. The entire classification and taxonomy are available as open access documents (in XML and flat-file formats) with this article. The taxonomy will be merged into the U.S. National Cancer Institute's Thesaurus, a curated, publicly available nomenclature and ontology that includes neoplasms and cancer-related terminology. Competing interests The author(s) declare that they have no competing interests. Authors' contributions This work represents the opinions of the author and does not represent the policy of the NIH or of any other U.S. Federal Agency. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 Neoplasia classification structure (XML version) Neoclxml.gz is a compressed (gzipped) XML file. The downloaded file should be renamed neoclxml.gz so that the .gz suffix can be recognized by unzip utilities. Unzip the file (using a free, open source utility such as gunzip.exe [23], or a proprietary utility such as Winzip). Once unzipped, the file should be renamed neocl.xml, so that it will have an .xml suffix. If the file is too large for viewing on your web browser, it can be viewed on plain-text word processors. Click here for file Additional File 2 Neoplasia classification with taxonomy (flat-file plain-text version) Neoself.gz is a compressed (gzipped) ascii flat-file. If the filename is changed during download, it should be renamed neoself.gz so that the .gz suffix can be recognized by unzip utilities. Unzip the file (using an open source utility such as gunzip.exe [23], or a proprietary utility such as Winzip). Once unzipped, the file is 19+ Mbytes in length. The expanded file should be renamed neoself.txt. Click here for file Additional File 3 Taxonomy validating Perl script The validating Perl script is xmlvocab.pl. Perl scripts will execute on any computer with a Perl interpreter. It requires the external taxonomy file named "neocl.xml" residing in the same subdirectory as xmlvocab.pl. Click here for file Additional File 4 Perl script for transforming taxonomy XML file to a plain-text flat file The Perl script neoself.pl transforms the XML database file (neocl.xml) to a flat file (neoself.txt). Click here for file Additional File 5 Perl script for extracting neoplasm codes and terms from UMLS The Perl script ca_mrrec.pl produces a file (neomrcxt.txt) containing all UMLS codes and terms with a neoplasm relationship. This script requires the external files MRCXT and MRCON (available at no cost from the National Library of Medicine) to reside in the same directory as ca_mrrec.pl. This script may take more than one-half hour to execute. Click here for file Additional File 6 Perl script for extracting UMLS codes for SNOMED-derived terms The Perl script snomout.pl produces a file (snomout.txt) containing all UMLS terms with a SNOMED-CT vocabulary relationship It requires the external file MRCXT (available at no cost from the National Library of Medicine) to reside in the same directory as snomout.txt. It produces an output file, snomout.txt that has a size of 1,895,054,040 bytes. This script may take more than one-half hour to execute. Click here for file Additional File 7 Perl script for extracting neoplasm concepts and terms from the SNOMED-CT subset of UMLS The Perl script ca_snrec.pl produces a file (neosnom.txt) containing all UMLS terms derived from SNOMED-CT having a neoplasm relationship. It requires the external file MRCON (available at no cost from the National Library of Medicine) and snomout.txt (produced by snomout.pl) [see Additional file 6 ], both residing in the same directory as ca_snrec.pl. This script may take more than one-half hour to execute. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535937.xml
515371	Fly Fights with Both Hands	null	Defending against attack is one of the most important challenges facing any organism. But while sticks and stones may break the bones of a lion, microscopic threats such as bacteria require different weapons. And it's not just we humans who have this problem—insects are prey to bacterial infections too. Their immune systems, however, rely on a far simpler set of defenses than those found in mammals. Exactly how one insect immune system recognizes bacteria, and how it fights off the invader, is the subject of a new study in this issue by Johann Deisenhofer and colleagues. The fruitfly, Drosophila , has long been known to use a set of molecular sentries called “peptidoglycan recognition proteins,” or PGRPs, that circulate in the fly's bloodstream. When a PGRP recognizes a bacterial invader, it triggers a cascade of events whose ultimate product is a group of antimicrobial compounds that attack and kill the bacteria. While the family of PGRPs has been extensively studied, exactly how they recognize their target bacteria has been less clear. At the cellular level, recognition requires contact, and the part of the bacterium the PGRP recognizes is, as its name implies, the peptidoglycan. A peptidoglycan is a special sort of molecular polymer found primarily on bacterial cell walls. Peptidoglycan forms when chains of sugar molecules (the glycans) are cross-linked by amino acids (the peptides) to form a meshwork that helps keep the bacterium from bursting under the osmotic strain of its contents. There are several types of peptidoglycans that differ in their precise sugar and amino acid constituents and in their ability to trigger the Drosophila defensive reaction. Deisenhofer and colleagues set out to determine whether this difference in triggering ability of particular peptidoglycans was linked to differences in the PGRPs that recognize them. To do this, they determined the three-dimensional structure of one PGRP, called PGRP-SA. They worked out not only the overall shape of PGRP-SA, but also which amino acids sat where on the convoluted surface of the protein. What they found on that surface was an extended groove down one entire side of the protein. To test whether this groove was indeed the recognition site for peptidoglycan, the group introduced a series of mutations to critical amino acids along the groove, testing each new form for its ability to bind peptidoglycan. Indeed, the binding and defense-triggering ability was worse for almost every mutant, demonstrating conclusively that the normal protein uses the groove to bind and recognize peptidoglycan. Then the team made a surprising discovery. They found that when PGRP-SA comes in contact with bacterial peptidoglycan, it begins to cleave the links between amino acids in the peptide portion of the peptidoglycan. This in itself is not so amazing—animals make plenty of peptide-cleaving proteins. But this protein has a difference, one which makes it unique in the animal kingdom. Stick model of the PGRP-SA residues chosen for mutational analysis To understand this difference, consider your two hands. They are mirror images of each other, alike yet not the same. No amount of twisting and turning will allow you to superimpose one exactly on the other—if you align the fingers, the knuckles will point in opposite directions, and if the knuckles point the same way, the fingers are all mismatched. This type of relationship between mirror images, called chirality (from the Greek for “hand”), is found in amino acids as well, a result of the three-dimensional geometry that radiates from their central atom. All the amino acids used by all known animal species are exclusively of the “left-handed” form, and the protein-digesting enzymes we make are designed specifically for these L-amino acids. Bacteria, however, link left-handed and right-handed amino acids together to form peptidoglycan. What Deisenhofer's team discovered was that unlike any other known animal enzyme, the Drosophila PGRP-SA was able to break apart this “L,D” (levo-dextro) linkage, making it, in their words, “the first eukaryotic protein exhibiting such an activity specific for peptide bonds existing only in prokaryotes.” What does it all mean? Deisenhofer and colleages' results are yet another demonstration that at the molecular level, understanding structure is the key to understanding function. They also show that when it comes to defense, it helps to be able to fight with both hands.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC515371.xml
524184	Distance, rurality and the need for care: access to health services in South West England	Background This paper explores the geographical accessibility of health services in urban and rural areas of the South West of England, comparing two measures of geographical access and characterising the areas most remote from hospitals. Straight-line distance and drive-time to the nearest general practice (GP) and acute hospital (DGH) were calculated for postcodes and aggregated to 1991 Census wards. The correlation between the two measures was used to identify wards where straight-line distance was not an accurate predictor of drive-time. Wards over 25 km from a DGH were classified as 'remote', and characterised in terms of rurality, deprivation, age structure and health status of the population. Results The access measures were highly correlated (r 2 >0.93). The greatest differences were found in coastal and rural wards of the far South West. Median straight-line distance to GPs was 1 km (IQR = 0.6–2 km) and to DGHs, 12 km (IQR = 5–19 km). Deprivation and rates of premature limiting long term illness were raised in areas most distant from hospitals, but there was no evidence of higher premature mortality rates. Half of the wards remote from a DGH were not classed as rural by the Office for National Statistics. Almost a quarter of households in the wards furthest from hospitals had no car, and the proportion of households with access to two or more cars fell in the most remote areas. Conclusion Drive-time is a more accurate measure of access for peripheral and rural areas. Geographical access to health services, especially GPs, is good, but remoteness affects both rural and urban areas: studies concentrating purely on rural areas may underestimate geographical barriers to accessing health care. A sizeable minority of households still had no car in 1991, and few had more than one car, particularly in areas very close to and very distant from hospitals. Better measures of geographical access, which integrate public and private transport availability with distance and travel time, are required if an accurate reflection of the experience those without their own transport is to be obtained.	Background The UK National Health Service has always aimed to provide health care for all. Although the importance of "fair access for all" (independent of the ability to pay, age, sex or area of residence) has been highlighted in recent policy documents [ 1 ], the meaning of 'fair access' is still debated [ 2 ]. Although there will always be variations in geographical access to health services, the extent of such variations and the relationship between distance to health services and the need for health care is unclear. If policy makers are to address inequities of access, more understanding is needed both of appropriate methods for measuring access and of the relationship between access to health services and health. Although 'fair access' can be characterised simply as 'providing the right service at the right time in the right place'[ 3 ], it is a complex concept covering the provision of services, the knowledge and opportunity to use them, and the measurement of need [ 4 ]. In the UK mergers of hospital trusts have highlighted tensions between the perceived safety, effectiveness and efficiency of larger specialist centres and the demand for more geographically accessible local care [ 5 , 6 ], revealing the lack of evidence on which to base decisions [ 7 ]. Geographical access – the distance which must be travelled in order to use health services – is one aspect of access which is often overlooked [ 2 ], but which presents barriers of cost, time and inconvenience. Although there is some evidence that increasing distance from health services inhibits the use of primary [ 8 ] and secondary care [ 9 ], and that it is associated with a range of poor health outcomes, from higher than expected numbers of deaths from asthma to lower than expected five year survival from cancer [ 10 , 11 ], few studies have attempted to quantify or set thresholds of poor access [ 12 , 13 ]. Furthermore, measures of geographical access can be difficult to compare. Rurality has often been used as a proxy for inaccessibility [ 14 ], as have dichotomous categorisations such as the presence or absence of a service provider in an area [ 15 , 8 ]. More complex measurements such as the straight line distance between populations (i.e demand points) and health service providers [ 16 , 17 ], or 'network distances' (which can include both road distance and travel time) [ 8 ] have added complexity, but the relationship between these measures is not clear. One assumption which is commonly made is that geographical inaccessibility of health services is essentially a rural problem, but there is little evidence demonstrating the differences in accessibility between rural and other areas. In any area, the greatest disadvantage is likely to be experienced by individuals without access to a car (including members of one-car households without daytime access). With the declining availability of public transport, it is likely that a private car is the only convenient way to travel in rural Britain [ 18 ]. Although car ownership is relatively high in rural areas, rates for the poor, the elderly and for women are far lower than average: the 2001 Census reports that more than two thirds of single-pensioner households, many of which comprise single women, do not have access to a car. Distance may therefore be a further burden on groups with a particularly high need for health care, raising issues of inequity. Furthermore, if geographical access to health services is a problem for some groups outside of traditional rural areas, then rural policies alone will not tackle the problem. In this paper, we aim to determine the geographical accessibility of health services, and the demographic and health related factors associated with it. We compare two measures of geographical access: straight-line distance and modelled drive-time along the road network to primary and secondary care throughout South West England. We investigate whether the most geographically inaccessible populations are in rural areas, describe the relationship between geographical access to hospital and population health, and investigate whether the populations furthest from health services have a greater need for health care due to age or deprivation. The study area is the former South West Region, comprising the counties of Avon, Cornwall and the Isles of Scilly, Devon, Dorset, Gloucestershire, Hampshire, the Isle of Wight, Somerset and Wiltshire. As defined in 1991 this area has a population of about 6 million, with a low proportion from ethnic minorities, and a higher than average proportion living in rural areas. Results Correlation between the access measures The straight-line and drive-time measures were highly correlated for both GP and hospital services (figure 1 ). Areas where residuals from the regression analysis of straight-line distance and drive-time to DGHs are more than two standard deviations from the norm were concentrated in coastal and rural wards of the far South West. Areas where residuals are negative indicate faster than expected drive times, wheras positive residuals indicate that drive time is slower than predicted by straight line distances (figure 2 ). The analysis was repeated, excluding wards along the boundary between the study area and neighbouring counties to check for edge-effects, but there was no difference in results. Figure 1 Correlation between straight line and drive-time measures to GP and hospital services Figure 2 Standardised residuals from the regression of drive time and straight-line distance to hospitals Distances to health services Distances to GPs were low, with a median distance of just 1 km to the closest practice (IQR 0.6 – 2.2). The calculation was repeated excluding branch surgeries (which tend to have limited opening hours), but this made little difference to the outcome, with a median distance to a main surgery of just 1.2 km. 95% of wards (98% of the population) were under 4.4 km, or 6.3 minutes, from their closest GP. The maximum distance to a GP was just 9.4 km (13.7 minutes). The median distance to a DGH was just less than 12 km (IQR 5.4 – 19.0), with a maximum of 50 km, corresponding to an estimated 13 and 48 minutes drive-time (table 1 ). Table 1 Access to DGHs and GPs 25 th centile Popn (%)* Median Popn (%) 75 th centile Popn (%) 95 th centile Popn (%) Maximum Straight line (km) DGH 5.4 2.40 (39.3) 11.6 3.97 (65.1) 19.0 5.15 (84.3) 29.0 5.92 (97.1) 50.1 GP surgery 0.6 2.24 (36.8) 1.0 4.17 (68.3) 2.2 5.39 (88.4) 4.4 5.96 (97.7) 9.4 Drive time ('minutes') DGH 7.1 2.38 (38.9) 13.4 3.93 (64.4) 20.5 5.17 (84.7) 31.6 5.93 (97.2) 48.3 GP surgery 1.0 2.19 (35.9) 1.7 4.00 (65.5) 3.4 5.28 (86.5) 6.3 5.89 (96.5) 13.7 Population in millions (percent of the total population) living in wards within this distance of their closest DGH and GP Remoteness and rurality For the purposes of this study, remoteness from health services was defined as over 5 km from a GP or over 25 km from a DGH. Access to primary care was good, with just 91 wards (6.3% of the total) remote from primary care. These areas have just 3% of the regional population. Of these the majority (63%) were ONS 'rural' areas. There were 162 wards which we classified as remote from hospitals (11% of the total, home to 6.5% of the region's population). All had drive-times to hospital of over 21 minutes; 81 (51%) were urban by the ONS classification, 69 (43%) were rural areas and the remaining eight (5%) were rural fringe. Four wards had no ONS urban / rural classification (table 2 ). Table 2 ONS rurality and remoteness from hospital N (%) Rural Rural fringe Not rural No classification Total Remote 69 (43%) 8 (5%) 81 (51%) 2 (1%) 162 (100%) Not remote 184 (14%) 146 (11%) 950 (74%) 6 (0.5%) 1286 (100%) Distance and the need for health care Deprivation scores ranged from -6.2 to 9.9, with a mean of -1.0, indicating that the study area had a slightly more affluent profile than the England and Wales average. The most affluent wards were in the middle of the range of straight-line distances from secondary care. The median deprivation score was 1.0 in the decile of wards closest to hospitals, decreased to a low of -2.2 in the 5 th decile, then rose steadily to -1.0 in the decile of wards furthest from hospitals, giving a slight 'U' shape to the relationship between deprivation and distance from health services (figure 3 ). Figure 3 Townsend deprivation score for deciles of wards by straight-line distance from DGH The proportion of over 65 year olds increased slightly with straight-line distance from hospitals: more remote wards had a slightly higher proportion of residents over the age of 65, but there was considerable variation within deciles of remoteness, and the observed difference was small. The proportion of the population under five years old in 1991 showed no clear trend with ward distance from hospital, but was slightly lower in more remote wards (figure 4 ). Figure 4 Age structure of wards by straight-line distance from DGH. average proportion of young (under 5) and elderly (over 65) population for deciles of Wards by straight line distance from DGH The age-standardised rate of LLTI was highest in the areas closest to hospitals. The LLTI rate decreased with increasing distance from hospital and then increased again in the most remote areas. Standardised rates of premature mortality showed no strong pattern with distance from a hospital, although median rates were high in areas close to hospitals and also slightly raised in the most remote areas. The proportion of households with no car was highest in the areas closest to hospital, but increased again in the decile of wards furthest from hospital. The same pattern was seen in the ownership of two or more cars – the lowest rates were found in areas either very close to or very far from hospitals (table 3 ). Table 3 Median values for health outcomes and car ownership for deciles of ward by straight-line distance from DGH Closest 2 3 4 5 6 7 8 9 Furthest LLTI SMR (0–64) 1.08 1.02 0.92 0.84 0.85 0.89 0.85 0.87 0.89 1.02 All-cause SMR (0–64) 1.08 0.99 0.93 0.91 0.87 0.93 0.88 0.91 0.96 0.94 Proportion of households with No car 34.2 29.3 25.4 20.4 20.3 21.0 20.1 20.7 20.2 23.1 Two or more cars 20.0 23.2 27.8 32.1 33.2 32.0 32.8 32.5 31.2 27.0 Discussion The impact of distance on the use of hospitals and other health care, and on health status, has not been well established. In the UK threshold distances of between 24 and 50 miles to specialist hospital services [ 19 , 20 ], 10 miles to screening services [ 21 ], 7 km (4 miles) to family planning clinics [ 22 ] and 2.5 miles to primary care [ 23 ] have all been used in reporting 'poor access', but there is little consensus and no strong theoretical or empirical basis for these choices. By international standards, distances to health services for our study population are low, averaging just 12 km to the closest hospital, but drive-times to hospital of up to 50 minutes are predicted by our model and there are groups who could be considerably disadvantaged by the travel distances we have reported. A variety of measures of geographic access of varying complexity and specificity exist and selecting an appropriate measure is not simple. Straight-line distances are widely used, easy to calculate and to compare and, in this study, they are closely correlated with the more complex drive-times. However, there is some evidence that areas of low correlation are concentrated in peripheral areas of the rural South West. In these areas straight-line distances underestimate true travel distance, reflecting sparse road networks and geographical barriers such as hills, rivers and coastline. Access to health services in these areas could be misrepresented by the use of the simpler measure, masking problems faced by these populations. Furthermore, neither measure used here reflects the experience of those without access to a private car. Travel to hospital and GP appointments is already known to be a problem for some groups in rural areas of the UK. Although informal systems of 'lift-giving' and more formal 'voluntary taxi' schemes often exist [ 24 ] these are not available everywhere [ 25 , 26 ], and it could be argued that a measure of travel by public transport is vital in determining accessibility for the most disadvantaged populations. Few studies have attempted this [ 27 - 29 ]., and composite measures, which include both public and private transport, are even less common [ 30 ]. Better measures of access, which integrate private and public transport, are required to reflect the experience of those on low incomes, and without their own transport. A surprising finding of this study was the relatively low proportion of wards remote from health care which are defined as 'rural'. Fewer than half of the wards remote from hospital and under two-thirds of those remote from primary care are classified as rural by the ONS. Analysis which concentrates on rural areas under the ONS definition, or even stretches this to include 'rural fringe' areas, will still miss over half of the wards which are remote from hospitals. There has been concern over the targeting of resources in concentrations of deprivation: the majority of deprived people live outside of these areas and are not reached by narrowly focused initiatives. Although the ward-level definition of rurality used here may class as 'urban' some small towns which many would consider essentially 'rural' when viewed at a larger scale (such as the Local Authority level), we conclude that caution should be exercised when evaluating and responding to poor access to health services, a high proportion of which occurs outside areas traditionally considered to be remote. In this study, we found no clear threshold at which need becomes greater or health status sharply declines. If anything, the converse is true with worse health status and greatest need in areas close to health services. Distance to health care was not associated with a high proportion of elderly or very young residents, but was related to deprivation. We found high deprivation in areas close to hospitals, relative affluence in more distant areas and an increase in deprivation in the most remote wards. Deprivation indices have been criticised for failing to represent deprivation in rural areas [ 31 ] and the relatively high proportion of rural areas in the most remote wards may hide even higher need in these areas. Further research using different measures of need and deprivation is indicated. Although the highest rates of morbidity and mortality were found in the areas closest to hospitals, there was some evidence of increasing rates in more remote areas. Rates of LLTI, particularly for those under 64, show an upwards trend in more remote areas. This supports previous findings that LLTI is higher in rural wards with the most dispersed populations [ 31 ], but it is not clear whether this reflects a true increase in morbidity or a perception of handicap of those living in such areas. The relationship between distance and all-cause premature mortality is less clear. The high levels of mobility which are often reported in populations living far from services were upheld by our study (expressed through high car ownership), but the areas most remote from hospitals begin to show a decrease in levels of car ownership. It is unlikely that this indicates less need for private transport, and may indicate a less wealthy population for whom travel is a potential problem. There are a number of important limitations to our study. We have explored only one region of England, a relatively affluent area with a very small ethnic minority population and an unusual 'peninsular' geography. Our findings need to be reproduced in other areas. We have limited our definition of access to simple geographical measures. Other aspects of accessibility include the quantity and quality of health services, and financial and cultural barriers to their use, and have not been explored here. The choice of SMRs and LLTI rates as health outcome indicators may have resulted in our inability to observe stronger relationships between geographical access and health: even over a six year period absolute numbers of deaths were low. More research is needed including the young and elderly and using a wider range of health status measures. Finally, the inter relationship between use of health care, need and access has been insufficiently explored. Conclusions This paper has provided a population-based estimate for access to both primary and secondary health care in South West England. We have shown that although geographical access to health services is generally good, remoteness from health services is an issue which affects both urban and rural areas. Studies concentrating purely on rural areas are therefore likely to underestimate the extent of geographical barriers to accessing health care. Areas which were furthest from hospitals did not have an especially old or young population, but there was some evidence of higher rates of LLTI and of deprivation in the most remote wards, indicating higher need for services in the areas furthest from them. Finally, the fact that almost a quarter of households in the decile of wards most remote from hospital services had no car in 1991 indicated a large number of people for whom travel is likely to be more difficult than implied by current measures of geographical access. Our understanding of the effect of distance on the use of services and on health outcomes is far from complete. Both the measurement of access and the understanding of need and deprivation require further exploration. The development of web-based public transport information systems may supply the data needed to enhance currently available measures of access by adding public transport travel times, likely to be relevant to access for the poorest and most deprived populations and the introduction of the Indices of Multiple Deprivation 2000 in England may present a clearer picture of the need for health care than traditional census-based indices [ 32 ]. This index contains a measure of geographical access to services, which has been of particular interest to rural populations and may provide a missing dimension to the measurement of deprivation. Linking geographical access with a wider range of health status measures and health care use in different populations is also vital if a clear picture of the impact of accessibility of health care is to be fully understood. Methods Measuring geographical access to health services A Geographical Information System (Arc/Info) and custom written programs were used to calculate two measures of access to health services. Access was calculated from all residential postcodes to primary care services (all main and branch General Practice (GP) surgeries) and secondary care services (acute hospitals (DGHs)) in the region. Data on main and branch GP surgeries (n = 1469) were obtained from all Family Health Services Authorities in 1998. Acute DGHs (n = 39) were defined as hospitals with general medicine and general surgery facilities and an Accident and Emergency department. DGHs were identified using the hospital year-books (1992–97) and hospitals were contacted to clarify their status in 1997 as necessary. The first access measure calculated was a widely used measure: the shortest straight-line distance between every residential postcode, the closest GP (both main and branch) and the closest DGH. A more complex measure of access was the shortest drive-time from each residential postcode to the closest GP and the closest DGH. This was modelled using estimated road-network travel speeds along the Bartholomew digital road network and associating these with residential postcode locations by the use of a travel time (drive-time) surface model. While including provision for congestion and slow travel through urban areas, this measure does not include any estimates for parking times or transfers between car and surgery or hospital. The methods are described in detail elsewhere [ 33 ]. The need for health care Proxy measures of the need for health care were calculated. The Townsend score, a widely used indicator of material deprivation, was calculated from 1991 census data. The variables used in the score are the percentage of economically active people over the age of 16 who are unemployed; the percentage of households which are overcrowded; the percentage of households with no car and the percentage of households not owning their own home. A log transformation is applied to the overcrowding and unemployment variables. The logged variables and the car ownership and owner occupation variables are standardised by creating z-scores for each value, and the four z-scores are summed to provide the final Townsend score. Scores are standardised to give a mean of zero for England and Wales: any scores greater than zero indicate relative deprivation, any less than zero represent relative affluence. The proportions of the population over 65 and under 5 years old – were taken from the 1991 Census Small Area Statistics (SAS). Health status was assessed using indirectly standardised rates of all-cause mortality and Limiting Long Term Illness (LLTI) for all those under 65 (premature mortality and morbidity). Data on LLTI were taken from the 1991 Census. The Office for National Statistics (ONS) provided data on all-cause mortality for the years 1991–1996. Data were aggregated over the six years, to minimise problems due to small numbers of cases in some wards. Assigning data to geographical areas Postcodes were allocated to 1991 Census wards using the 1991 and subsequent postcode to enumeration district directories. Travel times and distances were calculated for all residential postcodes and aggregated to ward level for analysis. The resident population of each ward was used to weight individual postcode times and distances to create a population-weighted average, as demonstrated in table 4 . ONS ward classifications were used to select 'rural' wards [ 34 ]. The ONS classifications are listed in table 5 . Only two categories: 'rural areas' and 'rural fringe', are unambiguously rural under this definition and we have defined than as rural here. All other wards were defined as urban. Table 4 Aggregating household level access data to wards Ward Postcode N Households from each postcode in Ward1 Time from PC to health services Households * time Ward1 PC1 10 10 100 Ward1 PC2 7 13 91 Ward1 PC3 2 11 22 Ward1 PC4 6 21 126 Sum (Ward1) 25 339 Population weighted average time for Ward1 ((hhds*time)/hhs) Table 5 The ONS ward classification ONS group Rural / urban classification Suburbia Urban Rural areas Rural Rural fringe Rural Industrial areas Urban Middling Britain Urban Prosperous areas Urban Inner city estates Urban Established owner occupiers Urban Transient populations Urban Metropolitan professionals Urban Deprived city areas Urban Lower status owner occupiers Urban Mature populations Urban Deprived industrial areas Urban Analyses To investigate if straight-line distance was a valid proxy for the more complex drive-time measure, the two were compared using Pearson correlation coefficients and a regression analysis of drive-time against straight-line distance. Areas where straight-line distance appeared to underestimate the drive-time more than expected were identified and mapped to investigate the extent of geographical clustering. Access to primary and secondary health services was described using median distances and inter-quartile ranges for both measures. To investigate the assumption that it is the residents of rural areas who are most disadvantaged by poor geographical access to health services, we first had to define poor access. Standard estimates of 'remoteness' from health services have not been established – there is no a priori definition of the distance regarded as 'remote from health services' and no consensus has been established in the literature on access to health services. The proportion of rural, rural fringe and urban wards which were 'remote' from health services under the definition of a straight-line distance of three, five or seven kilometres to a GP and 20, 25, 30 or 35 km to a hospital was therefore calculated (Table 6 ). We used an arbitrary cut-off point of a straight-line distance of 5 km to a GP and 25 km to a hospital, beyond which wards were classed as 'remote' from health services. These distances classified approximately 6% of the study population as remote from secondary care and 3% as remote from primary care. Table 6 ONS rurality and remoteness from primary and secondary care Rural Rural fringe Urban No classification Total wards All wards 253 (18%) 154 (11%) 1031 (71%) 10 (1%) 1448 (100%) GPs Remote (3 km) 117 (53%) 14 (6%) 84 (38%) 6 (3%) 221 (100%) Remote (5 km) 20 (53%) 4 (10%) 12 (32%) 2 (5%) 38 (100%) Remote (7 km) 5 (71%) 1 (14%) 0 (0%) 1 (14%) 7 (100%) Hospitals Remote (20 km) 126 (39%) 36 (11%) 158 (49%) 4 (1%) 324 (100%) Remote (25 km) 69 (43%) 8 (5%) 81 (51%) 2 (1%) 162 (100%) Remote (30 km) 30 (49%) 1 (2%) 28 (46%) 2 (3%) 61 (100%) Remote (35 km) 17 (59%) 0 (0%) 12 (41%) 0 (0%) 29 (100%) We then identified the proportion of remote wards that were rural under the ONS classification. To investigate relationships between distance to health services and the need for health care, straight-line distance to hospital was used to group wards into deciles and the deprivation score and the age profile of the population in each decile was described. Standardised rates for premature all-cause mortality and LLTI were used to indicate health outcomes for each decile of wards, and car ownership (as reported in the 1991 census) was used to indicate how easy travel would be for the population in each group. Authors' contributions HJ carried out the analyses and drafted the manuscript. HJ, PR, and DM collaborated in the formation of the research questions, the management of the study and the development of the paper. HJ and SB collected the data and calculated measures of deprivation, health and accessibility. DM designed the drive-time access measure. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524184.xml
544880	Successful recovery of infective endocarditis-induced rapidly progressive glomerulonephritis by steroid therapy combined with antibiotics: a case report	Background The mortality rate among patients with infective endocarditis, especially associated with the presence of complications or coexisting conditions such as renal failure and the use of combined medical and surgical therapy remains still high. Prolonged parenteral administration of a bactericidal antimicrobial agent or combination of agents is usually recommended, however, the optimal therapy for infective endocarditis associated with renal injury is not adequately defined. Case presentation Patient was a 24-years old man who presented to our hospital with fever, fatigue, and rapidly progressive glomerulonephritis. He had a history of ventricular septum defect (VSD). A renal biopsy specimen revealed crescentic glomerulonephritis and echocardiogram revealed VSD with vegetation on the tricuspid valve. Specimens of blood demonstrated Propionibacterium Acnes. The intensive antibiotic therapy with penicillin G was started without clinical improvement of renal function or resolution of fever over the next 7 days. After the short-term treatment of low dose of corticosteroid combined with continuous antibiotics, high fever and renal insufficiency were dramatically improved. Conclusion Although renal function in our case worsened despite therapy with antibiotics, a short-term and low dose of corticosteroid therapy with antibiotics was able to recover renal function and the patient finally underwent tricuspid valve-plasty and VSD closure. We suggest that the patients with rapidly progressive glomerulonephritis associated with infective endocarditis might be treated with a short-term and low dose of corticosteroid successfully.	Background Infective endocarditis has been classified as acute or subacute-chronic based on the clinical presentation and often presents extracardiac findings such as fever, anorexia, weight loss, malaise, and night sweats [ 1 ]. The prognosis of infective endocarditis has been shown to be strongly influenced by the complication of congestive heart failure and stroke [ 1 ]. Furthermore, glomerulonephritis, especially rapidly progressive glomerulonephritis, is also one of the complications associated with poor prognosis [ 2 ]. Infective endocarditis-induced rapidly progressive glomerulonephritis is treated with antibiotics alone, but it sometimes results in end-stage renal failure [ 2 ]. Although effective strategies to treat rapidly progressive glomerulonephritis have not been established, steroid therapy, immunosuppressive therapy, and plasmapheresis in addition to antibiotic therapy has been shown to be beneficial [ 3 ]. Here, we report a case of rapidly progressive glomerulonephritis associated with infective endocartitis in which the clinical symptoms were successfully improved by the treatment with short-term steroid therapy. Case Presentation A 24-year old man was admitted to our hospital because of macrohematuria and general malaise, along with insidious deterioration of renal function. The patient had been diagnosed as having ventricular septum defect (VSD) without complications and had been well six months before admission, when the patient presented a temperature of 38.0°C and a productive cough. One month before admission, the patient was admitted elsewhere because of fever, general malaise, and macrohematuria. The temperature was 38.6°C, the pulse was 120 beats per minute, and respirations were 24 times per minute. The blood pressure was 130/58 mmHg. On physical examination, the patient appeared acutely ill and pansystolic murmur of Levine III/IV was noted at the 4th left sternal border (LSB). The urine was positive for hematuria (+++) and protein (+++); the sediment contained 150–160 RBC/hpf and 10–15 granular casts/hpf. Laboratory test revealed 2.04 mg/dl of serum creatinine, 33.3 mg/dl of BUN, and 9.0 g/dl of hemoglobin. Echocardiogram demonstrated VSD and vegetation on the tricuspid valve associated with regurgitation. The patient was transferred to our hospital for evaluation of renal dysfunction. On the first hospital day, the temperature was 39.4°C and the physical examination demonstrated pansystolic murmur at the 4th LSB again, but no lymphoadenopathy and no localizing signs for a focus of infection. Laboratory findings were: anemia (Hb 8.6 g/dl, Ht 26.5%), presence of d-dimer (10.3 mg/ml), BUN 41 mg/dl, serum creatinine 2.49 mg/dl, proteinuria 0.96 g/24 h, 24 h Ccr 30.0 ml/min, hypocomplementemia (C3 29 mg/dl, CH50 16.5 U/ml), positive for inflammatory sign (CRP 2.8 mg/dl, ESR 104 mm/h), and positive for cryoglobulinemia (Table 1 ). Antibodies for DNA, RNP, Sm, and myeloperoxidase and proteinase 3 ANCA were normal (Table 1 ). Echocardiogram again showed vegetation on tricuspid valve and TR (Figure 1 ), suggesting the presence of right-sided bacterial endocarditis. However, there were no histories of tooth extraction, skin injury, or drug addiction and serology for viruses including HIV-1 was negative. Specimens of blood were obtained 5 times for culture and demonstrated Propionibacterium Acnes. A renal biopsy specimen showed typical crescentic glomerulonephritis (Figure 2A ) with interstitial inflammatory cell infiltration on PAS staining. Ten of eighteen glomeruli had cellular crescents without fibrocellular and fibrous crescents. Immune reactants such as C3 (Figure 2B ) and Ig M were found in peripheral capillary walls and in the mesangium. Table 1 Laboratory Values on Admission Hematocrit (%) 26.5 IgG (mg/dl) 2707 Hemoglobin (g/dl) 8.6 IgM (mg/dl) 768 White cells (per mm 3 ) 6100 IgA (mg/dl) 449 Total protein (g/dl) 6.8 C3 (mg/dl) 29 Total bilirubin (mg/dl) 0.3 C4 (mg/dl) 35 Aspartate aminotransferase (U/liter) 29 Rheumatoid factor (U/ml) 483 Alanine aminotransferase (U/liter) 29 Anti-dsDNA (U/ml) negative Lactate dehydrogenase (U/liter) 666 Anti-GBM (U/ml) negative Alkaline phosphatase (U/liter) 75 MPO-ANCA titer negative Creatine phosphokinase (U/liter) 34 Urea nitrogen (mg/dl) 41 Creatinine (mg/dl) 2.49 Urinary sediment Sodium (mmol/liter) 137 Erythrocytes +++ Potassium (mmol/liter) 4.7 Leukocytes - Calcium (mg/dl) 8.1 Cylinders + C-reactive protein (mg/dl) 2.8 Figure 1 Vegetation on tricuspid valve by echocardiography. Arrow denotes the vegetation. Figure 2 Crescentic glomerulonephritis induced by infective carditis on PAS staining and IF. A. PAS staining demonstrated circumferential and cellular crescent formation with interstitial nephritis. B. IF demonstrated C3 positive staining in mesangial area. Bacterial endocarditis complicated with rapidly progressive glomerulonephritis was diagnosed and the intensive antibiotic therapy with penicillin G was started, without clinical improvement of renal function (on day 7 serum creatinine 6.21 mg/dl) or resolution of fever over the next 7 days. The initial antibiotics were replaced with ampicillin and imipenem in addition to low dose of corticosteroid treatment initiated with intravenous methylprednisolone 0.5-g per day for 3 consecutive days followed by oral prednisolone 30-mg, 20-mg, and 10-mg per day each for 3 days, respectively. On day 16, laboratory test demonstrated normalization of inflammatory signs (CRP 0.5, ESR 11 mm/h), serum complements, and circulating immune complex and negative for blood culture and cryoglobulin. After the short-term treatment of low dose of corticosteroid combined with continuous antibiotics, high fever and renal insufficiency were dramatically improved (Figure 3 ). The antibiotic therapy lasted for 40 days until the day of the tricuspid valve-plasty and VSD closure. The patient was finally able to undergo the surgical operation, resulting in successful recovery from endocarditis-induced rapidly progressive glomerulonephritis. He is now well and laboratory test showed normal serum creatinine 0.72 mg/dl, while echocardiogram demonstrated mild regurgitation of tricuspid valve. Figure 3 Clinical course. PCG; penicillin G, ABPC; ampicillin, IPM; imipenem Discussion Therapy with a bactericidal antimicrobial agent or combination of agents is usually effective [ 1 - 3 ], although in some cases antibiotic therapy fails, resulting in end-stage renal failure requiring dialysis therapy. Here, we present a patient complicated with VSD who developed rapidly progressive glomerulonephritis accompanying right sided-subacute bacterial endocarditis caused by Propionibacterium acnes. Although Propionibacteium acnes is considered to be contaminant, it has been found to be a pathogen of infective endocarditis in spite of its weak virulence [ 4 ]. Furthermore, case-reports of shunt nephritis associated with Propionibacterium acnes were also reported [ 5 - 7 ]. Membranoproliferative glomerulonephritis is the lesion most frequently seen in shunt nephritis, but in some patients in whom untreated and inadequately treated bacteremia persists, mild renal involvement may progress to the development of severe impairment such as crescents and sclerotic glomeruli, possibly through the prolonged immune-mediated pathogenesis [ 8 ]. In the present case, the prolonged exposure to the weak pathogen resulted in the development of crescentic glomerulonephritis in association with circulating immune complexes and cryoglobulinemia. Moreover, in the present case, the antibiotic therapy alone was only able to suppress circulating bacteremia, but failed to decrease the size of vegetation and the nest of bacteria. However, the clinical improvement of our case was thought to be a delayed response to continued antibiotic therapy and the addition of anticoaglants [ 3 , 9 , 10 ]. Some case reports also showed that immunosuppressive therapies such as plasmapheresis, cyclophosphamide, and azathioprine with antibiotics could recover renal dysfunction of infective endocarditis-induced crescentic glomerulonephritis [ 3 , 11 ]. Recently, a short-tem and low dose of anti-inflammatory corticosteroid has also shown to be potentially effective in reducing the risk of death in patients with sepsis [ 12 , 13 ]. In conclusion, we suggest that patients with rapidly progressive glomerulonephritis associated with infective endocarditis might be treated with a short-term and low dose of corticosteroid successfully, in the case presenting the clinical and biological evidence of immune-mediated pathogenesis with the prolonged duration of the illness. Competing interests The author(s) declare that they have no competing interests. Authors' contributions Koya D and Shibuya K cared the patients, took the picture, and wrote the paper. Haneda M and Kikkawa R discussed the case. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544880.xml
514721	Hürthle cell carcinoma: diagnostic and therapeutic implications	Background Hürthle cell carcinoma is a variant of follicular cell carcinoma of thyroid. It may present as a low-grade tumour or as a more aggressive type. Prognosis depends upon the age of the patient, tumour size, extent of invasion and initial nodal or distant metastasis. Patient and methods The case of Hürthle cell carcinoma is reported in a 79-year-old man who presented with a rapidly increasing lump on the left side of his neck, having had a right hemithyroidectomy for colloid goitre 24-years-ago. Fine needle aspiration cytology confirmed the presence of Hürthle cells, raising the possibility of a Hürthle cell neoplasm. The patient underwent staging and surgery. Histology showed Hürthle cell carcinoma and the patient underwent adjuvant therapy. The literature on Hürthle cell neoplasms is reviewed. Conclusions Fine needle aspiration cytology may recognise Hürthle cell lesion but final diagnosis of carcinoma depends upon histological confirmation of vascular or capsular invasion. Staging and surgery in Hürthle cell carcinoma are similar to follicular carcinoma of thyroid with favourable outcome despite the controversy regarding the histological classification and adjuvant therapy. Elderly patients with Hürthle cell carcinoma need to be made aware of their poorer prognosis and should be offered more radical treatment.	Background The natural history of Hürthle cell carcinoma (HCC) is not well understood. It accounts for <5% of all differentiated thyroid malignancies. Hürthle cells are characterised by eosinophilic cytoplasm with trabecular/follicular growth pattern. [ 1 ]. Oncocytes are seen in follicular cell carcinoma but in HCC oncocytes represent more than 75% of cells, which exhibit a rather more trabecular growth pattern [ 2 ]. There is much debate regarding its clinical behaviour and little is known about the long-term survival of patients with HCC. Some studies have reported a relatively benign course while others have found the tumour to behave aggressively [ 3 - 6 ]. Most studies show that advanced age (>45), male sex, size of primary tumour (>4 cm), degree of invasion and recurrence are poor prognostic indicators [ 6 - 8 ]. Fine needle aspiration cytology is a good predictor of Hürthle cell neoplasm but is of little diagnostic value in evaluating HCC, since for a tumour to be deemed malignant one has to show vascular or capsular invasion [ 9 ]. Intraoperative frozen sections have a low predictive value. Udelsman et al found that in 96.4% cases with follicular neoplasm of thyroid, frozen section was neither informative nor cost-effective [ 10 ]. Well-encapsulated HCC run a favourable course while locally advanced HCC are associated with higher mortality and should be treated aggressively [ 4 , 11 ]. In a well-differentiated thyroid carcinoma death resulting from local disease is unusual and most die of distant metastases [ 12 ]. We report a case of a Hürthle cell carcinoma presenting in the left lobe of thyroid following a right hemithyroidectomy for a colloid goitre 24 years ago. Case presentation A 79-year-old male was referred in March 2003 with a lump on the left side of his neck. The patient had noted a sudden increase in the size of the lump over the preceding two months. He did not report any neck pressure symptoms, weight loss or anorexia. His past history included right partial thyroidectomy for a solitary nodule (colloid goitre) in 1978 and repair of abdominal aortic aneurysm in 1994. He had suffered myocardial infarction in 1995 and had an episode of acute coronary insufficiency in January 2003. His recent coronary angiograms showed an occluded left anterior descending artery and echocardiogram revealed good left ventricular function. He was a non-smoker and consumed alcohol in moderation. He had been taking warfarin, diltiazem MR, lisinopril, uniphyllin, glyceryl trinitrate tablets and buccal suscard. On examination he had left sided goitre extending superiorly into the posterior triangle and inferiorly into the retrosternal space, with variable consistency. The trachea was deviated to the right and there was cervical lymphadenopathy on the left side. Systemic examination was unremarkable and fine needle aspiration of thyroid gland showed presence of Hürthle cells. Computerised tomographic (CT) scan with contrast enhancement (figure 1 & 2 ) of the neck and thorax revealed large left sided thyroid goitre with significant mediastinal extension. It showed mixed attenuation with foci of calcification peripherally. There was a 3 cm complex mass on the left side of the neck, posterior to the carotid sheath structures and deep to the sternomastoid, indicative of lymph node metastases. Thyroid profile and routine blood investigations were unremarkable. Figure 1 Superior extension of left goitre with 3 cm diameter complex mass deep to sternomastoid, posterior to carotid sheath. Note the displacement of larynx to the right. Figure 2 Mediastinal extension of left goitre. Based on the above findings radical surgery was planned. On exploration of the neck we confirmed left goitre with intrathoracic extension and enlarged lymph nodes under the sternocleidomastoid close to the jugulodigastric muscle and surrounding the carotid sheath. There was no remnant thyroid tissue seen on the right side following the previous thyroid surgery. Left hemithyroidectomy with modified neck dissection (lymphadenectomy, preserving all vessels and nerves) was performed. Macroscopic examination of the thyroid lobe showed a well defined solid pale brown mass approximately 8 cm in maximum dimension, surrounded by a narrow rim of preserved thyroid tissue. The lymph node specimen comprised of several nodules of partly necrotic tissue. Microscopic examination showed the thyroid lobe containing a Hürthle cell neoplasm, which was mostly encapsulated, with foci of capsular and vascular invasion. The two lymph nodes revealed metastatic Hürthle cell carcinoma. [pT3, N1a, Mx], (Figure 3 & 4 ). Figure 3 Photomicrograph showing capsular invasion (Haematoxylin and Eosin ×200) Figure 4 photomicrograph showing Hürthle cell note the eosinophilic cytoplasm and prominent nucleoli (Haematoxylin and Eosin ×500). The patient had adjuvant therapy with oral radioiodine 131 (3060 MBq Sodium Iodine). He was put on a daily dose of 100 mcg of thyroxine. This was to be followed by a second dose of 5911 MBq of radioactive iodine six months from the time of the first dose. Discussion Hürthle cell carcinomas are heterogeneous neoplasms that display a wide range of biological behaviour and accounts for less than 5% of all differentiated thyroid malignancies. The term HCC should be restricted to tumours with more than 75% of oncocytic cells [ 2 ]. Oncocytes are seen in follicular thyroid cell carcinoma and in papillary thyroid cell carcinoma [ 13 , 14 ]. On one hand patients with HCC live for years with slow growing tumour and lymphatic metastases and on the other hand, patients die of highly aggressive tumour with haematogenous spread. Our patient had several indicators for poor prognosis such as his advanced age, male gender, large tumour size (8 cm), extra thyroid extension and nodal metastasis. Interestingly enough the patient had no pressure symptoms despite marked deviation of larynx, trachea and oesophagus, which may be due to previous right hemithyroidectomy. In elderly patients with sudden enlargement of neck mass and pre-existing thyroid conditions such as benign thyroid nodule, goitre (as in our case), Grave's disease or differentiated thyroid carcinoma, one has to bear in mind anaplastic thyroid carcinoma (ATC). In ATC local compression symptoms such as hoarseness, strider, dyspnoea and dysphagia occur as a rule [ 15 - 17 ]. In aggressive type of HCC haematogenous spread has been noted, but in ATC, at presentation patients are quite likely to have distant metastases involving lung, bone, brain and soft tissues [ 15 , 16 ]. Our patient had undergone fine needle aspiration cytology, which revealed Hürthle cells. Since the lesion was rapidly growing with mediastinal extension and nodal involvement, the patient underwent staging and left hemithyroidectomy with modified neck dissection. Histology confirmed HCC based on vascular and capsular invasion. Intraoperative frozen sections have low predictive value and are particularly not a sensitive test for diagnosing HCC therefore this was not carried out [ 10 ]. McIvor et al have clearly shown that FNAC can easily recognise the tumour as Hürthle cell lesion [ 9 ]. Cases with suspicious histology and over 50 years of age carry a high risk of cancer [ 18 ]. In the management of HCC the primary mode of treatment is surgical, ranging from hemithyroidectomy to total thyroidectomy. Larger tumours (>T2) require total thyroidectomy and lymphadenectomy if lymph nodes are involved [ 8 ]. Adjuvant radioiodine treatment or external beam radiotherapy is used for widely invasive carcinoma or locally advanced disease [ 8 ]. Several reports in literature have identified contra lateral foci of carcinoma in 40–70% of cases of HCC [ 11 , 19 ]. HCC is less responsive to radioactive iodine therapy [ 20 ] and taking into account the aggressive behaviour, it has been suggested that every Hürthle cell tumour greater than 2 cm should be treated by total thyroidectomy [ 21 ]. In 1990 they showed that recurrent disease was noted in 17% of patients treated with total thyroidectomy compared to 59% in cases where a more limited procedure was carried out [ 21 , 22 ]. Other authors support the role of total thyroidectomy as there is 15 to 35% incidence of multiple foci in HCC [ 23 ]. There are several reasons favouring the use of 131 I remnant ablation after near-total thyroidectomy [ 24 ]. First, presence of thyroid remnant can obscure 131 I uptake in cervical or lung metastases [ 25 , 26 ]. Second, distant (lung) metastases may be seen only on the post treatment whole body scan after remnant ablation [ 27 ]. Finally, remnant ablation may destroy residual normal follicular cells, which may become malignant [ 28 ] and any occult cancer that may recur years later. Radioiodine therapy has no overall effect on mortality but subgroup analysis has shown that those patients who receive radioactive iodine for adjuvant ablation of remnant thyroid tissue have lower mortality rate compared with patients who either did not receive treatment or in whom the indication was the presence of residual disease [ 29 ]. Radioiodine uptake in the elderly is much lower. Schlumberger and colleagues noted 131 I uptake at metastatic sites in only 53% of patients over 40 years of age, compared to 90% in patients below the age of 40 [ 30 ]. Univariate analysis indicated that older age and large tumour size predicted worse survival rates due to aggressive nature of the tumour (extra glandular invasion and multifocal disease). One recent series reviewed medical records of patients between the years 1944 and 1995. Of the 89 HCC cases studied, 29% had only undergone lobectomy as initial treatment and 50% had undergone partial resection. Of the three quarters of the patients in this series who received radioactive iodine, only 38% of patients with known metastases showed positive uptake [ 29 ]. Another study clearly suggested that treatment with 131 I to ablate the thyroid remnant and to treat residual disease were independent prognostic variables that favourably influenced recurrence, distant recurrence, and cancer death rates [ 24 ]. Our patient received radioactive iodine treatment in the postoperative period. He has been followed up with whole body scans (Fig 5 and 6 ), which indicate his response to adjuvant radioactive iodine therapy. He is on 125 mcg thyroxine in order to maintain a TSH level of less than 0.01 mIU/L and FT4 at the upper limit of normal (8–28 pmol/L). Figure 5 Whole body scan on November 3, 2003 following 131 I ablation therapy on 28 th October 2003, with 3060 MBq Sodium Iodine ( 131 I). Increased uptake is seen in the region of the thyroid bed. No abnormal accumulation was noted elsewhere. Figure 6 Whole body scan on 19 th April 2004 following 131 I ablation therapy on 13 th April 2004 with 5911 MBq Sodium Iodine ( 131 I). Two small focal area of uptake are seen in the thyroid bed. Low uptake focal area in the left lateral aspect of the neck, could possibly represent activity in a cervical node. Stojdinovic et al have treated 56 patients with HCC between the years 1940 and 2000 [ 8 ]. Of these patients 23(41%) had minimally invasive disease with no evidence of extra thyroid invasion (T2 N0 M0) and 33(56%) had widely invasive HCC. Primary mode of treatment was surgery ranging from lobectomy and isthumusectomy to total thyroidectomy with cervical lymphadenectomy in presence of lymph node involvement. Some patients received adjuvant radioiodine or external beam radiotherapy for widely invasive carcinoma. Study end points were relapse free survival and disease specific survival. They reported 8 years survival rate of 100% and 58% for low and high-risk cancers respectively. In their entire study cohort age was not found to predict the outcome but the most significant factor was widely invasive carcinoma. Khafif et al in their series (42 patients with HCC between 1957–1997) used radioiodine in patients with distant metastases; none had thyroid remnant ablation with radioactive iodine [ 4 ]. They reported an overall survival rate of 90.5% and noted that age, size of tumour and extent of resection adversely affected the prognosis. Hürthle cell lesion can be easily picked up on FNAC but to make a diagnosis of HCC one has to demonstrate vascular or capsular invasion. Intraoperative frozen sections have low predictive value and cases with advanced age (over 50), rapid enlargement of lump and palpable nodes should be regarded with high index of suspicion for presence of HCC. HCC or other differentiated carcinomas of thyroid in the elderly patients are generally more aggressive with less favourable prognosis compared to younger patients. They should be offered total thyroidectomy and selective lymph node dissection (when lymph nodes are involved) followed by ablative radioiodine therapy, provided they can withstand the above treatment. Coexisting medical disorders should be recognized and managed effectively prior to surgery [ 31 ]. Further research is needed to clarify the role of adjuvant radioiodine therapy in the management of HCC. Competing interests None declared. Authors' contribution MRH managed the patient, searched the literature and drafted the manuscript LI: did the histological study, and contributed to pathological aspects in the present study DG: Managed the patient, conceptualise the present report, edited the manuscript and coordinated after reviewing the manuscript	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514721.xml
514709	Rates of influenza vaccination in older adults and factors associated with vaccine use: A secondary analysis of the Canadian Study of Health and Aging	Background Influenza vaccination has been shown to reduce morbidity and mortality in the older adult population. In Canada, vaccination rates remain suboptimal. We identified factors predictive of influenza vaccination, in order to determine which segments of the older adult population might be targeted to increase coverage in influenza vaccination programs. Methods The Canadian Study of Health and Aging (CSHA) is a population-based national cohort study of 10263 older adults (≥ 65) conducted in 1991. We used data from the 5007 community-dwelling participants in the CSHA without dementia for whom self-reported influenza vaccination status is known. Results Of 5007 respondents, 2763 (55.2%) reported having received an influenza vaccination within the previous 2 years. The largest predictive factors for flu vaccination included: being married (57.4 vs. 52.6%, p = 0.0007), having attained a higher education (11.0 vs. 10.3 years, p < 0.0001), smoking (57.1% vs. 52.9%, p = 0.0032), more alcohol use (57.9% of those who drank more vs. 53.2% of those who drank less, p = 0.001), poorer self-rated health (54.1% of those with good self-rated health vs. 60.6% of those with poor self-rated health, p = 0.0006), regular exercise (56.8% vs. 52.0%, p = 0.001), and urban living (55.8% vs. 51.0%, p = 0.03). While many other differences were statistically significant, most were small (e.g. mean age 75.1 vs. 74.6 years for immunized vs. unimmunized older adults, p = 0.006, higher Modified Mini Mental Status Examination score (89.9 vs. 89.1, p < 0.0001), higher comorbidity (2.7 vs. 2.3 comorbidities, p < 0.0001). Residents of Ontario were more likely (64.6%) to report vaccination (p < 0.0001), while those living in Quebec were less likely to do so (48.2%, p < 0.0001). Factors retaining significance in a multivariate analysis included older age, higher education, married status, drinking alcohol, smoking, engaging in regular exercise, and having higher comorbidity. Conclusions The vaccination rate in this sample, in whom influenza vaccination is indicated, was low (55.2%). Even in a publicly administered health care setting, influenza vaccination did not reach an important proportion of the elderly population. Whether these differences reflect patient preference or access remains to be determined.	Background Influenza has the potential to cause serious complications (chiefly viral and bacterial pneumonia, with attendant mortality and morbidity), especially in the older adult (>= 65) population and in those with comorbid illness [ 1 - 3 ]. Influenza vaccination is demonstratively safe, effective [ 4 - 6 ], and cost-effective [ 7 , 8 ], and current guidelines in both Canada and the United States recommend that the vaccine be administered yearly to all individuals over the age of 65 and to all residents of Long Term Care Facilities [ 9 , 10 ]. Despite current evidence and recommendations, influenza vaccination rates remain suboptimal. Estimates of vaccine coverage in high-risk groups (including older adults) range from 10–40% in the UK [ 11 ], and 45–68% in the United States [ 8 , 12 , 13 ]. A 1993 Canadian study found that 57.5% of community-dwelling Albertans over the age of 65 had been vaccinated within the past 12 months [ 14 ]. A question about influenza vaccination included in the 1991 Statistics Canada General Social Survey found that 44.8% of randomly sampled community dwelling individuals aged 65 and older had been vaccinated during the 1990/91 flu season [ 15 ]. A Canadian study of vaccination rates among residents of long term care (LTC) facilities reported 79% coverage for the 1990/91 season, increasing to 83% in 1998/99 [ 16 ]. To our knowledge, no previous studies have examined predictors of vaccination in community-dwelling older Canadians. The objectives of our study were to determine the vaccination rate in this community-dwelling population of older Canadians and to identify factors predictive of influenza vaccination, in order to determine which segments of the older adult population should be targeted to achieve better coverage in influenza vaccination programs. Methods Sample/study population The Canadian Study of Health and Aging (CSHA) is a population-based representative sample drawn from Canadians over the age of 65, designed to study the prevalence, incidence, and risk factors for development of dementia [ 17 ]. A technical report providing details on sampling, design and measurement is available elsewhere [ 18 ]. Briefly, in 1991 and 1992, data were collected from 10,263 older adults: 9008 community-dwelling and 1255 in Long Term Care Facilities (LTCF). Those in the community were randomly selected from medicare lists (or the Enumeration Composite Record in Ontario), and institutionalized individuals were randomly selected from stratified random samples of institutions in each region. Individuals excluded were residents of the Yukon and Northwest Territories, those living on Aboriginal reserves or military bases, or those with life-threatening illnesses (examples cited include terminal cancer or conditions requiring life support). A Self-Administered Risk Factor Questionnaire (SARFQ) was completed by the non-demented community-dwelling participants in the study. The SARFQ included demographic questions and addressed issues of lifestyle, medical and family history, and medication use. From these questions a fitness/frailty scale was derived, that recognizes seven levels, from most fit (= 1) to most frail (= 7) [ 19 ]. A question on immunization history was asked, in which respondents were asked whether they had received the influenza vaccination, and if yes, approximately how many times they had had received it and the year of their most recent immunization. The sample population used in this study includes all participants who completed the SARFQ for whom influenza vaccination status could be determined based upon their answers. The derivation of the study population is shown in Figure 1 . Figure 1 Derivation of the study population. Statistical analysis Respondents were designated vaccinated if they reported having had an influenza vaccine within the 2 years preceding completion of the SARFQ. As the data were collected over a two-year period, whereas the question about the last vaccination asked for a date, we designated as unvaccinated those who did not report receiving a vaccination within the previous 2 years. Vaccinated and unvaccinated respondents were compared; as a first step crude analyses were done taking each variable individually (not adjusted for any others). Chi-squared test or Fisher's exact test was used for categorical variables and one-way ANOVA was used for continuous variables. Characteristics analyzed included age, gender, education, region of residence, marital status, smoking status, alcohol intake, self-assessed health status, exercise, diagnosis of dementia, Modified Mini-Mental State Examination (3MS)[ 20 ] score, number of comorbidities, and urban vs . rural residence. As a second step, a multivariate analysis done by stepwise selection of parameters found to be significant in the univariate analysis was then conducted. Analyses comparing responders vs . non-responders to the relevant SARFQ questions as well as comparing respondents who reported first-time vs . regular influenza vaccination were also undertaken. The possibility of interaction between variables was considered. We included in the multivariable model interaction terms of pairs of variables where interaction was plausible (between smoking and drinking, age and frailty according to the frail scale, and region of residence and urban/rural dwelling). Results Influenza vaccination status could be determined for 5007 (76.8%) of the 6521 non-demented, community-dwelling participants in the CSHA who completed the SARFQ. Of these, 2763 (55.2%) reported having received at least one influenza vaccination within the previous two years. The univariate analysis, in which each variable was examined individually without adjusting for the effects of others, showed several differences between the demographics, region of residence, lifestyle, and health status of the two groups (Table 1 ). Immunized older adults were on average more highly educated, and being married (living with a partner or spouse) was another significant predictor of vaccination. Urban dwellers in general, and residents of Ontario were significantly more likely to have been immunized than those living elsewhere in Canada. The lowest self-report vaccination rates were found in Quebec. Table 1 Univariate analysis of vaccinated vs. unvaccinated community dwelling older Canadians Risk Factor Vaccinated (n = 2763) Unvaccinated (n = 2244) OR (95% C.I.) P value Age mean (SD) 75.1 (6.5) 74.6 (6.7) 1.01 (1.00–1.02) 0.0056 Sex Male (%) 1162 (56.8) 883(43.2) 1.00 0.0527 Female (%) 1601 (54.1) 1361(45.9) 0.89 (0.80–1.00) Education mean years (SD) 11.0 (3.7) 10.3 (3.7) 1.05 (1.04–1.07) <0.0001 Region Atlantic (%) 471 (54.1) 399 (45.9) 1.00 Quebec (%) 453 (48.2) 486 (51.8) 0.79 (0.66–0.95) <0.0001 Ontario (%) 568 (64.6) 311 (35.4) 1.55 (1.28–1.88) Prairies (%) 540 (55.6) 431 (44.4) 1.06 (0.88–1.28) BC (%) 731 (54.2) 617 (45.8) 1.00 (0.85–1.19) Current Marital Status Not married (%) 1197 (52.6) 1078 (47.4) 1.00 0.0007 Married (%) 1566 (57.4) 1163 (42.6) 1.21 (1.08–1.36) Smoker No (%) 1232 (52.9) 1095 (47.1) 1.00 0.0032 Yes (%) 1495 (57.1) 1122 (42.9) 1.18 (1.06–1.32) Alcohol No (%) 1590 (53.2) 1399 (46.8) 1.00 0.0012 Yes (%) 1133 (57.9) 824 (42.1) 1.21 (1.08–1.36) Self-Rated Health Not Good/Very Poor (%) 502 (60.6) 326 (39.4) 1.00 0.0006 Very/Pretty Good (%) 2254 (54.1) 1912 (45.9) 0.77 (0.66–0.89) Regular Exercise No (%) 893 (52.0) 824 (48.0) 1.00 0.0014 Yes(%) 1815 (56.8) 1382 (43.2) 1.22 (1.08–1.36) 3MSE mean score (sd) 89.9 (6.4) 89.1 (6.4) 1.02 (1.01–1.03) <0.0001 No. of comorbidities mean (sd) 2.7 (1.7) 2.3 (1.6) 1.15 (1.11–1.20) <0.0001 Geography Urban (%) 2442 (55.8) 1934 (44.2) 1.00 0.0274 Rural (%) 320 (51.1) 306 (48.9) 0.83 (0.70–0.98) Of note, smokers were significantly more likely than non-smokers to have received the influenza vaccination. The same was true of individuals who consumed alcohol regularly when compared with those who drank rarely or not at all. Regular exercise was found to be a predictive factor for vaccination. Those who had received the vaccination had significantly more health problems, as did those who saw themselves as being in poorer health. In the multivariate analysis (Table 2 ) predictive factors for immunization that retained significance after adjusting for the effects of other variables include older age, higher level of education, being married, smoking, engaging in regular exercise, and having more co-morbid illnesses. Interaction terms describing interaction between age and frailty, smoking and drinking, and region of residence and urban/rural dwelling did not retain significance in the multivariable model, suggesting that there was no statistically significant interaction between these pairs of variables. Table 2 Multivariate analysis comparing vaccinated and unvaccinated community-dwelling older Canadians Risk Factor Odds Ratio 95% C.I. p value Older Age 1.02 1.01–1.06 <0.0001 More Education 1.05 1.03–1.07 <0.0001 Currently Married 1.29 1.14–1.47 <0.0001 Smoking 1.14 1.01–1.30 0.0401 Drinking 1.51 1.01–1.31 0.0358 Regular Exercise 1.25 1.10–1.42 0.0004 Region Atlantic 1.00 --- Quebec 0.94 0.77–1.14 <0.0001 Ontario 1.48 1.21–1.82 Prairies 0.99 0.82–1.20 BC 0.80 0.67–0.97 More Comorbidity 1.18 1.140–1.226 <0.0001 To investigate the impact of response bias, we compared those who had answered the influenza vaccination question and the 23.2% of the original sample who had not (Table 3 ). Non-responders were more likely to be rural residents and to drink alcohol regularly. Age, 3MS score, self-rated health, and number of comorbidities also differed between the groups: non-responders were more likely to be younger, healthier, and to have better self-rated health and higher 3MS scores. Several of the same characteristics (age, number of comorbidities, self-rated health, region of residence and urban vs. rural living) also differed between regular and first time vaccination users (Table 4 ). Repeat users of the influenza vaccine (i.e. those who reported at least two past vaccinations) were slightly older (75.3 vs . 74.1 years, p < 0.0001), had more comorbid illness (2.7 vs . 2.5, p = 0.0011), and poorer self-rated health (81.7% poor health vs . 77.9% good health, p = 0.045). Residents of Atlantic Canada (75.1% vs . 79.3% residing elsewhere, p = 0.0003) and those living in rural areas (79.3% of urban vs . 73.8% of rural-dwellers, p = 0.12) were less likely to be regular users. Table 3 Analysis of responders vs. non-responders to questions relating to influenza vaccination in the SARFQ Characteristic Value Responders (n = 5007) Value Non-responders (n = 1514) p value Age mean (SD) 74.9 73.4 <0.0001 Sex Male (%) 76.9 23.1 0.8416 Female (%) 76.7 23.3 Education mean (SD) 10.7 10.7 0.8164 Region Atlantic (%) 71.6 28.4 Quebec (%) 78.9 21.1 <0.0001 Ontario (%) 69.4 30.6 Prairies (%) 70.1 29.9 BC (%) 92.1 7.9 Current Marital Status Not married (%) 77.1 22.9 0.5742 Married (%) 76.5 23.5 Smoker No (%) 78.0 22.0 0.1888 Yes (%) 76.6 23.4 Alcohol No (%) 82.4 17.6 <0.0001 Yes (%) 75.8 24.2 Self-Rated Health Not Good/Very Poor (%) 82.4 17.6 <0.0001 Very/Pretty Good (%) 75.8 24.2 Regular Exercise No (%) 77.8 22.2 0.6263 Yes(%) 77.3 22.7 MMMSE mean (sd) 89.6 90.1 0.0071 No. of comorbidities mean (sd) 2.5 2.1 <0.0001 Geography Urban (%) 77.7 22.3 <0.0001 Rural (%) 70.7 29.3 SARFQ = Self-Administered Risk Factor Questionnaire in the Canadian Study of Health and Aging Table 4 Characteristics of regular vs. first time influenza vaccine users Risk Factor Regularly Vaccinated First Time Vaccinated P value Age mean years (SD) 75.3 (6.7) 74.1 (6.7) <0.0001 Sex Male (%) 79.5 20.5 0.29 Female (%) 78.9 22.0 Education mean years (SD) 11.0 (3.8) 10.8 (3.8) 0.27 Region Atlantic (%) 75.1 24.9 0.0003 Quebec (%) 83.7 16.3 Ontario (%) 81.5 18.5 Prairies (%) 78.2 21.8 BC (%) 75.5 24.5 Current Marital Status Not married (%) 79.2 20.8 0.45 Married (%) 78.1 21.9 Smoker No (%) 78.3 21.7 0.73 Yes (%) 78.8 21.2 Alcohol No (%) 78.4 21.6 0.83 Yes (%) 78.7 21.3 Self-Rated Health Not Good/Very Poor (%) 81.7 18.3 0.045 Very/Pretty Good (%) 77.9 22.1 Regular Exercise No (%) 80.2 19.8 0.086 Yes(%) 77.5 22.5 3MSE mean score (sd) 89.9 (6.4) 89.7 (6.4) 0.48 No. of comorbidities mean (sd) 2.7 (1.7) 2.5 (1.7) 0.0011 Geography Urban (%) 79.3 20.7 0.012 Rural (%) 73.8 26.2 Discussion Our data were collected in the early 1990s and may therefore not reflect current practice patterns. Although the Canadian Study of Health and Aging was not designed to study influenza vaccination, it does provide a large and unique data set of primarily community-dwelling older Canadians, and is therefore potentially useful in the examination of health-related risk factors and demographics that may influence decisions to vaccinate. Given the importance of influenza vaccination in the prevention of significant morbidity and mortality in populations at risk, the vaccination rate of 55.2% in our community-dwelling sample of older adults is concerning. People who were not vaccinated tended to be younger, non-smokers and to have fewer co-morbid illnesses. They were also found to have a lower level of education, not to be married and not to engage in regular exercise. These were the factors that retained statistical significance in the multivariable analysis, suggesting that these associations are unlikely to have arisen due to confounding by any of the variables investigated. Vaccination has previously been studied in the CSHA, but only in relation to its status as a potentially protective factor with respect to cognitive impairment [ 21 ]. The weight of evidence derived from re-examination of large databases is less than that derived from specifically designed trials, however this method still maintains an important role in epidemiological research. For example, evaluation of systematic problems can be used to help develop targeted efforts in improving vaccination rates. Our data do not include information on institutionalized older adults, where it might be expected that at-risk profiles would vary, and where vaccination rates are generally higher [ 16 ]. Another important source of potential error is our reliance on self-reported immunization status. However, self-report of influenza vaccination status in elderly outpatients has been found to be highly sensitive and moderately specific when checked against medical record documentation [ 22 ]. Other studies have identified predictive factors for vaccination in other countries [ 8 , 12 ]. An American study found that patients with more health conditions, higher rates of use of health care resources, and history of pneumonia were more likely to be vaccinated, while non-vaccinated individuals were older and more likely to have dementia or stroke [ 8 ]. An Iowa study identified a number of factors associated with the receipt of both influenza and pneumococcal vaccines: age over 70, self-owned residence, working, increased number of medical conditions, current prescription medication, and a physician visit within the past year. Geography (rural vs. urban living) was unrelated to vaccine receipt [ 12 ]. Our data suggest that there appears to be an important degree of targeting of vaccination resources within the older adult population to people who are less healthy. Even within these higher risk groups, however, immunization is incomplete. Perhaps such people are perceived by their health care providers as being at higher risk from influenza infection, and are thus more likely to be immunized. Similar explanations may account for the higher vaccination rates among smokers and those who consumed more alcohol. However, according to the current Canadian guidelines [ 9 ], influenza vaccination is indicated for all persons 65 years of age and over. If it is the case that those at the younger end of this age group are less likely to be vaccinated, as our study suggests, more must be done to ensure that vaccination is reaching its entire target population. Regular exercise was shown to be a factor predictive of influenza vaccination. Interestingly, this association may have been confounded by generally health-protective behaviour (which might be expected to be associated with both regular exercise, the explanatory variable, and vaccination, the outcome), given that individuals exhibiting healthy lifestyle choices (such as exercise) may have been more likely to have sought preventative health care and to have visited their health care providers more regularly, thus providing more opportunity to be vaccinated. Regular exercise did, however, retain significance in our multivariate analysis. The regional differences identified in our study may point to geographic differences in access to influenza vaccination, although the general milieu and level of awareness in both the medical community and society at large may also be significant. These regional differences suggest that certain areas may benefit from targeting of vaccination efforts. However, over and above questions of health policy and public health education, the question of access to vaccination is of vital importance if we are to achieve acceptable rates of coverage in this target population. The finding that rural residence is a negative predictor of vaccination is particularly concerning, and points to the larger equity issue of how uptake of influenza vaccination can be improved outside of major urban centers. However, our finding that rural vs. urban residence did not retain statistical significance in the multivariable regression model suggests that this crude association may be due to confounding by other factors. Our analysis comparing non-responders with those for whom influenza vaccination status was known on the basis of the SARFQ showed that response bias was indeed likely. Non-responders were more likely to drink more alcohol, a factor which was associated with being immunized. However, several factors which were associated with not being immunized, including younger age, good self-rated health, having fewer comorbidities, and residence in a rural area, were more prevalent among non-responders. This over-representation of a number of factors predictive of non-immunization among non-responders suggests that our results may well have been influenced by response bias. For example, we may have over-estimated the prevalence of vaccination in this population. The analysis comparing regular users of the influenza vaccine with those who reported first-time immunization in the SARFQ demonstrated a number of factors that differed between the two groups. The finding that younger age is associated with first time vaccination may be at least partially explained by those at the younger end of the > = 65 age group having spent less time in this "target group" for vaccination. As such, they would be more likely to be receiving the vaccination for the first time. The effect of age may also play a role in the finding that those with better self-assessed health and fewer comorbidities were more likely to be first-time users (as they may also have been younger). However, being in better health was also a predictive factor for non-vaccination within the previous two years, as was rural residence, suggesting that these factors may influence decisions to initiate as well as to sustain influenza vaccination over subsequent years. Conclusions Despite current recommendations and proven benefit, influenza vaccination rates in our Canadian sample were suboptimal; only 55.2% of older adults in our representative sample reported being vaccinated within the past two years. The predictive factors identified in this analysis may facilitate the study of targeting of older adults in future vaccination programs. Competing interests SM and HM have been involved in trials of Influenza vaccines funded by ID Biomedical. Neither has received any personal financial compensation for their involvement. MKA and KR have no competing interests. Author contributions MKA wrote the first draft, and reviewed each of the analyses. SM & KR reviewed the analyses and revised the manuscript. KR is a co-Principal Investigator of the CSHA. HM did the statistical analyses. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514709.xml
529311	Using animal models to determine the significance of complement activation in Alzheimer's disease	Complement inflammation is a major inflammatory mechanism whose function is to promote the removal of microorganisms and the processing of immune complexes. Numerous studies have provided evidence for an increase in this process in areas of pathology in the Alzheimer's disease (AD) brain. Because complement activation proteins have been demonstrated in vitro to exert both neuroprotective and neurotoxic effects, the significance of this process in the development and progression of AD is unclear. Studies in animal models of AD, in which brain complement activation can be experimentally altered, should be of value for clarifying this issue. However, surprisingly little is known about complement activation in the transgenic animal models that are popular for studying this disorder. An optimal animal model for studying the significance of complement activation on Alzheimer's – related neuropathology should have complete complement activation associated with senile plaques, neurofibrillary tangles (if present), and dystrophic neurites. Other desirable features include both classical and alternative pathway activation, increased neuronal synthesis of native complement proteins, and evidence for an increase in complement activation prior to the development of extensive pathology. In order to determine the suitability of different animal models for studying the role of complement activation in AD, the extent of complement activation and its association with neuropathology in these models must be understood.	Background Alzheimer's disease and complement activation A variety of inflammatory processes are increased in regions of pathology in the Alzheimer's disease (AD) brain [ 1 - 4 ]. There is a reciprocal relationship between this local inflammation and senile plaques (SPs) and neurofibrillary tangles (NFTs); both SPs and NFTs, as well as damaged neurons and neurites, stimulate inflammatory responses [ 5 ], and inflammatory processes exert multiple effects, some of which promote neuropathology [ 6 - 8 ]. Numerous retrospective studies have shown that long-term administration of nonsteroidal anti-inflammatory drugs (NSAIDs) to individuals with arthritis significantly reduces the risk for these individuals for developing AD [ 9 ]. These findings, together with the demonstration of elevated glial cell activation [ 10 - 12 ], complement activation [ 13 - 15 ], and increased acute phase reactant production [ 16 - 19 ] at sites of pathology in the AD brain, support the hypothesis that local inflammation may contribute to the development of this disorder [ 20 ]. Although a short-term trial of AD patients with the NSAID indomethacin suggested protection from cognitive decline [ 21 ], subsequent trials with other anti-inflammatory drugs have found no evidence for slowing of the dementing process [ 22 - 25 ]. These findings underscore the current perception of CNS inflammation as a "double edged sword" [ 26 , 27 ], with neuroprotective roles for some inflammatory components and neurotoxic effects for others [ 28 - 30 ]. The significance of complement activation, a major inflammatory mechanism, in AD is particularly problematic. The complement system is composed of more than 30 plasma and membrane-associated proteins which function as an inflammatory cascade. Complement activation promotes the removal of microorganisms and the processing of immune complexes. The liver is the main source of these proteins in peripheral blood, but they are also synthesized in other organs including the brain [ 31 ]. Protein fragments generated during activation of the system enzymatically cleave the next protein in the sequence, generating a variety of "activation proteins" with diverse activities (Table 1 ). Three complement pathways, the classical, alternative, and lectin-mediated cascades, have been identified (Fig. 1 ). Full activation results in the generation of C5b-9, the "membrane attack complex" (MAC), which penetrates the surface membrane of susceptible cells on which it is deposited and may result in cell death if present in sufficient concentration. The presence of early complement activation proteins [ 32 - 37 ] and of the MAC [ 38 - 42 ] has been demonstrated by immunocytochemical staining in the AD brain. Subsequent studies found that complement activation increases Aβ aggregation [ 43 , 44 ] and potentiates its neurotoxicity [ 45 ], attracts microglia [ 46 , 47 ], promotes microglial and macrophage secretion of inflammatory cytokines [ 48 , 49 ], and induces neuronal injury, and sometimes neuronal death, via the MAC [ 50 ]. These findings suggested that complement activation might contribute to the neurodegenerative process in AD. However, recent studies have also revealed neuroprotective functions for some complement activation proteins, including in vitro protection against excitotoxicity [ 51 , 52 ] and Aβ-induced neurotoxicity [ 53 ], as well as anti-apoptotic effects [ 54 , 55 ]. Further, C1q, the first complement protein to be deposited on cell membranes during activation of the classical complement sequence, may facilitate the clearance of Aβ by microglia [ 56 ], although this is controversial [ 57 ]. Understanding the role of complement activation in AD is of clinical relevance because some complement-inhibiting drugs are available, and others are being developed (see reviews by Sahu and Lambris [ 58 ], and Morgan and Harris [ 59 ]). Conditions for which these agents are currently being investigated include stroke [ 60 ], organ transplantation [ 61 ], glomerulonephritis [ 62 ], ischemic cardiomyopathy [ 63 ], and hereditary angioedema [ 64 ]. Modulation of CNS complement activation in experimental animal models of AD, both by treatment with complement-inhibiting drugs and by generation of AD-type pathology in complement-deficient animals, should be useful for obtaining a greater understanding of the role of this process in the development of AD-type pathology. Unfortunately, knowledge of the extent of complement activation in animal models is lacking. This paper will review (a) criteria for an optimal animal model to study this issue, (b) present knowledge about complement activation in animal models of AD, and (c) additional animal models which offer alternatives for addressing this question. Table 1 Biological activities of complement activation proteins, with relevance to AD. Name Biological activity C1q Enhances Aβ aggregation [43,44]; may facilitate Aβ clearance [56]; enhances Aβ-induced cytokine secretion by microglia [49] C3a Anaphylatoxin (increases capillary permeability) [155] ; protects neurons vs. excitotoxicity [52] C3b Immune adherence and opsonization [89] (may facilitate Aβ clearance by phagocytic microglia) C4a Anaphylatoxin (weak) [156] C5a Anaphylatoxin; protects neurons vs. excitotoxicity [51]; chemotaxic attraction of microglia [46,47]; inhibits apoptosis 54; increases cytokine release from Aβ-primed monocytes [48] C5b-9 Neurotoxicity [50]; sublytic concentrations may have both pro- and anti- inflammatory activities [157] Figure 1 Schematic diagram of classical, alternative, and lectin complement activation pathways. There is evidence for activation of the classical and alternative pathways in the AD brain. (Adapted from Sahu and Lambris, 2000 [58]). Criteria for an optimal animal model for studying AD-related complement activation While animal models of human disease generally have similar pathological findings to the human disorders, distinct differences remain. These models may be appropriate for studying some aspects of a disease process, while less suitable for others. To determine the significance of complement activation in the development of AD-type pathology, for example, some animal models may be of value primarily for investigating the relationship between early complement activation and SP and NFT formation, whereas others may be more relevant for studying the role of the MAC in neuronal loss. 1. Complete activation of complement Investigators at the Academic Hospital Free University in Amsterdam first reported the presence of early activation proteins in the classical complement cascade in the AD brain [ 32 - 34 , 36 , 37 ]. The MAC was not detected. However, further studies by other laboratories convincingly demonstrated the MAC, by a variety of techniques, in AD specimens [ 38 - 42 ]. The Dutch group has more recently reported detection of the MAC in brain specimens from subjects with dementia with Lewy bodies who met CERAD neuropathological criteria for AD [ 65 ]. The MAC has similarly been reported in SPs from subjects with Down's syndrome [ 66 ] and with familial British dementia [ 67 ], disorders in which typical AD-type neuropathology is present. An optimal animal model for studying AD-related complement activation should therefore have complete complement activation. 2. Association of complement activation proteins with neuropathology Complement proteins are detectable on or closely associated with SPs, NFTs, and dystrophic neurites in the AD brain. These findings are in agreement with in vitro studies indicating that Aβ and tau protein, the major components in SPs and NFTs, can fully activate human complement [ 42 , 68 - 71 ]. Although the above studies suggested that complement is activated principally by the aggregated forms of Aβ and tau, soluble, non-fibrillar Aβ may also be capable of activating complement [ 72 ]. In contrast to the robust staining of complement proteins in mature plaques, immunoreactivity to these proteins in diffuse plaques has generally been below the level of detection, though it has been reported in some studies [ 36 , 73 , 74 ]. Complement activation in the AD brain is increased primarily in regions containing extensive pathology (e.g., the hippocampus and cortex), and whether early complement components are also present in the diffuse plaques that develop in the AD cerebellum is controversial [ 74 , 75 ]. The above findings suggest that complement activation in an optimal animal model of AD should be associated with SPs and, in those models in which neurofibrillary pathology occurs, with NFTs. 3. Initiation of complement activation early in development of pathology How the increased complement activation in AD relates to the development of SPs and NFTs, and to neuronal loss, is unclear. Immunocytochemical staining for complement activation proteins in the aged normal human brain is generally faint, and may be below the level of detection [ 42 , 69 , 73 ]; of relevance is a recent report describing extensive neuron-associated C1q reactivity in a cognitively normal subject with neuropathological findings limited to diffuse cortical plaques [ 76 ]. Elderly "high pathology controls," lacking dementia but with increased numbers of entorhinal NFTs and neocortical Aβ deposits, have a slight increase in the percentage of C5b-9-immunoreactive plaques in comparison with aged normal subjects, though this percentage is far lower than in the AD brain [ 39 ]. A recent study in our laboratory [ 77 ] used enzyme-linked immunosorbent assay (ELISA) to measure the concentrations of two early complement activation proteins, C4d and iC3b, in brain specimens from AD and normal subjects. ELISA is more sensitive than immunocytochemical staining, though it provides no information regarding the cellular association of complement immunoreactivity. Increased concentrations of these early complement activation proteins were present in some aged normal specimens. These reports suggest that early complement activation may increase prior to the development of plaques and NFTs. Similar findings are desirable in an optimal animal model for studying AD-related complement activation. 4. Increased CNS production of native complement proteins Both mRNA expression and protein synthesis of native complement proteins are increased in the AD brain [ 78 - 80 ]. (Note: the distinction between detection of native complement proteins, vs. detection of complement activation proteins, has frequently been blurred. In some studies in which immunoreactivity to complement activation proteins (C3c, C4c, C4d) has been reported, the antisera used were also capable of detecting the respective native complement proteins (C3 or C4) [ 40 , 80 ]. Only when antisera are used whose immunoreactivity is limited to activation-specific neo-epitopes can complement activation be confirmed. The paucity of antisera which can detect complement activation proteins in experimental animal models is a significant obstacle to determining the extent of complement activation in these models.) In addition to neurons, complement proteins are synthesized by other cells in the CNS including microglia, astrocytes, oligodendrocytes, and endothelial cells [ 31 ]. The biological effects of these activation proteins are mediated by numerous regulatory proteins including CD59, clusterin, vitronectin, C1-inhibitor, C4-binding protein, decay-activating factor, and Factor H, which inhibit different steps in the complement cascade. All of these regulatory proteins are produced in the human brain, but less is known about their CNS synthesis in other species [ 31 ]. The status of some of these regulatory proteins in AD is unclear; for example, there are conflicting reports regarding the up-regulation of C1-inhibitor [ 81 , 82 ] and CD59 [ 41 , 82 , 83 ]. Thus, while an optimal animal model for studying AD-related complement activation should have up-regulated CNS synthesis of complement proteins, the alterations that should be present in complement regulatory proteins are less clear. 5. Alternative as well as classical complement activation Complement activation in the AD brain was initially thought to be limited to the classical pathway, but recent reports have also indicated increased concentrations of the alternative activation factors Bb and Ba, and Factor H, a regulatory factor for the alternative pathway, in the AD brain [ 84 , 85 ]. Alternative complement activation has also been reported in other familial dementias with pathologies similar to AD [ 67 ]. Therefore, while activation of the classical pathway is an absolute requirement for an optimal animal model of AD-related complement activation, an increase in the alternative pathway is also desirable. Complement activation in animal models of AD: present knowledge The examination of complement activation in experimental models of AD has been limited to mice and rats. The extent of complement activation and its relationship to the development of AD-type neuropathology have generally not been determined in these studies. APP/sCrry mouse Increased complement activation was induced by overproduction of transforming growth factor beta1 (TGF-β1) in transgenic mice expressing mutations in the human amyloid precursor protein (hAPP) gene. The APP mutations expressed in these mice have been associated with early-onset, familial AD [ 86 ]. The TGF-β1 overproduction resulted in a 50% reduction in Aβ accumulation in the hippocampus and cerebral cortex [ 87 ]. Because the production of soluble Aβ was unchanged, these results suggested that reduction in Aβ may have been due to its increased clearance by microglia. A subsequent study by the same investigators [ 88 ] found that the mRNA level of C3 in the cerebral cortex was 5-fold higher in APP/TGF-β1 mice than in APP mice at 2 months of age (prior to deposition of Aβ) and 2-fold higher at 12–15 months, when senile plaques are present. Thus, in this model, increased CNS synthesis of C3 precedes senile plaque formation. Because C3b, an activation protein produced by cleavage of C3, functions as an opsonin [ 89 ], the increased C3 levels together with the reduced Aβ deposition in the APP/TGF-β1 mice suggested a neuroprotective role for complement in this model. To investigate this possibility, the APP mice were crossed with mice expressing soluble complement receptor-related protein y (sCrry), a rodent-specific inhibitor of early complement activation [ 90 ]. APP/sCrry mice had a 2- to 3- fold increase in Aβ deposition in the neocortex and hippocampus at 10–12 months of age, together with a 50% loss of pyramidal neurons in hippocampal region CA3. The authors concluded that complement activation may protect against Aβ-induced toxicity, and may reduce the accumulation or promote the clearance of amyloid and degenerating neurons [ 88 ]. Neuroprotective functions (protection against excitotoxicity) have been demonstrated in vitro for C3a [ 52 ], and the increased neuronal loss in the APP/sCrry mouse may be due to decreased production of C3a as well as the opsonin, C3b. However, whether inhibition of complement activation in the AD brain would similarly result in increased neuropathology is unclear, because complement activation in AD is likely to be more extensive than in the APP mouse. Although no peer-reviewed articles have appeared in which the extent of complement activation in the APP mouse has been examined, two abstracts have dealt with this issue. Yu et al. [ 91 ] reported C3, C5, and C6 immunoreactivity to thioflavin-S-reactive plaques, whereas McGeer et al. [ 92 ] found only weak complement staining of plaques and slight upregulation of complement proteins. Significantly, neither study reported detection of the MAC. At least two factors, in addition to the lack of NFTs, mitigate against complement activation in the APP mouse being equivalent to that in AD: (a) the mouse complement system is functionally deficient, as mouse C4 lacks C5 convertase activity [ 93 ] and many mouse strains have low complement levels relative to other mammals [ 94 ], and (b) mouse C1q binds less efficiently to human Aβ than does human C1q, resulting in less activation of mouse complement than of human complement in the presence of human Aβ [ 95 ]. PS/APP mouse In addition to APP, mutations in the gene encoding for presenilin-1 (PS-1) have also been associated with familial AD [ 96 ]. The PS/APP mouse carries both of these transgenes and has been extensively used as a model for studying processes relating to the formation of SPs. Aβ deposition occurs more rapidly in these mice than in the single transgenic APP mouse [ 97 ]. In neither model does NFT formation occur. Aβ deposition in PS/APP mice is initially detected at 3 months of age, and increases with age; total Aβ burden peaks at one year of age, although the percentage of Aβ that is fibrillar (thioflavin-S reactive) increases up to 2 years of age. Matsuoka et al. [ 98 ] described the CNS inflammatory response to Aβ in these animals. Activated astrocytes and microglia increased in parallel with total Aβ and were closely associated with both diffuse and fibrillar plaques. C1q immunoreactivity was detected at both 7 and 12 months of age, co-localizing with activated microglia and fibrillar Aβ. These findings were similar to those in the AD brain in that complement activation was associated with SP formation. The extent of complement activation was not addressed in this study. APP (Tg2576)/C1q-deficient mouse Fonseca et al. [ 99 ] investigated the role of C1q in AD by crossing Tg2576 (APP) mice [ 100 ] and APP/PS1 mice with C1q knockout mice [ 101 ]. C1q immunoreactivity was associated with plaque formation in the APP Tg2576 animals, as previously reported by Matsuoka et al. [ 98 ]. In both the Tg2576/C1q - and APP/PS1/C1q - animals, lack of C1q did not alter either plaque density or the time course of plaque deposition. Neuronal cell numbers (NeuN + cells), assessed only in the Tg2576 (APP) mouse, were not changed by the absence of C1q; however, immunoreactivity to MAP-2 (a marker for neuronal dendrites and cell bodies) and synaptophysin (a marker for presynaptic terminals) in the hippocampus (region CA3) was increased 2-fold in the APP/C1q - animals, compared with APP mice. Microglial and astrocytic activation was significantly reduced in the APP/C1q - animals. These results were interpreted to suggest that in these animal models of AD, (1) early complement activation (as indicated by C1q deposition) in response to fibrillar Aβ deposition might be responsible for the chemotactic attraction of activated glial cells, and (2) the activated microglia, while unable to clear fibrillar Aβ, may have contributed to the loss of neuronal integrity indicated by reduced MAP-2 and synaptophysin staining in the APP mice. By recruiting activated microglia, complement activation could potentially contribute to neuronal injury even if full activation (MAC formation) does not occur. Postischemic hyperthermic rat model Coimbra and colleagues [ 102 ] described progressive neuronal loss in the hippocampus and cerebral cortex in rats subjected to common carotid artery occlusion to produce transient forebrain ischemia, as an animal model for stroke. The post-surgical hyperthermia which occurs spontaneously in these animals was suggested to promote the infiltration of microglia, whose secretory products increased the subsequent neuronal loss. A later study by the same group [ 103 ] found that subjecting the rats to post-surgical hyperthermia (38.5 – 40°C) increased microglial and astrocytic infiltration and accompanying neuronal loss, and resulted in the formation of AD-type pathology. Aβ-reactive diffuse plaques were detected in the cerebral cortex at 2 months post-surgery, with more compact plaques in the hippocampus and cortex by 6 months. Increased ubiquitin and phosphorylated tau immunoreactivity was observed at both time points, together with staining for C5b-9 in the somatosensory cortex. The MAC immunoreactivity co-localized with acid fuchsin staining, a marker for neuronal death [ 104 ]. Other complement proteins were not evaluated in these studies. This is apparently the only animal model of AD in which full complement activation has been reported. It is noteworthy that while both SPs and neurofibrillary pathology were present in these animals, the MAC apparently did not co-localize with these structures, unlike in AD. Acute lesioning Alterations in native complement mRNA and protein levels have been evaluated in the rat hippocampus following experimental induction of acute neuronal injury. These surgical and pharmacological procedures result in neuronal loss in the entorhinal cortex, and deafferentation of hippocampal neurons, similar to that which occurs in AD [ 105 ]. Selective damage to the rat hippocampus has been induced by surgical transection of the perforant pathway, which runs between the entorhinal cortex and the molecular layer of the dentate gyrus [ 106 , 107 ], systemic administration of the excitotoxin kainic acid [ 108 , 109 ], or injection of the neurotoxin colchicine into the dorsal hippocampus [ 109 ]. Surgical transection of the perforant pathway increased C1qB mRNA in the entorhinal cortex and hippocampus [ 106 ] and C9 immunoreactivity in the hippocampus [ 107 ]. Injection of kainic acid similarly increased C1qB and C4 mRNA expression and C1q immunoreactivity in the hippocampus [ 108 , 109 ]. Colchicine infusion into the dorsal hippocampus, which selectively damages granule cells of the dentate gyrus, produced elevated mRNA expression of hippocampal C1qB and C4 [ 109 ]. Though the acute neuronal damage in these studies differs from the chronic, progressive neurodegenerative process that occurs in AD, these results demonstrated that the neuronal response to injury includes upregulation of native complement protein synthesis. The significance of this upregulation, i.e. whether it promotes neuroprotection or neurotoxicity, was not addressed. Infusion of Aβ and C1q into rats Frautschy et al. [ 56 ] examined the effects of infusion of human C1q and oral administration of rosmarinic acid on glial cell proliferation (microgliosis and astrocytosis), plaque load, and memory (Morris water maze) in Aβ-infused rats. Rosmarinic acid inhibits both the classical and the alternative complement cascades, by covalent binding to newly formed C3b [ 110 ]; it also possesses anti-inflammatory [ 111 , 112 ], anti-oxidative [ 113 ], and anti-amyloidogenic properties [ 114 ]. Gliosis was greater with C1q and Aβ infusion than with Aβ alone. Plaque density was decreased by C1q infusion (note: this result differs from the in vitro study of Webster et al. [ 57 ], in which C1q was found to inhibit microglial phagocytosis of Aβ, and also from the recent study of Fonseca et al. [ 99 ] in which C1q deficiency had no effect on plaque density in APP mice), but, curiously, performance in the water maze worsened. Treatment with rosmarinic acid had the opposite effect; though plaque load increased, memory was improved. These findings were interpreted as suggesting that C1q and/or complement activation may, by promoting microglial activation, worsen memory independent of the clearance of Aβ. Additional animal models for studying AD-related complement activation TAPP and 3xTg-AD mice Mutations in the gene encoding for human tau protein have been linked to the development of frontotemporal dementia with parkinsonism [ 115 ]. By combining this mutation with the human APP and PS1 mutations associated with familial AD, animal models of AD have been produced in which NFTs as well as SPs are formed. Lewis et al. [ 116 ] crossed human APP swe mice (Tg2576) with mice expressing the transgene for a human tau mutation (JNPL3 mice) to generate a double mutant tau/APP mouse (the "TAPP mouse"). These mice develop SPs similar to APP mice (high numbers of plaques are present in older [8.5–15 months of age] mice, in the olfactory cortex, cingulate gyrus, amygdala, entorhinal cortex, and hippocampus), and older TAPP mice have NFTs, in association with increased astrocyte proliferation, in limbic areas. The plaques contain both Aβ 40 and Aβ 42 . Oddo et al. [ 117 ] injected the human transgenes for APP and mutated tau into embryos of PS1 "knock-in" mice, generating the "3xTg-AD" mouse which develops both SPs and NFTs in an age-related, region-specific manner. Aβ deposition in these animals precedes NFT formation, with extracellular Aβ (primarily Aβ 42 ) detected in the frontal cortex by 6 months of age, and in other cortical regions and hippocampus by 12 months. Many of the extracellular Aβ deposits are thioflavin-S-positive and are associated with reactive astrocytes. Phosphorylated tau initially appears in the hippocampus and subsequently in cortical regions; it is detected within neurons by 12–15 months and within dystrophic neurites at 18 months. Though Aβ immunoreactivity precedes that of tau, these proteins co-localize to the same neurons. The presence of NFTs as well as SPs suggests that the 3xTg-AD and TAPP models may be more relevant than APP or APP/PS-1 mice for studying the significance of complement activation in the development of AD-type pathology. Potential drawbacks for using these models for complement-related studies include, as discussed earlier, functional deficiencies in activation of mouse complement [ 93 ], decreased complement levels in common laboratory mouse strains [ 94 ], and the decreased efficiency of binding of mouse C1q by the human Aβ within the SPs in these animals [ 95 ]. It is not known whether a similar decrease in the efficiency of activation of mouse complement occurs when mouse C1q binds to human, rather than murine, tau protein. AD11 (anti-NGF) mouse Ruberti et al. [ 118 ] developed a mouse transgenic model, the AD11 mouse, in which neutralizing antibody to nerve growth factor (NGF) is secreted by neurons and glial cells. NGF exerts trophic effects on basal forebrain cholinergic neurons and is widely distributed in these neurons [ 119 ]; the local secretion of anti-NGF antibody in these mice results in marked loss of basal forebrain cholinergic neurons. Aβ-containing plaques, tau hyperphosphorylation, and NFTs are present at 15–18 months of age. CNS production of anti-NGF antibody increases with age in these animals, therefore pathology develops only in adult mice. Extracellular deposition of APP is widespread in the brain, including the cortex and hippocampus. Phosphorylated tau immunoreactivity is present in neurons and glia in the cortex and hippocampus, and intracellular NFTs, extracellular neurofibrillary deposits, neuropil threads, and dystrophic neurites are observed in the cortex. Behavioral abnormalities, including impaired object recognition and spatial learning, are associated with this neuropathology [ 120 ]. The Aβ-containing plaques in the AD11 mouse are of murine, rather than human, origin, allowing the problem of the poor efficiency of activation of mouse complement by human Aβ [ 95 ] to be overcome. However, it is unclear whether plaques in these animals contain Aβ in the β-pleated sheet conformation, which is thought to be the most effective conformation for activating complement [ 71 ]. The distribution of SPs and NFTs in this model is less similar to AD than for 3xTg-AD and TAPP mice, because in addition to the cortex and hippocampus, large numbers of APP-reactive structures are present in the neostriatum (where, in AD, plaques are primarily diffuse [ 121 ]), and in other areas of the brain. Despite these concerns, the AD11 mouse is attractive as a potential model for studying the significance of AD-related complement activation. Chlamydia pneumoniae -infected mouse C. pneumoniae is an intracellular, gram-negative or gram-variable bacterium long identified as a respiratory pathogen. It has more recently been demonstrated to be a causative agent in reactive arthritis [ 122 ] and to be associated with autoimmune disorders including multiple sclerosis [ 123 ] and atherosclerosis [ 124 ]. Some laboratories have also reported an association of this agent with AD [ 125 - 127 ], although this has not been confirmed by others [ 128 - 131 ]. A recent study by Little et al. [ 132 ] examined the hypothesis that experimental C. pneumoniae infection in BALB/c mice could produce AD-like pathology. Intranasal inoculation with C. pneumoniae resulted in deposition of Aβ 1–42 in the hippocampus, amygdala, entorhinal cortex, perirhinal cortex, and thalamus by 3 months post-inoculation. The majority of these Aβ deposits appeared similar to diffuse plaques, though a small number of them were thioflavin-S-reactive. NFTs were not detected. The authors suggested that soluble factors such as lipopolysaccharides, which are present in the cell wall of all Chlamydiae [ 133 ], may have been responsible for the altered amyloid processing which resulted in Aβ deposition. Because the Aβ within the SPs in these animals is of endogenous origin, and because other chlamydial species have been shown to activate complement [ 134 , 135 ], the C. pneumoniae -infected mouse may offer a novel infectious model for studying the relationship of complement activation to the development of Aβ-containing plaques. Aged dogs Old dogs, in particular the beagle, have been extensively investigated as a model for CNS Aβ deposition and associated age-related cognitive dysfunction. Aβ deposits are detectable in the brains of most older dogs [ 136 ]. The regional distribution of Aβ in the dog brain resembles that in humans, found initially in the prefrontal cortex, subsequently in entorhinal and parietal cortices, and lastly in occipital cortex [ 137 ]. Aβ 42 is the predominant type of Aβ deposited in plaques [ 138 ]. Canine plaques are nonfibrillar and do not contain neuritic elements; thus, they resemble diffuse Aβ deposits in the human brain, but not the mature plaques predominating in AD. The neuropathological findings in old dogs also differ from AD in that activated glial cells are rarely associated with Aβ deposits, and NFTs are not detected [ 136 , 139 ]. Age-related cognitive impairment, termed "canine cognitive dysfunction syndrome," occurs in some older dogs and correlates with Aβ deposition in the hippocampus and frontal cortex [ 140 , 141 ]. The endogenous nature of the deposited Aβ in old dog brain, and similarities between canine and human Aβ in their patterns of regional deposition, suggest that this model may be useful for studying the relationship between complement activation and plaque formation. Non-human primates Age-related formation of SPs has been reported in a variety of non-human primates including the cynomolgus monkey [ 142 ], rhesus monkey [ 143 ], chimpanzee [ 144 ], and marmoset [ 145 ]. Aβ within these plaques is predominantly Aβ 40 [ 146 ]. NFTs apparently do not form in the brains of most aged primates, with a few exceptions. The brain of the aged baboon contains phosphorylated tau protein [ 147 , 148 ], and an age-related accumulation of tau also occurs in the neocortex of the mouse lemur [ 149 - 151 ]. In this latter species, Aβ deposition occurs in the cerebral cortex and amygdala but is not age-dependent [ 151 ]. The mouse lemur appears to be the most promising primate species to date for studying the significance of AD-related complement activation because of the presence of NFTs as well as plaques. Other animal species Scattered reports of AD-type pathology in other species have also appeared. Adding trace amounts of copper to the water supply of cholesterol-fed rabbits results in Aβ deposition within SP-like structures in the hippocampus and temporal cortex, with associated learning deficits [ 152 ]. The neuropathology in the aged cat is similar to that in the old dog in that Aβ is deposited only as diffuse, Aβ 42 -containing plaques, and NFTs are not detected [ 138 ]. A report of AD-type pathology in an aged wolverine [ 153 ] described neuritic as well as diffuse plaques in the cortex and hippocampus, and intracellular NFTs containing phosphorylated tau protein in cortical and hippocampal neurons. Finally, the aged polar bear brain also contains both diffuse plaques and NFTs [ 154 ]. While the neuropathological findings in the aged wolverine and polar bear resemble AD more closely than in most species examined to date, their inaccessibility to laboratory researchers limits the usefulness of these species for studies of AD-related complement activation. Conclusions 1. Complement activation has been extensively studied in the AD brain. There is convincing evidence for activation of both the classical and alternative pathways, resulting in full activation as indicated by the presence of the MAC. Both aggregated Aβ (in SPs) and phosphorylated tau (in NFTs) are likely to be responsible for this activation. 2. Because complement activation generates both both neuroprotective and neurotoxic effects, the significance of increased complement activation in the development and progression of AD is unclear. 3. An optimal animal model for studying the significance of complement activation in the development of AD-type pathology would have complete activation of this process, with co-localization of complement activation proteins with SPs and with NFTs (if present). Other desirable features include early complement activation prior to the development of extensive neuropathology, increased CNS production of native complement proteins, and both classical and alternative pathway activation. 4. Surprisingly little is known about the extent of complement activation in animal models of AD. The postischemic hyperthermic rat [ 103 ] is the only animal model of AD in which full complement activation has been reported. The few studies with APP-transgenic mice have yielded conflicting results, with one investigation suggesting a neuroprotective role for complement activation [ 88 ], while another found that early complement activation (as indicated by C1q deposition) was associated with a loss of neuronal integrity [ 99 ]. Transgenic mouse models may be problematic for studies of AD-related complement activation because of inherent deficiencies in mouse complement activation and inefficient activation of mouse complement by the human Aβ present in the SPs in these animals. Other animal models in which SPs (and NFTs, if present) are of endogenous, rather than human, origin offer alternatives to transgenic mice for studying this issue. 5. The extent of complement activation and its association with neuropathology must be determined in animal models of AD to clarify the relevance of these models for investigating the significance of complement activation in the development of AD-type pathology. Abbreviations used Aβ, amyloid beta; AD, Alzheimer's disease; APP, amyloid precursor protein; CNS, central nervous system; MAC, membrane attack complex; mRNA, messenger ribonucleic acid; NFTs, neurofibrillary tangles; NGF, nerve growth factor; PS-1, presenilin-1; sCrry, soluble complement receptor-related protein y; SPs, senile plaque; TGF-β1, transforming growth factor beta1. Competing interests The author declares that he has no competing interests.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC529311.xml
509016	Graph-based iterative Group Analysis enhances microarray interpretation	Background One of the most time-consuming tasks after performing a gene expression experiment is the biological interpretation of the results by identifying physiologically important associations between the differentially expressed genes. A large part of the relevant functional evidence can be represented in the form of graphs, e.g. metabolic and signaling pathways, protein interaction maps, shared GeneOntology annotations, or literature co-citation relations. Such graphs are easily constructed from available genome annotation data. The problem of biological interpretation can then be described as identifying the subgraphs showing the most significant patterns of gene expression. We applied a graph-based extension of our iterative Group Analysis (iGA) approach to obtain a statistically rigorous identification of the subgraphs of interest in any evidence graph. Results We validated the Graph-based iterative Group Analysis (GiGA) by applying it to the classic yeast diauxic shift experiment of DeRisi et al., using GeneOntology and metabolic network information. GiGA reliably identified and summarized all the biological processes discussed in the original publication. Visualization of the detected subgraphs allowed the convenient exploration of the results. The method also identified several processes that were not presented in the original paper but are of obvious relevance to the yeast starvation response. Conclusions GiGA provides a fast and flexible delimitation of the most interesting areas in a microarray experiment, and leads to a considerable speed-up and improvement of the interpretation process.	Background Microarray experiments can provide a comprehensive picture of gene expression levels in biological samples. In a typical application they compare expression of several thousand genes under two different conditions (e.g. healthy vs. diseased tissue, wild type vs. mutant animals, drug-treated vs. control cells), using a small number of replicate experiments. Various techniques have been developed to rank genes according to their expression changes, e.g. based on the t-statistic [ 1 ] or the strong non-parametric RankProducts [ 2 ]. The resulting list of genes can then be restricted to those genes that fulfill a certain statistical criterion, usually an arbitrarily chosen maximum accepted false discovery rate. The main challenge to the biologist is contained in the next step of the analysis. It consists in identifying the biologically relevant expression changes, the "big picture" of the experiment. As microarray experiments tend to generate unexpected observations in areas outside the specialized expertise of the experimentalist, this can be quite difficult and time-consuming. A principled mechanism to identify the significant higher-level features of the experimental results would therefore be very useful. The biological interpretation process consists to a large extent of finding evidence connecting certain genes that are differentially expressed. This evidence can consist, e.g., of joint participation in some physiological process, physical interaction at the protein level, reported co-expression in earlier microarray experiments, a shared functional annotation, etc. This kind of evidence can intuitively be represented as a graph, and this feature is regularly used to visualize biological data, in the form of metabolic or signaling pathways or protein interaction maps. The task can then be described as the identification of subgraphs that as a whole show a statistically significant expression change. This would allow the biologist to focus her analysis on the most promising areas, without prior bias, while at the same time presenting the relevant evidence underlying each association for critical evaluation. Results and Discussion The Algorithm We have recently developed an approach, iterative Group Analysis (iGA) that identifies significantly changed functional classes of genes in a microarray experiment [ 3 ]. In contrast to similar approaches such as [ 4 - 8 ], the iGA method does not require a previous delimitation of a set of "differentially expressed genes", but uses an iterative calculation of p-values to determine the subset of class members that is most likely to be changed. Due to this feature, the iGA method is more sensitive in identifying functional classes that are slightly but consistently regulated, and works well on noisy data with small numbers of replicates, where the delimitation of gene lists can be overly restrictive. Here we extend this approach to the analysis of "evidence graphs", which offers much larger flexibility of the annotations that can be used and allows substantially improved visualization. Evidence graphs can be represented as bigraphs with two types of nodes, one for genes and one for the associated "evidence" (Fig. 1A ). For evaluation purposes we focused on two types of networks, one where the evidence consists of GeneOntology annotations (GO network) and one where the evidence comprises enzyme substrates (metabolic network). The construction of these networks from gene annotation files is fast and simple, but more unconventional networks are also straightforward (e.g. regulatory networks inferred from previous gene expression experiments, or "literature networks" based on co-occurrence of genes in publications). Before the calculation the bigraph is converted to a simple graph, eliminating the evidence nodes and introducing pairwise edges between all nodes that were connected to a common evidence node (Fig. 1B ). In addition to the graph, a complete list of genes sorted by differential expression is provided. All nodes without corresponding expression data are eliminated from the network. In the first step of the analysis, each gene node is assigned its rank in the list of genes, such that the node for the most strongly changed gene is labeled "1" (Fig. 1C ), the second most changed gene is labeled "2", etc. Then, local minima are identified in the graph, i.e. those nodes that have a lower rank than all their direct neighbors in the graph (Fig. 1D ). In the next step, subgraphs are iteratively extended from each of those local minima by including the neighboring node with the next highest rank ( m ) and, if present, all adjacent nodes of ranks equal or smaller than m . Hence, at each step of the extension process the newly extended subgraph is not adjacent to any outside node with a rank lower than m (Fig. 1E,1F,1G,1H ). At each extension step we thus obtain a subgraph with n members with a maximum rank m and can calculate the p -value for observing all n of n genes at rank m or better in a list of N total genes in the graph, , which follows easily from inserting these values in the cumulative hypergeometric equation, which is also used for iGA [ 3 ]. The extension process is continued until all nodes reachable from the local minimum are included or the subgraph reaches an arbitrary maximum size. After extending the subgraphs, for each local minimum the subgraph yielding the smallest p-value is selected as its "regulated neighborhood" and all local minima are sorted by increasing p-value of these regulated neighborhoods. The subgraphs at the top positions of the resulting list should contain the most relevant regions of the total evidence network. Comparison with Previous Approaches The method devised by Ideker et al. [ 9 ] for the determination of signaling and regulation circuits from a combination of protein interaction information and expression data could easily be extended to cover the more general case of microarray interpretation addressed by GiGA. This would only require the extension of "interaction" to include, e.g., participation in a shared cellular process or shared functional annotation. Their approach uses aggregate z-scores to evaluate the quality of each subgraph. This requires a relatively complex parameter estimation procedure followed by simulated annealing. In contrast, the rank-transform of the data that is the basis of GiGA allows non-parametric p-value calculations and is thus much faster (and computationally less demanding) than the method described by Ideker et al. A disadvantage of both methods is that they both do not guarantee to find the optimally scoring subgraph. The recently released commercial Pathway Analysis software from Ingenuity Systems seems to be based on a similar concept as GiGA, i.e. the determination of regulated subgraphs in annotation networks. However, it just classifies genes as changed ("focus genes") or unchanged based on an arbitrary selected significance cut-off and thus discards most of the relative-change information (gene ranks) used by GiGA . Therefore, this method is difficult to apply to very noisy or unreplicated experiments where a reliable delimitation of the "changed" genes becomes impossible. Experimental Case Study To validate the GiGA approach, we used the yeast diauxic shift experiment by DeRisi et al. [ 10 ]. In this classical study the authors examined the response of yeast cells to glucose depletion in the growth medium. As the biology of this process is extremely well understood and the functional annotation of the yeast genome is very comprehensive, we could use this dataset to test the ability (and reliability) of GiGA to identify the relevant subgraphs of interest. In their original publication, DeRisi et al. highlight the following changes during starvation: Rechanneling of metabolites into the tricarboxylic acid (TCA) and glyoxylate cycle, increase in aerobic respiration (cytochrome c oxidase and reductase), gluconeogenesis, and carbohydrate storage (glycogen and trehalose biosynthesis). In contrast, 95% of ribosomal proteins, as well as tRNA synthetases and translation elongation/initiation factors were strongly down-regulated. Twenty hours after the initial inoculation of the sample about 20% of all genes showed at least a two-fold change in expression. Table 1 and 2 show the result of an iterative Group Analysis of these expression data. Seven time points after inoculation were examined. Significant expression changes become apparent at 13.5 hours. It can be seen that all the processes identified manually by DeRisi et al. are already apparent at this level of analysis. In addition, iGA highlights the up-regulation of sugar transporters at the early stages of starvation, obviously a desperate attempt by the cells to take up the last remaining sugars from the medium. It also highlights the induction of heat shock proteins and the repression of ribosome biosynthesis processes in the nucleolus, which were not discussed in the original paper. What is missing in these lists, however, is the automatic identification of the interrelationships between the identified processes, which would be particularly informative in more realistic applications with a less well understood biology. This connectivity between genes and functional classes is provided by GiGA. Table 3 summarizes the results for the 20.5 hour time point, using two different networks, one for GeneOntology classes, and one for enzyme substrates, extracted from the SwissProt catalytic activity descriptors of yeast proteins. For both up- and down-regulated genes, the most significant subgraph is widely separated from the next best one, contains the largest number of genes, and comprises almost all the processes detected by iGA and the original authors. (We here restricted the size of subgraphs to a maximum of 40 genes to keep navigation of the results simple.) Figures 2 and 3 show the automatically generated visualization of the corresponding most significant subgraphs and the associated annotation. It is obvious how GiGA highlights the functional connections between different enzymes. In Fig. 2 , the association between small and large ribosomal subunits, nucleolar rRNA processing and translational elongation is faithfully reconstructed. In Fig. 3 , which uses the enzyme substrate network (which we considered to be more challenging for the algorithm), the interplay between the TCA cycle, the overlapping glyoxylate cycle, and all the relevant protein complexes of the respiratory chain are readily apparent. The agreement with the manual interpretation by DeRisi et al. extends down to the single gene level, while at the same time adding additional, obviously relevant connections, e.g. from the TCA cycle to the ATP synthase complex. Fig. 3 also shows the second best subgraph, which contains the cytochrome c oxidase subunits together with two catalase genes that may be involved in the detoxification of the hydrogen peroxide generated by the respiratory burst induced by starvation. The performance of GiGA (as well as iGA) is best appreciated when compared to the results of an extensive expert interpretation of the same data. Table 1 to 3 , and Fig. 2 and 3 show how both techniques succeed in detecting and condensing exactly the genes and processes that were considered relevant by the expert biologists when first interpreting the same data [ 10 ]. GiGA effectively summarizes the original publication in three subgraph pictures (Fig. 2 and 3 ). This is even more astonishing when considering that these results are achieved for each single time-point separately (see, e.g., Tab. 1 and 2 ). This reveals the stability of the approach towards the measurement variance inherent in any unreplicated microarray experiment. It is important to be aware that each of the highlighted subgraphs has to be carefully evaluated for its biological relevance. On the one hand, the sheer number of possible subgraphs in an evidence network creates a major multiple-testing problem, which means that some of the detected associations may be due to chance. Random permutations of the expression data – which can be generated by the GiGA software – can give an idea of the expected false-discovery rate. On the other hand, functional annotations are at present notoriously unreliable and spurious edges may affect the details of the results. Also, not all genes within a detected subgraph will necessarily show a strong expression change, because sometimes less strongly affected genes may connect those genes that do change. Such a relation is for example expected for many transcription factors and their targets [ 9 ]. Nonetheless, GiGA is able to direct the user to the most interesting areas in the evidence network. The GiGA method is not restricted to use with exhaustively annotated genomes. It can work on a wide variety of "evidence" to build the necessary network, including hypothetical predicted functions or associations. It is even possible to apply GiGA to metabolomics results, which are characterized by the absence of any significant amount of annotation – usually only exact molecular masses and their differential abundance are known. In that case, an evidence network can be built from the measured masses themselves, linking compounds m 1 and m 2 whenever their mass difference (Δ m = \| m 1 - m 2 \|) can be explained by a common biochemical transformation (e.g. dehydrogenation: Δ m = 2* mass ( hydrogen )). The set of relevant transformations can easily be collected from any biochemistry textbook. In addition, one can introduce edges for condensation reactions between observed masses, i.e. if m 1 + m 2 = m 3 + mass(H 2 O) then edges between m 1 and m 3 , and m 2 and m 3 are added to the evidence network. We are currently developing the application of GiGA to this kind of data in the Sir Henry Wellcome Functional Genomics Facility at the University of Glasgow ; data not shown). Conclusions The present analysis of a biologically well-understood test case demonstrates the reliable performance of GiGA. The method automatically identifies all relevant physiological processes, puts them into context, summarizes them in an intuitive format, and associates them with the underlying evidence (Fig. 2 and 3 ). It can be applied to experiments with very small numbers of replicates (a single time point in the diauxic shift test case) and can be used with any available functional annotation, including protein interaction networks, co-expression data or literature mining results, as well as in areas beyond microarray analysis. For visualization, we have used the graph-layout software aiSee, but output files suitable for a variety of graphical tools can easily be generated by slight modifications in the implementation. GiGA can be used as a stand-alone tool, but we expect that it will be most useful when integrated into existing microarray analysis software, and for that reason the GiGA algorithm is freely available without restrictions. Methods Yeast gene expression data for the diauxic shift experiment were obtained from the Stanford Microarray Database . GeneOntology annotations were obtained from Affymetrix . The enzyme substrate networks were built based on information contained in the annotation of the yeast proteome in the SwissProt database . The GiGA algorithm has been implemented as a Perl script and compiled as a Windows command line executable. These files are available (together with a manual and example files) as Additional files 1 to 8 . Authors' contributions RB devised and implemented the GiGA technique and drafted the manuscript. AA and PH supervised the project. All authors read and approved the final manuscript. Supplementary Material Additional File 1 GiGA program. For use from the Windows command line. Click here for file Additional File 2 GiGA source code. Click here for file Additional File 3 GiGA manual. Describes the use of GiGA applied to the example data (Additional files 4 to 6). Click here for file Additional File 4 Gene expression data. Sorted list of genes, based on expression during the yeast diauxic shift. Click here for file Additional File 5 Evidence network. List of gene pairs connecting genes whenever their gene products are enzymes that share a common substrate. Based on annotation derived from SwissProt. Click here for file Additional File 6 Genenames file. Contains descriptive names of the yeast genes contained in Additional file 4. Click here for file Additional File 7 Example output in text format. List of significantly affected subgraphs detected in the experimental data (Additional file 4) using GiGA with default settings. Click here for file Additional File 8 Example output in graph-description language format. Contains the same results as Additional file 7, but in a format that can be visualized and explored using graph-layout software, e.g. aiSee, which is freely available for academics at . Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC509016.xml
524190	Methods to find out the expression of activated genes	This review deals with the methods of identifying genes that have been activated by inner or outer impulses. The activation and subsequent expression of a gene can be detected by its transcription into a corresponding messenger ribonucleic acid (mRNA). Principles of the methods for identification of individual activated genes, as well as groups of activated genes are described, the former methods being mostly based on subtractive hybridization and serial analysis of gene expression (SAGE), the latter on microarrays. Examples of gene activation by the hormone 17beta-estradiol (E2) are given.	Introduction In previous reviews, methods for the measurement of receptors and their interactions with other transcription factors and genes were described [ 1 - 3 ]. In this review, gene activation is discussed with a particular emphasis on the methods enabling detection of the activated, turned-on, genes. The action of the hormone 17beta-estradiol (E2) is taken as an example of the function of many other small-molecule compounds in gene activation and in the expression of the activated gene. The life of humans and animals is influenced by the activity of a series of genes that are kept in a silent state, or are activated, depending on the temporary needs of the body. This switching on and off of each gene is executed by an assembly of transcription factors forming a transcription initiation complex (TIC). Examples of such transcription factors are estrogen receptors (ER-alpha, ER-beta, and possibly other isomers) that, before being incorporated into a TIC, have to be activated by E2. This hormone itself is synthesized, when an initial signal is given, by virtue of an activation of a series of appropriate genes. Via ER, E2 has manifold biological effects. Biological targets of E2 are, inter alia, blood vessel walls [ 4 - 8 ], blood platelets [ 9 ], bone [ 7 , 10 - 12 ], breast cancer cells [ 13 ], central nervous system [ 7 , 14 , 15 ], retinal pigment epithelium [ 16 ], synthesis of clotting factors [ 17 ]. It is evident that E2 is associated with many biological effects and that many genes must be involved. Consequently, ER must be able to bind to DNA segments, called response elements, in the neighborhood of various genes. The response elements participate, together with other transcription factors, in the formation of TICs that are specific for each gene. An important problem, currently studied in many laboratories, is to find out which genes are activated in various circumstances. The methods that solve this problem are based on a comparative (differential) approach. A test (target) sample, containing active genes is compared with a control sample in which the genes have not been activated. Using this approach, the active genes are singled out among the multitude of inactive genes. However, the comparisons may reveal the opposite of activation, i.e., downregulation of genes. Generally, the activity of a gene is characterized by its transcription into mRNAs as the first step leading to the synthesis of specific proteins. Non-activated genes in the control tissue do not produce any corresponding mRNAs. In most methods, the mRNAs prepared from the test and control tissue are each reverse transcribed into the corresponding complementary deoxyribonucleic acid (cDNA), in order to enable a substantial increase of the material for analysis by polymerase chain reaction (PCR) [ 2 ]. As most methods do not operate with full mRNA transcripts, but with shorter sequences, the allocation of such sequences to known (or unknown) genes has to be found by advanced computer programs and gene databases. The methods used for the identification of active genes are sketched below. Included are even methods that have not yet been used for the identification of E2-activated genes. It has to be mentioned that only principles, not technical details are dealt with in this review. Neither the techniques of cloning or of identification of genes by sequencing are described here. The readers who are not familiar with these techniques are advised to consult appropriate textbooks [e.g., [ 18 ]]. The dedicated computer programs and databases that are needed for the identification of sequences or genes will not be described here either. These can be found in the references quoted below. It will only be mentioned here that the large databases are GenBank and Celera . Activated (expressed) genes can be found by comparison of gene contents in the test and control tissues. There are essentially two approaches for finding activated genes: (i) an individual identification, or (ii) an identification of expression profiles after hybridization to a set of known gene fragments (probes) attached to chips in microarrays. Individual identification This approach means that genes are identified individually, even if several genes can eventually be picked up after cloning. There are several methods that can be used. Differential display Differential display seems to be the technically simplest method. Its name stems from the end-point that is a comparison of a side-by-side display of the test and control preparations by electrophoresis. In its basic form, total RNA of the test and control samples is separately subjected to reverse transcription into cDNA that, in turn, is PCR-amplified using arbitrarily chosen primers. The products are applied to a gel electrophoresis and the band(s) that are specific for one of the preparations are cut from the gel, further amplified by PCR (using the same primers) and eventually sequenced [ 19 ]. In a more advanced version, mRNAs of the test and control cells are separately reverse transcribed to cDNA (Fig. 1 ). Each transcription is carried out in the presence of a oligo(dT) primers, directed to the poly(A) tail at the 3' terminus of the mRNA and constructed as 5'(NMT11)3' where N can be guanine (G), adenine (A), thymine (T), or cytosine (C), and M is G, A, or C [ 20 - 22 ]. The primers with G residues are superior to those having one C residue. Those ending in A or T are the least efficient. With use of an arbitrary decamer as the second primer, a PCR is carried out to amplify the transcript in order to obtain a sufficient working material. This is usually done in the presence of a radioactive nucleotide. Other methods are commonly used, such as silver staining. Amplified DNA fragments are separated on a denaturing polyacrylamide gel, the test preparation side by side with the control. Each band differing from those seen in the control electrophoresis is then used for sequencing, subcloning, or as a probe for cDNA library screening. Large amount of results can be obtained depending on the variation in N and M nucleotides. In spite of the basic simplicity of the procedure, the time and workload can be considerable, depending on the number of NM combinations tried. Figure 1 Principle of a differential display. Test and control mRNA are separately reverse transcribed in the presence of anchored oligo(dT) primers containing nucleotides N and M in various combinations (see the text). The same primer and an arbitrary decamer are then used as primers in a PCR. The products are subjected to electrophoresis (PAGE). An additional band (see arrow) in the test sample represents a gene that had not been activated in the control sample. A11 and T11 denote eleven A and T molecules, respectively. Subtractive hybridization with hydroxylapatite separation The test mRNA is reverse transcribed into cDNA [ 23 ]. This is hybridized with the mRNA of the control sample (Fig. 2 ). A portion of the test cDNA (corresponding to the activated gene) does not find any complementary part in the mRNA of the control sample and remains non-hybridized as a single-stranded cDNA (ss-cDNA). This can be isolated by chromatography on a hydroxylapatite column. The hybridization of the isolated ss-cDNA with control mRNA followed by another chromatography can be repeated to increase the purity of the isolated product [ 23 ]. A cDNA library is produced and the subtracted sequence eventually identified. Alternatively, a second hybridization of the isolated ss-cDNA is carried out with the original test mRNA giving rise to a cDNA-mRNA hybrid which, after conversion to double stranded cDNA, is inserted into a vector, a cDNA library is constructed and several specific cDNA clones are isolated, leading to the identification of several genes [ 24 ]. Figure 2 Flow-sheet of subtractive hybridisation with hydroxylapatite separation. Test mRNA is reverse transcribed into a cDNA. This is hybridized with control mRNA. The non-hybridized portion of the single-stranded sequence of test cDNA is separated by chromatography on hydroxylapatite (HAP) and further processed. In another variant [ 25 ], the test and control mRNAs are both reverse transcribed into cDNA. cDNA of the test sample is hybridized with cDNA of the control sample. The non-hybridized part of the test cDNA is a single-stranded DNA that is separated by hydroxylapatite. The single-stranded DNA is cloned into a vector to produce a subtracted library. Clones with a strong hybridization signal to the subtracted probe are selected and sequenced. Subtractive suppression hybridization with PCR Isolation of a single-stranded test cDNA is not needed in this method. mRNAs of the test and control samples are prepared and each is reverse transcribed into cDNA. Each transcript is digested with the enzyme Rsa I to obtain shorter, blunt-ended fragments. The test cDNA is divided into two portions (see Fig. 3 ). One of them is ligated with Adapter A, the second with adapter B. Each portion is hybridized with an excess of control cDNA. A mixture of hybridization products is formed (Fig. 3 ). A tiny fraction of cDNA remains unhybridized, single-stranded. This is a fragment that may be called specific, or differentially expressed, or subtracted. It originates from the gene that had been activated. It is absent in the control sample. This specific fragment is bound either to Adapter A or B in the two portions. In the second hybridization, the portions are mixed. After annealing, a small amount of the specific fragment is obtained double-stranded. It contains Adapter A on the one end and Adapter B on the other. After adding primers specific for the Adapters, the ends are filled and the specific fragment is amplified by PCR to make sure that sufficient amounts are available for a further processing. Cloning, sequencing and comparing with a gene database establish the identity of the gene(s) [ 26 , 27 ] [ – "PCR-Select Subtraction kit"]. In contrast to the above methods, the primers for PCR amplification are clearly defined, avoiding thus problems with random primers. This method was used in a number of studies, such as the identification of genes upregulated in rats by E2 and progesterone treatment [ 28 ]. A predecessor of this technique is the "representational difference analysis" [ 29 , 30 ]. Figure 3 Outline of subtractive suppression hybridisation with PCR. Test cDNA and control cDNA are digested with Rsa I. The test cDNA sequences are divided into two halves, one of them being ligated with Adapter A (empty squares), the second one with Adapter B (filled squares). Each half is hybridized with control cDNA. The single-stranded (non-hybridized) sequences of both halves (denoted by asterisks) are annealed in a second hybridization step, primers to the Adapters are added and, after PCR, cloning and gene identification are carried out. Expressed sequence tags (EST) To describe the EST method, the following example is given. cDNA libraries were prepared by reverse transcription from mRNAs of the tissues to be examined [ 31 ]. The libraries were converted to plasmids, transfected into Escherichia coli and plated. Hundreds of clones were picked at random. These were subjected to sequencing, followed by computer matching to known genes listed in the GenBank database. The average length of a sequence was 397 bases; ESTs longer than 150 bases were found to be most useful for similarity searches and mapping. Subtractive hybridization (see above) was used to isolate the ESTs specific for one of the libraries. For example, a fibroblast cell line cDNA library was hybridized with a hippocampus library; the common sequences were removed and the specific hippocampus sequences remained. Using the EST method, more than 2000 human brain genes were identified [ 32 ]. Serial Analysis of Gene Expression (SAGE) The SAGE allows serial analysis of gene expression, an analysis of thousands of transcripts. It is based on the assumption that a short nucleotide sequence 10 base pairs (bp) – a tag – contains sufficient information to uniquely identify a transcript. In this respect SAGE differs from the EST approach. The principle of SAGE is as follows: mRNA is reverse transcribed into cDNA with use of a biotinylated primer, the cDNA is cleaved with a restriction endonuclease and the 3' portions are then isolated by binding to streptavidin beads [ 33 ]. In another version( ) (Fig. 4 ), mRNAs are captured prior to reverse transcription on oligo(dT) magnetic beads. Double stranded cDNAs are synthesized and digested with the restriction endonuclease Nla III that cleaves most transcripts at least once. The part attached to the magnetic bead is further processed. The reaction mixture is divided into two portions. The portions are ligated via a restriction site R to an adapter A and B, respectively, each consisting of 40 bp. Taking advantage of the restriction sites R, both portions are cleaved with the restriction enzyme Bsm FI in the distance of 14 bp. In this way "tags" are formed. Out of these 14 bp, 4 bp are a non-specific segment GTAC. These tags are blunt-ended with the Klenow fragment of DNA polymerase I. The two separate pools of tags are ligated together via a blunt-end ligation to produce "ditags". The ditags, flanked by the adapters A and B, are amplified by PCR with use of primers for A and B. The adapters are removed by the enzyme Nla III and the ditags are concatenated. The resulting concatemers (a series of linked ditags) are cloned into a plasmid vector to create a SAGE library. Individual clones are then sequenced. SAGE is carried out for each sample to be compared. Figure 4 Flow-sheet of SAGE. mRNAs are captured on oligo(dT) magnetic beads (open ovals). Double stranded cDNAs are synthesized. They are digested with Nla III. The product is divided into two halves. These are ligated to 40 bp adapters AR and BR, respectively. Both adapters contain a sequence R that is a recognition site for the restriction enzyme Bsm FI. This cuts a 14 bp sequence 3' of the site, forming a 10 bp tag. After cleavage with Bsm FI, the tags are ligated to form a product containing a ditag (the points of ligation are denoted by filled circles). This is amplified using primers complementary to A and B. The AR and BR adapters are cut away with Nla III to release a ditag. These are ligated to form concatemers containing multiple ditags. The concatemers are cloned and sequenced. Thanks to the concatenation, many tags can be detected in a single clone [ 33 ]. As each tag is supposed to uniquely identify a transcript, SAGE can generate a comprehensive profile of gene expression. Indeed, many unique transcripts were identified with use of SAGE tags [ 34 ]. The method is particularly useful for detecting genes of low level of expression or in rare tissues (e.g., early embryo) [ 35 , 36 ]. In addition, the amount of individual tags provides quantitative estimates of gene expression [ 37 ]. Still, the specificity of detection of genes with use of the short tags is not absolute. There are two main problems [ 38 ]. The first one is that many SAGE tags have no match to known sequences in databases. These tags may represent so far unidentified genes, but their shortness makes it difficult to characterize the genes. The second problem is that the SAGE tags may find multiple matches in the databases [ 39 , 40 ]. Therefore, attempts have been made to increase the specificity by prolongation of the tags by various methods. One such method is called GLGI (Generation of Longer cDNA fragments from SAGE tags for Gene Identification) [ 34 , 38 , 40 ]. The main feature of this method is the use of a SAGE tag as the sense primer for the PCR of a segment of cDNA. An anchored oligo(dT) serves as an antisense primer. In this manner a cDNA "tag" of up to several hundred bases is created. However, this method does not seem to improve the specificity of SAGE because even "non-specific" tags are co-amplified. Better of seems to be another variant of SAGE, the LongSAGE [ 41 ]. This is based on the use of tags 21 bp (out of which 4 represent a restriction site), tags longer than those in SAGE. The prolongation of tags is achieved by the use of the restriction endonuclease Mme I. The longer tags increase the power of identification of genes, while not diminishing the sensitivity of SAGE given by the use of PCR and concatenation. Theoretical calculations showed that >99.8% of the 21 bp tags were expected to occur only once in a genome. SAGE was used for the investigation of differences in gene expression in various health conditions. In the studies of breast tumors [ 37 ], global gene expression profiles in breast carcinoma cells were compared with those in normal mammary epithelial cells. The patterns of gene clusters in normal tissue were distinctly different from those of tumors of different stage and histological grade. The most dramatic change occurred at the normal-to-in situ carcinoma transition. This change can be an important marker for an early diagnosis. In another study, several genes regulated by estrogen or tamoxifen were identified in an estrogen-dependent breast cancer cell line. One of them was studied closer. It appeared to play a significant role in estrogen-promoted cell growth [ 42 ]. Gene profiles – microarrays The DNA microarray analysis is used to identify profiles of expressed genes in a given tissue and time. Thousands of known cDNA sequences or oligonucleotides are imprinted on a solid support, sometimes called a chip (e.g., a microscope slide or a nylon membrane), using application robots. Typically, individual spots are 100–300 micrometers in size and are spaced about the same distance apart [ 43 ]. More than 30,000 sequences can be fitted on the surface of a chip. These sequences serve as probes. Alternatively, the probes are synthesized in situ (60-mers) [ 44 ]. By hybridization, test (target) sequences (cDNAs or cRNAs) are bound to the cognate probes. The basic approach is the comparison of degree of hybridization in the control and test preparation. There are two basic techniques for the detection of hybridization. The control and test preparations are placed on a single chip, or, separately, on two chips. In the single chip technique [ 18 ], mRNAs from the control and test cells/tissues are separately reverse transcribed. During the transcription processes two different fluorescent dyes (e.g., Cy3 – green, Cy5 – red) are incorporated into the control and test cDNAs, respectively. The labeled molecules are mixed and hybridized to the cDNA array. There is a competition for each probe on the chip between the control and test mRNAs. The test cDNAs are selectively bound to some probes, the control cDNAs may be bound to other probes. With use of fluorescence scanning it is possible to distinguish the hybrids with control sequences (exhibiting, e.g., green fluorescence) from the hybrids with test sequences (e. g., red) [ 45 ]. Alternatively, the dyes may be reversed, and the control and test cDNAs may be labeled with the red and green dye, respectively. The hybrids that arise when the control and test cDNA occur in equal amounts may show a yellow fluorescence. The black spots indicate no hybridization (Fig. 5 ). One of the commercial companies utilizing this approach is Agilent . Figure 5 Model of a microarray. In a single-chip technique reverse transcription from mRNAs to cDNAs is separately carried out for the test and control cell preparations. During the transcription one of the fluorescent dyes (e.g., Cy3 – green and Cy5 – red) are incorporated into the cDNAs of each preparation. A mixture of these two preparations is then hybridised to the corresponding gene-representing sequences on a chip. The activated genes of the control sample exhibit green color, those of the test sample provide red spots, equally bound cDNAs can be visualized by yellow spots, no hybridization remains black. Using a variant of the method [ 46 , 47 ], certain groups of activated genes could be defined as predictors of the clinical outcome of breast cancer. Up to 5000 genes were tested for up-regulation (red) or down-regulation (green) in up to 100 patients with various degrees of disease progression. Correlations of disease grades with gene expression profiles were established, and a strategy was provided to select patients who would benefit from adjuvant therapy. In the two-chip technique, mRNAs of the test and control tissues/cells is reverse transcribed into a double-stranded cDNA from which a cRNA is prepared. In the course of the cRNA synthesis biotin molecules are incorporated [ 48 ]. The control and test cRNAs are separately hybridized to two identical chips. The binding is detected by staining with a fluorescent dye coupled to streptavidin. Signal intensities are used to calculate the relative cRNA abundance for the genes represented on the array. For comparisons of the intensities on both chips advanced computer programs have to be used. A combination of single-chip and two-chip techniques was applied in a study [ 51 ] where two chips and two fluorescent dyes were used. Commercial systems are available from several sources. For example, Affymetrix (GeneChip) [ ] produce chips by a photolithographic method in which thousands of different oligonucleide probes are synthesized in situ on the chip [ 49 ]. A compact technique has been introduced by the Febit company [ 50 ]. In a single benchtop instrument called Geniom a light-activated oligonucleotide microarray synthesis takes place, as well as addition of biotin-labeled cRNA sample, hybridization and fluorescence detection after incubation with streptavidin-phycoerythrin [ 50 ]. Other systems for microarray production, target preparation, hybridization and result evaluation are offered by Amersham Biosciences and Clondiag Chip Technologies . As a rule, more than one gene is activated, and a spectrum of genes is discovered either occurring sporadically or in clusters [ 49 ]. For example, when a diseased tissue was compared with a healthy one, an expression profile, a disease fingerprint, was identified [ 49 ]. In the case of breast tumors, a molecular portrait of each tumor was obtained [ 52 ], or, molecular profiling (a set of gene clusters) provided predictions of responses to adjuvant treatment [ 46 , 53 ]. Gene activation in breast cancer cells in the presence of E2 included, apart from the known estrogen-responsive genes, a series of novel genes expressing growth factors and components of the cell cycle, adhesion molecules, enzymes, signaling molecules and transcription factors [ 48 ]. Gene expression patterns of breast carcinomas allowed to distinguish tumor subclasses [ 54 ]. E2 caused up-regulation of 250 genes in vascular endothelial cells that could be prevented by an inhibitor [ 55 ]. In an experimental encephalomyelitis a markedly enhanced gene activation by E2 was noted [ 56 ]. Sometimes a technically easier macroarray is used, e.g., on a 96-well plate [ 57 ]. Obviously, the choice of gene sequences to be used as probes must be very selective in this case. This approach has been adopted by the SuperArray Bioscience Corporation offering selected profiles of genes in the macroarray format for various areas (e.g., cancer, cell cycle, cytokine and inflammatory response, etc.). Quite often the gene identification obtained by an array is confirmed by other methods such as Northern blot analysis [ 58 ], or real-time PCR [ 43 , 58 ][ ]. A negative identification can be achieved by the use of siRNA (small interfering RNA – SuperArray Corp.). siRNAs are short RNA duplexes between 15 to 21 nucleotides in length. Once transfected into cells, a siRNA targets the mRNA containing an identical sequence and degrades it in a catalytic manner. The degraded message is no longer functional in translation (the biosynthesis of protein) and thus in the expression of the corresponding gene. SuperArray Corp. provides a line of validated populations of siRNAs in the form of SureSilencing siRNA kits. Conclusions The methods described above can suit two purposes. The single-gene methods can detect and identify new, previously unknown, genes, whereas microarrays can handle a great number of known genes to establish profiles of their expression. SAGE seems to have advantages over hybridization-based methods for the studies of gene expression, such as differential display and subtractive hybridization. SAGE is superior to the EST approach in providing high efficiency in identifying the genes that are expressed at low levels and that represent a majority of genes in the human genome [ 36 ]. Microarray techniques usually detect activation of a multitude of genes – a gene profile – that differs from the profile in control tissues/cells and thus – in medicine – may have a diagnostic and/or prognostic value. However, the microarray techniques usually require commercially produced chips as well as specialized equipment and advanced, powerful, computing facilities. Thus they are hardly affordable for small or medium-size laboratories unless they have substantial financial resources. A big question at another level remains so far unanswered: which is the biological "chain of commands" in a given tissue and time resulting in the activation of genes enabling the biosynthesis of cornerstones for gene activation, such as ligands (e.g., E2), receptors (e.g., ER) and other transcription factors, the entire machinery leading to gene activation and expression.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524190.xml
544894	GOLD.db: genomics of lipid-associated disorders database	Background The GOLD.db (Genomics of Lipid-Associated Disorders Database) was developed to address the need for integrating disparate information on the function and properties of genes and their products that are particularly relevant to the biology, diagnosis management, treatment, and prevention of lipid-associated disorders. Description The GOLD.db provides a reference for pathways and information about the relevant genes and proteins in an efficiently organized way. The main focus was to provide biological pathways with image maps and visual pathway information for lipid metabolism and obesity-related research. This database provides also the possibility to map gene expression data individually to each pathway. Gene expression at different experimental conditions can be viewed sequentially in context of the pathway. Related large scale gene expression data sets were provided and can be searched for specific genes to integrate information regarding their expression levels in different studies and conditions. Analytic and data mining tools, reagents, protocols, references, and links to relevant genomic resources were included in the database. Finally, the usability of the database was demonstrated using an example about the regulation of Pten mRNA during adipocyte differentiation in the context of relevant pathways. Conclusions The GOLD.db will be a valuable tool that allow researchers to efficiently analyze patterns of gene expression and to display them in a variety of useful and informative ways, allowing outside researchers to perform queries pertaining to gene expression results in the context of biological processes and pathways.	Background The excessive consumption of high calorie, high fat diets and the adoption of a sedentary life style have made obesity and atherosclerosis major health problems in Western societies. In the USA, over 50% of the population are over-weight (BMI > 25) and close to 25% are considered obese (BMI > 30) [ 1 , 2 ]. As a consequence, a large fraction of the population is at risk to develop a broad range of common, life-threatening diseases including non-insulin dependent diabetes, various hyperlipidemias, high blood pressure and atherosclerosis. Vascular disease including coronary heart disease and stroke is currently the major cause of death in the United States and in other industrialized nations. At the root of obesity and atherosclerosis is an excessive deposition of neutral lipids. Adipose tissue accumulates predominantly triglycerides, whereas macrophages along the blood vessel wall mainly accumulate cholesterol and cholesteryl esters. Accordingly, a detailed understanding of the molecular mechanisms that govern the balance between lipid deposition and mobilization is fundamentally important for the prevention and improved treatment of disease. In addition to the apparent environmental components involved in the pathogenesis of disorders related to lipid and energy metabolism, a large number of studies have provided undisputed evidence that susceptibility genes contribute around 50% of the phenotype. These genes encode products involved in the cellular uptake, synthesis, deposition and/or mobilization of lipids. However, characterization of many if not most of these genes and their products remains rudimentary. Deficiencies in the current level of understanding extend to key enzymes such as important triglyceride hydrolases in adipose tissue [ 3 ] or cholesteryl ester hydrolases in macrophages, hormones, signal transduction pathways, and the regulation of the transcription of relevant genes. While medical molecular biology traditionally associates single genes and gene products with diseases, a growing body of evidence suggests that several common disease phenotypes arise from the delicate interaction of many genes as well as gene-environment interactions. To elucidate the development of obesity and atherosclerosis, it will be necessary to analyze patterns of gene expression and relate them to various metabolic states. To discover novel genes, processes and pathways that regulate lipid deposition and mobilization, a departure from hypothesis-driven research and turn to a discovery-driven approach is necessary. The application of high-throughput technologies and genome-based analysis will provide the tools for the analysis of gene-gene and gene-environment interactions in a systematic and comprehensive manner. To facilitate genomic research we have initiated the development of a system for storing, integrating, and analyzing relevant data needed to decipher the molecular anatomy of lipid associated disorders. In order to provide a reference for pathways and information of the relevant genes and proteins in an efficiently organized way, we have created the Genomics Of Lipid-Associated Disorders database (GOLD.db). The GOLD.db integrates disparate information on the function and properties of genes and their protein products that are particularly relevant to the biology, diagnosis management, treatment, and prevention of lipid-associated disorders. Construction and content The main goal of the GOLD.db was to provide biological pathways with image maps and visual pathway information. For each element in the pathway, specific information exists including structured information about a gene, protein, function, literature, and links. The GOLD.db provides also the possibility to map gene expression data individually to each pathway. Additionally, analytic and data mining tools, reagents, protocols, references, and links to relevant genomic resources were included in the database. The GOLD.db was implemented in Java technology [ 4 ]. Hence, the pathway editor, as well the web application are platform independent. The web application of GOLD.db is build in Java Servlets and JavaServer Pages technology based on the Model-View-Controller Architecture [ 5 ]. For the implementation, the freely available struts framework [ 6 ] was used. This code can be easily deployed in any Servlet Container. We used the Servlet Container Tomcat (also freely available at [ 7 ]) which is accessible from all web browsers. Oracle 9i was used as database management system. The interface between the Java and the Database management system was established using Java database connectivity (JDBC) 2.0. Therefore, migration to other freely available DBMSs like mySQL can be easily done. For additional storage and communication between the pathway-editor components, the markup language XML containing structured, human readable information, was used. The provided pathways can be downloaded as Scalable Vector Graphics (SVG) [ 8 ], a standard for describing two-dimensional graphics in XML, and can be visualized in this format on the client side with the web browser using a plug-in for SVG. For tracking the repository of the reagents like clone resources and libraries which can be used for microarray studies, we have developed a relational database. Information about the vector, the sequence and length of the clone insert, primers for the PCR amplification, tissue, organism, accession number, library, container, storage information, date and person and access to other clone bases (e.g. IMAGE Consortium) can be stored. Users of the GOLD.db can list these clones and get all the information about each available clone. With restricted access, clone information or even clone lists can be uploaded and selection lists can be created and deleted. The input mask is designed in such way that the user can choose one of the elements of the created selection lists. In order to deal with the huge amount of data associated with large scale studies and to perform sequence based and microarray analysis, several bioinformatic tools were integrated or can be downloaded. Sequence similarity search against databases can be performed with BLAST (Basic Local Alignment Search Tool) [ 9 ], FASTA [ 10 ] or HMM (Hiden Markov Models) [ 11 ] on a 50 CPU Myrinet Cluster. The sequence retrieval system SRS (LION Bioscience AG, Heidelberg, Germany) was included to enable rapid, easy and user friendly access to the large volumes of diverse and heterogeneous data [ 12 ]. The latest version of the PathwayEditor for the construction of biological pathway diagrams can be downloaded. For microarray analysis the platform independent JAVA tools ArrayNorm [ 13 ] for normalization of microarray data and Genesis [ 14 ] for clustering and analysis of large scale gene expression datasets were made available. Utility and discussion Pathways In order to construct the biological pathways of interest, we have developed a pathway editor [ 15 ] and an extended version to map gene expression data (pathway mapper). This drawing tool provides the possibility to draw elements – typically representing a gene as part of the pathway – and the connection between those elements. The benefit of this tool is that information can be appended to each element via an input mask. This information can be accessed by clicking on the corresponding element in the image map within the pathway mapper or when saved and uploaded via the web interface to the GOLD.db. To design this pathway service as flexible as possible, features are provided for the remove, up- and download of relevant pathways (image maps) including the underlying additional information of the elements. However, this service is on a restricted basis to prohibit unauthorized access. Since some pathways tend to become very detailed an option to search for genes or gene accession number, respectively, within the pathways was built in. The pathway editor is executable as a standalone application and is available from [ 16 ]. Currently annotated pathways are the insulin signaling pathway, the IGF-1 pathway and the adipogenesis regulatory network. Other pathways of lipid metabolism will follow in the near future. Available KEGG pathways can also be adapted with the pathway editor based on the provided XML files [ 17 ] and uploaded in the same way. All relevant KEGG pathways for different organisms are provided. Moreover, pathways from BioCarta were made available within the GOLD.db and HTML files [ 18 ] were parsed to provide additional meta-information of the pathway elements. For each element in the pathways a specific information field exists. The field includes structured information about a gene, protein, function, literature, and links to well-curated and annotated databases. Besides the gene name and the symbol name – for human the HUGO symbols and gene names and for mouse the MGI nomenclature were used – RefSeq numbers for the transcript and the protein as well as a link to SwissProt/UniProt and LocusLink is available. For the elements of the KEGG pathways a link to the provided enzyme or product information was given. The description, localization and classification of the factors are entered by the annotator in plain text and are accessed in the same format. The references used to generate the content of the database entries can be appended, including a link to the PubMed entry. There is also the possibility to create a list of reference entries for the pathway. If a clone for a specific gene is available in the clone resources, the clone name will be displayed automatically and a link with optional information about this clone is provided. Mapping of gene expression data sets to pathways Through the integration of several types of biological information deeper insights into the molecular mechanisms and biological processes can be gained than just by the analysis of one type of experimental results. In the GOLD.db it is possible to map gene expression data (for instance results of microarray studies) to the corresponding elements of the available pathways similar to previous efforts [ 19 ]. Either an individual or a provided gene expression data set can be used to visualize the gene expression at different experimental conditions sequentially or all at once in the context of a pathway. If an element (gene) of the pathway is included in the data set, the related symbol in the image map is color coded according to the relative gene expression or the log ratio in two color microarray experiments, respectively. As key for the mapped relation the RefSeq number [ 20 ] is used. Hence, only those elements in the data set file are mapped, where the RefSeq number in the data set is specified. For the KEGG pathways each element classified by the enzyme classification number (EC) is virtually subdivided into different corresponding RefSeq entries, since one EC is represented by one or more RefSeq entries. Curated gene expression data sets Analysis of gene expression patterns in animal and cell models for lipid-associated disorders will help to understand the fundamental gene relations and regulatory mechanisms responsible for the development of obesity related diseases. The huge amount of data associated with the analysis of large scale gene expression analysis raises the demand of tools for storing, processing and retrieving complex information. Although a number of studies have been published and despite the requirements of some journals to deposit microarray data in public databases like GEO or ArrayExpress , it is still very difficult for researchers to obtain the original data. Web sites with Supplementary information are not maintained and/or not further developed. Hence, a database with a large collection of curated datasets will be enormously valuable for the community. Approaches to upload and retrieve gene expression data were pursued within the GOLD.db. Large scale gene expression data sets can be uploaded in form of tab delimited text files (Stanford file format) [ 21 ] as used for cluster analysis programs together with additional information about the experimental conditions and the citation for already published data sets. Within those data sets the search for specific genes is possible to provide integrated visualization of gene expression levels in different studies and experimental conditions. Example for using GOLD.db: regulation of Pten during adipocyte differentiation Recently, it was shown that insulin sensitivity, energy expenditure, and thermogenesis were enhanced in adipose-specific Pten-deficient (AdipoPten-KO) mice. Body and adipose tissues weight in these mice were significantly lower than those of control mice in spite of a larger food intake [ 22 ]. We addressed the question how is the expression of the Pten gene regulated during adipocyte differentiation in different models and experimental setups and in which pathways is PTEN involved. The workflow for the analysis is described in Figure 1 . Pten (phosphatase and tensin homolog deleted on chromosome 10) is known as tumor suppressor gene and is a protein and lipid phosphatase with the major substrate phosphatiylinositol 3,4,5-triphosphate (PIP3), as indicated in the annotated insulin signaling pathway within the GOLD.db. In fact, Pten regulates negatively the insulin signaling pathway in 3T3-L1 adipocytes [ 23 ]. Figure 1 Various result tables from using GOLD.db to address the question how is PTEN regulated during adipocyte differentiation ( top left: result of search in SRS for phosphatase and tensin homolog; top right: pathways, in which PTEN is involved; bottom left: relative gene expression levels of PTEN in different datasets; bottom right: PTEN dependent cell cyle pathway with mapped gene expression levels) During adipocyte differentiation cyclin dependent kinase inhibitors, like p21 leads to a hypophosphorylation of the Retinoblastoma protein (Rb) which allows binding to the E2F transcription factor, causing cells to permanently exit the cell cycle – a required step in adipocyte differentiation called mitotic clonal expansion – before entering the terminal differentiation state. pRb interacts physically with adipogenic CCAAT/enhancer-binding proteins and positively regulates transactivation by C/EBPβ and therefore plays a pivotal role in adipocyte differentiation [ 24 , 25 ]. Hence, since a) PTEN is expressed during adipogenesis (Figure 1 ), b) is involved in the regulation of Rb [ 22 ], a major player in adipogenesis, and c) is an important component in cell cycle arrest and apoptosis (Figure 1 ), it can be postulated that PTEN plays an important role in fat cell development. Thus, using recently identified key player for food intake and weight control and using the GOLD.db, it is possible to address relevant questions and generate testable hypotheses on the molecular mechanisms of fat cell development. Conclusions The vast quantity of gene expression data generated in genomic studies presents a number of challenges for their effective analysis and interpretation. In order to fully understand the changes in expression that will be observed, we must correlate these data with phenotype, genotype, metabolism and other information including the tissue distribution and time course expression data gleaned from previous studies. The goal of our work was the development of a specialized database and tools that allow researchers to efficiently analyze patterns of gene expression and to display them in a variety of useful and informative ways, allowing outside researchers to perform queries pertaining to gene expression results in the context of biological processes and pathways. The uniqueness of the GOLDdb database we have developed is threefold: 1) the inclusion of annotated pathways, 2) the availability of curated datasets and 3) the possibility to map experimental data on biological pathways. The upcoming challenges will be to include data from functional analysis and proteomics data, which will give us new opportunities in understanding mechanisms of different applications and lipid-associated disorders in particular. Availability and requirements The GOLD.db database should be cited with the present publication as a reference. Access to GOLD.db is possible through the world wide web at . The pathway editor and the clone tracker are available free of charge to academic, government, and other nonprofit institutions. Author's contributions HH was responsible for the content, the annotation process, webdesign, and processing of data sets. MM was responsible for the implementation of the database and web application as well as the relational database for the clone tracker. BM and JH had implemented the mapping of expression data to pathways. GS is involved in providing of sequence analysis tools and server software. DMS has annotated the insulin signaling pathway. ZT was responsible for the design of the study and for overall project coordination.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544894.xml
528782	The InDeVal insertion/deletion evaluation tool: a program for finding target regions in DNA sequences and for aiding in sequence comparison	Background The program InDeVal was originally developed to help researchers find known regions of insertion/deletion activity (with the exception of isolated single-base indels) in newly determined Poaceae trn L-F sequences and compare them with 533 previously determined sequences. It is supplied with input files designed for this purpose. More broadly, the program is applicable for finding specific target regions (referred to as "variable regions") in DNA sequence. A variable region is any specific sequence fragment of interest, such as an indel region, a codon or codons, or sequence coding for a particular RNA secondary structure. Results InDeVal input is DNA sequence and a template file (sequence flanking each variable region). Additional files contain the variable regions and user-defined messages about the sequence found within them (e.g., taxa sharing each of the different indel patterns). Variable regions are found by determining the position of flanking sequence (referred to as "conserved regions") using the LPAM (Length-Preserving Alignment Method) algorithm. This algorithm was designed for InDeVal and is described here for the first time. InDeVal output is an interactive display of the analyzed sequence, broken into user-defined units. Once the user is satisfied with the organization of the display, the information can be exported to an annotated text file. Conclusions InDeVal can find multiple variable regions simultaneously (28 indel regions in the Poaceae trn L-F files) and display user-selected messages specific to the sequence variants found. InDeVal output is designed to facilitate comparison between the analyzed sequence and previously evaluated sequence. The program's sensitivity to different levels of nucleotide and/or length variation in conserved regions can be adjusted. InDeVal is currently available for Windows in Additional file 1 or from .	Background Gaps caused by insertion and deletion events (indels) are often important features of DNA sequence data, which is widely used in phylogenetic studies [ 1 - 13 ]. Although some authors consider indels to be potentially misleading [ 1 , 14 ], others consider indels to be important characters [ 8 , 9 , 13 , 15 , 16 ] and have argued that treating them as missing data can weaken phylogenetic analyses [ 3 , 7 , 10 , 17 ]. Even though it is generally accepted that indels that cannot be unambiguously positioned make confident homology assessment impossible and, therefore, regions that contain them should be excluded from phylogenetic analyses [ 3 , 4 , 6 , 8 ], it has been proposed that even these regions are valuable if properly coded [ 15 ]. Phylogenetic estimation depends on accurate character homology assessment (sequence alignment) [ 4 , 8 , 18 ], which is made more difficult by the presence of indels [ 2 - 4 , 6 , 8 , 15 ]. Indel occurrence is context-dependent, and it has been repeatedly reported that indels tend to be found clustered into specific length-variable regions [ 1 , 3 , 5 , 6 , 8 , 19 - 21 ]. Accurate assessment of these regions (proper alignment and recognition of relative indel rate, reversals, parallel events, and multiple, overlapping events) is aided by comparison among many sequences from various taxonomic levels [ 4 , 8 , 19 ]. Sequence comparisons are complicated by the ambiguities gaps introduce into alignments. Finding a target region recognized from one alignment within another can be time consuming and difficult to perform accurately. Because of the limits of computer screen size and human analytical ability, alignments of hundreds of sequences are difficult to evaluate, even when they can be prepared. Poaceae is one of the largest families of flowering plants and is economically important [ 22 ]. Lower-level phylogenies within the family often make use of the largely non-coding plastid sequence between the trn L (UAA) 5' exon and trn F (GAA), hereafter called trn L-F [ 17 , 23 - 28 ]. As of Apr 2, 2004, the NCBI Entrez Nucleotides database [ 29 ] contained 505 Poaceae trn L-F sequences. Comparing indel regions across these sequences can reveal patterns in indel behavior and aid researchers in creating accurate alignments. A discussion of the indel regions in these sequences is being prepared separately for publication. The program InDeVal was originally developed to help researchers find known indel regions (with the exception of isolated single-base indels) in newly determined Poaceae trn L-F sequences and simultaneously compare them with 533 previously determined sequences (those mentioned above, plus 28 determined by SDSH). It is supplied with input files designed for this purpose. More broadly, the program is applicable for finding specific target regions (referred to as "variable regions") in DNA sequence. A variable region is any specific sequence fragment of interest, such as an indel region, a codon or codons, or sequence coding for a particular RNA secondary structure. The LPAM algorithm, which was specifically designed for InDeVal, is used to find sequence (referred to as "conserved regions") flanking the variable region in the analyzed sequence. InDeVal can find multiple variable regions simultaneously (28 indel regions in the Poaceae trn L-F files). The program's sensitivity to different levels of nucleotide and/or length variation in conserved regions can be adjusted. Implementation Input files InDeVal uses three types of input files: one conserved region file, separate variable region files for each variable region, and a DNA sequence file (Table 1 ). The conserved region file contains a template of sequence immediately flanking the variable regions (regions of interest). A variable region file contains messages that indicate which permutation of the variable region is in the analyzed sequence. The sequence file contains a set of sequences to be analyzed with InDeVal. All files are plain text (ASCII). Conserved region and variable region files are in InDeVal-specific formats. Detailed instructions for creating them are in the help files. The sequence file is in FASTA-format. A conserved region file contains at least one template, created by taking a representative sequence, removing the variable regions, and replacing them with variable region file names ( Additional file 2 ). Multiple (15) templates were used in the Poaceae trn L-F sequences to accommodate single, large deletions that spanned otherwise conserved regions. Treating them the same as the other indels would have resulted in a few large, difficult-to-interpret variable regions. InDeVal performance is improved by designing templates with conserved regions at least 20 bases long on either side of each variable region. However, the program still functions using templates with only one conserved region, conserved regions only 4 bases long, and variable regions flanked by other variable regions (such as clearly distinguishable, adjacent indel regions varying at different rates, which are found in Poaceae trn L-F). Although InDeVal parameters can be adjusted to reflect different degrees of nucleotide and length variation in conserved regions, it is a good idea to use a representative sequence for template design, especially if some conserved regions are short. Additional templates can be designed to accommodate distantly related taxa. (In the Poaceae trn L-F files, separate templates were designed for Pooideae, Ehrhartoideae, and the PACCAD clade). A variable region file contains the sequence variations the researcher expects in the region and some output information about each variation ( Additional file 2 ). In the Poaceae trn L-F files, the output information is the list of taxa with each variation. For coding sequence, the output information could be the amino acids for which the variations code. If the user is interested only in knowing if a particular variation is present, the output could be simply "Yes". Variants not found in the variable region file are also reported by InDeVal. The program works if the variable region file is completely blank (variable regions are all bases found between template conserved regions), but it obviously cannot provide output messages in this case. Symbols can be used in variable region sequence to draw attention to specific features of interest. They are ignored during alignment, but are displayed in the Variable Region Sequence List Box ( Additional file 2 ). The Poaceae trn L-F variable region files include spacing that emphasizes repeat motifs, hyphens to indicate that an entire variable region has been deleted, and stars to indicate possible inversion sites. InDeVal can help the user create these input files, comparing new sequences to those already included and indicating what adjustments should be made. Sequence files are in a less stringent FASTA-format ( Additional file 2 ) and can be in either orientation. They can include spaces, numbers, capital or small letters, IUPAC ambiguity symbols, and carriage returns without disrupting InDeVal function. A conserved region file (TemplatePtrnLF.txt) and 28 variable region data files, designed from 533 Poaceae trn L-F sequences, are provided with InDeVal. These files are based on the first author's critical examination (to be published elsewhere) of various alignments of these sequences using the web-based programs BLAST 2 SEQUENCES [ 30 ] and/or CLUSTAL W multiple alignments [ 31 ]. Aligning a sequence with a given template InDeVal begins by sorting a template's conserved regions by length. It then searches for each conserved region in the analyzed sequence (using LPAM – see below), proceeding from longest to shortest. Found conserved regions are used to limit the search space for future searches (Figure 1 ). Sometimes conserved regions are not found or are found at multiple locations within their subsequence. The program records this information and proceeds to the next longest sequence. A conserved region that cannot be aligned is recorded as "not found." A conserved region with multiple possible alignments is recorded as "found", but none of its possible alignments is used to limit the searches for the remaining regions. This template alignment algorithm preserves the ordering found in the template, giving priority to the alignment suggested by the longer conserved regions, which are assumed to be more reliable. Using the alignment of longer regions to reduce the search space for smaller regions minimizes the probability of finding ambiguous or incorrect alignments. Length-Preserving Alignment Method (LPAM) InDeVal uses the LPAM algorithm, designed specifically for InDeVal and described here for the first time, to align the conserved regions within their subsequences. LPAM divides a conserved region into overlapping "words", strings of sequence of a user-defined length. That is, given a 4-base word length, the sequence "caatgt" would be represented as "caat", "aatg", "atgt". (Conserved regions shorter than the word length are found only if they match exactly.) LPAM searches the analyzed sequence for each of these words and notes multiple finds, single finds, and missing words. Each word is allowed one "vote" for a possible alignment, i.e., for the base in the analyzed sequence that begins the conserved region. If the word occurs once, it casts its vote for that alignment. If it occurs multiple times, its vote is divided equally among the possibilities. Words that are not found cast no vote. The alignment that receives the most votes is assumed to be the most probable. LPAM permits a (user-definable) degree of length variation. A word's possible alignments are evaluated according to whether or not they agree with the most probable alignment for the region as a whole. Suppose, for example, that the most probable ungapped conserved region alignment would begin at base 53. If the tolerance is set to 3, LPAM will allow the first word to begin at any base from 50 to 56, preferring the possibility closest to 53. If the first word in the conserved region sequence has no acceptable alignment, LPAM searches for the first word that does have an acceptable alignment. All bases preceding this first found word are aligned with no gaps. In the above example, if the first and second words are not found and the third word aligns beginning at base 57, the first two bases of the conserved region will be aligned with bases 55 and 56. Each subsequent base is aligned similarly. If the word it begins has no acceptable alignment, the base is aligned according to the previous acceptable word. If the word it begins has multiple acceptable alignments, the one closest to the previous alignment is chosen. Note that the range of acceptable alignments remains constant; it is a function of the initial voting and does not depend on any of the choices made in aligning individual bases. The aligned bases from the analyzed sequence are compared with the template conserved region sequence, and a percentage similarity is established. If this is greater than the user-defined cut-off value, the conserved region is listed as found. InDeVal then checks the most probable alignment that clearly differs from the first. (In the example above, this excludes any alignment beginning at bases 50–56.) If this alignment also yields a percentage similarity greater than the cut-off value, neither is selected, but the region is still considered to be "found" for purposes of measuring similarity to the template (see below). The LPAM algorithm is able to align conserved regions despite both point mutations and indels. However, deletions in the first or last word of a conserved region are indistinguishable from point mutations (and are interpreted as such). LPAM proved effective in the Poaceae trn L-F sequences, where conserved regions are by definition length-conserved. In 533 sequences, there were only three instances where an indistinguishable mismatch was caused by an indel instead of a point mutation and resulted in misalignment. It is important to design templates so that the ends of conserved regions are not in sequence prone to indel activity. The alignment suggested by LPAM is not necessarily locally maximal, i.e., there are cases where a slight adjustment of the alignment would produce a higher percentage of matched bases. Furthermore, a number of factors can prevent LPAM from finding a correct alignment. Repetitive sequence, resulting in multiple matches, can be very problematic. InDeVal may not be appropriate for analyzing repetitive sequence, and if it is used for this purpose, templates must be designed with great care. If every word contains a point mutation, the conserved region cannot be found. Indels can cause problems if they are dispersed so that several words are disrupted. In general, LPAM will correctly align regions if the mutations are clustered and the first and last base can be assigned correctly (which will be the case if they have no indels between them and the undisrupted word that will position them). Though hypothetical situations where LPAM would create incorrect alignments are easy to envision, in practice, the algorithm proved able to reliably determine where the conserved regions of a sequence begin and end. This is sufficient for the purposes of InDeVal. It is important to note that even when LPAM does not find a conserved region alignment, a possible alignment is displayed, and the user can rearrange it (to a limited extent) and interpret it. Choosing the best template Once conserved regions have been found, InDeVal calculates a "found conserved region score" for each template. Each matched conserved region is given a value equal to the number of bases in that region in the template (regardless of how many individual bases actually matched), and these values are summed to give a score for the template. This prevents a template with a major deletion from being selected over a template with more matching conserved regions (as could happen if the deletion template were a better match at the base-per-base level). Templates with higher scores are ranked above those with lower scores (Figure 2 ). If two templates have roughly equal found conserved region scores (to within a user-selected tolerance), the one with the higher "score percentage" (calculated by dividing the score by the total number of conserved region bases in the template) is ranked higher. This ensures that sequences that actually do have deletions will be matched with deletion templates, even though the templates (theoretically) have an equal number of bases in found conserved regions (Figure 2 ). All the templates are ranked in this way and listed in the Template List Box in order. The highest ranked template is automatically selected, and its variable regions are analyzed. The user has the option of comparing the sequence to a template other than the one selected by InDeVal. Finding the correct variable region sequence Each variable region is defined as the sequence between two specific conserved regions. If both flanking conserved regions are found, the bases between them are recorded, and the appropriate variable region file is searched for a matching string of bases. Sequence between two found conserved regions is classified as a confused region if the template indicates that it should contain one or more conserved regions that were not found. Because the variable regions in a confused region are not clearly delimited, InDeVal searches the entire region for each variable region sequence in each applicable variable region file, and all potential matches are recorded. The user is able to select from among these possibilities, rearrange the display to reflect them, and study different sequence interpretations. In this manner, InDeVal is able to deal satisfactorily with most short conserved regions that cannot be found because of point mutations. In the Poaceae trn L-F file, one- and two-base conserved regions were incorporated into adjacent variable region files because they were impossible to find. A short conserved region will be misaligned if a point mutation prevents it from being found in the proper place and a perfect match is found nearby. If this occurs, it is possible to set the program to disregard conserved regions of this size. (This situation has only been observed for 4- and 5-base conserved regions). The region can then be satisfactorily parsed using the confused region algorithm. Results and discussion InDeVal has two output windows (Table 1 , Additional file 2 ). The Sequence Analysis Window displays the analyzed sequence broken into conserved and variable regions and lists template information. From this window, the user can load template and sequence files, export the analysis to a text file, set LPAM parameters and warning message options, convert the sequence to its complement prior to analysis, and choose a template other than the one selected by InDeVal. For conserved regions, the template sequence is listed so that it can be compared directly with the analyzed sequence. For variable regions, if the analyzed sequence matches one of the variants in the variable region file, the message for that variant is displayed. The Variable Region Analysis Window displays the variable region file name, variants proposed by InDeVal, and the sequence and length of the line currently selected in the Sequence Analysis Window (which can be any line from any sequence). The user can request displays of some or all of the sequences in the variable region file. A text file can be created once the user is satisfied with the display in the Sequence Analysis Window. This file gives information about the sequence and template, lists the positions of the various conserved and variable regions, and shows the entire analyzed sequence. The user can choose whether or not variable region file information is displayed. InDeVal for the Windows platform is available in Additional file 1 . The source code for InDeVal in Microsoft Visual Basic 6.0 is available in Additional file 3 . InDeVal analyzes only DNA sequences, but a version for protein sequences could be created using the source code. It would have to be altered to recognize amino acid sequence (all non-nucleotide letters are presently ignored), and an algorithm to recognize frame-shifts would be helpful. Conclusions InDeVal is a program designed to search DNA sequence for target regions and to display information about them. It can find multiple target regions simultaneously and is relatively robust when challenged by conserved region variation and differences among analyzed sequences in length, spanned region, and format. InDeVal differs from other alignment software in that it breaks the analyzed sequence into user-defined units and emphasizes regions that are of most interest to the user. This makes it possible to quickly compare specific features among many hundreds of sequences. An advantage of InDeVal is that, while it can be used to quickly skim and classify hundreds of sequences, it displays all surrounding sequence. Therefore, if at any time questions arise about the initial classification, the researcher can recreate the InDeVal alignment, instantly find the area in question, and study it for alternative explanations. Availability and requirements Project name: InDeVal Project home page: Operating system(s): Windows Platform Programming language: Microsoft Visual Basic 6.0 Other requirements: None License: None Any restrictions to use by non-academics: No restrictions Authors' contributions SDSH planned and oversaw the project, collected the data, prepared the data files, tested the program, and wrote the documentation. JAH suggested the project and designed, wrote, and debugged the program. Both authors read and approved the final manuscript. Supplementary Material Additional File 1 InDeVal 1.0 installation version InDeVal is currently available for the Windows platform. Instructions for use can be found in the help files, which are included with the program and are accessible from the Sequence Analysis Window. An installation version of InDeVal 1.0 with accessory files can be obtained by clicking on the link below or by visiting . Running the installation program installs (in addition to system files) an InDeVal program directory that contains the Indeval.exe file, the 7 InDeValHelp.txt files, the InDeValOptions.txt file, the InDeValParams.txt (parameters) file, and a Templates subdirectory containing TemplatePtrnLF.txt and a Vr subdirectory with 28 Poaceae trn L-F variable region files. The program is archived using WinZip ® 9.0. WinZip is available from . The zipped package has 3.2 MB. Click here for file Additional File 2 Annotated illustrations of InDeVal files and displays This file (386 kB) is a 9-page pdf that contains annotated illustrations and explanations of InDeVal files and displays. The structure and format of the conserved region, variable region, sequence, and output files is illustrated. Annotated screenshots of the Sequence Analysis Window and Variable Region Analysis Window are also included. These illustrations, combined with the InDeVal help files, serve as the InDeVal Manual. Click here for file Additional File 3 InDeVal source code The source code for InDeVal in Microsoft Visual Basic 6.0 can be obtained by clicking on the link below or by visiting . The files are archived using WinZip ® 9.0. The package (61 kB) includes the 32 code files, the InDeVal icon and the bitmap from which it was constructed, and InDeValSourceCodeHelp.txt, a file with advice on orienting within the source code files. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC528782.xml
551533	Variation in tissue-specific gene expression among natural populations	The expression of a selected suite of 192 metabolic genes in brain, heart and liver in three populations of the teleost fish Fundulus heteroclitus was examined. Only a small subset (31%) of tissue-specific differences was consistent in all three populations, indicating that many tissue-specific differences in gene expression are unique to one population and thus are unlikely to contribute to fundamental differences between tissue types.	Background The regulation of gene expression varies extensively among tissues, individuals, strains, populations and species [ 1 - 6 ] and variation in gene expression has a genetic basis [ 7 , 8 ]. Despite such biological variance, differences in gene expression are used to describe cancers [ 9 - 12 ], heart failure [ 13 , 14 ] and metabolic diseases [ 15 ]. It is common for these pathologies to be associated with changes in tissue-specific gene expression or changes in metabolic gene expression. For example, many different cancers have unique tissue-specific patterns of gene expression [ 16 ], and thyroid cancers are associated with increases in aerobic metabolic gene expression [ 17 ]. Although tissue-specific gene expression patterns are often used as a method to identify functionally relevant genes, how conserved these differences are among outbred individuals and among populations has not been well documented. It is possible that many of these changes represent polymorphism among individuals or populations and are not specifically associated with disease. To address this we used a well established system (tissue-specific gene expression) and genes with well defined function and tissue-specific distributions (metabolic genes). Given the high variance in gene expression among individuals and populations, our goal was to examine the conservation of tissue-specific gene expression among populations of the same species. Specifically, we assessed the among-population variance of tissue-specific patterns of gene expression (in brain, heart and liver) in the teleost fish Fundulus heteroclitus . A cDNA microarray was used to measure levels of expression in normal healthy male fish for 192 genes involved in central metabolic pathways. We used this compact array in order to impose a high degree of technical and biological replication (24 replicates for each of three tissues from nine individuals with two samples per array). Also, this array was used because metabolic genes are essential, are known to have tissue-specific expression, especially in fish, and are often misused as controls with little characterization of variation in expression among individuals or tissues. Analysis of variance (ANOVA) was used as a statistical test to determine which genes were differentially expressed among tissues and populations. Tissue-specific patterns of gene expression were compared among populations. As expected, we detected extensive variation in gene expression among tissues. Unexpectedly, only a fraction (31%) of tissue-specific differences was conserved between all populations. Results Variation among Variation among individuals within groups was high (groups included the nine tissue-by-population groupings; Figure 1 ). Nearly half of genes (92 genes, 48%) were differentially expressed ( p < 0.05) among individuals within populations and tissues (Figure 1 ), and inter-individual differences ranged over fivefold. Variation among tissues Although variation among individuals was high, added variation due to tissues was significant. Considering 192 genes and a p -value of 5%, one would expect less than 10 false-positive differences among tissues under the null hypothesis. We detected 76% of genes (146 of 192 genes) differentially expressed among brains, hearts and livers (ANOVA, p < 0.05). Selecting the α level at which differences between treatments are considered significant is problematic because of the large number of comparisons performed. As such, we present a volcano plot to illustrate the range of expression differences between tissues and associated p -values (Figure 2 ). When α is set at 0.01, 0.001 or at the Bonferroni-corrected value (2.6 × 10 -4 ), the proportion of significant genes is 67% (129 genes), 50% (96) and 39% (75), respectively. Significant differences in expression ranged from less than 1.2-fold to nearly 16-fold (Figure 2 ). The predominant pattern of tissue-specific expression can be described by expression significantly different in the liver compared to the other two tissues (Figure 3 ). Many expected tissue-specific patterns emerged. For example, the brain-specific fatty-acid-binding protein was typically more highly expressed in the brain than in other tissues ( p = 0.005), hepatocyte nuclear factor 4-alpha (a transcription factor) was more highly expressed in liver than in other tissues ( p < 0.001), and two genes involved in glycerolipid metabolism -lipoprotein lipase and phopholipase XIII A2 - were more highly expressed in liver than other tissues ( p < 0.001 for both genes). Liver-specific expression accounted for 61% of the expression differences among tissues (Figure 4 ). Heart-specific and brain-specific expression accounted for 24% and 15% of differences among tissues, respectively. Regardless of population, expression patterns were typically most similar between heart and brain, and least similar between liver and heart (Figure 5 ). There were 67 genes printed on the array that code for proteins involved in oxidative phosphorylation, and 88% (59 genes) were differentially expressed between tissues (genes highlighted in green, Figure 3 ). Of differentially expressed oxidative phosphorylation genes, only 10% (six genes) were expressed more highly in the liver than in other tissues, whereas the remaining 90% (53 genes) had lower expression in the liver compared to brain or heart. Variation among taxa A small proportion of genes (six genes, 3%) differed in expression among populations ( p < 0.05). However, it should be noted that although the split-plot design is powerful for detecting differences between split-plot factors (tissues), it is considered to have low power for detecting differences between blocks (populations) [ 18 ]. As such, it is likely that 3% is an underestimate of true among-population differences in gene expression. Indeed, two-way ANOVA (data not shown), which has higher power for detecting population differences but is less valid than the split-plot model for testing individual and tissue differences, detected among-population differences in expression for 18% of genes at p < 0.05, or 6.3% of genes at p < 0.01. Each tissue contributed a similar number of genes differentially expressed among populations. Surprisingly, differences among tissues in gene expression were not consistent across all three populations. More than one-third (37%) of the genes differentially expressed between tissues were significant in only one of the three populations (Figure 6 ). Population-specific differences were distributed among the three populations; Georgia had 40% of the population-specific genes, and New Jersey and Maine had 34% and 26%, respectively. A proportion of these inconsistencies could be due to false-positive or false-negative differences between tissues in individual populations. However, statistically significant interaction between tissue and population was detected for many (30%) of these inconsistencies (see Additional data file 1). A relatively small proportion of tissue-specific genes (31%) have consistent expression patterns in all three populations (Figure 6 ; also see Additional data file 1 for details). This subset of genes also reflects the different metabolic status of brain, heart and liver; most of the genes involved in oxidative phosphorylation were more highly expressed in brain and heart than in liver (Figure 7a , Table 1 ), and most of the genes involved in fatty-acid metabolism, glycerolipid metabolism, steroid metabolism and detoxification were more highly expressed in liver. The majority of the tissue-specific genes were not consistent among populations (a subset of these genes are illustrated in Figure 7b , Table 1 ). Quality control Variation among technical replicates was low, and permutation tests indicated that the ANOVA model was robust. Sample coefficients of variation (CVs (standard deviation/mean) × 100), which estimate technical variance due to replicate spots (six spots per hybridization), repeated measures (two hybridizations per dye), and dye (two dyes per sample), were calculated for each gene of each of the 27 samples. CVs less than 5% accounted for 95% of sample/genes, respectively. Of the many comparisons performed (differences among tissues, populations, interaction), permutation tests results agreed with ANOVA results (the same comparisons identified as significant or not significant) for 99.1% of comparisons, suggesting that our ANOVA model was robust. Discussion Considerable variation occurs among the 27 samples (three tissues from each of three individuals from three populations) used to measure inter-individual and tissue-specific variation in gene expression. We are able to precisely describe the patterns of gene expression for 192 metabolic genes because of the low experimental variation; for 95% of the replicate measures of gene expression the standard deviation is less than 5% of the mean. Notably, gene expression is statistically different for many genes among individuals within a population for a tissue (48%), between tissues (76%), and between populations (3%). For genes with tissue-specific expression, only a fraction (31%) had expression patterns consistent across all three populations. These data do not specifically identify tissue-specific differences that are inconsistent across populations, but rather emphasize that tissue-specific differences detected can vary from one population to another. When measured from a single population, highly significant differences in tissue-specific expression do not necessarily represent genes relevant to general functional or morphological differences between tissues. Variation among individuals Variation in gene expression among healthy male individuals raised under controlled laboratory conditions was high. Nearly half of the metabolic genes (48%) were differentially expressed among individuals within a population for any one tissue (Figure 1 ), with fold differences ranging from 1.2- to 5-fold and p -values ranging down to 10 -7 . Differences in gene expression among individuals are unlikely to be due to common reversible environmental factors that affect physiological performance (acclimation effects) since all individuals used in this study were housed in a common environment and fed the same food for at least two months. However, the differences could be due to irreversible developmental effects or genetic variations that affect gene expression. Regardless of this, if these differences are heritable or due to developmental plasticity, they represent variation one would expect to find among outbred organisms, including humans. Other studies that have measured inter-individual differences in gene expression have also detected high levels of variation in a variety of taxa. Among crosses of different yeast strains a large number of differences in expression (6% of genes varying more than twofold) were detected between morphotypes [ 1 ]. A previous study of the same Maine and Georgia Fundulus populations assayed here detected 18% of genes differentially expressed among healthy individuals [ 3 ]. Although inter-individual variance in gene expression seems prevalent, our observation that 48% of genes are differentially expressed among individuals is high. This may reflect the greater precision of these measurements as a result of extensive technical replication (24 replicate measures per sample) as coefficients of variation for technical replicates was less than 5% for 95% of the genes. Indeed, using similar methods and tools, a concurrent study assessing variation in Fundulus also detected a very high proportion of genes (94%) differentially expressed among individuals [ 19 ]. Alternatively, since our array is heavily biased toward metabolic genes, detected variance may also reflect a greater variation in metabolic gene expression. We could speculate that the high variation in metabolic genes reflects a greater allowable variation. That is, there may be less selective pressure to constrain metabolic variation either because varying the amount of an enzyme does not affect metabolism or variation in metabolism is phenotypically acceptable. One could test this by using an array with more comprehensive representation of the genome and comparing variances of different gene classes defined by function. Considering the high inter-individual variation detected, the data presented here underscore the importance of including biological replicates within treatment groups in order to ascribe differences in expression to treatment rather than to inter-individual variation. Statistically, an analysis of variance can be used to examine the effects of technical and biological variation, and these tests have proved powerful for detecting significant differences in gene expression [ 3 , 4 ], even differences as small as 1.2-fold. The cost of resources in microarray experiments should no longer excuse lack of biological and technical replication. Often, microarray experiments pool individual samples within treatment groups to capture biological variation. However, this approach only estimates an average level of expression and fails to estimate biological variation. When only small quantities of RNA can be extracted from samples, one can estimate biological variation by pooling multiple independent samples [ 20 ]. A variety of factors can contribute to differences in gene expression among individuals. Pritchard et al . [ 21 ] proposed that differences in immune status may explain the 3.3% difference in gene expression among genetically identical mice. Sex explained a large portion of among-individual variation in gene expression in Drosophila , whereas genotype was less of an influence, and the influence of age was weak [ 4 ]. Furthermore, this type of variation can be biologically relevant. For example recent work in Fundulus indicates that most inter-individual variation in metabolism can be accounted for by differences in metabolic gene expression [ 19 ]. Variation among tissues Another important source of biological variation in gene expression is differences in expression among different tissues; 76% of genes were differentially expressed between brain, heart and liver, and expression in the liver was the most distinct compared to heart and brain. In this study, genes printed on our array are primarily enzymes functional in central metabolic pathways such as fatty-acid metabolism, glycolysis and oxidative phosphorylation. Of the oxidative phosphorylation genes differentially expressed between tissues, 92% were more highly expressed in heart or brain than in liver (Figure 3 ). The primary purpose of the heart is to act as a pump, and contraction is highly dependent on oxidative metabolism [ 22 ]. The metabolic rate in the brain is 7.5 times the average rate in the rest of the body [ 23 ]. High metabolic demand in the brain supports pumping of ions across neuronal membranes during action potentials and metabolism is primarily oxidative. Mitochondria are the principal sites for oxidative phosphorylation, and are most numerous in heart, brain and skeletal muscle cells. The liver, in contrast, is much more functionally diverse, as it is involved in carbohydrate storage, synthesis of proteins, glucose, fatty acids, cholesterol and lipids, and metabolism of xenobiotics and endogenous compounds, and has a relatively low respiration rate. Accordingly, transcripts of genes functional in oxidative phorphorylation appear to represent a much smaller portion of the cell's RNA transcripts in liver tissues than in the heart or brain. In addition, genes involved in fatty acid and phospholipid synthesis were more highly expressed in liver than the other tissues. Differences in expression among tissues detected using our array appear to reflect differences in the metabolic status of brain, heart, and liver. Because data presented here support well established patterns of metabolism, they suggest that measuring mRNA expression using microarrays accurately reflects changes in proteins and their phenotypic effect. Many microarray studies have used expression levels of 'housekeeping' genes as an internal control for comparisons among arrays, individuals and treatments. Housekeeping genes may be defined as those that are involved in routine cellular metabolism and always expressed in all cells. Accordingly, many, if not most, of the genes studied here could be considered housekeeping genes. Nearly half of these genes were expressed at different levels between individuals, with fold differences ranging from 1.2- to 5-fold and p -values ranging down to 10 -7 . Lee et al . [ 24 ] applied ANOVA to screen four previously published datasets for housekeeping genes across a variety of biological contexts. They found that all genes that are commonly used as controls had fold changes ranging from greater than 2.0 to more than 300 within at least one dataset, and coefficients of variation were concordantly high, reflecting high variance in expression of these genes. It appears that upon application of ANOVA, statistically significant differences in expression of housekeeping genes can be detected among individuals and across different biological contexts, and scaling for differences among arrays using expression levels of these genes ought to be approached with caution. Although genes differentially expressed among tissues reflect their different metabolic requirements, it should be noted that the purpose of the current study was not to comprehensively identify suites of genes responsible for functional differences between tissues. The relatively small number of printed probes was useful for a high degree of technical replication, and obviously represents a small portion of the expressed genes. However, this approach shows that highly significant differences in gene expression among tissues may be apparent but not consistent among closely related taxa. Therefore, highly significant differences in gene expression found only within a single population may not necessarily represent genes relevant to general functional or morphological differences between tissues. Variation among taxa Although the pattern of metabolic gene expression among tissues reflects established patterns of tissue-specific metabolism, there is additional variation due to population. It should be noted that the split-plot statistical design is not as powerful for detecting among-block differences (among populations) as for detecting differences among split-plot factors [ 18 ]. We detected 3% of genes (6 of 192) differentially expressed among populations. This proportion is similar to that detected in a previous study [ 3 ] in which 2.6% of genes were differentially expressed between Maine and Georgia Fundulus hearts. Similarly, approximately 1% of genes were differentially expressed in brain tissue among inbred strains of mice [ 2 ]. Differences in gene expression are to be expected among taxa (phylogenetically distinct groups of organisms which may include strains, populations or species), with the majority of differences most likely to be attributable to random genetic drift. For more distantly related groups, one would expect expression patterns to be more divergent than for closely related groups. Indeed, expression patterns between humans and chimpanzees are more similar than those between humans and orangutans, and similar results were obtained from comparisons among three mouse species [ 5 , 6 ]. An unexpected finding is that the tissue-specific differences depend on which population was assayed. Differences in gene expression are expected between tissues because of functional divergence and between populations because of neutral genetic divergence. In addition, one might expect that the number of genes significantly different between populations would depend on the tissue. One might also expect tissue-specific differences to be consistent in all taxa. Yet our data indicate that tissue-specific expression patterns are not fixed within a species. The genes for which expression is significantly different between tissues are not all the same in all three populations. Of the 128 genes that have tissue-specific patterns of expression in any population, 37% are tissue-specific in only one of the three populations and 32% are found in only two of the three populations. Overall, it would appear that only 31% of tissue-specific differences in gene expression are consistent among all populations of F. heteroclitus . One needs to be careful about this interpretation, however. Our emphasis was not to specifically identify genes that have significant interaction between tissue and population. Rather, we emphasize that genes detected as tissue specific will vary from one population to another, and most microarray studies measure treatment-specific expression patterns in only one population of test organism. Because inter-individual variation is high, it is probable that inclusion of more replicate individuals in each group would increase the sensitivity of ANOVA, and the number of genes that distinguish tissues consistently in all populations may change. The consistent tissue-specific differences still support expectations based on the metabolic requirements of each tissue (for example, genes involved in oxidative phosphorylation were more highly expressed in heart and brain, and those involved in fatty-acid and lipid metabolism were more highly expressed in the liver; Figure 7a ). Accordingly, those differences in expression that are consistent across several groups of organisms are most likely to account for functional and morphological differences among tissues, emphasizing that this type of comparative approach may be powerful for testing the biological relevance of other functional traits. For example, expression differences between diseased and non-diseased tissues may vary among mouse strains, so that the subset of differences that are consistent across strains are more likely to be functionally related to the diseased state. Our data suggest that many of the differences in gene expression detected between experimental groups may be of little functional importance because they vary among taxa. We suggest that patterns of expression that are consistent in different populations are more likely to be functionally important. Elucidation of adaptively important variation, such as variation related to antibiotics, pesticides or temperature adaptation, may also benefit from such a comparative approach that screens for conserved patterns. However, there is the possibility that partitioning of genetic polymorphisms among populations may allow distinct groups of organisms to reach different physiological or biochemical solutions to the same biological challenges. For example, patterns of polymorphism in a gene that regulates coat color in mammals indicated recent directional selection and was associated with coat color in one pocket mouse population, but not in a second population [ 25 ]. Other loci were probably responsible for adaptive variation in coat color in the second population. Conclusions These data indicate high variation in metabolic gene expression among individuals and thus expression of these housekeeping genes is unreliable as an internal control or as a method of normalization across samples. Second, concordance between tissue-specific expression patterns and established metabolic functions of brain, heart and liver indicate that measuring mRNA levels accurately reflects physiological status. Furthermore, since many metabolic genes differ in expression among brain, heart and liver, those studies using whole organisms need to rule out whether changes in expression reflect differences in the proportions of various tissues among samples. Finally, studies seeking to identify patterns of gene expression related to physiological states, such as disease or toxic stress, must consider both variation between individuals and differences between populations. Because of this biological variation, not all differences between treatments in any one population of test organism are likely to be generally relevant. We suggest that conserved patterns of treatment-specific gene expression among taxa are most likely to be functionally related to the physiological state in question. Methods and materials Animals and maintenance Teleost fish Fundulus heteroclitus were collected from the field by seine and minnow trap in June 2003, transported to the University of Miami RSMAS laboratory under controlled temperature and aeration conditions, and acclimated to common conditions (20°C, 15 parts per thousand salinity) in recirculating 100-gallon tanks for at least two months before experiments. Fish were sacrificed by cervical dislocation and tissues were excised and stored in RNAlater (Ambion) at -20°C. Fish were collected at Wiscasset, Maine; Stone Harbor, New Jersey, and Sapelo Island, Georgia. Only healthy male fish were used for the following experiments. Microarrays Microarrays were printed using 192 cDNAs from a F. heteroclitus cardiac library encoding essential proteins for cellular metabolism [ 26 ]. These cDNAs were a subset of over 40,000 expressed sequences in our online database Funnybase [ 27 ]. These 192 cDNAs were amplified with amine-linked primers and printed on 3-D Link Activated slides (Surmodics) using a SpotArray Enterprise piezoelectric microarray printer (PerkinElmer Life Sciences) at Louisiana State University. Slides were blocked following slide manufacturer protocols. The suite of 192 amplified cDNAs was printed as a group in six spatially separated replicates. Four hybridization zones of these six replicate arrays were printed per slide, with each zone set separated by a hydrophobic barrier. Hybridization experimental design Microarray analyses were applied to three tissues (brain, heart and liver) from three individuals collected from three populations of F. heteroclitus . Each of these 27 samples was measured four times, twice with Cy3 and twice with Cy5 (Figure 8 ). In addition, since a hybridization zone covered six replicate printed arrays, total experimental replication per sample per gene was 24-fold. A total of 108 hybridizations were performed (27 × 4), and Cy3-Cy5 hybridizations were balanced (although incompletely) among tissues and populations in a sheet-loop design (Figure 8 ). Sample preparation RNA was extracted from tissue homogenate in a chaotropic buffer using phenol/cholorform/isoamyl alcohol. All reagents were from Sigma unless otherwise noted. Tissues were removed from RNAlater, blotted dry, and homogenized using an electric homogenizer in 400 μl chaotropic buffer (4.5 M guanidinium thiocyanate, 2% N-lauroylsarcosine, 50 mM EDTA pH 8.0, 25 mM Tris-HCl pH 7.5, 0.1 M β-mercaptoethanol, 2% antifoam A). An equal volume of 2 M sodium acetate (pH 4.0) was added to the homogenate, followed by 400 μl acidic phenol (pH 4.4), and 120 μl chloroform/isoamyl alcohol (23:1). The mixture was kept at 4°C for 10 min then centrifuged at 4°C at 16,000 g for 20 min. Supernatant was removed and combined with 400 μl isopropanol, stored at -20°C for 30 min, then centrifuged at 4°C at 16,000 g for 30 min. The remaining RNA pellet was rinsed twice with 400 μl of 70% ethanol, then further purified using the Qiagen RNeasy Mini kit (Qiagen) following the manufacturer's protocols. Purified RNA was quantified spectrophotometrically, and RNA quality was assessed using the Agilent 2100 Bioanalyzer. RNA was stored in 1/10 volumes 3 M sodium acetate and 2.5 volumes 100% ethanol at -20°C. RNA for hybridization was prepared by amplification using a modified Eberwine protocol [ 28 ]. The Ambion Amino Allyl MessageAmp aRNA Kit was used (according to manufacturer's protocols) to copy template RNA by T7 amplification following incorporation of a T7 promoter, resulting in amplified template in the form of antisense RNA. Amino-allyl UTP was incorporated into targets during T7 transcription, and resulting amino-allyl antisenseRNA was coupled to Cy3 and Cy5 dyes (Amersham Biosciences). Hybridization Labeled aRNA aliquots of the two individual samples for each hybridization (18 pmol each of Cy3 and Cy5) were vacuum dried together and resuspended in 12 μl hybridization buffer (final concentration of each labeled sample = 1.5 pmol/μl). Hybridization buffer consisted of 5 × SSPE, 1% SDS, 50% formamide, 1 mg/ml poly(A), 1 mg/ml sheared herring sperm carrier DNA, and 1 mg/ml BSA. Slides were washed in sodium borohydride solution according to Raghavachari et al . [ 29 ] to reduce autofluorescence. Following rinsing, slides were boiled for 2 min and spin-dried in a centrifuge at 800 rpm for 3 min. Samples (12 μl) were heated to 90°C for 2 min, quick cooled to 42°C, applied to slide (hybridization zone area was 350 mm 2 ), and covered with a coverslip. Slides were placed in an airtight chamber humidified with paper soaked in 1 × SSC buffer and incubated 12-18 h at 42°C. Following hybridization, slides were scanned using the Packard Bioscience ScanArray Express microarray scanner (PerkinElmer Life Sciences). Resulting .tiff images were imported into spot grids built in ImaGene (Biodiscovery) for each array, and spot signals were collected as fluorescence intensities for each dye channel. Data processing and statistical analysis Raw data were first sum normalized [ 30 ], which involves summing the total signal from each replicate array to the same value. Then spatial bias on each array was smoothed using a lowess transformation in MAANOVA Version 0.93-2 for R [ 31 ]. Other methods of normalization have also been proposed [ 32 - 34 ]. Log 2 values of lowess-transformed sum-normalized data were used for all subsequent statistical analyses. MIAME-compliant data [ 35 ] have been submitted to the Gene Expression Omnibus as accession number GLP1224. Data were analyzed in a split-plot ANOVA design with population as blocks and tissues as split-plot factors using scripts written in MatLab Version 6 (The MathWorks). MatLab code is available upon request from the authors. Nested within tissue-by-population samples were technical replicates. Replicate spots within hybridization (six), replicate hybridizations per labeling (two) and replicate labelings per sample (two; Cy3 and Cy5) represent the three levels of technical variance nested within the tissue-by-population sample. The ANOVA structure is presented in Figure 9 and Table 2 , and the model can be written as: y = grand mean + population + tissue + population-tissue interaction + individual in population + tissue-by-individual within population + dye within individual + hybridization within dye + spot within hybridization where y is the normalized log 2 expression and individual in population and tissue-by-individual within population are random effects. To test for differences among multiple means (for example, among population and tissue groups), and to correct for multiple comparisons, the T-method [ 36 ] was applied. The T-method calculates the minimum significant range defined as MSR = Q α[ kv ] × SE where the critical value Q α[ kv ] is the studentized range [ 37 ], k = number of groups in the comparison (for example, if comparisons are among tissues then k = 3), v = degrees of freedom of MS tissue-by-individual within population , and SE is the standard error among tissue-by-individual samples within populations. The T-method following ANOVA was used to identify genes differentially expressed among tissues in each population. These data were then used to contrast tissue-specific and population-specific expression patterns. Robustness of ANOVA data was tested using a permutation test; means for the 27 biological samples were randomly permuted 1,000 times between population and tissue and test statistics were recalculated for differences among populations, tissues and tissue-by-population interaction. Agreement between ANOVA and permutation test results would indicate the robustness of the ANOVA model. Finally, in order to graphically illustrate expression similarity among tissues, expression distance between samples was calculated as the sum of differences of log 2 expression values over all genes, and neighbor-joining trees of global similarity of expression patterns among tissues (L, liver; H, heart; B, brain) were constructed [ 38 ] for each population. Additional data files The following additional data are available with the online version of this paper. Additional data file 1 lists the results from statistical analyses for all genes. Listed for each gene are p -values associated with statistical tests for differences in expression between populations, tissues, tissue-by-population interaction, and among individuals within populations. Also listed are mean expression for each sample, and columns comparing differences in expression between tissues within each population. Final columns tabulate whether a tissue difference was detected for each comparison, whether this difference was consistent between populations, and whether significant interaction was detected for that gene. Supplementary Material Additional data file 1 The results from statistical analyses for all genes Click here for additional data file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC551533.xml
539062	Gene Signatures Predict Interferon Response for MS Patients	null	Multiple sclerosis (MS) can be an unpredictable disease. It develops when the body's immune system attacks healthy nerve cells and disrupts normal nerve signaling. Patients experience a wide range of symptoms—including tingling, paralysis, pain, fatigue, and blurred vision—that can appear independently or in combination, sporadically or persistently. Although symptoms appear in no particular order, flare-ups are common in the majority of patients. MS flare-ups are commonly treated with beta-interferon. Adverse effects are not uncommon, and, more importantly, a sizable proportion of patients show a reduced response, or no response at all. Given the variability of the disease and treatment response, being able to predict how a particular patient is likely to respond to interferon would help doctors decide how close to monitor the patient or even whether to consider alternative treatments. In a new study, Sergio Baranzini et al. describe a computational model that can predict a patient's therapeutic response to interferon based on their gene expression profiles. Immune cells typically secrete interferons to fend off viruses and other pathogens. Interferons stem viral infection by inhibiting cell division in neighboring cells—thus preventing the virus from reproducing—and triggering pathways that kill the infected cells. It's thought that interferon therapy may relieve symptoms associated with MS by correcting imbalances in the immune system that lead to disease. Interferon therapy produces changes in the gene expression profile of targeted cells—that is, it inhibits or activates certain genes—which in turn alters the cells' activity. Blood samples were taken from 52 patients with relapsing-remitting MS (marked by acute flare-ups followed by partial or full recovery), and their RNA was isolated from a class of immune cells called peripheral blood mononuclear cells. After patients started interferon therapy, blood was taken at specific time points over the course of two years. Baranzini et al. measured the expression level of 70 genes—including a number involved in interferon interactions and immune regulation—at each time point. Expression levels of three genes in beta-interferon responders (red) and non-responders (blue) The authors used statistical analyses to search for gene expression profiles that were associated with patients' therapeutic outcomes. They looked for patterns in analyses of single genes, gene pairs, and gene triplets, and found their model's predictive accuracy increased with gene number. They also looked for genes that showed different expression patterns over the two years based on patient response, time passed, and patient response over time. These analyses identified genes that increased activity independently of clinical response (interferon can activate genes that have no effect on disease), as well as genes that were associated with a good or poor response. Some of these genes were also the best predictors of patient response before therapy was started. This approach can predict the probability of a good or poor clinical response with up to 86% accuracy. Baranzini et al. offer hypotheses to explain how the observed gene activity might produce the differential responses to therapy—for example, a poor response may stem from downstream signaling events rather than from problems with drug metabolism. But the authors caution that the mechanisms connecting these genetic signatures to specific outcomes—and the mechanisms that produce a positive interferon response—have yet to be established. For now, these patterns should be thought of as markers. Still, these results suggest that doctors could one day tailor MS patients' treatments to their molecular profile, and perhaps take some of the uncertainty out of this capricious disease.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC539062.xml
529265	Psychological trauma and evidence for enhanced vulnerability for posttraumatic stress disorder through previous trauma among West Nile refugees	Background Political instability and the civil war in Southern Sudan have resulted in numerous atrocities, mass violence, and forced migration for vast parts of the civilian population in the West Nile region. High exposure to traumatic experiences has been particularly prominent in the Ugandan and Sudanese of the West Nile Region, representing an indication of the psychological strain posed by years of armed conflict. Methods In this study the impact of traumatic events on the prevalence and severity of posttraumatic stress disorder (PTSD) in a random sample of 3.339 Ugandan nationals, Sudanese nationals, and Sudanese refugees (1.831 households) of the West Nile region is assessed. Results Results show a positive correlation between the number of traumatic events and the number of endorsed PTSD symptoms. Of the 58 respondents who experienced the greatest number of traumatizing experiences, all reported symptoms which met the DSM-IV criteria for PTSD. Conclusions There is a clear dose-effect relationship between traumatic exposure and PTSD in the studied populations with high levels of traumatic events. In this context, it is probable that any individual could develop PTSD regardless of other risk-factors once the trauma load reaches a certain threshold.	Background The debate about the impact of traumatic life events on psychiatric disorders has a long tradition in psychiatry. The introduction of posttraumatic stress disorder (PTSD) into the Diagnostic and Statistical Manual of Mental Disorders (DSM-III [ 1 ]) manifested the general recognition that a chronic condition consisting of characteristic symptoms including involuntary intrusions of the past, avoidance behavior and a condition of general hyperarrousal can be caused by traumatic exposure and must be viewed as mental disorder. Consequently, the original conceptualization of PTSD was based on the implicit assumption that the traumatic event is the main agent for the development of PTSD [ 2 ]. The initial idea was that traumatic events could cause PTSD in anyone regardless of pre-trauma vulnerability. Contrary to this assumption, the following research showed that the development of chronic PTSD is rather the exception than the rule after the experience of a traumatic event. Community studies in the US showed that whereas more than 50% of the population reported the experience of a traumatic event, the prevalence of PTSD was not higher than 7.8% [ 3 ]. Among the different events studied, rape seemed to be the most adverse experience, with about 50% of victims developing chronic PTSD. But even studies that researched PTSD in those who experienced events considered to be most adverse, like torture in prison, found PTSD prevalence rates under 50% [ 4 ]. The realization that traumatic exposure is not a sufficient determinant of PTSD has stimulated vast research into risk and protection factors for the development of PTSD [ 5 , 6 ]. These studies show that pre-trauma developmental vulnerability (adverse childhood, psychiatric history, etc.), peri-traumatic factors (like peri-traumatic dissociation) [ 6 ], posttraumatic factors (like social support) as well as genetic factors [ 7 ], mediate the development of PTSD, although effect sizes were generally small. A popular and intuitively plausible assumption in this context is the dose-response model of PTSD. This hypothesis predicts that the probability for the development of PTSD after the experience of a traumatic event mainly depends on the severity of trauma exposure. Some studies tried to test this hypothesis by relating the objective severity of the traumatic event to symptoms of PTSD. However, the empirical evidence for this model is scarce, with some findings supporting this hypothesis but many failing to confirm a relationship with meaningful effect sizes [ 8 , 9 ]. The probability of detecting a relationship between trauma exposure and PTSD depends on the range and variance of traumatic exposure that is present in the population studied. Studies investigating the relationship between the objective severity of single events and PTSD are restricted to a narrow variance of traumatic exposure. Community studies that assess trauma exposure across different types of traumatic events should be more adequate to examine a dose-effect hypothesis. From a worldwide perspective, even community studies in industrialized countries are restricted to a relatively narrow range of trauma exposure. In contrast, community studies in civil populations affected by war enable the examination of a much wider range of traumatic exposure. These populations present a continuum of subjects ranging from individuals without any history of traumatic events to victims with a history of high numbers of severe events that are rarely to be found in communities without a history of war. Studying a community sample of Cambodian refugees who had fled the Pol Pot regime, Mollica [ 10 ] actually confirmed a clear linear relationship between the number of traumatic events and symptoms of PTSD and depression. Other studies with refugee populations are in line with this result [ 11 - 14 ]. These studies suggests a specification of the dose-response model; i.e., that it is not the severity of a single traumatic event that is linearly related to symptoms of PTSD, but the severity of previous cumulative trauma exposure. Consequently, it can be hypothesized that each individual who has experienced or is experiencing traumatic events will develop PTSD after reaching a certain threshold of traumatic exposure. As this threshold is probably very high, a large number of subjects exposed to a large variance of traumatic events is necessary to test this hypothesis. We examined the dose-response relationship in the context of a large survey in the West-Nile regions of Sudan and Uganda. The study included Ugandan nationals with a quite peaceful development in the last decade, as well as Sudanese nationals living in the Southern Sudan war region and Sudanese refugees who had fled to Uganda. Among these groups we expected a sufficient variance of traumatic exposure to test for the specified dose-response hypothesis, including an adequate number of subjects who had to experience a series of extremely severe traumatic events. Cumulative trauma exposure was estimated by assessing the number of different traumatic event types experienced or witnessed so far. We considered this measurement to be more reliable than assessing the frequency of traumatic events as many survivors of civil wars reported countless exposures to specific traumatic events. To examine the impact of recent traumatic exposure, we also assessed the traumatic event types experienced or witnessed in the last year. Methods As part of a study designed to better understand the impact of forced migration on fertility, mortality, violence and traumatic stress among Sudanese nationals living in southern Sudan and Ugandan nationals and Sudanese refugees living in northern Uganda, we interviewed 3371 individuals from 1842 households in the Ugandan and Sudanese populations in the West Nile. Interviews were structured and were administered in the native languages of Lugbara or Juba Arabic. The study's design involved a multi-stage sampling design. The full training of the interviewers took two months. The project objectives and the rationale behind the structure of the survey instrument as well as that of each question in the questionnaire were discussed in detail. Great attention was also paid to issues such as initial contacts, maintaining a professional attitude while in the field, avoiding influencing the respondent, and reducing interviewer and courtesy biases. The importance of collecting information by means of standardized questions so that the same question was asked to all respondents is stressed and questioning and probing skills were developed. Supervisors were instructed separately on data collection guidelines, their roles and their responsibility to ensure data quality. Keeping in mind the sensitive nature of some of the questions regarding violence and trauma and the fact that the team members were from the study population and probably had experiences similar to the respondents, a workshop on sexual and gender-based-violence was conducted by a consultant to the UNICEF office in Kampala, before the survey. The aim of this workshop was to increase awareness and sensitivity of the team towards respondents and their experiences. Another consultant to the project reviewed the team's interviewing skills and the project's data quality control measures just before the start of the survey. Problem areas were identified and remedied. Data were complete and analyzed for N = 3179 respondents: 2,540 (75 %) of the respondents were women (15–50 years of age) and 831 (25%) were men (20–55 years of age). Details of the sampling, translation and assessment procedures, as well as the socio-demographic characteristics of the populations, have been described elsewhere [ 15 ]. Traumatic events were assessed using a checklist consisting of possible war and non-war related traumatic event types (i.e. witnessing or experiencing injury by a weapon or gun, beatings/torture, harassment by armed personnel, robbery/extortion, imprisonment, poisoning, rape or sexual abuse, beatings, abduction, child marriage, forced prostitution/sexual slavery, forced circumcision, etc.). The checklist was compiled after interviews with key informants (security personnel, doctors, community leaders, women's representatives) and 30 respondents from all three populations about their personal history of stressful events. Following these interviews, the single events obtained in these studies were rated as being potentially traumatic by experts. The following pilot checklist was pre-tested among further 44 Ugandans and Sudanese in areas not selected for the survey and modified according to the suggestions of the respondents. A primary item analysis based on inter-item correlations led to the exclusion of some events that were obviously not directly related to traumatic histories, e.g. the experiencing of witchcraft. Events included 19 experienced events and 12 witnessed events. Respondents were asked for each event type if they had experienced or witnessed such an event ever (i.e., lifetime experience) and if it happened in the past year . PTSD in respondents was assessed using the Posttraumatic Stress Diagnostic Scale (PDS), modified for assessment by trained lay interviewers [ 16 ]. The PDS is a self-report measure widely-used in industrialized countries as a screening instrument for the diagnosis and severity of PTSD based on DSM-IV Criteria. Confidentiality was assured and it was explained that researchers were not working for any UN or Ugandan government organization. Informed consent was obtained using a standardized form explaining the potential risks of participation and explaining that no compensation would be provided. Informed consent forms were signed by the respondent and a witness; fingerprints were taken from illiterate respondents. No financial incentives were provided and respondents were informed that no improvements in living conditions were to be expected as a result of participating in the survey. Respondents were provided with referrals to counseling services provided by NGOs where available. Results As no major clustering effects were expected in this large sample, statistical analyses were carried out on unweighted data. To examine the relationship between continuous PTSD symptoms and the number of event types reported, we correlated the PDS score and its subscales, intrusion, avoidance and arousal with the number of event types. The number of event types in life correlated with the frequency of intrusions (r = .49), hyperarousal (r = .41) and avoidance (r = .47), all P < 0.001. The PDS sumscore correlated significantly (P < 0.001) with the number of traumatic events in the past year (r = .45) and for the whole life (r = .49; see figure 1 ). Overall, 31.6% of the male and 40.1% of the female respondents (N = 3179) fulfilled the DSM criteria for a PTSD-diagnosis. We divided the whole population studied in the survey into different groups based on the number of traumatic event types reported, separately for the events reported for last year and in life . The initial division was made as follows: the first group consisted of respondents endorsing 0–3 event types, the second group consisted of individuals endorsing 4–7 event types. Each following group endorsed an additional four or more event types. Because the number of individuals in the groups of 12–15, 16–19, 20–23 and 24–27 event types was very small for the analyses of events reported last year ( n = 38, 14, 8, 13, respectively), these groups were merged to two groups of 12–19 and 20–27 event types. Figure 2 shows the number of individuals and the prevalence of PTSD in these groups, separately for the groups based on the events reported for the whole life and for last year. The presentation indicates a near linear rise for increasing psychological strain with the number of traumatic event types ranging from 23% in respondents who reported three or fewer traumatizing experiences to 100% prevalence of PTSD in those who report 28 or more traumatic event types in their past. Figures related to traumatic event types in the past year display and even more pronounced increase of PTSD symptoms with significantly higher prevalence rates for the first three categories of numbers of events (Figure 2 ). Figure 2 Prevalence of PTSD and number of individuals in groups of respondents. In this figure respondents are pooled on the basis of number of traumatic event types reported for whole life and last year. Discussion High prevalence rates of PTSD have been reported for three different population groups in the West Nile: Sudanese nationals (44.6%), Sudanese refugees (50.5%) and Ugandan residents (23.2%) [ 15 ]. Here we show that the exposure to traumatic events and the number of different types of traumatic experiences in particular can account for the different proportion of PTSD cases. The prediction of increased PTSD prevalence with increasing number of traumatic events is consistent with other studies investigating victims of organized violence [ 11 - 14 ]. As demonstrated, the number of traumatic events correlated equally strong with avoidance and with re-experiencing symptoms but coefficients were weaker, although still significant, for the hyperarousal cluster. These results are in agreement with [ 17 ], who also found a strong correlation between cumulative trauma and symptoms of re-experiencing and avoidance. Contrary to these findings, Mollica [ 10 ] could not find a correlation with avoidance symptoms. Problems in the translation of the avoidance items in the PTSD instruments might be responsible for this difference, as subtle modification in the translation process may turn PTSD avoidance criteria (like "less interest in important activities" or "feeling as if future plans will not come true") into unspecific depressive items that are unrelated to a traumatic experiences. Typically, even severe single traumatic event produce PTSD in not more than half of those affected. Therefore, PTSD is not an inevitable consequence of potentially traumatizing events. Results from this study, however, suggest that there may be no ultimate resilience to ward off PTSD or that a psychobiological breaking point exists for even the most resistant individual. In the three population groups that were surveyed, each respondent experiencing 28 or more different traumatic event types developed the full set of symptoms of PTSD. This cumulative trauma threshold identified in this study is very high and affected only a small minority of persons even in a war-torn population. Nevertheless, if the cumulative exposure to traumatic events is high enough, these results indicate that anybody will develop chronic PTSD. We conclude that there is no ultimate resilience to traumatic stress and that the repeated occurrence of traumatic stress has a cumulative damaging effect on the mental health of the victim. In these conditions, the effect of pre-trauma factors is reduced to the modulation of the probability of exposure to traumatic events itself. The factors that determine who is exposed to many traumatic events and who manages to flee to secure places may depend on pre-trauma psychological factors. Further studies with war-populations should examine whether the exposure to traumatic events only depends on uncontrollable external factors or whether individual factors contribute to a person's ability to seek safe places. Conclusions High levels of trauma exposure is found in populations affected by civil war. We show that PTSD, the major psychological consequence of war events, is linearly correlated with traumatic exposure, thus explaining the high prevalence rates of PTSD generally found in war-torn societies. These findings highlight the need for reducing the frequent exposure to traumatic events by preventing wars, controlling the violence in wars, and providing safe and stable living environments for refugees. At the same time, the presence of high numbers of PTSD cases requires the implementation of individual and community based treatment programs. Given very limited resources in refugee communities, these centers must be created to provide short-term care and must be manageable by local personnel [ 18 , 19 ]. The provision of appropriate mental health assistance is necessary to break the vicious cycle of violence and psychological morbidity. Competing interests The author(s) declare that they have no competing interests. Authors' contributions FN, MS, UK & TE designed the study. FN, MS and UK composed the set of instruments. UK was responsible for original instrument translation and data collection, FN, MS, CK and TE for the validation part. FN and CK performed the data analysis. CR, TE and FN drafted the original manuscript and all authors revised and approved the final manuscript. Figure 1 Scatterplot of number of traumatic event types for whole life and severity of PTSD symptoms. A number randomly chosen in the interval between -.05 and +0.5 was added to both the abscissa and the ordinate to visualize overlapping points. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC529265.xml
535894	A national survey of the prevalence of schistosomiasis and soil transmitted helminths in Malaŵi	Background Past estimates have put the prevalence of schistosomiasis between 40% and 50% in the Malawi population overall based on studies undertaken ten years or more ago. More recent surveys in known high risk areas find similar levels. However control measures, changing ecology and migration may have led to changes in the prevalence of schistosomiasis in different parts of Malawi. A national schistosomiasis and soil-transmitted helminth (STH) survey was undertaken to measure the distribution, prevalence and intensity of infection in November 2002. Methods A school was selected randomly from a random sample of 30 Traditional Authorities stratified by six distinct ecological zones, and 1,664 year 3 pupils (9–10 year olds) were questioned about recent illnesses and "red urine". Samples of urine and faeces were examined for the presence of eggs using the standard Kato-Katz technique for soil-transmitted helminths and intestinal schistosomiasis and urine samples using the filtration technique for Schistosoma haematobium . Results The prevalence of Schistosoma mansoni is 0.4% (95% CI 0–1.3%), S. haematobium 6.9% (95% CI 1.9 – 11.9%), hookworm 1.3% (95% CI 0.4–2.3%), Ascariasis 0.5% (95% CI 0.1–1.0%) and trichuriasis 0% in year 3 pupils (modal age 10 years of age). Intensity of infection is low for all infections except for 2.5% who have high intensity S. haematobium infection. The "red urine" question is 67% sensitive and 80% specific for positive S. haematobium microscopy. Conclusions The reduction in prevalences may be real as a result of recent control measures, or false if historical results were based on surveys of high risk populations. Another explanation is that this survey used an unrepresentative sample of schools. Detailed analysis suggests this is unlikely. Recommendations include the use of a 30% positive threshold for the "red urine" screening question to be used in schoolchildren in high prevalence areas. This survey, based on a national probability sample excluding the northern region lakeside area, finds much lower overall prevalence and intensity of schistosomiasis and STHs than previous estimates based on selected surveys. Disease control featuring chemotherapy may be having a profound effect. The localised nature of the distribution of the infections means that control programmes may work best if undertaken at district level or below. "Red urine" questionnaire surveys may help identify hot spots.	Background Chronic infections with soil-transmitted helminths and schistosomiasis are common in Malaŵi and cause considerable morbidity [ 1 ]. The Ministry of Health wishes to establish the distribution of these infections prior to a reassessment of national policy regarding their control. All water bodies in Malawi are considered to be potential transmission sites for schistosomiasis [ 2 ]. The most recent national estimate puts the prevalence of schistosomiasis between 40% and 50% of the population [ 3 ]. This is based on surveys carried out prior to 1991. Numerous local surveys have been conducted more recently in different parts of the country [ 4 - 7 ]. In a survey carried out by Randall et al in 1996 in northern Malawi, S. mansoni and S. haematobium were detected in 27% and 20% of schoolchildren respectively [ 8 ]. A survey performed by the Bilharzia Control Programme in Mangochi reported in 1999 showed between 60–80% of school children living along the lakeshore has S. haematobium [ 9 ]. This level of infection was confirmed in a later survey performed by the Lakeshore Schistosomiasis Control Project, which found the prevalence amongst school aged children (5–15 years old) to be as high as 80% along the lake shore, with some areas having 100% infection rates [ 2 ]. In Malawi there is little contemporary information available on helminth infection. Furthermore, fragmented control measures, changing ecology [ 10 ] (such as considerable water resource development and deforestation in the last 10 years) and migration may have led to marked changes in the prevalence of schistosomiasis within regions in Malawi since previous studies were done. A national schistosomiasis and soil-transmitted helminth (STH) survey was undertaken to measure the distribution, prevalence and intensity of infection in November 2002. Results from the survey are reported in this paper. Methods Methods follow recent WHO advice to allow comparison with other international studies [ 11 ]. This includes a symptom and illness questionnaire and urine and faecal sampling. The use of questionnaires as a rapid assessment tool for urinary schistosomiasis has been evaluated in multi-country studies [ 4 ]. Seven areas (with the number of Traditional Authorities in each) represent the seven distinct ecology and soil conditions in Malaŵi (figure 1 ):- Northern Region highland (23) Northern Region lakeside Central Region highland (62) Central Region lakeside (17) Southern Region highland (51) Southern Region lakeside (31) Urban (3) The northern region lakeside had recently been surveyed by the Karonga Prevention Study [ 8 ]. To save expense this ecological area was not resurveyed. Hence only six areas were included in this survey. Study population Primary-school children were chosen as the target population. There are several reasons for conducting the survey in this age group. Children consistently have the highest prevalence and transmission of schistosomiasis and soil-transmitted nematodes (except hookworm); treatment via schools on a national scale is feasible [ 12 ] (and primary school enrolment is reasonably high in Malawi) and extremely cost-effective [ 13 ]. Third-year school classes (9–10 year olds) are commonly surveyed and this was the age-group for this survey. Sample frame and selection The sample was selected using a stratified multi-stage cluster design [ 14 ]. The six ecological zones constituted the strata. School children were clustered within schools which were clustered within Traditional Authorities (TA). Five Traditional Authorities (TA) were randomly selected from the total in each ecological zone using a computer generated random numbers technique. A list of all primary schools in these TAs was compiled by contacting the relevant District Education Officer and one school was selected at random. Private schools were included though the vast majority were government schools. In the three urban districts of Blantyre, Lilongwe and Mzuzu five schools were selected. A total of 30 schools were surveyed between 25 September and 14th November 2002. A census of all children attending standard 3 who were present on the day of the survey was taken and all were enrolled. Thus between 36 and 71 third year schoolchildren (aged 8–10 years) in 5 schools selected at random from each of six ecologically distinct regions of the country were sampled (Figure 2 ). A total of 1664 children were enrolled. Questionnaire All children were administered a standard symptom questionnaire by a class teacher in their local language [see Additional file 1 ]. The questionnaire was completed before specimen collection and was designed to assess general health without biasing responses towards helminthiasis in particular. Children were asked about symptoms during the past month [e.g. cough, itch, headache, fever] and whether they had particular illnesses during the past month [e.g. malaria, diarrhoea, skin, eye disease, bilharzia]. Specimen collection methods School children were asked to provide a stool sample and at least 10 ml of urine (collected mid-morning). Stool samples were stored in a cool box during transfer to the laboratory, and then kept at 4°C until processed within 24 hrs of collection. Urine and faecal analysis Stool samples were tested for the presence of eggs using the standard Kato-Katz technique [ 15 ] for soil-transmitted helminths and intestinal schistosomiasis. Samples with ascaris more than 50,000 eggs per gram (epg), hookworms more than 4,000 epg, trichuria more than 10,000 epg and S. mansoni more than 400 epg were defined as high intensity. Urine samples were tested using the filtration technique for S. haematobium . Infection with S. haematobium was classified as high intensity if at least 50 eggs per 10 ml of urine were detected or there was visible haematuria [ 16 ]. Data analysis Data were double entered into EpiInfo and then exported into STATA v7 for initial analysis. SPSS was used for chi squared, paired-t and McNemar tests. EpiInfo 2002 was used to calculate adjusted prevalence rates for ecological zones and national prevalence. The prevalences and 95% confidence intervals were computed incorporating the stratification by ecological zones. Sampling weights were computed as the product of the number of TA's in an ecological zone times the number of schools in the selected TA. For the urban zone, weights were adjusted for the fact that there were only three TA's and two schools were selected from two of them (Complex sample frequency analysis, Epi Info 2002, CDC Atlanta). The relationship between the prevalence and intensity of S. haematobium infection in schools was investigated using Pearson product moment correlation coefficient (SPSS). Results The sample A total of 1,664 children were surveyed. Between 36 and 71 children were included from each school. There were 1,662 completed questionnaires, 1,546 stool samples and 1,638 urine specimens collected. Just over half (50.7%) of the sample was female. The mean age (for 1,658 children where age was recorded) was 10.7 years [range 6–17 years]. The mean age of boys was 10.9 years and of girls 10.5 years. Prevalence and intensity of infection Schistosomiasis A total of 140 children from 19 different schools had S. haematobium detected by urine microscopy, of which 45 had high intensity infection. The prevalence in schools ranged between 0 (11 schools) and 43.1% (a school in the southern lowlands). Twenty one children from eight different schools had visible haematuria. Only 8 children from 2 different schools had S. mansoni detected by stool microscopy and none were high intensity infections. Overall, the national prevalence of S. mansoni was found to be 0.4%. The national prevalence of S. haematobium was found to be 6.9% (95% CI 1.9–11.9%) (Table 1 ). 2.5% of these were high intensity infections [range 0–23.3%]. There was a clear association between prevalence and intensity (r = 0.9, n = 30, p = <0.01); schools with high prevalence of S. haematobium also had high intensity of infection. There was a wide range of ages in standard 3, with mode of 10 years of age. The observed prevalence was higher in older children (see figure 3 ) and boys (9.0%) compared with girls (8.2%). Soil transmitted helminths The number of children with detectable helminth infestation was low. Twenty-six children from 13 schools had hookworm detected from stool, none of which were high intensity. Eighteen children from nine different schools had ascaris eggs detected and no trichuris eggs were detected at all. There was no obvious pattern of co-infection with schistosomiasis and soil-transmitted helminths across the schools sampled; in 16 schools there was concordance with either both or neither type of infection detected; in 14 schools there was no concordance. The prevalence of STHs was much lower than that of schistosomiasis (Table 1 ). Only 1.8% [95% C I 0.6–3.1%] of children had evidence of infection with any of hookworm, ascaris or trichuriasis on stool microscopy. There were no heavy intensity STH infestations detected. Prevalence and intensity of infection by ecological region Schistosomiasis S. haematobium infection was detected in all ecological areas (Table 2 ). The highest rates were in the Southern Lowland (23.2%) and the Central lowland (9.5%); the other four areas examined all showed a prevalence of between 2% and 7.4%. There was a wide range in the prevalence of S. haematobium infection in all areas. S. mansoni was rarely detected, and only found in the Southern Highlands (1.3%). To assess the sampling method used to select schools we classified the selected schools into high and low expected prevalence of S. haematobium based on ecology and known previous prevalence and compared our survey results with these expectations. Of the 12 expected high prevalence schools seven were found to be high and 5 unexpectedly low. Of the 18 expected low prevalence schools 12 were found to be low and 6 unexpectedly high. Unexpected results were found in the expected low prevalence as much as in the expected high prevalence schools using the McNemar test for symmetry actual (p = 0.29). We also compared the historical results for Malawi districts compiled by the London School of Hygiene and Tropical Medicine for the Schistosomiasis Control Initiative in 2001 (personal communication with Simon Brooker) with our results. Prevalence has fallen overall by 24% (95% confidence intervals of 16–32%). A formal paired t-test finds this reduction is true for low as well as high prevalence schools (t = 6.4 with 29 degrees of freedom and highly statistically significant). Soil transmitted helminths There was also a low prevalence of STHs in all areas, though urban areas had the highest rates for any STH (6.6%) and hookworm was most prevalent in the Northern Highland area (2.6%). There were no STHs detected in the Central lakeshore schools. Questionnaire responses Total number of respondents was 1,662 (Table 3 ). It is possible to calculate the sensitivity and specificity of the "red urine" question in respect of the microscopic presence of S. haematobium (Table 4 ). The sensitivity of the "red urine" question in respect of S. haematobium being seen by microscopy is 67% and the specificity 80%. At a national average prevalence of 8.5% the "red urine" question has a positive predictive value of 24% and a negative predictive value of 96%. At the higher prevalence (22.5%) found in the Southern Lowlands zone the positive predictive value is 49% and the negative predictive value is 89%. Discussion Prevalence and distribution The overall prevalence of schistosomiasis is well below expectations. Based on previous studies in Malawi the overall prevalence of schistosomiasis was thought to be between 40% and 50% in the population. Of course these were surveys of selected populations, perhaps undertaken in the season of high transmission. Certainly the disease is localised in Malawi. In the study conducted in the northern lakeshore area in 2000 by Randall et al [ 8 ], schoolchildren from four schools were screened and there was a wide range of prevalence: 5%–57% on S. haematobium and 6%–42% in the case of S. mansoni . The school located closest to rice fields had a high prevalence of schistosomiasis. In another recent survey carried out at the Cape McClear peninsula in the Southern Lake zone results showed markedly different S. haematobium and S. mansoni prevalence amongst adjacent villages situated on the lake shore and inland (Paul Bloch, personal communication). Our survey, representative of all schoolchildren in the country, and undertaken just before the rainy season, suggests far lower levels of 7% for S. haematobium and 0.4% for S. mansoni . The finding highlights two important features of schistosomiasis in Malawi. Firstly, these infections are highly localised and secondly they are not common in general. As expected higher prevalence rates are found in the Southern lake/lowlands zone. S. mansoni does not seem to be a generalised problem in any particular ecological area in the country. The overall prevalence of STHs is lower than expected. The survey conducted in Karonga District (Randall et al., 2000 [ 8 ]) finds a higher prevalence of STHs as well as of S. mansoni and S. haematobium . The differences could be due to seasonal variation (the survey in Karonga district was conducted after the rainy season). This possibility is supported by a low prevalence found in Karonga district in samples collected in November 2001, prior to the rains (unpublished observation, MA Perez). Supporting evidence is seen in the prevalence rates found in urban areas where rains would not affect transmission as much as in rural areas. Similar results are found in previous studies in urban Blantyre: S. mansoni (0%), Ascaris (18%) and Hookworm (0.4%) [ 7 ]. Another explanation for the higher prevalence rates found in Karonga District could be the different methodology used to process the stools. The Kato-Katz method [ 16 ] was used during our survey in order to calculate intensity of infection. Due to time and personnel constraints, only one Kato-Katz slide was read from each sample, and this could have affected the sensitivity of our testing [ 17 , 18 ]. On the other hand the survey in Karonga District was done using a commercial kit (Parasep ® , Intercep Ltd) in order to concentrate the stools and achieve high diagnostic sensitivity. Because of the dissimilar laboratory methods we have been unable to include results for the Northern Lakeland Ecological zone in the estimate of overall prevalences for the whole country. The northern lakeshore zone has tended to have high prevalences. The impact of omitting this zone is likely to underestimate the national results. We intend to survey the zone using our own laboratory methods when funds are available to complete the national picture. Although this survey is designed to be representative of this age group in Malawi, it is limited in scale by the budget ($12,000). This is an important limitation when one considers that the distribution of schistosomiasis in populations is likely to be highly localized [ 19 ]. The population sampled is representative of the general population of school children in standard three with respect to age and sex distribution. However, this survey does not represent children in Malawi who do not attend school or who were absent on the day of the survey due to illness, or who live in the northern lakeland zone. Despite the introduction of free primary education, the net enrolment rate in Malawi is 78%. There is little difference between poor and non-poor households in regard to the proportion of primary school-aged children in school [ 20 ]. However, poor children are likely to drop out before reaching standard 5, especially girls in rural areas. (Malawi poverty reduction strategy p.7). Children enrolled but absent due to sickness may have higher rates of schistosomiasis. In future if localised surveys in areas of low school enrolment are contemplated surveyors should consider including children who are not attending school or are absent due to illness. Drug treatment has been widely if intermittently available. The National Schistosomiasis Control Programme does not have documented evidence of where and when universal drug treatment has been offered to communities. Nor does it have details of where and when Praziquantel has been available for prescription to symptomatic individuals at health centres and hospitals. Supplies and universal treatment programmes have tended to depend on intermittent interest and funding from NGOs. However Praziquantel seems to have been widely available (Kayuni SA. National Survey to find out difficulties people face in taking regular anti-schistosomiasis drugs in Malawi. Fourth Year Student Project. College of Medicine, Blantyre, 2003) and may have had an effect. In this survey of 264 people living in high risk villages in eleven randomly selected districts 71% reported receiving anti-schistosomiasis treatment. Several NGOs such as Save the Children (US) in Mangochi and World Vision in Chikwawa in recent years have studied helminth infections in schoolchildren. Save the Children (US) surveyed and treated schoolchildren in the entire Mangochi District after finding a prevalence of Ascaris lumbricoides (1.9%), Hookworm (24%), and S. mansoni (16%) in four schools. The last survey done in 2000 showed an overall prevalence of A. lumbricoides (0.4%) and Hookworm (1.5%) on the four schools after two consecutive courses of treatment (Mary Mukaka, Save the Children, personal communication). Similar interventions may have had an effect on the prevalence of infections in other parts of the country. Another possible explanation for the unexpectedly low prevalence results could be due to a sampling method that selected a low preponderance of schools in high risk areas. However the results of the analysis looking at expected and actual prevalence found in each school suggest this did not happen. Unexpected results were found in the expected low prevalence as much as in the expected high prevalence schools using the McNemar test for symmetry actual (p = 0.29). The reason for the comparative low prevalence rate of S. haematobium now is not due to asymmetric sampling of low risk schools; it is due to a marked trend of a reduced prevalence across the board. It is not just the high risk schools that have lower prevalence now. Low risk schools have lower rates as well. Compared to the historical results for Malawi districts compiled by the London School of Hygiene and Tropical Medicine for the Schistosomiasis Control Initiative in 2001, prevalence has fallen overall by 24%, and this reduction is true for low as well as high prevalence schools. Our interpretation is that the results reflect a widespread reduction in schistosomiasis prevalence and not a biased sample in the current survey which inadvertently missed schools in high risk areas. However bias may still be an explanation for the change. The widespread reduction in prevalence could be due to bias in the selection of high risk schools in the historic studies rather (than in the selection of low risk schools in our survey). Our sampling fraction of 30 of a possible 3,900 schools is sufficient to give statistical confidence in our overall prevalence result of 6.9% (95% CI 1.9 – 11.9%) for S. haematobium so long as our schools and Traditional Authorities were chosen randomly, which they were. Questionnaire results Of the questions asked, only the red urine question is sufficiently sensitive and specific to be used as a screening test. It is a cheap way of assessing schistosomiasis prevalence and as such may be useful to identify hot spots within districts. Implications for the control programme Past control programmes, despite their fragmentary application in Malawi, seem to have achieved much in recent years. High intensity infections are now rare, and it is of course these that are thought to cause morbidity. There remain some high prevalence locations but these are uncommon now. The survey findings suggest that future approaches to control can focus on the identification of hot spots through local surveys. These should be undertaken in districts in high prevalence ecological zones, probably through "red urine" questionnaire surveys of school children. Because of the low positive predictive value of the "red urine" question, a low threshold should be used to categorise each school. The categorisation as proposed in Table 5 takes this into account and provides a simple way to identify those schools that require a urine microscopy survey. Conclusions The highly effective drugs now available and the control methods used in Malawi seem to have reduced the prevalence of both schistosomiasis and soil helminths to low levels. There remain some localised areas of high prevalence for which local control measures will be required. It seems that the epidemiology of these diseases makes local control at district level or below the preferred public health approach. Competing interests The author(s) declare that they have no competing interests. Authors' contributions BS conceived the survey and provided logistical advice. CB designed the survey and undertook the final analysis and first drafting. BP supervised the survey and undertook the initial analysis. PM undertook the randomization and led the survey team. MP supervised the quality control of the laboratory work. All contributed to the final report. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 questionnaire; the questionnaire in Chichewa used in the schools to find out about the health of the pupils surveyed. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535894.xml
555755	Recognition and binding of mismatch repair proteins at an oncogenic hot spot	Background The current investigation was undertaken to determine key steps differentiating G:T and G:A repair at the H- ras oncogenic hot spot within the nuclear environment because of the large difference in repair efficiency of these two mismatches. Results Electrophoretic mobility shift (gel shift) experiments demonstrate that DNA containing mismatched bases are recognized and bound equally efficiently by hMutSα in both MMR proficient and MMR deficient ( hMLH1-/- ) nuclear extracts. Competition experiments demonstrate that while hMutSα predictably binds the G:T mismatch to a much greater extent than G:A, hMutSα demonstrates a surprisingly equal ratio of competitive inhibition for both G:T and G:A mismatch binding reactions at the H- ras hot spot of mutation. Further, mismatch repair assays reveal almost 2-fold higher efficiency of overall G:A repair (5'-nick directed correct MMR to G:C and incorrect repair to T:A), as compared to G:T overall repair. Conversely, correct MMR of G:T → G:C is significantly higher (96%) than that of G:A → G:C (60%). Conclusion Combined, these results suggest that initiation of correct MMR requires the contribution of two separate steps; initial recognition by hMutSα followed by subsequent binding. The 'avidity' of the binding step determines the extent of MMR pathway activation, or the activation of a different cellular pathway. Thus, initial recognition by hMutSα in combination with subsequent decreased binding to the G:A mismatch (as compared to G:T) may contribute to the observed increased frequency of incorrect repair of G:A, resulting in the predominant G G C → G T C (Gly → Val) ras- activating mutation found in a high percentage of human tumors.	Background Several different DNA repair systems have evolved within all living cells to correct mispaired or damaged nucleotide residues generated either by endogenous events or by exposure to exogenous mutagenic agents [ 1 , 2 ]. The frequency of mutational events varies widely within the genome, and specific sites harboring increased frequency of mutation are now defined as 'hot spots' of mutation. The human ras protooncogene family contains three such hot spots – codons 12, 13, and 61. Factors contributing to these and other hot spots of mutation are still largely unknown, despite much investigation, but now appear to have several different contributions, such as type of DNA damage, genomic location, surrounding sequence, cell cycle position, efficiency of the optimal DNA repair pathway, and involvement of alternate repair and other cellular pathways. DNA mismatch repair (MMR) is a repair system that corrects mispaired nucleotides and insertion/deletion loops (IDLs), resulting from replication, recombination, or repair errors. Consequences of defects in this DNA repair pathway are evidenced by microsatellite instability (MSI), elevated mutation frequency throughout the genome, enhanced recombination events, as well as tolerance to cytotoxic effects of alkylating agents as evidenced by decreased apoptosis. Deficient MMR is, in turn, associated with hereditary nonpolyposis colorectal cancer (HNPCC), as well as other types of sporadic tumors in humans and animal models [ 3 - 6 ]. Although the contribution of MMR to highly mutable genomic sites other than microsatellite sequences is largely unknown, several specific genetic mutations involved in neoplastic progression, including ras- activating mutations, have been reported as frequent occurrences in HNPCC and other MSI tumors [ 7 , 8 ]. DNA MMR is conserved amongst highly divergent species, reflecting the essential role of this DNA repair process [ 9 , 10 ]. The MMR system in eukaryotes is more complex and has different mismatch-specific repair efficiencies than that of E. coli [ 11 - 13 ]. Several human homologs of bacterial MutS and MutL proteins have now been identified [ 14 - 19 ]. The primary human homologs for MutS that play instrumental roles in MMR include hMSH2, hMSH6, and hMSH3 [ 20 , 21 ]. The hMutSα heterodimer (hMSH2 and hMSH6) has been demonstrated to recognize and bind DNA mispairs and short IDLs [ 14 , 21 - 24 ]. The hMutSβ heterodimer (hMSH2 and hMSH3) preferentially recognizes and binds IDLs of up to 12 nucleotides. Human cells lacking hMSH2 protein expression also lack its cognate partners, hMSH3 and hMSH6 (due to decreased stability). This lack is associated with defective MMR, microsatellite instability, and is associated with a high percentage of HNPCC [ 25 ]. MutL homologs that are most relevant to human MMR are hMLH1 and hPMS2, which form the hMutLα heterodimer [ 9 , 26 ]. Similar to the observed instability of individual MutS homologs, lack of hMLH1 protein expression results in the lack of hPMS2 protein, which in turn results in microsatellite instability, defective MMR, and is also associated with a high percentage of HNPCC [ 9 , 10 , 15 , 25 , 27 - 32 ]. The MutLα heterodimer is thought to act as a molecular matchmaker between the MutSα-DNA complex and downstream enzyme activities responsible for subsequent identification, excision, and replacement of the incorrect base [ 9 , 19 ]. Biochemical interactions and genetic studies have further implicated proliferating cell nuclear antigen (PCNA), exonuclease I (EXO1), replication protein A (RPA), replication factor C (RFC), and DNA polymerase δ as active participants in the MMR pathway [ 19 , 27 , 29 , 30 , 33 - 35 ]. Recently, the differential requirement of specific proteins associated with MMR have been identified for both 3'-nick directed and 5'-nick directed MMR by the use of an in vitro model [ 35 , 36 ]. We have previously demonstrated that the efficiency of correct mismatch repair within the cell can differ significantly, depending on exact type of mismatch, site-specific location, phase of cell cycle, or cell type [ 23 , 24 , 37 , 38 ]. Within this report, we have focused on a more precise understanding of the differences between MMR protein interactions with a G:A (least repaired) or G:T (best repaired) mismatch located at H- ras codon 12 to better understand molecular events leading to activating mutations at this site. Our current results indicate that initial recognition and subsequent binding for repair signaling by hMutSα may be two separably measurable steps in the MMR pathway that can significantly affect downstream cellular events. Results Specificity of hMutSα protein binding to DNA mismatches at H-ras codon 12 To determine that the binding complex recognizing mismatches at codon 12 of H- ras in HCT116 + Ch.3 nuclear extract is composed of hMutsα, DNA binding reactions were conducted using [ 32 P]-G:T oligos incubated with nuclear extracts from these MMR proficient cells. In confirmation of hMutSα binding to a G:T mismatch at this oncogenic site, the hMutSα :DNA complex was efficiently interrupted by goat anti-hMSH6 and with rabbit anti-hMSH2 (Figure 1 , lanes 3 and 5), but not by BSA, nonspecific goat IgG, or nonspecific rabbit IgG (lanes 1, 2, and 4, respectively). As further evidence of specific hMutsα:DNA mismatch binding, addition of 1.5 mM ATP completely disrupted the gel shifted band (results not shown) [ 24 , 30 ]. Thus, MMR protein binding to a G:T mismatch located at H- ras codon 12 within the nuclear environment appears to be primarily, if not exclusively, hMutSα. Comparison of hMutSα binding 'avidity' to a G:T mismatch located at H-ras codon 12 within MMR competent and deficient nuclear extracts The HCT116 cell line completely lacks both 3' and 5' nick-directed MMR, putatively due to lack of hMLH1 expression, although these cells normally express hMSH2 and hMSH6. To determine if hMutSα recognition and binding of a mismatch at the H- ras codon 12 hot spot might undergo alteration within HCT116 cells, in conjunction with the lack of hMutLα, binding competition experiments were performed. Figure 2 is a comparison of the amount, or 'avidity', of hMutSα binding affinity within MMR proficient (HCT116 + Ch. 3) and within MMR deficient (HCT116) nuclear extracts. These gel shifts demonstrate virtually identical mismatch specific binding avidity of hMutSα to [ 32 P]-G:T-oligo. As well, 50X molar excess of unlabeled G:T-oligo, which does not completely inhibit hMutSα gel shift of the radioactively labeled mismatch, demonstrates a similar degree of competitive inhibition of mismatch specific binding in both MMR proficient and MMR deficient nuclear extracts (Fig. 2 , lanes 3 and 5). These results provide strong evidence that there is no discernable difference in hMutSα binding avidity to a G:T mismatch at H- ras codon 12, despite that HCT116 + Ch. 3 is MMR proficient and HCT116 lacks expression of hMutLα and is MMR deficient. Comparison of competitive inhibition of hMutSα binding avidity to G:T versus G:A mispairs at H-ras codon 12 We have previously determined that the G:A mismatch at H- ras codon 12 is accurately repaired back to G:C much less frequently than G:T [ 37 ]. In addition, we have previously demonstrated that recognition and binding of this poorly repaired mismatch is also very weak within HeLa nuclear extracts [ 23 ]. We therefore asked if competitive inhibition by several different concentrations of these mismatches would reveal more specific mechanisms of these intriguing differences at this same location. Firstly, using increasing concentrations of unlabeled G:T oligo as a competitor for [ 32 P]-G:T-oligo, we observe the expected highly competitive inhibition of the radioactive G:T mismatch gel shift band in the presence of as little as 25X cold G:T-oligo (Figure 3a ; compare cold G:T fold increase lanes "0" up through "100"). In comparison, the radioactive G:T-oligo gel shift band is much less competitively inhibited by increasing concentrations of cold G:A oligo (Figure 3a ; compare cold G:A fold increase lanes "0" up through "100"), further confirming the observed preference of hMutSα for a G:T mismatch as compared to a G:A mismatch at the H- ras codon 12 hot spot of mutation. Comparison of the densitometric band intensities reveal that even 25X cold G:T oligo can successfully decrease hMutSα binding to radioactive G:T by 80%, but that 25X cold G:A decreases hMutSα binding to radioactive G:T by only 10%. An alternate approach to more precisely define the extent of this phenomenon was to measure competitive inhibition of radioactively labeled G:A-oligo (rather than of the avidly bound [ 32 P]-G:T-oligo) by incubation with unlabeled G:A-oligo, to determine if equal ratios of competitive binding might occur for G:T and G:A, despite the very different avidities of hMutSα for each mismatch. This would provide an indication of equal recognition of each type of mismatch by hMutSα, despite unequal binding avidity for these two mismatches. For these experiments, it was necessary to extend incubation periods from 30 min. to 2.5 hr, as sufficient concentrations of gel shifted bands can be detected with the G:A oligo only after extended incubation (unpublished observations). As expected, binding of hMutSα to [ 32 P]-G:A-oligo is approximately 100 fold weaker in intensity than for [ 32 P]-G:T-oligo after the extended incubation (Figure 3b ; compare "Intensity" of lanes 1 and 3 on densitometric graph). However, 100X molar excess of cold G:A-oligo competitively inhibits MMR protein binding to [ 32 P]-G:A-oligo by 89%, which is similar to the same concentration of cold G:T-oligo competitive inhibition of MMR protein binding to [ 32 P]-G:T-oligo (97%), as determined by comparison of the densitometric intensity of each band. All of the above experiments have been repeated with similar results. MMR at H-ras codon 12 hot spot of mutation Table 1 contains the results of a MMR assay designed to score nick-directed MMR as correct, incorrect or unrepaired [ 39 ]. Positive control experiments demonstrate that MMR at the H- ras codon 12 sequence within the pUC19 vector is nick-directed, yielding a high degree of correct repair by MMR proficient E. coli (DH5α); G:T → G:C = 95%, G:A → G:C = 78%, in agreement with our previously published results using a different plasmid vector [ 23 ]. Additionally, inserting the same H- ras 69 mer into a pUC18 vector to determine effects of reverse orientation of the oligo and 'nicked' strand resulted in an almost identical frequency of correct repair at codon 12; G:T → G:C = 100%, G:A → G:C = 78%. These results confirm that the mismatch is recognized and repaired by nick-directed bacterial MMR and that there is no strand bias for MMR during high copy number bacterial replication of this plasmid. Background repair after direct transformation into NR9161 E. coli (MMR deficient) was consistently between 41–44% for both mismatches (data not shown), and is favorably comparable to an average of 50 – 60% background repair efficiency by this and other strains of MMR deficient E. coli (personal communication with Roel Schaaper). Results within Table 1 were determined by incubating plasmids containing a G:T or G:A mismatch at the H- ras codon 12 middle base pair location containing a 5' nick on the "T" or "A" strand (as described in methods) with MMR proficient (HCT116 + Ch. 3) or MMR deficient (HCT116) nuclear extracts. Efficiency of correct, incorrect, and total (correct + incorrect) MMR for each mismatch was subsequently calculated (as described in methods) [ 39 ]. Surprisingly, total repair of the G:A mismatch (30%; G:A → G:C + T:A) was almost twice as high as for G:T (18.2%; G:T → G:C + A:T). This phenomenon was consistently observed within the several different experiments required to obtain the total results depicted in Table 1 , and was statistically significant (p < 0.05). In direct contrast, correct (nick-directed) repair of G:A → G:C (60%) was significantly low when compared to correct (nick-directed) repair of G:T → G:C (96%) (p < 0.005). Inadvertent nicks in the "G" (correct) strand of the H- ras 69 mer or pUC plasmid during preparation could not have contributed to the increased total, or increased incorrect repair (G:A → T:A) results for the G:A mismatch because the efficiency of incorrect repair would then be similar for both the G:T and the G:A mismatch. Instead, MMR results within Table 1 were found to be consistently different for the two mismatches, and are the compiled results of several different mismatch-containing plasmid preparations and subsequent MMR assays. In addition, MMR proficient E . coli nick-directed repair results are consistently high, for both mismatch-containing plasmid preparation subsequently used for each nuclear extract MMR assay (data not shown). As well, MMR deficient E. coli repair ratios (correct to incorrect) are 1:1 after transformation of unmethylated plasmids (prepared by replication in GM2929; E. coli dam - dcm - ). In combination, all of the above control experiments demonstrate a lack of any significant unintentional misrepair events on either strand of DNA, for either site-specific mismatch plasmid preparation. Therefore, the results within Table 1 suggest that increased total repair (correct + incorrect) combined with increased incorrect repair of G:A, as compared to the decreased total repair and increased correct MMR repair of G:T, may result in increased mutational events when a G:A mismatch occurs at this oncogenic site. Somewhat surprisingly, MMR deficient nuclear extracts did not repair either mismatch at codon 12 above background, indicating a complete lack of alternate DNA repair activity that can correct either mismatch at this oncogenic site. Therefore these results indicate that the observed repair efficiencies in MMR proficient nuclear extracts are due solely to MMR activity, rather than to a combination of MMR and other DNA repair pathways. Alternatively, it is possible that the methods used for nuclear extract preparation or the in vitro MMR assay might cause an artifactual decrease in the activity of other DNA repair pathways. Discussion and conclusions Previous investigations of ours have revealed significantly decreased nick-directed G:A → G:C repair at codon 12 of H- ras , as compared to G:T → G:C repair at this location [ 23 , 37 ]. Decreased MMR of G:A has also been observed within different sequences by other investigators, although molecular mechanisms for these differences remain obscure [ 12 ]. The current investigation was undertaken to determine key steps differentiating G:T and G:A repair pathways at the H- ras oncogenic hot spot within the nuclear environment. Our results suggest that, firstly, the MMR pathway is the primary, if not the only, DNA repair pathway that can recognize and is subsequently responsible for the repair of both mismatches at this oncogenic location within the cell. Secondly, although 'recognition' of either G:T or G:A by hMutSα is likely equal, and is essential for initiation of MMR, the avidity – or strength – of hMutSα binding to either mismatch is not equal, and may play a significant role in the decision of whether to activate nick-directed correct MMR, or the activation of an alternate response pathway. In support of this concept, and in correlation with weak hMutSα binding to G:A, we have observed significantly decreased nick-directed (correct) MMR of G:A → G:C (60%), as compared to nick-directed MMR of G:T → G:C (96%). We have however, consistently measured almost twice the efficiency of total repair of G:A (to either G:C or T:A) as compared to total G:T repair (to G:C or A:T) at this oncogenic site. This phenomenon appears to be due to increased non nick-directed incorrect repair of G:A → T:A (rather than to replication without repair, which results in a mixture of G:C and T:A). This raises the possibility that (a) initial recognition of a mismatch is essential but separable from (b) differential binding of MMR proteins to specific mismatches, which in turn is directly correlated either with nick-directed correct MMR, or with alternate events other than nick-directed MMR. In support of this concept, Wang, et. al. has recently demonstrated by atomic force microscopy that E. coli MutS-DNA complexes exist in two conformations [ 40 ]. The initial recognition of a mismatch by MutS results in a localized kink in the DNA conformation and is termed the initial recognition complex (IRC). The second step is required for MMR and is a further conformational change in which the localized kink in the DNA becomes unbent. This is called the ultimate recognition complex (URC) and may be the conformation required for ATP activation and subsequent MMR activity. It is also possible that either differential binding kinetics (not measurable by gel shift), or very low avidity of binding or a different molecular binding mechanism (undetectable by gel shift) may contribute to subsequent cellular events specific to the G:A mismatch. These possibilities are currently being investigated. An additional hypothesis has been proposed by Junop, et. al., placing E. coli MutS in the role of 'authorizing' different repair events [ 41 ]. Also, it is now well documented that while hMutSα recognizes DNA damage other than mismatches, the MMR pathway does not appear to play a direct role in the repair of these damaged bases, but rather is associated with initiation of cell cycle and/or apoptotic events and therefore is described as a "sensor of genetic damage" within this context. The molecular mechanisms contributing to the various cellular activities associated with the well documented hMutSα recognition and differential binding to different DNA structures is not yet clear. Although there does not appear to be a consistent physical size or shape of hMutSα recognizable structures that trigger different pathways of cellular activity, DNA sequence context does appear to play a role [ 9 , 12 ]. Figure 4 is the model summarizing our hypothesis. This concept is compatible with our current set of experimental data, and with the recently described two-step conformational alteration of the E. coli MutS-DNA recognition complex [ 40 ]. Briefly, hMutSα appears to equally recognize both G:T and G:A mismatches at the codon 12 hot spot of mutation, but binds more avidly to G:T, or the "URC" conformation. The stronger, or alternate conformational, binding of hMutSα to G:T results in increased accuracy of MMR, but may decrease overall repair efficiency. The less avidly bound G:A mismatch, or the "IRC" conformation, is repaired with increased total efficiency as compared to G:T, but accuracy is sacrificed. These results agree with, and build upon our previous experimental results, and also agree with the presence of specific mutations predominantly found in human tumors [ 23 , 24 , 37 ]. It has now been demonstrated by several different investigators that, although any base pair other than G:C at the H- ras codon 12 location is activating, the majority of human tumors containing mutations at this site are G:C → T:A transversions [ 42 - 44 ]. Our observed increased overall MMR of G:A, combined with a high ratio of incorrect MMR of G:A → T:A at this oncogenic location correlates well with the predominant G G C → G T C (Gly → Val) ras- activating mutation found in a high percentage of human tumors [ 42 , 44 , 45 ]. Thus the demonstration of a significant increase in the frequency of both total repair and incorrect repair of the G:A mismatch at the H- ras codon 12 oncogenic hot spot of mutation, combined with a complete lack of rescue by other DNA repair pathways, is biologically relevant. Methods Nuclear extracts, oligonucleotides, and site-specific mismatched plasmids Human colorectal carcinoma cell lines HCT116 and HCT116 + Ch. 3 were cultured in Iscove's Modified Dulbecco's Medium supplemented with 10% fetal bovine serum (FBS) at 37°C, 5% CO 2 . HCT116 + Ch. 3 cell line also received 0.4 mg/ml Geneticin (G418). Both HCT cell lines were kind gifts of C. Richard Boland; UCSD. Nuclear extracts were prepared as described previously [ 46 ]. Synthetic oligonucleotides of 69 bases containing the coding strand sequence 5'- AATTC ACGGAATATAAGCTGGTGGTGGTGGGCGCCG G CG GTTGGGCAAGAGTGCGCTGACCATCCAG G -3', as well as complementary noncoding oligomers, were obtained from Operon (Alameda, CA). The above 69 mer oligo is a portion of the coding sequence of human H- ras DNA. The bolded underlined G represents H- ras codon 12 middle G position, plus an additional 30 bases of H- ras sequence both 5' and 3' of codon 12, with an Eco R1 site 5' and a Bam H1 site 3' (restriction enzyme recognition nucleotides in italics). The wild type sequence (coding strand containing codon 12 middle G) was 5'-phosphorylated. Mismatch containing noncoding strand sequences (either T or A opposite codon 12 middle base G) were not phosphorylated, thus providing a 5' nick in the strand containing the incorrect base after ligation into the pUC19 plasmid. All other reagents were purchased from Sigma unless otherwise noted. Preparation of site-specific mismatched oligonucleotides and plasmids Complementary 69 mer oligos containing the wild type sequence at codon 12; middle base (coding strand G) and one mismatched base (noncoding strand T or A), as described above, were annealed in an equimolar ratio at a final concentration of 0.2 μg DNA per μl of annealing buffer (1 mM Tris-HCL, 1 mM MgCl 2 , pH 7.5). Ligation of annealed mismatched oligos to Eco R1/Bam H1 digested pUC19 DNA was accomplished at a 20:1 molar ratio of 69 mer to pUC19 DNA (~3–6 ng DNA per μl of ligation solution), using T4 ligase, per manufacturer's recommendations (Invitrogen Corp.; Carlsbad, CA). The pUC19 vector used for ligation of mismatch-containing H- ras oligo for subsequent measurement of MMR within nuclear extracts was grown in GM2929 ( E. coli dam - dcm - ). The pUC19 and pUC18 plasmids used as positive controls for correct bacterial mismatch repair were grown in DH5α ( E. coli dam + ). Electrophoretic gel mobility shift assays Gel shift assays were performed using nuclear extracts from either HCT116 or HCT116 + Ch. 3 cells and [ 32 P]-dATP-labeled-69 mer duplexes by a fill-in reaction, using [α- 32 P]-dATP and Klenow polymerase, per manufacturer's protocol (Invitrogen). Each nuclear protein-DNA binding assay was performed using 0.8 – 1.0 × 10 5 cpm homoduplex or heteroduplex 69 mer and 5–10 μg protein from nuclear extract solution in an equal volume of 2X gel shift reaction buffer, resulting in a final concentration of 1 mM MgCl 2 , 0.5 mM EDTA, 0.5 mM DTT, 50 mM NaCl, 10 mM Tris-HCL, pH 7.5, 0.1 mg/ml poly [dI:dC], 5 mg/ml BSA, 6% glycerol, in a final volume of 15 – 20 μl. In addition, each reaction contained 100-fold molar excess (100X) unlabeled homoduplex 69 mer, unless otherwise noted. Incubations were for 30 min. at 37°C, unless otherwise noted. Gel shifts of each reaction were electrophoresed at 20 mA within a 4.8% nondenaturing acrylamide gel in 0.5 × TBE buffer, at room temperature. Radioactively labeled oligomers were visualized by autoradiography of the dried gel. Site-specific mismatch repair assay within human nuclear extracts An in vitro mismatch repair assay was performed essentially as described by Thomas et. al. [ 39 ], except for modifications as described below. For each reaction, 14 fmols of pUC19 plasmid containing site-specific mismatched H- ras 69 mer, with a nick on the same strand and 36 bases 5' of the incorrect base, was incubated with 50 μg of HCT116 + Ch. 3 or HCT116 nuclear extract in a final volume of 25 μl containing 30 mM HEPES buffer, pH 7.9, 100 μM each of dATP, dCTP, dGTP and dTTP, 200 μM each of CTP, GTP and UTP, 4 mM ATP, 40 mM creatine phosphate, 100 ng/μl creatine phosphokinase, 0.5 mM DTT, and 7 mM MgCl 2 . Negative control experiments were performed in the above solution without nuclear extract. Each reaction was incubated at 37°C for one hour, as described [ 39 ]. Plasmid DNA containing the H- ras insert was recovered using Wizard SV Miniprep System per manufacturer's directions (Promega; Madison, WI), and subsequently used to transform E. coli strain NR9161 ( MutL - ). In addition, DH5α (MMR proficient E. coli ) were transformed directly with the ligated and nicked plasmid as a positive control for correct mismatch repair [ 37 ]. Subsequently, plasmid DNA was purified from ampicillin-resistant bacterial colonies and digested with Nae I , which recognizes a single restriction digestion site unique to the wild type H- ras sequence at codon 12 middle base pair (G:C) [ 23 ]. After electrophoresis in 1% agarose, banding patterns of resulting DNA fragments were analyzed to score plasmids as having correct, incorrect, or no repair [ 23 , 37 , 38 ]. MMR efficiency of each human nuclear extract assay, above MMR deficient bacterial background repair, was determined by the following equations [ 39 ]: Total repair efficiency = 100 × (1 - [fraction of unrepaired plasmids incubated with human nuclear extract results / fraction of unrepaired untreated plasmids from direct transformation of NR9161]). Correct repair efficiency = 100 × {(fraction of correctly repaired incubated with human nuclear extract) - [(fraction of correct repair by NR9161 direct transformation) × (1 - fraction of total repair efficiency)]}. Incorrect repair efficiency = Total repair efficiency - Correct repair efficiency. Statistical comparisons of G:T and G:A results were conducted by a non-parametric equivalent to the Student's t-test for differences between proportions. Statistical significance was indicated for those comparisons with a P < 0.05 assuming a 2-tailed distribution [ 47 ]. Authors' contributions ME designed and carried out the MMR assay studies and helped draft the manuscript. HH designed and carried out all other experiments. AS performed all densitometry measurements and helped draft the manuscript. MF and SB helped carry out the MMR assays. KW conceived of the study, participated in its design and coordination and wrote the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555755.xml
555741	Pharmacokinetics of isoflavones, daidzein and genistein, after ingestion of soy beverage compared with soy extract capsules in postmenopausal Thai women	Background Isoflavones from soybeans may provide some beneficial impacts on postmenopausal health. The purpose of this study was to compare the pharmacokinetics and bioavailability of plasma isoflavones (daidzein and genistein) after a single dose of orally administered soy beverage and soy extract capsules in postmenopausal Thai women. Methods We conducted a randomized two-phase crossover pharmacokinetic study in 12 postmenopausal Thai women. In the first phase, each subject randomly received either 2 soy extract capsules (containing daidzin : genistin = 7.79 : 22.57 mg), or soy beverage prepared from 15 g of soy flour (containing daidzin : genistin = 9.27 : 10.51 mg). In the second phase, the subjects received an alternative preparation in the same manner after a washout period of at least 1 week. Blood samples were collected immediately before and at 0.5, 1, 2, 4, 6, 8, 10, 12, 24 and 32 h after administration of the soy preparation in each phase. Plasma daidzein and genistein concentrations were determined by using high performance liquid chromatography (HPLC). The pharmacokinetic parameters of daidzein and genistein, i.e. maximal plasma concentration (C max ), time to maximal plasma concentration (T max ), area under the plasma concentration-time curve (AUC) and half-life (t 1/2 ), were estimated using the TopFit version 2.0 software with noncompartmental model analysis. Results There were no significant differences in the mean values of C max /dose, AUC 0–32 /dose, AUC 0-∝ /dose, T max , and t 1/2 of genistein between both preparations. For pharmacokinetic parameters of daidzein, the mean values of C max /dose, T max , and t 1/2 did not significantly differ between both preparations. Nonetheless, the mean AUC 0–32 /dose and AUC 0-∝ /dose after administration of soy extract capsules were slightly (but significantly, p < 0.05) higher than those of soy beverage. Conclusion The bioavailability of daidzein, which was adjusted for the administered dose (AUC/dose), following a single oral administration of soy beverage was slightly (but significantly) less than that of soy extract capsules, whereas, the bioavailability adjusted for administered dose of genistein from both soy preparations were comparable. The other pharmacokinetic parameters of daidzein and genistein, including C max adjusted for the dose, T max and t 1/2 , were not different between both soy preparations.	Background Menopause is associated with estrogen deficiency and its accompanying symptoms such as accelerated bone loss and atherosclerosis. Hormone replacement therapy (HRT) has traditionally been used for treatment of menopausal disorders. Estrogen helps to maintain bone density, relieve menopausal symptoms, as well as influence emotional state [ 1 ]. Estrogen replacement therapy after menopause therefore improves the health and quality of life for women. However, the Women's Health Initiative (WHI) randomized controlled trial has recently found that although estrogen-alone hormone therapy reduces the risk of hip and other fractures in healthy postmenopausal women with prior hysterectomy, it significantly increases the risk of stroke (but has no significant effect on the risk of coronary heart disease, breast or colorectal cancer) [ 2 ]. In addition, long-term estrogen replacement therapy in postmenopausal women who have a uterus might has the disadvantage of being tissue agonists for endometrial tissue, which increases the incidence of endometrial cancer [ 3 ]. Although adding progestin to estrogen can be used to prevent the development of endometrial cancer, this combination may cause some unwanted side effects, i.e. breast cancer, venous thromboembolism, stroke and coronary heart disease [ 3 ]. Thus, alternative therapies, which include natural products such as phytoestrogens and herbs, offer attractive options because they may protect against breast and endometrial cancer, have fewer side effects and still provide health benefits [ 4 ]. The most common forms of phytoestrogens are the isoflavones. Among the food consumed by humans, soybeans contain the highest concentration of isoflavones. Isoflavones have a chemical structure resembling that of estrodiol-17β, the most potent mammalian estrogen. The major isoflavones, namely, genistein and daidzein, have several features in common with estradiol-17β, including an aromatic A ring with hydroxyl group in the same plane at a distance similar to that in estradiol (Figure 1 ) [ 5 ]. Indeed, isoflavones appear to have both estrogenic and antiestrogenic effects, like selective estrogen receptor modulators (SERMs), depending on the target tissue [ 6 ]. Therefore, rather than classifying soy isoflavones as "estrogens", they should be judged more correctly as natural SERMs. Reproductive cells, especially those of the breast and uterus, are rich in estrogen receptor α (ERα), whereas other cells (such as those in the bone) have greater amounts of estrogen receptor β (ERβ) than ERα. This differential distribution of the two types of estrogen receptors, and the greater affinity of the isoflavones for ERβ in relation to ERα, suggests that the isoflavones have different effects in different tissue [ 7 ]. Figure 1 Structures of isoflavones. Despite the beneficial effects of isoflavones on postmenopausal health are still controversial, there are some researches, including epidemiological studies, suggest that isoflavones may help to alleviate postmenopausal symptoms and protect against chronic diseases such as hormone-dependent cancer (e.g. breast and endometrial cancer), cardiovascular diseases and osteoporosis [ 8 - 11 ]. Thus, in terms of both health promotion and chronic disease prevention, the potential public health impact of daily soy consumption could be important, especially in postmenopausal women. Although many commercial soy capsules containing isoflavone extract are now available in many Western countries, soybeans and soy food (such as tofu, soy flour, soy milk, etc.), which provide the main sources of isoflavones, are consumed in significant amounts in Asian countries because they are inexpensive and high in quality protein. The purpose of this trial was to compare the pharmacokinetics of plasma isoflavones, daidzein and genistein, in postmenopausal Thai women after a single dose of orally administered commercial soy extract capsules and soy beverage. Methods Study design This study was a single dose, randomized two-phase crossover study with a washout period of at least one week. It was approved by the Medical Ethics Committee of the Faculty of Medicine, Chiang Mai University and was in compliance with the Helsinki Declaration. Subjects A total of 12 postmenopausal Thai women, who ranged in age from 46–61 years (average 52.83 ± 3.88 years), participated in this study. Their mean weight and height was 52.23 ± 6.38 kg and 1.54 ± 0.06 cm, respectively. The body mass index (BMI) of each subject was within 18–24 kg/m 2 (22.06 ± 1.83 kg/m 2 ). Their serum follicle-stimulating hormone concentrations were more than 20 IU/l and the average level was 69.60 ± 31.21 IU/l. All had to be in good health on the basis of medical history and physical examination. Routine blood tests including complete blood count (CBC) with differential, blood urea nitrogen (BUN), creatinine (Cr) and liver function test (LFT) had to be within the normal limit. Subjects had to give both verbal and written information regarding the study. Signed informed consent was obtained prior to entry. Exclusion criteria included subjects with known premenopausal status (<12 months since the last spontaneous menstrual bleeding and a serum follicle-stimulating hormone concentration ≤ 20 IU/l) as well as those with a known history of chronic renal, liver, pulmonary or cardiovascular diseases, recent cigarette smoking, substance abuse or addiction, use of antibiotics within the previous 6 weeks, consumption of more than 2 alcoholic drinks/day, regular use (more than 1 dose/week) of over-the-counter or prescribed medications, and malignancy. Isoflavone preparations The isoflavone preparations used in this study were commercial soy extract capsules and soy beverage. Soy extract capsules were purchased from Canada, whereas, the soy beverage was prepared from mixing 15 g of commercial soy flour (Doi Khum ® , Chiang Mai, Thailand, lot no. 19GODE) with 300 ml of hot water. The concentration of isoflavones in each preparation was measured as aglycones (daidzein and genistein) and β-glycoside conjugates of isoflavones (daidzin and genistin) that were analyzed by high performance liquid chromatography (HPLC) as described below. Quantification of isoflavones in soy preparations Two soy preparations, soy flour and soy extract capsules, were chosen as isoflavone sources. The sample extractions and concentration determinations were modified from the method described by Nakamura et al. [ 12 ]. Two hundred mg of soy flour or 300 mg of powder from the soy capsule (1 capsule) was placed in a centrifuge tube. Ten ml of 80% methanol in water was added to the centrifuged tubes, and sonicated for 30 min. Isoflavonoids were extracted for 24 h at an ambient temperature. One ml of the mixture was centrifuged, and 10 μl of clear supernatant was diluted with mobile phase (100 times for soy flour and 400 times for soy extract capsule) and spiked with 20 μl of internal standard (IS, 100,000 ng/ml fluorescein and 50,000 ng/ml chloramphenicol for quantification of aglycones and β-glycosides, respectively). Five μl of the mixture was injected into the HPLC system. Separation was performed isocratically at 50°C. The flow rate of the mobile phase was maintained at 1 ml/min and the analytes were detected by UV absorption at 259 nm. The mobile phase for the quantification of aglycones consisted of 5 mM phosphoric acid in methanol/acetonitrile (100:85, v/v), whereas, that for quantification of β-glycoside conjugates comprised 5 mM phosphoric acid in methanol/acetonitrile (80:20, v/v). The Isoflavone contents of unknown samples were determined by using a calibration curve of the peak height ratios of isoflavones and IS versus respective isoflavone concentrations with the use of linear regression. All samples were analyzed within the same day, in which intraday assay validation was performed. Dosage and drug administration Subjects were admitted to the Clinical Pharmacology Unit of the Faculty of Medicine, Chiang Mai University at 6:30 a.m. after an overnight fast of at least 8 h. They were randomized to receive either 2 soy extract capsules with 300 ml of water, or 300 ml of soy beverage at 7:00 a.m. They remained upright and fasted for 2 h after soy product administration. Water and lunch were served at 2 h and 4 h, respectively after dosing. Blood samples were collected at different time points (see below). After the blood sample collection at 12 h postdose, the subjects were discharged from the Clinical Pharmacology Unit and asked to come back again on the next day to give blood samples at 24 and 32 h postdose. While waiting for blood sample collections, the subjects were allowed to perform all of their daily activities, except moderate to high degrees of exercises. After a washout period of at least 1 week, the subjects received the alternative preparation and the blood samples were collected in the same manner. Identical food and fluid were served during the 2 study periods. The subjects were required to refrain from drinking caffeine containing beverages and alcohol, and instructed to consume no soy products (except those given in this study) from the time of screening until the end of the research. Blood sample collection Venous blood samples (7 ml/each) for determination of soy isoflavones were collected predose and then exactly 0.5, 1, 2, 4, 6, 8, 10, 12, 24 and 32 h after administration. Samples were obtained from the forearm by venipuncture through an indwelling intravenous catheter (BD Insyte ® ) and collected in a heparinized vacutainer (BD Insyte ® ). The blood collecting tubes were centrifuged at 2,500 rpm for 20 min and the plasma samples were separated and frozen at -20°C until analyzed. Determination of plasma isoflavone concentrations The assay was modified from the solid phase extraction procedure and HPLC technique as previously described by Thomas et al [ 13 ]. Briefly, an aliquot (125 μl) of plasma was transferred to a 1.5 ml plastic vial and treated with 0.25 ml of β-glucuronidase/sulfatase mixture from Helix pomatia (Sigma G-0876) to hydrolyze glucuronide and sulfate conjugates of genistein and daidzein. The enzyme mixture was made up freshly and contained 0.15 g of ascorbic acid in 10 ml of 0.2 M acetate buffer, pH 4.0, and 500 μl of β-glucuronidase/sulfatase from Helix pomatia . To allow for complete hydrolysis in the plasma samples, 0.75 ml of 0.2 M ammonium acetate buffer was added, and the tubes were capped and heated overnight in a water bath (15–18 h, 37°C). The tubes were removed from the water bath and allowed to cool to room temperature. After enzymatic hydrolysis, plasma samples were spiked with 5 μl of internal standard (IS, 100,000 ng/ml fluorescein in 80% methanol). After vortex mixing for 30 sec and centrifugation at 13,000 rpm for 5 min, sample purification was performed by using a solid phase extraction cartridge placed in the vacuum box. The cartridges were preconditioned with 2.50 ml each of methanol, water and 175 mM phosphate buffer, respectively. The samples were loaded into the cartridges and allowed to flow freely, then the cartridges were washed with 0.1 ml of 20% methanol in water, 0.1 ml of 175 mM phosphate buffer and 0.1 ml of 20% methanol in water, consecutively. After drying the cartridge by slow suction, isoflavones and IS were eluted with 2.0 ml of 20% methanol in ethyl acetate. The eluents were dried by a SpeedVac concentrator. The residues were dissolved in 50 μl of the mobile phase, and 10 μl of samples were injected into a HPLC system. The mobile phase consisted of 4 mM perchloric acid in methanol/acetonitrile (115:85, v/v). The flow rate was maintained at 1 ml/min and the analytes were detected by UV absorption at 259 nm, and the column was maintained at 40°C. The retention time for daidzein and genistein was 7.699 and 13.031 minutes, respectively. The lower limit of quantitation was 5 ng/ml. Plasma concentrations of daidzein and genistein were determined using a calibration curve and linear regression of the seven known isoflavone concentrations versus the peak height ratios of isoflavones and IS. The %CV of intraday precision for plasma daidzein concentrations ranged from 0.85–9.45%, whereas, the %CV of interday ranged from 3.90–9.96%. The %CV of intraday and interday precision for plasma genistein concentrations ranged from 1.29–3.65%, and 4.37–6.84%, respectively. The %deviation in intraday and interday assay for plasma daidzein concentrations ranged from 4.35–5.84% and -6.11–4.22%, respectively. On the other hand, the %deviation in intraday and interday assay for plasma genistein concentrations ranged from -5.00–1.98% and -7.79%–1.52%, respectively. Data analysis and statistical methods Pharmacokinetic parameters The maximal plasma concentration (C max , ng/ml) and time to maximal plasma concentration (T max , h) were obtained directly by the visual inspection of each subject's plasma concentration-time profile. The areas under the plasma concentration-time curve from time 0–32 (AUC 0–32 , ng.h/ml) and 0-∝ (AUC 0-∞ , ng.h/ml) as well as half-life (t 1/2 , h) were determined by non-compartmental analysis. The slope of the terminal log-linear portion of the concentration-time profile was determined by least-squares regression analysis and used as the elimination rate constant ( K e ). The elimination t 1/2 was calculated as 0.693/ K e . The AUC from time zero to the last quantifiable point (AUC 0–32 ) was calculated using the trapezoidal rule. Extrapolated AUC from time t to infinity (AUC t-∞ ) was determined as Ct/ K e . Total AUC was the sum of AUC 0–32+ AUC 32-∞ . In this study, the sampling time was continued for more than 3 half-lives, therefore, the AUC 0-32 was sufficient to cover at least 80% of the total AUC. The calculation was performed by using the TopFit software version 2.0 for personal computer. Statistical analysis The pharmacokinetic parameters were presented as mean ± SD. The differences between the mean values of C max /dose, T max , t 1/2 , AUC 0–32 /dose and AUC 0-∝ /dose of the two isoflavone preparations were statistically analyzed by using the paired t -test. However, the results from the analysis were not different regardless of whether the statistical comparison was performed by using the paired t -test (parametric method) or Wilcoxon's sign rank test (non-parametric method). Results Isoflavone contents in soy preparations The quantification of isoflavone contents demonstrated that both soy flour and soy extract capsule preparations contained predominantly daidzin and genistin (the form of β-glycoside conjugates), whereas, aglycones were rarely found and isoflavones in other forms were not measured. Isoflavone contents in both soy preparations are shown in Table 1 . Table 1 Isoflavone contents in soy preparations used in this study. Daidzin Genistin Total 1 Soy flour Mg/g 0.62 ± 0.005 0.70 ± 0.01 1.32 ± 0.01 % 47 53 100 Soy extract capsule Mg/capsule 3.90 ± 0.04 11.29 ± 0.17 15.18 ± 0.21 % 26 74 100 1 Summation of daidzin and genistin contents. Pharmacokinetics of daidzein and genistein in healthy postmenopausal Thai women The Mean plasma daidzein and genistein concentration-time curves after a single dose of both orally administered soy preparations are shown in Figure 2 , 3 . The individual and mean pharmacokinetic parameters of daidzein and genistein following oral administration of soy beverage and soy extract capsules are shown in Tables 2 , 3 . Figure 2 Mean plasma daidzein concentration-time curves after a single dose of orally administered soy beverage and soy extract capsules. Figure 3 Mean plasma genistein concentration-time curves after a single dose of orally administered soy beverage and soy extract capsules. (Note: the orally administered dose of genistein from soy extract capsules was approximately two times higher than that from soy beverage). Table 2 Individual and mean pharmacokinetic parameters of daidzein following a single dose of orally administered soy beverage (B) and soy extract capsules (C). Subj No. C max (ng/ml) C max /Dose 1 AUC 0–32 (ng.h/ml) AUC 0–32 /Dose 1 AUC 0-∝ (ng.h/ml) AUC 0-∝ /Dose 1 T max (h) t 1/2 (h) B C B C B C B C B C B C B C B C 1 67.42 60.73 7.27 7.80 354.62 650.40 38.25 83.49 496.21 722.51 53.53 92.75 6 6 6.06 8.04 2 142.17 56.33 15.34 7.23 1142.38 931.84 123.23 119.62 1206.09 1024.24 130.11 131.48 6 1 6.74 6.02 3 64.95 76.70 7.01 9.85 1118.40 1117.20 120.65 143.41 1704.60 1171.78 183.88 150.42 8 6 19.80 6.85 4 132.30 124.43 14.27 15.97 1336.29 1356.49 144.15 174.13 1402.26 1427.11 151.27 183.20 6 6 6.70 6.91 5 77.66 93.20 8.38 11.96 1293.52 1251.35 139.54 160.64 1483.69 1561.72 160.05 200.48 8 10 9.94 10.50 6 105.65 142.82 11.40 18.33 1409.24 1378.43 152.02 176.95 1533.26 1450.55 165.40 186.21 10 6 6.21 6.86 7 160.86 125.37 17.35 16.09 1576.65 1669.28 170.08 214.28 1761.36 1721.18 190.01 220.95 4 6 8.48 5.97 8 102.77 74.08 11.09 9.51 1158.07 1090.36 124.93 139.97 1227.07 1149.04 132.37 147.50 6 6 6.85 6.76 9 45.34 105.77 4.89 13.58 300.89 849.39 32.46 109.04 340.35 876.40 36.72 112.50 4 6 3.46 4.40 10 52.98 84.74 5.72 10.88 401.82 579.95 43.35 74.45 533.03 761.42 57.50 97.74 1 6 4.82 4.38 11 99.98 87.38 10.79 11.22 1291.80 1546.66 139.35 198.54 1455.80 1705.46 157.04 218.93 8 10 6.59 7.63 12 103.60 120.64 11.18 15.49 578.60 909.87 62.42 116.80 665.06 971.69 71.74 124.74 4 6 6.49 5.71 Mean 96.31 96.02 ND 10.39 12.33 996.86 1110.94 ND 107.54 142.61* 1150.73 1211.93 ND 124.14 155.57* 5.92 6.25 7.68 6.67 SD 36.18 27.71 3.90 3.56 455.69 342.26 49.16 43.94 504.94 354.46 54.47 45.50 2.43 2.26 4.14 1.65 1 Pharmacokinetic parameters adjusted for oral administered dose. ND = statistical analysis was not performed due to differences in oral administered doses between both preparations, *Statistical significance (p < 0.05) compared to corresponding values of soy extract capsules. Table 3 Individual and mean pharmacokinetic parameters of genistein following a single dose of orally administered soy beverage (B) and soy extract capsules (C). Subj No. C max (ng/ml) C max /Dose 1 AUC 0–32 (ng.h/ml) AUC 0–32 /Dose 1 AUC 0-∝ (ng.h/ml) AUC 0-∝ /Dose 1 T max (h) t 1/2 (h) B C B C B C B C B C B C B C B C 1 90.41 151.49 8.60 6.71 954.82 1950.40 90.85 86.42 1040.69 2248.09 99.02 99.61 6 6 8.46 9.92 2 68.08 124.90 6.48 5.53 693.31 1821.17 65.97 80.69 739.89 1941.50 70.40 86.02 1 8 7.62 7.42 3 149.53 301.34 14.23 13.35 1801.34 4392.34 171.39 194.61 1989.66 5048.92 189.31 223.70 8 8 8.25 9.37 4 212.94 516.77 20.26 22.90 1979.53 5541.88 188.35 245.54 2140.25 6175.55 203.64 273.62 6 6 8.10 8.93 5 96.95 220.38 9.22 9.76 1281.58 2478.42 121.94 109.81 1423.11 3072.33 135.41 136.12 8 10 9.73 12.40 6 96.57 301.62 9.19 13.36 1277.75 2776.28 121.57 123.01 1483.87 2972.50 141.19 131.70 10 6 10.10 7.74 7 223.77 371.31 21.29 16.45 3184.25 5260.83 302.97 233.09 4022.02 6003.94 382.69 266.01 4 8 12.60 9.58 8 110.60 193.01 10.52 8.55 948.75 2217.44 90.27 98.25 998.31 2370.24 94.99 105.02 6 6 5.07 8.41 9 54.86 183.88 5.22 8.15 425.80 1481.47 40.51 65.64 527.69 1500.60 50.21 66.49 4 6 4.41 4.80 10 69.95 214.54 6.66 9.51 773.90 1856.73 73.63 82.27 789.07 1940.05 75.08 85.96 6 6 5.34 6.63 11 105.13 217.57 10.00 9.64 1128.81 3121.71 107.40 138.31 1177.74 3208.73 112.06 142.17 6 8 6.32 5.27 12 117.61 345.31 11.19 15.30 871.60 2591.36 82.93 114.81 914.46 2631.99 87.01 116.61 4 6 5.35 5.05 Mean 116.37 261.84 ND 11.07 11.60 1276.79 2957.50 ND 121.48 131.04 1437.23 3259.54 ND 136.75 144.42 5.75 7.00 7.61 7.96 SD 53.80 110.68 5.12 4.90 746.04 1372.13 70.98 60.79 949.02 1599.34 90.30 70.86 2.34 1.35 2.44 2.28 1 Pharmacokinetic parameters adjusted for oral administered dose. ND = statistical analysis was not performed due to differences in oral administered doses between both preparations. The mean plasma daidzein concentration-time curve after a single dose of both orally administered soy preparations revealed a biphasic pattern. The first peak of plasma daidzein concentration was reached approximately 1 h after ingestion of both preparations, whereas, the second peak attained higher plasma concentrations at 5.92 ± 2.43 h for soy beverage, and 6.25 ± 2.26 h for soy extract capsules. The mean maximal plasma daidzein concentrations (C max ) were 96.31 ± 36.18 ng/ml and 96.02 ± 27.71 ng/ml for soy beverage and soy extract capsules, respectively. Pharmacokinetic analysis of the plasma concentration-time curves showed that the elimination t 1/2 was 7.68 ± 4.14 h for soy beverage and 6.67 ± 1.65 h for soy extract capsules. The AUC 0–32 was 996.86 ± 455.69 and 1110.94 ± 342.26 ng.h/ml for soy beverage and soy extract capsules, respectively, whereas, the AUC 0-∝ was 1,150.73 ± 504.94 ng.h/ml for soy beverage and 1,211 ± 354.46 ng.h/ml for soy extract capsules. The mean plasma genistein concentration-time curve after a single dose of both orally administered soy preparations also demonstrated a biphasic pattern, but the first peak of plasma genistein concentration after ingestion of soy extract capsules was less evident. The mean T max for the second peak of plasma concentration was 5.75 ± 2.34 h for soy beverage and 7.00 ± 1.35 h for soy extract capsules. The mean C max was 116.37 ± 53.80 ng/ml for soy beverage and 261.84 ± 110.68 ng/ml for soy extract capsules. Pharmacokinetic analysis of the plasma concentration-time curves showed the elimination t 1/2 of 7.61 ± 2.44 h for soy beverage and 7.96 ± 2.28 h for soy extract capsules. The AUC 0–32 was 1276.79 ± 746.04 and 2957.50 ± 1372.13 ng.h/ml for soy beverage and soy extract capsules, respectively, whereas, the AUC 0-∝ was 1,437.23 ± 949.02 ng.h/ml for soy beverage and 3,259.54 ± 1,599.34 ng.h/ml for soy extract capsules. After oral administration of both soy preparations, the pharmacokinetic parameters of daidzein were statistically compared, and the mean values of C max /dose, T max , and t 1/2 did not significantly differ between both preparations. Nonetheless, the mean values of AUC 0–32 /dose and AUC 0-∝ /dose after administration of soy extract capsules were slightly (but significantly, p < 0.05) higher than those after soy beverage intake (Table 2 ). For pharmacokinetic parameters of genistein, there were no significant differences in the mean values of C max /dose, AUC 0–32 /dose, AUC 0-∝ /dose, T max , and t 1/2 between both preparations (Table 3 ). Discussion In this study, the pharmacokinetics of plasma daidzein and genistein were evaluated in 12 postmenopausal Thai women after a single dose of orally administered soy beverage and soy extract capsules. These women were enrolled from a pool of volunteers after they had been screened for medical history, BMI, serum follicle-stimulating hormone concentration and blood tests. Since the design of this study was similar to that of the bioequivalence testing, 12 subjects were enrolled according to the minimum number of subjects stipulated by the Canadian and European guidelines for bioequivalence testing. The soy extract product available in Thailand was not selected as the study preparation because only minimum isoflavone content was labeled without any details about the proportion of daidzin and genistin content. The Canadian soy extract preparation was used as an alternative because the exact total of isoflavone content (18.2 mg/capsule) as well as daidzin and genistin content (9.1 mg/capsule, each) were declared. Nonetheless, our quantification of isoflavones in this preparation revealed that the amount of daidzin and genistin in each capsule was 3.90 ± 0.04 and 11.29 ± 0.17, respectively. Therefore, the orally administered dose in this study was calculated according to our quantification, but not by the amount declared. Initially, we tried to measure the total isoflavone contents in each preparation by using the acid hydrolysis method [ 14 , 15 ]. Briefly, either 1 g of soy flour or powder from the soy extract capsules was dispersed in a mixture of 10 ml of 10 M HCl and 40 ml of 96% ethanol (containing 0.05% butylated hydroxy toluene, BHT) followed by refluxing at 100°C for 2 h. The mixture was cooled to room temperature and the ethanol lost during the refluxing was replaced. One ml of this mixture was centrifuged. Ten μl of clear supernatant was diluted 100 times and determined by the HPLC method. The principle of this assay is to hydrolyze β-glycoside conjugates of isoflavones to aglycones, and the detection of total aglycones reflects the total isoflavones. Unfortunately, we failed to measure total isoflavones as aglycones after acid hydrolysis. The recovery of daidzein and genistein was very low (approximately 20–30%) as compared to the other published data, even though we had tried to vary many factors such as the concentration of HCl, duration of refluxing time, temperature during reflux, amount of BHT added, etc. However, since isoflavones are present predominantly as β-glycoside conjugates (e.g. daidzin and genistin) in most commercially available soy products such as soybean or soy flour (with the exception of fermented soy products) [ 14 ], we used a specific HPLC condition for measuring daidzin and genistin, and another condition for measuring aglycones (daidzein and genistein) without acid hydrolysis. Isoflavones in other forms (malonyl glycoside and acetyl glycoside conjugates) were not determined because their commercial standards were not available. Glycitein and its derivatives were also not determined, due to a much smaller amount found in soybeans [ 16 ]. Since both soy preparations consist of different proportions of daidzin : genistin (approximately 1:1 for soy flour, 1:3 for soy extract capsules), the appropriate amount of daidzin content in each preparation was therefore considered first for pharmacokinetic comparison. In this study, each volunteer was assigned to receive 2 capsules of soy extract (daidzin : genistin = 7.79 : 22.57 mg) to compare with soy beverage prepared from 15 g of soy flour (9.27 : 10.51 mg). These dosages caused an approximately equal amount of daidzin between the two preparations, whereas, the genistin content in soy extract capsules was approximately two-fold higher than that of soy beverage. The oral administration of these dosages resulted in the plasma concentrations of daidzein and genistein being high enough and convenient for measurement by the HPLC method. In addition, preparing soy beverage from 15 g of soy flour in 300 ml of water was practical and created an acceptable concentration for consumption. From previous studies, the bioavailability of isoflavones was investigated and compared among various soy food and beverages. So far, there has been no study that compares bioavailability of isoflavones from soy extract versus natural soy food or beverages. Our purpose was to investigate pharmacokinetics of daidzein and genistein after ingestion of soy beverage compared to soy extract capsules in postmenopausal Thai women. Daidzein and genistein contents in soy food can vary and depend on the raw material and processing conditions used to produce a particular food product. Furthermore, in each type of soy food, there are different forms of isoflavones in differing amounts. However, based on the equivalent dose of isoflavones, the administration of different soy food has shown no difference in isoflavone bioavailability [ 17 ]. In our study, the bioavailability, adjusted for dosage (determined by AUC 0–32 /dose and AUC 0-∝ /dose) of genistein after ingestion of both soy preparations, was not significantly different. In contrast, the bioavailability of daidzein following ingestion of soy extract capsules was significantly greater than that following ingestion of soy beverage. This might be the result of at least 2 possibilities. Firstly, the food matrix of soy flour may alter the bioavailability of daidzein. Dietary factors such as fiber and carbohydrate have been associated with differences in the metabolism of daidzein to equol [ 18 - 20 ]. Urinary recovery of equol is higher following the ingestion of tempeh when compared with homogeneous soymilk and textured vegetable protein. This suggests that the combination of a food matrix might protect daidzein from degradation until it reaches the large intestine where it can be metabolized to equol by the microflora [ 20 ]. Secondly, in this study, we only determined the isoflavone contents in both soy preparations as the forms of β-glycoside conjugates (daidzin and genistin) as well as aglycones (daidzein and genistein). However, soy flour and soy extract capsules might have contained some malonyl glycoside and acetyl glycoside conjugates of isoflavones, which were not measured in this study. We postulate that the proportions of malonyl glycoside and acetyl glycoside conjugates of daidzein in soy extract capsules may be greater than those in soy flour, resulting in better bioavailability after these conjugates are converted and absorbed as daidzein from the gastrointestinal tract. The mean T max of daidzein and genistein from both soy preparations in this study was shorter than the values of 8–11 h (after ingestion of β-glycoside conjugates) as reported in previous studies [ 21 , 22 ]. This difference might result from variation in age, race, uptake rates, hydrolysis of glycosides by gut bacteria or gut wall enzymes, further metabolism (for example glucuronides within the liver), etc. The second peak demonstrated in the plasma concentration-time curves of daidzein and genistein possibly resulted from enterohepatic recirculation of the glucuronide and sulfate conjugates of isoflavones excreted in bile [ 22 ]. The elimination t 1/2 of daidzein and genistein in this study was comparable to the values of 6–8 h from other studies [ 20 - 22 ]. It has been suggested that a daily intake of soy isoflavone extract containing 50/50 mg of genistin and daidzin [ 23 ] or 76 mg of isoflavones [ 24 ] can significantly decrease hot flushes in the group treated with soy products over the placebo. A meta-analysis of 38 clinical trials, which examined the relationship between soy protein intake and serum lipids, have shown that the consumption of soy in men and women is associated with a significant decrease in serum cholesterol, LDL and triglyceride levels [ 25 ]. In a randomized, double-blind, placebo-controlled trial, which examined the effects of dietary soy supplements containing 118 mg of isoflavones on the lipid profiles of men and postmenopausal women with relatively normal cholesterol levels, the LDL/HDL ratio decreases in the isoflavone treatment groups without any change in total cholesterol [ 26 ]. In addition, it has been found that those postmenopausal women with greatest phytoestrogen consumption have the highest bone mineral density (BMD) at the hip and spine. Subjects with the highest intake of isoflavones also have significantly lower levels of serum PTH, osteocalcin and urinary N-telopeptide [ 27 ]. Besides, isoflavone supplementation (61.8 mg of isoflavones) for four weeks shows potentially beneficial effects on bone metabolism and serum lipids in perimenopausal women in a randomized controlled trial [ 28 ]. Another trial has demonstrated that continuous dietary intake of isoflavones (37.3 mg/day) for ten weeks may inhibit postmenopausal osteoporosis [ 29 ]. Therefore, in our opinion, the total dose of isoflavones that benefits menopausal health is up to approximately 100 mg/day. Since our study revealed that the bioavailability of genistein from soy beverage and soy extract capsules was similar, and the bioavailability of daidzein was slightly (although statistically significant) lower than that of soy extract capsules, the amount of soy beverage, which provides the isoflavone bioavailability equivalent to soy isoflavone capsules containing 50/50 mg of daidzein and genistein, should be equal to approximately 5 cups/day (15 g of soy flour/cup). If one consumes other soy food or prepares a soy beverage in a higher concentration, the daily volume of consumption would be reduced. This inexpensive soy beverage is an appropriate alternative food supplementation compared to the more expensive soy extract capsules and HRT for postmenopausal Thai women, and is in line with Thailand's present socioeconomic status. However, the different proportion of daidzein and genistein in various soy preparations might affect beneficial outcomes for postmenopausal women. In this aspect, clinical studies should be investigated further. Conclusion The bioavailability of daidzein, which was adjusted for the administered dose (AUC/dose) following a single oral administration of soy beverage, was slightly (but significantly) less than that of soy extract capsules, whereas, that of genistein from both soy preparations was comparable. There was also no difference in other pharmacokinetic parameters of daidzein and genistein, including C max adjusted for dose, T max and t 1/2 between both soy preparations. Competing interests The author(s) declare that they have no competing interests. Authors' contributions EA performed the quantification of isoflavones and statistical analysis. ST supervised data collection and analysis, and drafted the manuscript. NR supervised the quantification of isoflavones. SP initiated the research question and participated in the selection of patients eligible for the study. CS participated in the design of the study and drafted the manuscript. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555741.xml
524027	Ketogenic diets and physical performance	Impaired physical performance is a common but not obligate result of a low carbohydrate diet. Lessons from traditional Inuit culture indicate that time for adaptation, optimized sodium and potassium nutriture, and constraint of protein to 15–25 % of daily energy expenditure allow unimpaired endurance performance despite nutritional ketosis.	Introduction In the opinion of most physicians and nutrition scientists, carbohydrate must constitute a major component of one's daily energy intake if optimum physical performance is to be maintained [ 1 ]. This consensus view is based upon a long list of published studies performed over the last century that links muscle glycogen stores to high intensity exercise. It has also been reinforced by the clinical experience of many physicians, whose patients following low carbohydrate formula or food diets frequently complain of lightheadedness, weakness, and ease of fatigue. During the time that this consensus view of the necessity of carbohydrate for vigorous exercise was forming, the last pure hunting cultures among the peoples of North America finally lost out in competition with expanding European cultural influences. Between 1850 and 1930, the routine consumption of carbohydrates spread north from the U.S. Plains States through central Canada, where the indigenous peoples had heretofore made at most seasonal use of this nutrient class. However the last of these groups to practice their traditional diet, the Inuit people of the Canadian and Alaskan Arctic regions, were luckily observed by modern scientists before their traditional dietary practices were substantially altered. The reports of these early scientists imply that the Inuit people were physically unhampered despite consuming a diet that was essentially free of identifiable carbohydrate. Given this juxtaposition of clinical research results favoring carbohydrate against observed functional well-being in traditional cultures consuming none, it is an interesting challenge to understand how these opposing perspectives can be explained. This paper will review the observations of early explorer scientists among the Inuit, track the controversy that they stimulated among nutritionists in the last century, and utilize some of the forgotten lessons from the Inuit culture to explain how well-being and physical performance can be maintained in the absence of significant dietary carbohydrate. The origins of carbohydrate supremacy Until the development of agriculture over last few millennia, our human ancestors' consumption of dietary carbohydrate was opportunistic. As some groups adapted to hunting and fishing for their sustenance, they were able to move into temperate and then arctic regions, where limited access to wild grain, nuts, and fruit dictated sustained dependence upon fat and protein as primary sources of dietary energy. With the development of agriculture came the ability to grow and store grain, allowing societies to remain in a stable physical location, build permanent dwellings, and potentially stimulating the development of written language (those early stone tablets would have been difficult to transport from camp to camp on a dog sled). Starting from locations in the Middle East and Asia, cultures based upon agricultural wheat and rice spread over 5 millennia to dominate Europe, Africa, and the Americas. With its ability to support a non-nomadic life style, greater population density, and permanent communities; there were clear advantages of agriculture-based societies over those based upon hunting and fishing, particularly as agricultural communities built the infrastructure to support trade and transport. Given its success in this competition of cultures (and by implication, the competition of their diets), it is an easy assumption that a grain-based diet is functionally superior to one based upon the meat and fish (fat and protein) of the hunting societies that they superseded. As the science of nutrition developed in the early 20 th Century, numerous comparative studies were undertaken to assess differences between diets. Although there were some advocates of low carbohydrate diets (eg, the Banting diet of the 19 th Century, promoted for weight loss and diabetes control), the prevailing premise for these studies was that carbohydrate was a necessary nutrient for optimum human health and function. Among studies confirming this view, a classic was the 1939 study by two Danish scientists, Christensen and Hansen [ 2 ]. They did a crossover study of low carbohydrate, moderate carbohydrate, and high carbohydrate diets, each lasting one week. At the end of each diet, the subjects' endurance time to exhaustion on a stationary bicycle was assessed. Compared to the mean endurance time on the low carb diet of 81 minutes, the subjects were able to ride for 206 minutes after the high carb diet. During the Second World War, another oft-cited study was performed, this time examining the practicality of pemmican (a mixture of dried meat and fat) as a light-weight emergency ration for soldiers. This experiment by Kark et al [ 3 ] involved abruptly switching soldiers in winter training in the Canadian Arctic from standard carbohydrate-containing rations to pemmican. This study only lasted 3 days, as the soldiers rapidly became unable to complete their assigned tasks, which included pulling loaded sleds 25-miles per day through deep snow. With the resurgence of biomedical science in the 1960's came development of the percutaneous needle biopsy, facilitating assessment of intra-muscular fuel stores and metabolism. This led to the concept of muscle glycogen as the limiting fuel for high intensity exercise [ 4 ] and to the nutritional strategy of carbohydrate loading [ 5 ]. The clear consensus that developed from this research was that fat had limited utility as a fuel for vigorous exercise, and that humans are physically impaired if given a low carbohydrate diet. The hunter's counterpoint – practical observations on ketogenic diets Although high-carbohydrate diets might be more effective in short-term tests of high-intensity exercise, there are multiple clues in the published literature that the debilitating effects of ketogenic diets are overstated. Not only is there the demographic evidence that whole populations of people lived for millennia as hunters, but there are many reports of Europeans crossing over to live within the cultures of these hunting societies without apparent impediment. One of the earliest documented demonstrations of physical stamina during a ketogenic diet was the Schwatka 1878–80 expedition in search of the lost Royal Navy Franklin expedition. The Schwatka expedition, sponsored by the New York Herald and the American Geographical Society, departed from the west coast of Hudson's Bay in April of 1879 with 4 Caucasians, 3 families of Inuits, and 3 heavily laden dog sleds. Totaling 18 people, they started out with a month's supply of food (mostly walrus blubber) and a prodigious supply of ammunition for their hunting rifles. After covering over 3000 miles on foot over ice, snow and tundra, all 18 members of the original party plus their 44 dogs returned to Hudson's Bay in March of 1880. Once their initial provisions were depleted, the expedition's only source of additional food was hunting and fishing, as there were no other sources of supply along their route. The leader of this expedition, Lt. Frederick Schwatka, was a graduate of both West Point and Bellevue Hospital Medical College. His summary of the expedition was published as a news article in the New York Herald in the Fall of 1880, but his written diary was lost for 85 years until its discovery and publication by the Marine Historical Association of Mystic CT in 1965 [ 6 ]. This fascinating 117-page saga describes how Schwatka, a frontiersman and U.S. Army surgeon, collaborated with his Inuit guides to accomplish a remarkable feat of physical endurance. In one notation, Schwatka provides an interesting insight into his weaning from their initial supply of carbohydrate-containing food. " When first thrown wholly upon a diet of reindeer meat, it seems inadequate to properly nourish the system, and there is an apparent weakness and inability to perform severe exertive fatiguing journeys. But this soon passes away in the course of two or three weeks." This observation, written a century before the current author first came to grips with the issue of "keto-adaptation", offers an early clue to resolve the dichotomy between impaired performance with low carbohydrate diets in the laboratory and their lack of debilitating effects when taken among people practiced in their use. That Schwatka was not impaired by his prolonged experience eating meat and fat is evidenced by his diary entry for the period 12–14 March 1880, during which he and an Inuit companion walked the last 65 miles in less than 48 hours to make a scheduled rendezvous with a whaling ship and complete his journey home. Twenty-six years later, a Harvard-trained anthropologist named Vilhjalmur Stefansson entered the Arctic with the purpose of studying the Inuit language and culture. Having been born in 1879 in Manitoba and grown up in North Dakota, it is unlikely that Stefansson was aware of the Schwatka expedition or its reported technique of extended dogsled travel while living by hunting. However when separated from his expedition and thus his source of supply over the winter of 1906–7, Stefansson was taken in by a group of Inuit on the Canadian Arctic coast. With the arrival of spring in June of 1907, he both spoke their language and had acquired their skill of living and traveling by dogsled on a hunter's diet. For the next decade, Stefansson traveled extensively over the arctic mainland and among the islands to the north. During this period, he was away from the outposts of European settlement for periods of up to 18 months at a time, and in the remote regions of the Canadian Arctic he lived with groups of Inuit for whom he was the first European they had met. Stefansson wrote extensively about these experiences in both the scientific literature and in books for the lay public [ 7 ]. One of the main themes of his writing was the adaptation of the Inuit culture to survive as nomadic groups in the arctic on a diet consisting solely of the products of hunting and fishing. Coming as it did in the same time period that the science of nutrition was blossoming with the discovery and characterization of vitamins (eg, the first vitamin to be chemically defined was thiamin by Funk in 1911), Stefansson's claim that one could live and function well on the products of just one food group caused tremendous controversy [ 8 ]. Subjected to great criticism and even scorn, Stefansson agreed to recreate the Inuit diet under scientific observation. Therefore, for the calendar year of 1929 he and a colleague from his arctic explorations ate a diet consisting of meat and fat for 12 months. This experiment, supervised by Dr. Eugene DuBois, was conducted at Bellevue Hospital in New York. For the first 3 months of this study, the two explorers were under constant observation to guarantee dietary compliance, after which they were allowed more freedom of movement but with frequent tests to document that they remained in ketosis. This study was reported in multiple peer-reviewed publications, the primary reports being published in the Journal of Biological Chemistry in 1930 [ 9 , 10 ], As noted by DuBois [ 8 ], the study results were essentially "negative", in that both subjects survived the 12 months in apparent good health, having no signs of scurvy (which was predicted to occur within the first 3 months) or other deficiency diseases. It is interesting to note from the careful observations published from the Bellevue study that Stafansson ate relatively modestly of protein, deriving between 80–85% of his dietary energy from fat and only 15–20% from protein [ 9 ]. This was, and still remains, at odds with the popular conception that the Inuit ate a high protein diet, whereas in reality it appears to have been a high fat diet with a moderate intake of protein. In his writings, Stefansson notes that the Inuit were careful to limit their intake of lean meat, giving excess lean meat to their dogs and reserving the higher fat portions for human consumption [ 11 ]. It is also interesting to conjecture that the vigorous defense of his arctic observations by Stefansson may have led indirectly to the development of the carbohydrate loading hypothesis. Stefansson was a polarizing influence in the field of nutrition, and his advocacy of pemmican as an emergency ration for troops during the Second World War led directly to the Kark study quoted above, which in turn was a predecessor to many comparative dietary trials performed in Europe and the U.S. in later decades. Modern ketogenic diet performance studies There was a resurgence of interest in very low calorie ketogenic diets for weight loss in the 1970's, followed closely by the complications (including sudden death) associated with the Liquid Protein diet popularized in 1976. However, the fatigue and apparent cardiac dysfunction caused by this collagen-based fad diet stood in stark contrast to the published experience of arctic explorers such as Schwatka and Stefansson. In addition, physicians who monitored patients following very low calorie diets observed wide variations between the exercise-tolerance of these individuals. Given that the elegant research on the metabolism of total fasting by Dr. George Cahill and colleagues had demonstrated that full adaptation of nitrogen, fat, and carbohydrate metabolism required a number of weeks [ 12 ], it seemed reasonable to hypothesize that exercise tolerance would take more than a week to recover after removal of carbohydrate from the diet. This view was supported by the subsequent discovery of the prescient adaptation quote from Schwatka's diary [ 6 ] noted above. To test this hypothesis, the current author (under the mentorship of Drs. Ethan Sims and Edward Horton at the University of Vermont) undertook a study of subjects given a very low calorie ketogenic diet for 6 weeks in a metabolic research ward [ 13 ]. The protein for this diet, along with a modicum of inherent fat, was provided by lean meat, fish, and poultry providing 1.2 grams of protein per kg of reference ("ideal") body weight daily. In addition, mindful that the natriuresis of fasting could reduce circulating blood volume and cause secondary renal potassium wasting, the subjects were prescribed 3 grams of supplemental sodium as bouillion and 25 mEq (1 g) of potassium as bicarbonate daily. Treadmill performance testing of these subjects included determinations of peak aerobic power (VO 2 max) after a 2-week weight maintenance baseline diet, and again after 6 weeks of the ketogenic weight loss diet. Endurance time to exhaustion was quantitated at 75% of the baseline VO 2 max. This endurance test was repeated again after one week of weight loss and finally after 6 weeks of weight loss. Other than these tests, the subjects did no training exercise during their participation in this study. To compensate for the fact that the average subject had lost over10 kg, the final endurance treadmill test was performed with the subject carrying a backpack equivalent in weight to the amount lost. The energy expenditure data (expressed as oxygen consumption) and exercise times across this 8-week inpatient study are shown in Table 1 . That these subjects'peak aerobic power did not decline despite 6 weeks of a carbohydrate-free, severely hypocaloric diet implies that the protein and mineral contents of the diet were adequate to preserve functional tissue. As can be noted, endurance time to exhaustion was reduced after one week of the ketogenic diet, but it was significantly increased over the baseline value by the 6-week time point. However the interpretation of this endurance test is confounded by the fact that the oxygen cost (ie, energy cost) of the treadmill exercise had significantly decreased following the weight loss, and this occurred despite the subjects being made to carry a backpack loaded to bring them back to their initial exercise test weight. Table 1 Exercise parameters of Vermont study [13] Baseline Week 1 Week 6 VO 2 max (LPM) 2.49 -- 2.49 Exercise VO 2 (LPM) 1.88* 1.71 1.50* Endurance time (min) 168 + 130 249 + LPM, liter per minute week 6 < baseline, P < 0.05 + week 6 > baseline, P < 0.01 This question of improved efficiency notwithstanding, it is clear that our subjects experienced a delayed adaptation to the ketogenic diet, having reduced endurance performance after one week followed by a recovery to or above baseline in the period between one and six weeks. Given the reduced energy cost of the exercise despite the backpack, the extent of this adaptation cannot be determined from this study. To explain this improved exercise efficiency, we can speculate that humans are more efficient carrying weight in a modern backpack than under their skin as excess body fat. It is also possible that these untrained subjects became more comfortable with prolonged treadmill walking by their third test, and therefore improving their overall efficiency. Given the uncertainties of this study caused by the subject's weight loss and potential for improved technique with multiple tests, the current author undertook a second study under the mentorship of Dr. Bruce Bistrian at MIT in Cambridge MA [ 14 , 15 ]. The diet employed in this followup study was patterned after that consumed by Stefansson during his year in the Bellevue study (and thus presumably close to that traditionally consumed by the Inuit) with the intention that the subjects would be in ketosis without weight loss. This second study utilized competitive bicycle racers as subjects, confined to a metabolic ward for 5 weeks. In the first week, subjects ate a weight maintenance (eucaloric) diet providing 67% of non-protein energy as carbohydrate, during which time baseline performance studies were performed. This was followed by 4 weeks of a eucaloric ketogenic diet (EKD) providing 83% of energy as fat, 15% as protein, and less than 3% as carbohydrate. The meat, fish, and poultry that provided this diets protein, also provided 1.5 g/d of potassium and was prepared to contain 2 g/d of sodium. These inherent minerals were supplemented daily with an additional 1 g of potassium as bicarbonate, 3 grams of sodium as bouillon, 600 mg of calcium, 300 mg of magnesium, and a standard multivitamin. The bicyclist subjects of this study noted a modest decline in their energy level while on training rides during the first week of the Inuit diet, after which subjective performance was reasonably restored except for their sprint capability, which remained constrained during the period of carbohydrate restriction. On average, subjects lost 0.7 kg in the first week of the EKD, after which their weight remained stable. Total body potassium (by 40 K counting) revealed a 2% reduction in the first 2 weeks (commensurate with the muscle glycogen depletion documented by biopsy), after which it remained stable in the 4 th week of the EKD. These results are consistent with the observed reduction in body glycogen stores but otherwise excellent preservation of lean body mass during the EKD. The results of physical performance testing are presented in Table 2 . What is remarkable about these data is the lack of change in aerobic performance parameters across the 4-week adaptation period of the EKD. The endurance exercise test on the cycle ergometer was performed at 65% of VO 2 max, which translates in these highly trained athletes into a rate of energy expenditure of 960 kcal/hr. At this high level of energy expenditure, it is notable that the second test was performed at a mean respiratory quotient of 0.72, indicating that virtually all of the substrate for this high energy output was coming from fat. This is consistent with measures before and after exercise of muscle glycogen and blood glucose oxidation (data not shown), which revealed marked reductions in the use of these carbohydrate-derived substrates after adaptation to the EKD. Table 2 Exercise parameters of MIT EKD study [15] VO 2 max (LPM) Exercise VO 2 (LPM) Exercise RQ Endurance time (min) Baseline 5.1 3.18 0.83 147 EKD-4 5.0 3.21 0.72* 151 LPM, liter per minute * P < 0.01 Examining the results of these two ketogenic diet performance studies together indicates that both groups experienced a lag in performance across the first week or two of carbohydrate restriction, after which both peak aerobic power and sub-maximal (60–70% of VO 2 max) endurance performance were fully restored. In both studies, one with untrained subjects and the other with highly trained athletes who maintained their training throughout the study, there was no loss of VO 2 max despite the virtual absence of dietary carbohydrate for 4–6 weeks. This whole-body measure of oxidative metabolism could not be maintained unless there was excellent preservation of the full complement of functional tissues including skeletal muscle (and mitochondrial) mass, circulating red cell mass, and cardiopulmonary functions. The possibility raised by the first study of improved endurance time after keto-adaptation was not substantiated by the second study employing highly trained athletes without the complicating variable of major weight loss. It is thus likely that the increased endurance time in the Vermont study was due to improved efficiency (ie, less hobbling from a backpack than from an equal weight of internal body fat) and/or improved acclimation to the endurance test procedure. Such acclimation would not be expected in the second study, as the highly trained bicycle racers were well conditioned to the stationary ergometer at the start of the study. It is also worth noting that the bicycle racers remained weight stable (excepting the half kilogram of reduced muscle glycogen) across the 4 weeks of the EKD, which was equi-caloric with the baseline diet. Although 4 weeks is a relatively short period to assess small differences in energy efficiency between diets, this observation implies that there was no great reduction in the efficiency of energy metabolism after keto-adaptation. As a final note in this section, neither the Vermont study nor the MIT study has been refuted in the 2 decades since their publication. Understandably given the expense of human metabolic ward studies and the orthogonal conclusions of these two studies, neither study has been corroborated by a similar human study. However two subsequent animal studies examining physical performance after keto-adaptation have yielded results consistent with those presented above [ 16 , 17 ]. Resolving the performance paradox There are three factors that can help us explain the paradox presented by studies showing superior performance with high carbohydrate diets versus the present author's two studies noted above. Adaptation The most obvious of these is the time allotted (or not) for keto-adaptation. In this context, the prescient observation of Schwatka (that adaptation to "a diet of reindeer meat" takes 2–3 weeks) says it all. None of the comparative low-carbohydrate versus high-carbohydrate studies done in support of the carbohydrate loading hypothesis sustained the low carbohydrate diet for more than 2 weeks [ 5 ], and most (including the classic report of Christensen and Hansen [ 2 ]) maintained their low-carbohydrate diets for 7 days or less. There are to date no studies that carefully examine the optimum length of this keto-adapataion period, but it is clearly longer than one week and likely well advanced within 3–4 weeks. The process does not appear to happen any faster in highly trained athletes than in overweight or untrained individuals. This adaptation process also appears to require consistent adherence to carbohydrate restriction, as people who intermittently consume carbohydrates while attempting a ketogenic diet report subjectively reduced exercise tolerance. Sodium and potassium The second factor differentiating the author's studies from many others is optimized mineral nutriture, which has benefits for both cardiovascular reserve in the short term and preservation of lean body mass and function over longer time periods. The Inuit people lived much of the year on coastal ice (which is partially desalinated sea water), and much of their food consisted of soup made with meat in a broth from this brackish source of water. When they went inland to hunt, they traditionally added caribou blood (also a rich source of sodium) to their soup. With these empirically derived techniques, the Inuit culture had adapted the available resources to optimize their intakes of both sodium and potassium. When meat is baked, roasted, or broiled; or when it is boiled but the broth discarded, potassium initially present in the meat is lost, making it more difficult to maintain potassium balance in the absence of fruits and vegetables. Because our research subjects were accustomed to eating meat, fish, and poultry prepared as something other than soup, we chose to give them most of their sodium separately as bouillon and a modest additional supplement of potassium as potassium bicarbonate. With these supplements maintaining daily intakes for sodium at 3–5 g/d and total potassium at 2–3 g/d, our adult subjects were able to effectively maintain their circulatory reserve (ie, allowing vasodilatation during submaximal exercise) and effective nitrogen balance with functional tissue preservation. An example of what happens when these mineral considerations are not heeded can be found in a study prominently published in 1980 [ 18 ]. This was a study designed to evaluate the relative value of "protein only" versus "protein plus carbohydrate" in the preservation of lean tissue during a weight loss diet. The protein only diet consisted solely of boiled turkey (taken without the broth), whereas the protein plus carbohydrate consisted of an equal number of calories provided as turkey plus grape juice. Monitored for 4 weeks in a metabolic ward, the subjects taking the protein plus carbohydrate did fairly well at maintaining lean body mass (measured by nitrogen balance), whereas those taking the protein only experienced a progressive loss of body nitrogen. A clue to what was happening in this "Turkey Study" could be found in the potassium balance data provided in this report. Normally, nitrogen and potassium gains or losses are closely correlated, as they both are contained in lean tissue. Interestingly, the authors noted that the protein only diet subjects were losing nitrogen but gaining potassium. As noted in a rebuttal letter published soon after this report [ 19 ], this anomaly occurred because the authors assumed the potassium intake of their subjects based upon handbook values for raw turkey, not recognizing that half of this potassium was being discarded in the unconsumed broth. Deprived of this potassium (and also limited in their salt intake), these subjects were unable to benefit from the dietary protein provided and lost lean tissue. Also worthy of note, although this study was effectively refuted by a well-designed metabolic ward study published 3 years later [ 20 ], this "Turkey Study" continues to be quoted as an example of the limitations of low carbohydrate weight loss diets. Protein dose The third dietary factor potentially affecting physical performance is adjusting protein intake to bring it within the optimum therapeutic window for human metabolism. The studies noted herein [ 13 - 15 , 20 ] demonstrate effective preservation of lean body mass and physical performance when protein is in the range of 1.2 – 1.7 g/kg reference body weight daily, provided in the context of adequate minerals. Picking the mid-range value of 1.5 g/kg-d, for adults with reference weights ranging from 60–80 kg, this translates into total daily protein intakes 90 to 120 g/d. This number is also consistent with the protein intake reported in the Bellevue study [ 9 ]. When expressed in the context of total daily energy expenditures of 2000–3000 kcal/d, about 15% of ones daily energy expenditure (or intake if the diet is eucaloric) needs to be provided as protein. The effects of reducing daily protein intake to below 1.2 g/kg reference weight during a ketogenic diet include progressive loss of functional lean tissue and thus loss of physical performance, as demonstrated by Davis et al [ 21 ]. In this study, subjects given protein at 1.1 g/kg-d experienced a significant reduction in VO 2 max over a 3 month period on a ketogenic diet, whereas subjects given 1.5 g/kg-d maintained VO 2 max. At the other end of the spectrum, higher protein intakes have the potential for negative side-effects if intake of this nutrient exceeds 25% of daily energy expenditure. One concern with higher levels of protein intake is the suppression of ketogenesis relative to an equi-caloric amount of fat (assuming that ketones are a beneficial adaptation to whole body fuel homeostasis). In addition, Stefansson describes a malady known by the Inuit as rabbit malaise [ 8 ]. This problem would occur in the early spring when very lean rabbits were the only available game, when people might be tempted to eat too much protein in the absence of an alternative source of dietary fat. The symptoms were reported to occur within a week, and included headache and lassitude. Such symptoms are not uncommon among people who casually undertake a "low carbohydrate, high protein" diet. Conclusions Both observational and prospectively designed studies support the conclusion that submaximal endurance performance can be sustained despite the virtual exclusion of carbohydrate from the human diet. Clearly this result does not automatically follow the casual implementation of dietary carbohydrate restriction, however, as careful attention to time for keto-adaptation, mineral nutriture, and constraint of the daily protein dose is required. Contradictory results in the scientific literature can be explained by the lack of attention to these lessons learned (and for the most part now forgotten) by the cultures that traditionally lived by hunting. Therapeutic use of ketogenic diets should not require constraint of most forms of physical labor or recreational activity, with the one caveat that anaerobic (ie, weight lifting or sprint) performance is limited by the low muscle glycogen levels induced by a ketogenic diet, and this would strongly discourage its use under most conditions of competitive athletics. List of abbreviations VO 2 max – maximum aerobic capacity RQ – respiratory quotient EKD – eucaloric ketogenic diet Competing interests None declared.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524027.xml
549591	SARS Transmission Pattern in Singapore Reassessed by Viral Sequence Variation Analysis	Background Epidemiological investigations of infectious disease are mainly dependent on indirect contact information and only occasionally assisted by characterization of pathogen sequence variation from clinical isolates. Direct sequence analysis of the pathogen, particularly at a population level, is generally thought to be too cumbersome, technically difficult, and expensive. We present here a novel application of mass spectrometry (MS)–based technology in characterizing viral sequence variations that overcomes these problems, and we apply it retrospectively to the severe acute respiratory syndrome (SARS) outbreak in Singapore. Methods and Findings The success rate of the MS-based analysis for detecting SARS coronavirus (SARS-CoV) sequence variations was determined to be 95% with 75 copies of viral RNA per reaction, which is sufficient to directly analyze both clinical and cultured samples. Analysis of 13 SARS-CoV isolates from the different stages of the Singapore outbreak identified nine sequence variations that could define the molecular relationship between them and pointed to a new, previously unidentified, primary route of introduction of SARS-CoV into the Singapore population. Our direct determination of viral sequence variation from a clinical sample also clarified an unresolved epidemiological link regarding the acquisition of SARS in a German patient. We were also able to detect heterogeneous viral sequences in primary lung tissues, suggesting a possible coevolution of quasispecies of virus within a single host. Conclusion This study has further demonstrated the importance of improving clinical and epidemiological studies of pathogen transmission through the use of genetic analysis and has revealed the MS-based analysis to be a sensitive and accurate method for characterizing SARS-CoV genetic variations in clinical samples. We suggest that this approach should be used routinely during outbreaks of a wide variety of agents, in order to allow the most effective control.	Introduction During infectious disease outbreaks, due to either new or established agents, extensive information gathering is required to enable identification of the source, transmission routes, and the effect of containment policies. It is becoming increasingly clear that traditional approaches based on travel and contact tracing are not sufficient for tracking an outbreak. New sequence-based techniques for pathogen detection and identification have the potential to become perhaps the most important component of these investigations, as demonstrated in the recent worldwide effort in fighting the epidemic of severe acute respiratory syndrome (SARS). The discovery of the SARS coronavirus (SARS-CoV) as the etiological agent for SARS was a major breakthrough [ 1 ], which was quickly followed by the successful sequencing of the whole genome of the virus [ 2 , 3 ]. Genome sequence comparison between this new coronavirus and the three known classes of coronavirus revealed a similar genome structure, but minimum homology at the amino acid level, strongly suggesting that the SARS-CoV was a new class of coronavirus [ 2 , 3 , 4 ]. A comparative sequence analysis of 14 SARS isolates from different countries suggested a moderate genetic diversity among the SARS isolates and thus implied a slow evolution of the SARS-CoV genome [ 5 ]. Furthermore, sequence variation analyses of SARS-CoV isolates demonstrated that common genetic variations in the SARS-CoV genome could be used as “molecular fingerprints” to partition the viral isolates into different genetic lineages, track the transmission of a specific viral lineage, and infer the origin of infection [ 5 , 6 ]. Therefore, the characterization of SARS-CoV's genetic variations is not only instrumental for understanding its genetic diversity and genome evolution but also important for tracking its transmission and understanding its epidemiological pattern in human populations. Direct sequence analysis of a pathogen in a large number of clinical samples, particularly at a population level, is generally cumbersome, technically challenging, and expensive. Therefore, a rapid, sensitive, high-throughput, and cost-effective screening method would greatly facilitate large-scale characterization of genetic variation of pathogens at a population level. A mass spectrometry (MS)–based method for detecting single nucleotide polymorphisms has been routinely used as a high-throughput method for genotyping human samples, and, thus, we sought to extend this methodology to detect pathogen sequence variations. Here, we demonstrate the high sensitivity of the MS-based analysis in detecting SARS-CoV sequence variations and apply it to analyzing both cultured viral isolates and uncultured tissue samples of SARS-CoV. Methods Patients and Samples The SARS-CoV samples used in the sensitivity study were previously described [ 7 ]. Briefly, five in vitro samples were generated by spiking 200-μl human whole-blood samples with SARS-CoV virus obtained from a Vero E6 cell culture of an anonymous Singapore patient. Vero cell cultured viral isolates were obtained from 13 Singapore patients: the presumed index case (patient Sin2500) of the Singapore SARS outbreak, whose date of illness onset was 25 February 2004; three primary contacts (patients Sin2677, Sin2774, and Sin2748), whose dates of onset were 9 March 2004 (Sin2677 and Sin2774) and 14 March 2004 (Sin2748); one secondary contact (patient Sin2679), who was believed to have contracted SARS from index patient Sin2500 through another primary contact not included in this study and whose date of illness onset was 15 March 2004; and another eight patients (Sin842, Sin845, Sin846, Sin847, Sin848, Sin849, Sin850, and Sin852), who were believed to be the fifth- or six-generation contacts of index patient Sin2500 (based on contact tracing records) and whose dates of illness onset ranged from 2 April to 14 April 2004. All the patients fitted the World Health Organization case definition for probable SARS [ 8 ]. The virus was cultured in vero cells following isolation from respiratory samples (three endotracheal tube swabs, three throat swabs, one nasal swab, two nasopharyngeal aspirates, and four lung tissues) obtained from the patients between 0 and 11 d after onset of symptoms. Uncultured lung tissue samples were also obtained from patients Sin842, Sin848, Sin849, and Sin852. Bronchoalveolar lavage material was obtained from a serologically confirmed German patient with SARS who had been traveling on the same flight as an early Singaporean patient with SARS who was later hospitalized in Germany [ 9 ]. Virus could not be isolated from the sample owing to its inappropriate storage. RNA Extraction and cDNA Synthesis For the spiked human blood samples, RNA was extracted from 200 μl of blood into 50 μl of water using a HighPure RNA kit (Roche, Basel, Switzerland). For the clinical samples, RNA was extracted into 30 μl of water using a QiAmp viral RNA mini kit (Qiagen, Valencia, California, United States). RNA samples were reverse transcribed into cDNA using 2 μl of RNA as template, a SuperScript kit (Invitrogen, Carlsbad, California, United States), and 13 sequence-specific primers [ 10 ]. All cDNA products were purified by ethanol precipitation and then resuspended in 20 μl of water. Real-Time Quantitative PCR Analysis Real-time quantitative PCR analyses of the RNA samples from spiked human blood were performed using a Lightcycler Sars-CoV quantification kit (Roche). Each analysis was done using 1 μl of RNA and in accordance with the manufacturers' instructions. Single Nucleotide Variations of SARS-CoV Twenty-one single nucleotide variations (SNVs) of SARS-CoV were analyzed in this study. Eight of these were identified from our previous sequence analysis of five Singapore SARS-CoV viral isolates and represent variations [ 5 ], and 13 SNVs were identified from a more recent genome sequence analysis of additional Singapore SARS-CoV viral isolates. The former eight SNVs were used in evaluating the sensitivity of MS-based genotyping analysis in detecting SARS-CoV viral genotypes, and the later 13 SNVs, as well as five of the former eight SNVs, were used to genotype the patient-derived samples. SNV Analysis A primer extension genotyping assay was designed for each SARS-CoV SNV using SpectroDesigner software (Sequenom, San Diego, California, United States) and analyzed using the MassARRAY system (Sequenom) and the recommended protocol for the MADLI-TOF (matrix-assisted laser desorption ionization/time-of-flight) MS-based genotyping analysis [ 11 ]. One microliter of cDNA (equivalent to 0.1 μl of RNA from 1 μl of spiked human blood) was used as template in each analysis. Statistical Analysis of Sensitivity The analytical detection limit of the combined RNA preparation/RT-PCR/MALDI-TOF MS system was determined by probit analysis [ 12 ] using the Statgraphics Plus 5.0 software package (Statistical Graphics, Jena, Germany). Results Sensitivity for Detecting SARS-CoV SNVs Prior to determining viral sequence variants in clinical samples, we measured the sensitivity of the MS-based assay for detecting SARS-CoV sequence variations by analyzing eight SNVs in in vitro human blood samples spiked with SARS-CoV. The viral RNA copy numbers in five spiked human blood samples were quantified by real-time PCR analysis and determined to be 1.64 × 10 6 (SB1), 3.84 × 10 3 (SB2), 2.17 × 10 3 (SB3), 6.21 × 10 2 (SB4), and 1.20 × 10 2 (SB5) copies per microliter ( Figure S1 ). The three samples (SB3, SB4, and SB5) with the lowest virus copy numbers were used to determine the sensitivity of the MS-based assay. We were able to successfully type the virus in 16 of the 16 analyses of the SB3 sample (equivalent to 217 RNA copies per reaction), 12 of the 15 analyses of the SB4 sample (equivalent to 62 RNA copies per reaction), and six of the 15 analyses of the SB5 sample (equivalent to 12 RNA copies per reaction). According to probit analysis, this corresponds to a 95% probability of detection when at least 75 copies of viral RNA (95% confidence interval, 59–107 copies) are present per reaction, and still a 50% detection chance at 38 copies of virus RNA per reaction (95% confidence interval, 29–53 copies). All 64 sham spiked samples gave negative genotype calls. Sequence Variant Determination of SARS-CoV in Viral Isolates Having demonstrated the high sensitivity of the MS-based analysis, we typed 18 SARS-CoV SNVs in the cultured viral isolates from nine Singapore patients, including five early Singapore cases (Sin2500, Sin2774, Sin2748, Sin2677, and Sin2679) and four later Singapore cases (Sin842, Sin848, Sin849, and Sin852). Of the 18 SNVs analyzed, nine were detected in at least two SARS isolates ( Table 1 ). Assuming that a common sequence variation originated through a single mutation in a host and was then propagated by subsequent infection to others, the sharing of a specific sequence variant by different viral isolates suggests that either these viral isolates share a common ancestor or they have direct ancestor–descendant relationship. Sequence variants at the nine common SNV sites were used to reconstruct the molecular relationship among the nine viral isolates. The pattern of the shared variants among the nine viral isolates ( Table 1 ) clearly indicated two major molecular lineages of isolate ( Figure 1 ). One lineage includes the four early isolates from patients Sin2500, Sin2774, Sin2748, and Sin2677, and the other includes the early isolate from patient Sin2679 and the four later isolates. The first lineage is defined by the sequence variant T:C:T at SNV positions19,084, 23,174, and 28,268, whereas the second lineage is defined by the sequence variant C:T:C. Both the variants are distinct from the presumed ancestral sequence variant C:C:C observed in the Urbani viral isolate. The later Singapore isolates were also differentiated from the early isolates by the sequence variant pattern at SNV positions 22,549 and 23,735. Figure 1 The Molecular Relationship among 13 Singapore SARS-CoV Isolates Based on the Genotype Sharing Pattern of the Viral Isolates Table 1 Genotypes of Five Early and Four Later Singapore SARS-CoV Isolates in 18 SNVs a SNV positions are numbered according to their nucleotide positions in the genome sequence of the Urbani isolate NA, not analyzed In addition, the sequence variant sharing pattern at SNVs 28,008, 548, 1,727, and 13,347 can further differentiate the four fifth- and six-generation isolates into three different sub-lineages ( Figure 1 ) defined by three distinct variants: C:T:T:C in the isolate from patient Sin842, T:T:T:C in the isolate from patient Sin849, and T:C:C:T in the isolates from patients Sin852 and Sin848 ( Table 1 ). To further confirm this three-sub-lineage pattern observed in the later Singapore isolates, we typed another four later Singapore isolates, from patients Sin845, Sin846, Sin847, and Sin850, at five critical SNVs (19,084, 28,008, 548, 1,727, and 13,347). The detected sequence variations in the four new isolates supported the three-sub-lineage pattern in the later Singapore isolates ( Table 2 ; Figure 1 ). The second and third sub-lineages are more closely related to each other than to the first one, as all the members of the latter two sub-lineages show the variant T at SNV position 28,008, whereas the two members of the first sub-lineage show the variant C. Table 2 Genotypes of Two Early and Eight Later Singapore SARS-CoV Isolates at Five SNV Positions The molecular relationship among the viral isolates derived from MS-based viral sequence analysis is consistent with the one derived from the whole-genome sequence analysis of the same isolates [ 13 ], clearly demonstrating that a small subset of commonly shared variances can be used as “molecular signature” to differentiate and thus track viral isolates. Direct Sequence Variation Determination of SARS-CoV in Primary Lung Tissue Samples We also typed the 18 SNVs in four uncultured lung tissue samples from patients Sin842, Sin848, Sin849, and Sin852, and compared their sequence variations with their matched cultured isolates ( Table 3 ). Table 3 Comparison of Paired Direct Tissue Samples and Vero Cell Cultured Isolates of SARS-CoV Of the 72 direct sequence variation comparisons (18 SNVs in four sample pairs), nine differences were identified. Six of these were due to heterogeneous sequences in primary tissue samples. For example, the cultured isolate from patient Sin849 showed the sequence variant T:T:C at SNV positions 548, 1,727, and 13,347, a subset of the heterogeneous T/C:T/C:T/C variant observed in the matched lung tissue sample from the same patient. Direct comparison of the MS spectrums of the three different sequence variants (T, T/C, and C) at SNV position 1,727 in Figure 2 clearly ruled out the possibility of variant miscall. More interestingly, the heterogeneous variant T/C:T/C:T/C in the primary lung tissue sample revealed both of the two existing variants of T:T:C and C:C:T seen in other cultured and lung tissue samples ( Table 3 ). Figure 2 MS Spectrums of the Three Distinct Genotypes at SNV Position 1,727 The T example is from the cultured viral isolate from patient Sin849, the T/C example is from the uncultured lung tissue sample from patient Sin849, and the C example is from the cultured viral isolate from patient Sin852. In addition, another three sequence variant differences were observed between the paired cultured and tissue samples from patient Sin848 at SNV positions14,807, 26,205, and 26,509, where the tissue sample showed the sequence variant C:C:T, but the matched culture isolate showed the variant T:T:C ( Table 3 ). In all three SNV positions, the cultured isolate showed novel sequence variants, whereas the primary lung tissue showed the Urbani isolate's variants (see Table 3 ). Confirmation of the Singapore Origin of a German SARS-CoV Isolate The application of tagging and thus tracking of SARS-CoV strains using viral lineage- and/or strain-specific sequence variants was further demonstrated in our investigation of a clinical sample from a German patient. This patient stayed in Hanoi, Vietnam, before sharing an airplane flight with an early Singapore SARS patient on his way to New York via Frankfurt [ 9 ]. German health authorities assumed that this patient was infected somewhere in Hanoi, Vietnam, although the possibility of him being infected by a Singapore SARS-CoV strain during his flight to New York could not be ruled out. We genotyped the virus directly from a brochoalveolar lavage specimen from the German patient at four SNV positions (19,084, 23,792, 26,428, and 27,111) that were distinctive for the early Singapore SARS-CoV isolates (see Table 1 ). The sequence of the isolate from the German patient was determined to be T:C:G:A at these SNV positions. The detection of the variant T at SNV position 19,084 in the isolate from the German patient strongly suggested that this patient was indeed infected by an early Singapore SARS-CoV strain, as the T sequence variant at position 19,084 was detected only in the early Singapore isolates [ 5 , 6 ] and is clearly different from the C variant at position 19,084 observed in the Vietnam-originated isolates [ 6 , 14 ]. Furthermore, this German isolate's sequence variant, T:C:G:A, was detected only in the isolate from the Singapore primary case Sin2748, suggesting that this German patient was probably infected by a SARS-CoV strain originating from or closely related to the SARS-CoV strain of the Singapore primary case Sin2748 and not from the Hanoi outbreak. Indeed, the T:C:G:A variant is also present in the Frankfurt-1 isolate [ 15 ], which was contracted from the early Singaporean patient traveling on the same flight as the German patient. Discussion Through the application of this MS mini-sequencing approach to the SARS outbreak in Singapore we have demonstrated the precision that pathogen sequence data can add to an epidemiological investigation. We analyzed 18 SARS-CoV SNVs and determined the molecular relationship among the viral isolates from 13 Singapore SARS patients (see Figure 1 ) from different stages of the Singapore outbreak. The molecular relationship among the patients' viral isolates derived from our MS-based viral sequence analysis is not consistent with the current understanding of the clinical transmission relations between these patients. According to contact tracing records, patient Sin2500 was believed to have been the index case of the Singapore SARS outbreak and to have introduced the SARS-CoV virus into the Singapore population following a visit to the Hotel M (Hong Kong) [ 5 , 6 ]. Patients Sin2774, 2748, and Sin2677 were believed to have been infected directly by the index case Sin2500, and patient Sin2679 was believed to have been infected by the index case through another, unidentified, primary patient [ 5 ]. Given the pattern of sequence variations observed in the viral isolates, in order for this presumed clinical transmission relationship among these five patients to be correct, one has to assume that during viral transmission from index case Sin2500 to secondary case Sin2679 via an unidentified primary case (two human-to-human transmissions), two reverse mutations at SNV positions 19,084 and 28,268 and one novel mutation at SNV position 23,174 occurred. Although this is not impossible, it is unlikely considering the observed mutation rate among the post–Hotel M SARS-CoV isolates [ 5 , 16 , 17 ]. A more parsimonious explanation would be a single-mutation scenario in which, instead of contracting the virus from the presumed index case Sin2500, the secondary patient Sin2679 and all the later Singapore cases were infected by a virus strain from the Hotel M cluster through another, as yet unidentified, route, and during this transmission, a novel mutation occurred at SNV position 23,174. Thus, patient Sin2679 or another unidentified Singapore patient from whom patient Sin2679 contracted SARS should be the index case of all the late-generation Singapore SARS patients. Further viral genetic characterization of additional Singapore SARS cases, especially early-generation ones, may shed light on this hypothesis. Unidentified secondary SARS-CoV infection routes from Guangdong to Hong Kong were also suggested by the genetic characterization of SARS-CoV isolates, although none of these contributed substantially to the subsequent Hong Kong outbreak [ 6 , 18 ]. A further application of MS-based viral sequence variation analysis in tracking the virus strain and thus the transmission of SARS-CoV was demonstrated by our confirmation of the Singapore origin of a SARS-CoV isolate from a German patient. Travel and contact tracing records for this German patient indicated more than one potential exposure to SARS-CoV, and because virus could not be cultured from the patient, it was difficult to pinpoint the origin of his infection by classical sequencing methods that require virus enrichment by culture. By genotyping the four SNV positions that showed unique variants in the early Singapore SARS-CoV isolates, we confirmed that this German patient was indeed infected by an early Singapore virus strain, most likely in a hitherto unnoticed aircraft transmission event from an early Singaporean patient who was later hospitalized with SARS in Germany [ 9 ]. Therefore, our results clearly demonstrate the usefulness of the sequence variation information as molecular fingerprint in “tagging” SARS-CoV viral strains. Direct viral sequence variation analysis of uncultured lung tissue samples identified cases of heterogeneous viral sequences in single patient samples. As the SARS-CoV virus is a single-strand RNA virus, the discovery of different sequences in a single tissue sample suggests the presence of multiple viral sequence variants, or quasispsecies, within the host when the sample was retrieved. Our result has further confirmed a recent observation of SARS-CoV quasispecies in individual patients [ 19 ] and is consistent with observations in other viral infections. Furthermore, direct comparison between cultured and uncultured samples from the same patient confirmed the existence of the heterogeneous viral sequences only in the uncultured tissue sample, which suggests that of the two initial variants in the human host, only one survived in the vero cell culture. This raises a concern that viral genetic characterization in cultured viral isolates may not capture the whole sequence variation spectrum of a virus in a patient population. Our study has clearly demonstrated the advantages of MS-based genetic analysis as a method for large-scale viral genetic characterization in clinical samples. Firstly, MS-based analysis has high sensitivity, providing successful detection of virus more than 95% of the time at virus concentrations as low as 75 RNA copies per reaction (equivalent to a detection sensitivity of 10 3 –10 4 RNA copies per milliliter), which is close to the detection limit of real-time RT-PCR based diagnostic tests (demonstrated to be 5–85 copies of viral RNA per reaction) [ 7 , 20 , 21 , 22 ] and within the concentration range reported for SARS-CoV in respiratory and plasma samples [ 9 , 20 , 22 , 23 , 24 ]. Typical RT-PCR sequencing usually requires as many as 1,000 copies of template, as large PCR fragments are typically amplified with less efficiency than the small fragments (about 100 bp) that are commonly used in MS-based analysis. Secondly, our detection of heterogeneous viral sequences in single clinical samples demonstrated the accuracy of the MS-based assay in characterizing SARS-CoV sequence variations. Thirdly, MS-based assay requires only a small amount of starting material for genetic characterization, 0.1 μl of RNA per reaction in the present study. Currently, MS-based genotyping analysis of human genetic variation is routinely done in a multiplex fashion, where multiple single nucleotide polymorphisms are genotyped simultaneously in a single assay. It is thus conceivable that the development of a multiplexing MS-based SNV assay for SARS-CoV could further reduce the required amount of starting material, which would be especially beneficial for analyzing uncultured clinical samples, in which viral materials are often limited. Multiplexing analysis of MS-based assays also greatly reduces the cost of analysis to about US$0.10–$0.20 per analysis (depending on the level of multiplexing), which is much cheaper than conventional sequence analysis, whose cost is typically a few dollars per analysis. Therefore, MS-based sequence variation analysis is a sensitive, accurate, cost-effective, and high-throughput method for confirming putative variations and characterizing known variations in clinical samples, especially for large-scale population studies. MS-based sequence variation analysis is complementary to the identification of new sequence variation by direct sequence analysis and is particularly suitable for investigating agents for which there is already extensive sequence information. Direct sequence analysis is still the “gold standard” for identifying new sequence variations, but it is inefficient and is not necessary for characterizing known sequence variations in a large number of samples. A combination of initial characterization of genome sequence by direct sequence analysis in a subset of samples and the subsequent analysis of informative genetic variations via a MS-based approach is more efficient and suitable for large-scale population investigations. The genome sequences of a wide variety of pathogens and strains are being rapidly accumulated. In bacterial pathogens, strain sequence information is frequently limited to a relatively small number of genes (where using all of them simultaneously might be appropriate), whereas in viruses such as influenza, extensive genomic knowledge is accumulating. Such accumulation of both partial- and whole-genome sequence information for pathogens will further extend the usefulness of this approach. In summary, our reassessment of the SARS-CoV transmission route in Singapore using MS-based viral sequence variation analysis highlighted the limitation of conventional epidemiological analysis based on travel and contact tracing, and the importance of informing clinical and epidemiological investigation of pathogen transmission by genetic analysis. With its demonstrated high throughput [ 25 ], sensitivity, accuracy, and cost effectiveness in determining viral sequence variations, MS-based genetic analysis can greatly facilitate the large-scale epidemiological investigations of SARS-CoV and other agents of infectious disease, and may allow for real-time investigation in outbreak situations. Supporting Information Figure S1 Detection of SARS-CoV by Real-Time Quantitative PCR in Spiked Human Blood Samples The x-axis denotes the cycle number of the quantitative PCR assay, and the y-axis denotes fluorescence intensity (F2) over the background level. RNA standards were as follows: 1.05 × 10 6 copies per reaction (line a), 1.01 × 10 5 copies per reaction (line b), 9.4 × 10 3 copies per reaction (line c), 8.9 × 10 2 copies per reaction (line d), and 1.07 × 10 2 copies per reaction (line e). The virus loads determined in the five spiked human blood samples were as follows: SB1, 1.64 × 10 6 copies per reaction; SB2, 3.84 × 10 3 copies per reaction; SB3, 2.17 × 10 3 copies per reaction; SB4, 6.21 × 10 2 copies per reaction; and SB5, 1.20 × 10 2 copies per reaction. NTC, non-template control. (244 KB DOC). Click here for additional data file. Accession Numbers The GenBank ( http://www.ncbi.nlm.nih.gov/Genbank/index.html ) accession number for the Frankfurt-1 isolate is AY291315 and for the Urbani isolate is AY278741. Patient Summary Background Molecular biology (studying the makeup and function of molecules) is increasingly being used to track outbreaks of infectious diseases. For example, molecular biology can help to identify the cause of a disease (such as a virus, bacterium, or parasite) and understand how it spreads. Why Was This Study Done? These researchers had previously used molecular biology techniques to study different strains of the SARS virus, which causes the often fatal disease called severe acute respiratory syndrome. They had found that different strains could be distinguished from each other on the basis of specific genetic “fingerprints.” They now wanted to find quick and easy ways to determine the identity of particular viral strains found in sick patients. What Did the Researchers Do? They used a molecular biology technique called mass spectrometry. They took samples from patients (such as blood samples and nasal swabs) and determined whether they could detect the SARS virus, identify specific strains, and distinguish between them. What Did They Find? They found that mass spectrometry is a useful tool for detecting the SARS virus and for distinguishing between different strains. They also found that they could use this tool to help understand how the SARS virus had been transmitted between specific patients. What Are the Limitations? For this technique to work, the researchers needed pre-existing information about genetic differences between strains. This means that detailed DNA sequencing is necessary to find these differences in the first place and to discover new ones as the virus evolves. What Next? The authors suggest that combining initial genetic sequencing of the different strains with the mass spectrometry technique to analyze subsequently large numbers of samples is the most efficient and cost-effective approach. More Information Online Public access Web pages on SARS from Science Magazine: http://www.sciencemag.org/feature/data/sars/ News article on the SARS genome on the Genome Network News Web site: http://www.genomenewsnetwork.org/articles/05_03/sars_3.shtml Genome Institute of Singapore (GIS): http://www.gis.a-star.edu.sg/homepage/default.jsp GIS's press release on differences between SARS strains: http://www.gis.a-star.edu.sg/homepage/gismediapress.jsp?pid=19	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC549591.xml
549585	Asthma Severity and Prevalence: An Ongoing Interaction between Exposure, Hygiene, and Lifestyle	Why are the prevalence and severity of asthma increasing? Platts-Mills looks at the key studies that can help to anwer this important question	Over the last hundred years, there have been major triumphs in medicine related to public health, vaccination, and the introduction of new medicines. However, over the same period, several diseases have increased in prevalence and/or severity. In some cases, the causes of the increase have become obvious—the increases in lung cancer, coronary artery disease and type 2 diabetes, for example, are not considered to be a mystery. On the other hand, a large group of diseases related broadly to “inflammation” have also increased. For these, a wide range of hypotheses about causation have been proposed. Type 1 diabetes, rheumatoid arthritis, and inflammatory bowel disease have increased since 1980 [ 1 ]. Some analyses of the increase in hay fever and asthma would suggest a similar time course, and this parallelism of the time frame has been taken to suggest that there could be a common cause. Indeed, there is a proposal that these diseases are all related to some changes in “cleanliness” or “hygiene” that have resulted in decreased activation of a common control mechanism. Specifically, this control has been ascribed to T regulator cells, which produce interleukin-10 (IL-10) or transforming growth factor-beta. Change in Living Conditions The major changes in rural areas, tropical villages or in Europe pre-1900 that could be related to the change in immune responses include: decreases in helminth infection; physical proximity to farm animals [ 2 ]; exposure to those mycobacteria that are commonly found in the soil; bifidobacteria colonization of the gut; as well as decreased prevalence of Hepatitis A infection [ 3 ]. The implication in each case is that asthma has increased secondary to an increase in inflammation or an increase of the allergic response that is closely associated. This assumes that hay fever, asthma, and other diseases have increased in parallel, which is probably not true. Hay fever became a problem in Northern Germany and England as early as 1900. Clear evidence for this view comes from Noon's description of the development of immunotherapy for hay fever in 1911, the studies on hay fever prevalence by Ratner and Silverman in New York in 1935, and the recognition that hay fever was a major community problem in New York in 1946. More recently, Harold Nelson analyzed all the studies on hay fever published in the United States and found a prevalence of ~15% in 1960, with no convincing evidence of an increase since then (H. Nelson, personal communication). If seasonal hay fever was epidemic in 1960, then the subsequent increase in asthma has to be seen in a different light. The best estimates for the start of the asthma epidemic are around 1960 for such diverse populations as army recruits in Finland [ 4 ], school children in Birmingham in the United Kingdom, and African American children in Charleston, South Carolina, United States [ 5 ]. In each of these studies, the increase in asthma symptoms or disease has been greater than tenfold. However, the absolute values of the change have been dramatically different in New Zealand and Scotland (from ~2% up to 20%) compared with Finland (from 0.2% to 4%) [ 6 ]. Furthermore, some countries have experienced much greater increases in hospitalization than others. Clear evidence for increases in mortality has only come from the United Kingdom, New Zealand, and the United States, countries with a high prevalence of symptoms and hospitalization. Several hypotheses have been proposed to explain the increase in asthma ( Box 1 ). At this point, our questions about the increase in asthma are: 1) Is an increase in allergy or hay fever a necessary precursor for the increase in asthma, a parallel event, or separate? 2) Why did the increase in asthma have such a consistent time course throughout the Western world in countries where changes in infectious diseases have occurred very differently? and 3) Why is asthma more common and more severe in some countries than in other countries that have had an equal scale of increase (i.e., tenfold) over the same time course? Box 1. Hypotheses about the Cause of the Increase in Asthma: Arguments For and Against Hypothesis 1: Increased exposure to perennial allergens, e.g., dust mites For: (A) Housing changes: houses are built more tightly and are more well insulated; more furnishings; fitted carpets, and (B) more time spent indoors. Increased exposure leads to increased sensitization. Against: Increases in asthma have been seen in countries where dust mites are not present in homes, and in the Netherlands the concentration of mite allergens has actually decreased by as much as tenfold over the past 15 years Hypothesis 2: Changed immune responsiveness is due to cleanliness For: (A) Bacterial and other infections have decreased due to improved hygiene, immunization, and antibiotics, and (B) changed gut flora (antibiotics, diet, etc.). Change from Th1 to Th2 leads to increased allergy. Against: In New York City, the increase in seasonal hay fever occurred 30 years before the increase in asthma, and in Africa, children whose families move into cities, including informal settlements, have experienced increases in infections, wheezing, and diagnosed asthma Hypothesis 3: Loss of a lung-specific protective effect, 1960–2000 For: (A) Changing diet leads to a changed inflammatory response, and (B) decline in physical exercise. Increased wheezing among allergic children. Against: This is unlikely to be just be a lung-specific effect The Persistent Association between Specific IgE, Total IgE, and Asthma Long before the first case control or prospective study, the association between allergy and asthma was obvious in case series. These earlier studies reported skin testing with an amorphous extract called “house dust,” but it was not until the identification of dust mites that the association was clarified [ 7 ]. Indeed, as late as 1978, there were significant doubts that allergens played a role in asthma. Using extracts of Dermatophagoides pteronyssinus (dust mites), the strong association between this allergen and asthma was established in many parts of the world, with odds ratios as high as 6 or even 10 [ 7 , 8 ]. The possibility of a causal relationship was further supported by bronchial challenge studies and avoidance experiments [ 9 ]. The cohort in Poole, Dorset was, until recently, the only study with household measurements of allergen and the results strongly suggested that exposure in the child's own house was the primary determinant of sensitization [ 8 ]. Subsequently, studies from other parts of the world provided evidence about other indoor allergens, particularly cats, dogs, and the German cockroach [ 10 , 11 ]. These studies showed that perennial exposure to allergens was an important cause of inflammation in the lungs and associated nonspecific bronchial hyperreactivity ( Figure 1 ). In most of the case-control and prospective studies, sensitization to seasonal pollens has not been significantly associated with asthma [ 7 ]. This is an important issue; if allergy is associated with asthma for genetic reasons or because of some common immunological feature, it is not clear why the association should be with perennial allergens only. Figure 1 Sensitization, Inflammation, and Wheezing Increases in prevalence/severity of asthma (reversible airway obstruction) could occur because of changes in different parts of the hypersensitivity and inflammatory response in the lungs. The Roman numerals refer to the three hypotheses in Box 1 . Since assays for total serum IgE became available in the 1970s, it has been clear that patients with asthma have, on average, higher total IgE than patients with hay fever or no allergy. Indeed, by 1980 this was considered an established fact in textbooks of immunology. It was assumed that the increased total IgE related to allergen-specific responses. However, in some studies, the association between total IgE and asthma was stronger than the association between asthma and specific IgE. In 1989, Burrows et al. went further and suggested that specific IgE correlated with hay fever, while total IgE correlated with asthma [ 12 ]. The implication was that IgE has a complex relationship with asthma that is not dependent on specific allergens. The strength of the association between asthma and total IgE raises questions that have not been resolved. Do specific IgE antibody responses contribute to or even push total IgE? If so, do the IgE antibody responses to some allergens have more effect than others? This question is relevant both to attempts to explain major differences in total IgE between countries and to studies on acute asthma. In emergency room and hospital studies, the geometric mean total serum IgE of patients with asthma is often greater than 200 IU/ml higher than values found in population-based studies. Recent work from Heymann et al. and Green et al. on patients hospitalized for asthma has suggested that the interaction between rhinovirus and allergy occurs predominantly among patients with total IgE > 200 IU/ml [ 13 ]. Thus, the different properties of allergens could influence both the prevalence and severity of asthma. However, the properties of the dominant allergens do not explain the overall increase in prevalence, which has occurred in countries with very different houses, climates, and traditions of domestic pet ownership. The Paradoxical Relationship between Cat Ownership and Sensitization: Significance for Prevalence or Severity Antigen exposure is considered to be a primary requisite for immune responses, and allergen-specific responses are no exception. There are many examples of allergens that are not significant in areas where the allergen is not encountered. For example, the pollen of olive trees is not relevant in northern countries, dust mite allergens are not significant in the northern part of Scandinavia or the mountain states of the United States, and cockroach allergens are not significant in suburban areas of the United States, the United Kingdom, or New Zealand. For dust mites, there is a wide range of evidence that increased exposure increases sensitization. Homes in Sweden, Berlin, the United Kingdom, and New Zealand have progressively higher concentrations of mite allergen and progressively higher prevalence of sensitization to mite allergens [ 7 ]. But there is now evidence that increasing exposure to cats does not lead to a higher prevalence of allergy [ 14 , 15 , 16 ]. On a population basis, the effect may be profound; sensitization to cats among school-age children is generally ~10%, while mite sensitization is often as high as 30% ( Figure 2 ). This effect cannot be ascribed to inadequate exposure, since all estimates of the quantity of cat allergen inhaled are higher than for dust mites. Furthermore, the quantity of cat allergen found in schools or even in houses without cats is sufficient to sensitize at-risk children [ 11 ]. Figure 2 Contrast between Exposure to Dust Mite or Cat Allergens and the Relevant Immune Responses The dashed line indicates the approximate value of 20 mg Fel d 1/g floor dust or the presence of a cat. In fact, children raised in a house with a cat were less likely to be sensitized to cats [ 15 ]. Initially, it seemed possible that the effect was an example of reverse causation. However, the effect has been observed in countries where a large proportion of families keep cats, and very few families report choosing not to own a cat because of asthma in the family. Furthermore, the presence of a cat in a house in New Zealand does not decrease IgE antibody response to dust mites—in other words, tolerance to cats can be cat-specific [ 17 ]. Understanding how the response to cat allergen is controlled could provide an insight into how both the prevalence and the titre of IgE antibody responses in general are (or should be) controlled. It seems inevitable that the primary control is by T cells specific for cat allergens. Indeed, there is already excellent evidence that injection of peptides derived from the cat allergen Fel d 1, which give T cell responses, can be used for immunotherapy [ 18 ]. Studying overlapping peptides of Fel d 1, we identified a striking response to two peptides at the terminal end of Chain 2. Furthermore, both allergic and tolerant individuals respond to these peptides by producing IL-10 and interferon gamma [ 19 ]. The implication is that Fel d 1 inherently induces control, and that this control influences both allergic and non-allergic responses to cat allergen ( Figure 3 ). In keeping with this, IgE antibody responses to cat allergen are not quantitatively as high as those to dust mites [ 17 ]. Figure 3 Mechanisms of T Cell Control over the Hypersensitivity Response to Cat Allergen The response to cat allergen, Fel d 1, includes T cells that produce high levels of IL-10 that have some of the features of the T regulator one cells (Tr1). However, these cells appear to be a feature of both allergic and non-allergic responses, implying that the whole response to cat allergen is controlled. Some of the effects of cat exposure may need to occur early, but other studies suggest strongly that the different responses can change following a prolonged change in exposure late in life. One of the central assumptions of the cleanliness hypothesis is that regulation of immune responses is, at least in part, non-specific. It is assumed that helminth infection, mycobacteria, Hepatitis A, endotoxin, and early-life infections can create a milieu that leads to a decrease in allergic responses in general. Likewise, we might expect that exposure to high concentrations of cat allergen, which can induce IL-10 producing cells, should have a general effect. There is some evidence for this hypothesis. In Detroit, the presence of two or more animals in the house tended to reduce allergy in general. In Sweden, the presence of a cat in the house is associated with decreased sensitization to cat, and also to birch and dog. However, in the United States and New Zealand, the presence of a cat in the home has no effect on the prevalence or titre of IgE antibody to dust mites [ 15 , 17 ]. Thus, it is clear that under some circumstances the tolerance to cats can occur in highly atopic individuals and be cat-specific. This phenomenon is not in keeping with any version of the cleanliness hypothesis. It has been proposed that animals in the home could have a non-specific effect because they shed or encourage endotoxin. However, recent studies on airborne endotoxin found that the presence of cats had no effect, and that there were significantly lower endotoxin levels in houses with cats compared to dogs. The Possible Role of Lifestyle Changes Over the last half of the 20th century, there have been major changes in diet and physical activity. The most obvious result of these changes is an increase in obesity, which has reached epidemic proportions in the United States. Since 1994, there have been multiple reports of an association between elevated body mass index (BMI) and asthma [ 20 ]. In 1996, we suggested that changes in physical activity could be related to bronchospasm [ 21 ]. Thus, there are three distinct but strongly interrelated aspects of lifestyle that could be relevant to the prevalence and severity of asthma: diet, physical activity, and obesity. It is much easier to document BMI than diet or physical activity, and although some of the obesity data are convincing, they are not consistent and certainly not comparable to the association between obesity and diseases such as type 2 diabetes in childhood. In a typical study, Camargo et al. found that the prevalence of wheezing was 13% among the heaviest quintile and 7% among the lowest quintile for BMI [ 22 ]. Comparable values for type 2 diabetes would be 5% and 0.1%. Studies using questionnaires have attempted to ask whether there is a relationship between wheezing and physical activity. However, the track record for questionnaires on this subject is very poor. Westerterp and his colleagues have reported two observations: first, that general activity contributes more to energy consumption than “aerobic exercise” does and second, that many subjects who initiate an exercise program (such as a twice weekly visit to the gym) overcompensate so that they actually decrease overall activity [ 23 ]. Recently, we have documented a decrease in activity among children (age ~4 years) in the United States Head Start program (a child-development program that aims to increase the school-readiness of young children in low-income families) who have a history of wheezing [ 24 ]. Although there are several possible explanations for this result, it seems clear that decreased activity can be present before elevated BMI. The next question to ask is whether physical activity would have a beneficial effect on asthma? Fredburg concluded that full expansion of the lungs had a more potent effect on bronchial smooth muscle than isoprenaline [ 25 ]. Togias and his colleagues have shown that prolonged shallow breathing (=20 minutes) can lead to increased non-specific bronchial reactivity [ 26 ]. It is obvious that expansion of the lungs is decreased during prolonged periods of sitting down. However, periodic expansion of the lungs occurs with sighs. Recent data from our group shows that “sigh rates” while seated are very variable but are significantly lower when watching a video than while reading [ 27 ]. Thus there is a real possibility that some forms of childhood behavior—TV, videos, computer games, etc.—might be associated with sigh rates low enough to increase non-specific bronchial hyperreactivity. A different explanation for the effects of physical activity comes from evidence that physical activity can be “anti-inflammatory.” This evidence relates to several different models though not at present to the lungs. In some studies, obesity appears to be a risk factor for wheezing among non-allergic children. However, in most studies, the association between allergen sensitization and asthma has been found in obese and non-obese children equally [ 28 ]. Obviously, some obese individuals are unfit and become breathless on exercise. In addition, these individuals may have sleep-disordered breathing. Thus, there are other conditions that are easily confused with exercise-induced asthma or nocturnal asthma. Taken together, there are excellent reasons for asking whether lifestyle changes have contributed to the increased prevalence or severity of asthma. However, it seems unlikely that this effect occurs on normal lungs, so the hypothesis has to be that decreased physical activity in patients who are allergic can allow persistent or increased severity of wheezing. Conclusions Although the explanation for the increase in asthma is not yet clear, it is possible to put forward a model that includes elements of each of the three main hypotheses. Children raised in the tropics, on farms, or in villages such as those in Africa or Papua New Guinea have exposure to endotoxin or infections sufficient to interfere with the development of allergen-specific IgE antibody responses. Once water supplies are clean, and major infectious diseases have been controlled, allergic diseases will appear. However, asthma appears to be associated with perennial, i.e., indoor exposure, and may be more common or more severe in countries where mites or cockroaches are the major source of allergens. Even with indoor allergen exposure, wheezing may remain transient or mild, provided prolonged outdoor play is normal. It is the combination of the control of infectious diseases, prolonged indoor exposure, and a sedentary lifestyle that is the key to the asthma epidemic and, in particular, the key to the rise in severity. Using this analysis, the severity of asthma in North American cities becomes much easier to explain. Children in New York, Atlanta, Philadelphia, and Washington, D.C. spend long hours indoors, have high exposure to mite, cockroach, and/or rodent allergens, and have very low levels of physical activity. In conclusion, it appears that these combined factors are the key to the asthma epidemic and, in particular, the key to the rise in severity. We clearly need to develop ways to increase prolonged physical activity, both among patients with asthma and in the general population. We also need to investigate whether prolonged moderate activity is beneficial in the treatment of asthma and/or is “anti-inflammatory.” What is equally clear is that the current obsession of the medical profession with the pharmaceutical management of asthma (as well as other lifestyle-related diseases) does not address the reasons why the disease has become so common and so severe.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC549585.xml
515365	Continued Colonization of the Human Genome by Mitochondrial DNA	Integration of mitochondrial DNA fragments into nuclear chromosomes (giving rise to nuclear DNA sequences of mitochondrial origin, or NUMTs) is an ongoing process that shapes nuclear genomes. In yeast this process depends on double-strand-break repair. Since NUMTs lack amplification and specific integration mechanisms, they represent the prototype of exogenous insertions in the nucleus. From sequence analysis of the genome of Homo sapiens, followed by sampling humans from different ethnic backgrounds, and chimpanzees, we have identified 27 NUMTs that are specific to humans and must have colonized human chromosomes in the last 4–6 million years. Thus, we measured the fixation rate of NUMTs in the human genome. Six such NUMTs show insertion polymorphism and provide a useful set of DNA markers for human population genetics. We also found that during recent human evolution, Chromosomes 18 and Y have been more susceptible to colonization by NUMTs. Surprisingly, 23 out of 27 human-specific NUMTs are inserted in known or predicted genes, mainly in introns. Some individuals carry a NUMT insertion in a tumor-suppressor gene and in a putative angiogenesis inhibitor. Therefore in humans, but not in yeast, NUMT integrations preferentially target coding or regulatory sequences. This is indeed the case for novel insertions associated with human diseases and those driven by environmental insults. We thus propose a mutagenic phenomenon that may be responsible for a variety of genetic diseases in humans and suggest that genetic or environmental factors that increase the frequency of chromosome breaks provide the impetus for the continued colonization of the human genome by mitochondrial DNA.	Introduction Insertion of new sequences into nuclear DNA has a major impact on its architecture and is an important mechanism for the evolution of eukaryotic genomes. Moreover, when targeted to gene loci, these insertions can be mutagenic, and in humans this process contributes to a number of diseases ( Deininger and Batzer 1999 ; Neil and Cameron 2002 ; Nelson et al. 2003 ). The frequency of the insertion events and the site of integration are therefore critical factors influencing genomic stability. In humans these two aspects have been investigated for mobile elements, including long and short interspersed elements and retroviruses ( Li et al. 2001 ; Batzer and Deininger 2002 ), but much less is known about nuclear DNA sequences of mitochondrial origin (NUMTs), which have been found associated with diseases in humans ( Willett-Brozick et al. 2001 ; Borensztajn et al. 2002 ; Turner et al. 2003 ). DNA fragments of mitochondrial origin, originating from both coding and noncoding regions, are found as sequence fossils in the nuclear genomes of various eukaryotes ( Blanchard and Schmidt 1996 ). However, de novo integrations have been recently detected in yeast and humans ( Ricchetti et al. 1999 ; Yu and Gabriel 1999 ; Turner et al. 2003 ), and insertion of NUMTs in the nuclear genome has been found to be an ongoing process in yeast ( Ricchetti et al. 1999 ). We and others have previously shown that NUMTs integrate in the nuclear genome during the repair of double-strand breaks (DSBs) in yeast growing mitotically ( Ricchetti et al. 1999 ; Yu and Gabriel 1999 ). In these studies, sequences of mitochondrial origin were the main or the exclusive type of DNA able to integrate at an induced DSB. Before the sequencing of the human genome was completed, occasional reports described sequences of mitochondrial origin located in the nucleus ( Tsuzuki et al. 1983 ; Perna et al. 1996 ), in one case in vivo ( Zischler et al. 1995 ). More recently, sequence analysis performed on the first human genome draft revealed the presence of between 280 and 296 NUMTs ( Mourier et al. 2001 ; Tourmen et al. 2002 ). According to one study, it appears that only one third of NUMTs were integrated as new sequences, whereas the remaining two thirds originated as duplications of preexisting NUMTs ( Hazkani-Covo et al. 2003 ). Another report suggests that most NUMTs arose from independent insertion events ( Bensasson et al. 2003 ), thereby raising questions regarding the real insertion rate of these sequences in the human genome. Moreover, it has been suggested that most NUMTs have been inserted in a primate ancestor ( Tourmen et al. 2002 ; Bensasson et al. 2003 ). Thus, the rate and the effects of colonization of the human genome by DNA fragments of mitochondrial origin remain unclear, and the presence of such sequences has not been fully investigated in humans. With the availability of the human genome sequence coupled with significant discoveries on the evolution of Homo sapiens, experimental approaches that compare individuals within this species and its closest relative, the chimpanzee, can be undertaken ( Chen et al. 2001 ). In the present study, we demonstrate the presence in vivo of NUMTs in the human and in the chimpanzee genomes, using genome-wide sequence analysis combined with direct evaluation of DNA samples of individuals. Moreover, we show a significant degree of insertion polymorphism of NUMTs in human populations. We also provide a comprehensive analysis of NUMTs that have specifically colonized the human genome, and we determine the fixation rate of these sequences in H. sapiens. Furthermore, we correlate these findings with observations of the insertion of NUMTs involved in human diseases. We show that human-specific NUMTs , unlike those in yeast, preferentially integrate in known or predicted genes and can therefore be mutagenic, thereby generating genetic alterations in humans. Results NUMTs Are Mitochondrial Sequences Residing in the Human Nuclear Genome From a blastn search on the database of H. sapiens published by the public consortium ( Lander et al. 2001 ), using as query the human mitochondrial (mt) DNA sequence ( Anderson et al. 1981 ), 211 NUMTs were found. We observed that approximately 93% of NUMTs represent insertions of a single DNA fragment, whereas 7% consist of multiple, unrelated mtDNA fragments, similar to the pattern frequently found in yeast ( Ricchetti et al. 1999 ). We observed that NUMTs ranged in size from 47 to 14,654 bp with a sequence identity to the human mtDNA of 78%–100%. Previous analyses were based on less complete genome sequencing (84% complete for the most recent study; Bensasson et al. 2003 ), while our study was performed on a 99% complete sequencing of the euchromatic genome of H. sapiens. Our updated analysis (data not shown) reveals that the majority of these NUMTs correspond to those previously documented ( Mourier et al. 2001 ; Tourmen et al. 2002 ). However, our study draws attention to NUMTs that are highly identical to mtDNA, and some of these sequences did not appear in earlier analyses. Indeed most of the NUMTs in our study are shorter than 100 bp (see Table 1 ), whereas former investigations focused more on longer NUMTs. Although the presence of shorter NUMTs has been reported, these have not been published ( Tourmen et al. 2002 ). Combining the lowered size threshold and the more complete database, we were able to identify 23 new NUMTs (labeled with an asterisk in Table 1 ). Most of the new sequences are of the highest interest for our studies of the acquisition of NUMTs by H. sapiens and of insertion polymorphism in humans (see below). Table 1 PCR Amplification and Sequence Analysis of NUMTs from Humans and Chimpanzees Upper part, scheme of the PCR strategies; thin line, chromosomal DNA; thick line, NUMT; arrowed lines, PCR primers PCR amplifications were done either with primers A + B or A + C. In NUMT code names, the first number represents the chromosome number, and the second the NUMT size; an asterisk to the right indicates NUMTs described in this paper. “Percent” indicates the percentage of identity of the NUMT to the human mtDNA as scored by blastn. Missing values in the “no NUMT” column indicate that the PCR was done with the strategy A + B, rather than A + C (see upper part of table). “Amplified in Chimp” indicates whether the NUMT did amplify (+) or not (−) in the chimpanzee genome. Where BLAST output search results from a database in April 2003 did not fit with our PCR and sequence analysis, we chose as indicated in footnotes “a” and “c.” a January 2003 b Strategy A + B c July 2001 BLAST results, which were consistent with our sequencing d These two lines indicate the 5′ (up) and the 3′ (dw) portion of the same NUMT e This NUMT is present also on Chromosome 9; separate analysis of these two NUMTs was not possible by PCR and sequencing (mixed products) f NUMT is specific to chromosome Y; it is absent from the eight-female sample g This is the same as “f,” but an additional locus without NUMT is present on the X chromosome h This NUMT, described previously (Zischler et al. 1995), was renamed here ND, no amplification was detected More information on these NUMTs is available in Table S1 To determine whether sequences of mitochondrial origin were actually integrated in the human nuclear genome, and were not a result of contamination of DNA library preparations, we selected 42 NUMTs for analysis in human samples. Our choice included the 36 NUMTs with the highest identity (91% to 100%) to the mtDNA, one NUMT having the longest stretch of DNA with high identity (88%), one NUMT corresponding to the highly variable region of the mtDNA (D-loop) ( Cann and Wilson 1983 ), and four NUMTs randomly chosen with identity from 79% to 90% ( Table 1 ). Each of these NUMTs was amplified by PCR from DNA obtained from 21 human donors (eight females, 13 males) representing different ethnical groups. Our pool consisted of ten Caucasians, seven Africans (including four Pygmies), two Japanese, and two Chinese ( Table 2 ). To amplify chromosomal NUMTs and avoid amplification of the mt chromosome, we used a primer located in the upstream flanking region, in combination with a primer located either in the 3′ region of the NUMT or in the downstream flanking region (see upper part of Table 1 , primers A + B or A + C, respectively). In the former case, PCR amplification served as a supplementary control for bona fide mtDNA integration at the locus, while in the latter, PCR amplification was followed by sequencing to assay for the presence of the NUMT. Forty-one out of 42 loci tested amplified a fragment of the expected length ( Table 1 ), while one locus (14–1023 [Chromosome 14; size = 1023 bp]) amplified a fragment not containing the NUMT in all individuals tested. Eighteen loci were analyzed further in one or more individuals by sequencing the amplified fragment to verify whether these DNA fragments included the sequence of mt origin (see Table 1 ). The expected sequence was indeed present in all cases tested (except at polymorphic loci, described later, and at insertions in the Y chromosome, present only in males). Table 2 Insertion Polymorphism of NUMTs Displaying Distinct Lineage Characteristics in Humans For each NUMT is indicated the PCR amplification containing (+/+) or not containing (−/−) NUMT; (+/−) indicates it is heterozygous. NUMT 11-541 was identified previously (Zischler et al. 1995) a Locus has been sequenced b Both allelic forms have been sequenced In summary, results from two amplification strategies and from sequencing demonstrated that these NUMTs were indeed present at the expected chromosomal location and that they are bona fide mt sequences residing in the human nuclear genome. Insertion Polymorphism of NUMTs in Human Populations The colonization of human populations by various NUMTs revealed striking disparities. Thirty-five NUMTs were present in homozygous form in all individuals tested (see Table 1 ). Interestingly, NUMTs 1-74, 2-53, 12-89, and 18-192 were present in only a few individuals either as homozygous or heterozygous loci (see Figure 1 ; Tables 1 and 2 ). A more limited heterogeneity was observed for NUMTs 13-75 and 2-132, where only two and one individual, respectively, were heterozygous. In total, six out of 41 NUMTs showed insertion polymorphism. Integration of NUMTs was further confirmed by sequencing both positive and negative samples (see Table 2 for the samples tested). The sequences of these six NUMTs are shown in Figure 2 . NUMT 11-541, whose insertion polymorphism was previously described ( Zischler et al. 1995 ; Thomas et al. 1996 ), was reanalyzed here ( Tables 1 and 2 ). By comparing the flanking sequences of individuals carrying a NUMT with those of individuals who have no NUMT, it is possible to identify the junction sites and to also clarify the mechanism by which NUMTs were inserted. This analysis was done for NUMTs 2-132 and 18-192, in which the junction sites ( Figure 2 ) show microhomology between the invading NUMT and the chromosomal end, and sometimes addition of a few nucleotides. Both the presence of microhomology and the addition of short sequences also accompanied the insertion of NUMTs in the yeast genome ( Ricchetti et al. 1999 ), and they are hallmarks of the DSB repair mechanism non-homologous end-joining (NHEJ) ( Critchlow and Jackson 1998 ). This suggests that NHEJ may also account for the insertion of NUMTs in humans. Figure 1 Polymorphism of NUMTs 18-192, 1-74, and 2-53 The polymorphism of NUMTs 18-192, 1-74, and 2-53 as revealed by PCR amplification and electrophoresis of the products on 2% agarose gels. For each locus, the upper arrow indicates the fragment that contains the NUMT, and the lower arrow indicates the fragment that does not contain the NUMT. The individual tested is indicated above. The (+/+) are homozygous positive, (+/−) are heterozygotes, and (−/−) are homozygous negative. Figure 2 Sequence Insertion Polymorphism of Six NUMTs Sequence of NUMTs 1-74, 2-53, 2-132, 12-89,13-75 and 18-192 are indicated in lower case and the flanking sequences in capital letters. Underlined letters represent nucleotides homologous to both the mt and the chromosomal sequences (microhomology). Bold and italicized letters correspond to nucleotide additions, following the NUMTs insertion, which are absent from the −/− individuals. The individuals sequenced are indicated in Table 2 . In all cases the sequence corresponded to the one available on the human genome public Web sites. Boxes represent exon sequences. In 12-89, the exon sequence would extend till the stop codon (taa). Interestingly, one or more of these six NUMTs were detected among individuals within each ethnic group, indicating that their insertion in the nuclear genome occurred soon after the origin of modern humans and that they represent the most recent integrations of our studied cases. Despite the limited sampling size (42 alleles), the frequency of alleles carrying the insertion varies greatly according to the NUMT (calculated from Table 2 ): 98%, 95%, 48%, 29%, and 21% for NUMTs 2-132, 13-75, 2-53, 18-192, and both 12-89 and 1-74, respectively. Moreover, allele frequencies among different ethnic groups are not equal. For example, NUMTs 1-74 and 18-192 are poorly represented among Caucasians and Asians and are more frequent in non-pygmy Africans. NUMT 12-89, unlike other NUMTs, is poorly represented in non-pygmy Africans. As a result, each NUMT presents a unique population fingerprint. Acquisition of NUMTs by H. sapiens To evaluate which NUMTs are specific to humans, we amplified by PCR the 42 loci described above on chimpanzee DNA (Pan troglodytes). For each locus, one to three chimpanzee individuals were analyzed. Forty-two out of 42 primer pairs successfully amplified the target site also in chimpanzees because of the high sequence identity of the two genomes (average 98.7%) ( Chen et al. 2001 ). Only the regions flanking the previously described NUMT 11-541, which is considered separately in our investigation, did not amplify in chimpanzees. Locus 14-1023, where no NUMT was identified in humans, also showed no insertion in the chimpanzee. Surprisingly, only 14 loci contained the NUMT (see Table 1 ). All of these NUMTs were also found in all human individuals tested, indicating that they were present in the common ancestor of human and chimpanzee. On the contrary, 27 NUMTs absent from the chimpanzee genome represent recent acquisitions in H. sapiens. The distribution of these NUMTs in the human chromosomes is shown in Figure 3 . All NUMTs whose presence was not found in all human individuals fell in this category. From our data, 24 out of 27 NUMTs specific to humans had greater than 94% of sequence identity to the human mtDNA, and three out of 27 NUMTs had sequence identity of 91%–92%. This higher level of identity is consistent with the idea that NUMTs specific to humans are recent insertions (for the calculation of the insertion time of NUMTs, see Materials and Methods ). Similar values of identity to the mtDNA were also found for the recent insertions of mt sequences in the yeast genome ( Ricchetti et al. 1999 ). Seven out of 14 NUMTs present both in humans and in chimpanzees have lower levels of identity to the mtDNA (between 79% and 90%), as expected; however, the remaining seven NUMTs have 94%–96% identity to the mtDNA (see Table 1 ), indicating that the level of identity per se is not a rigorous criterion for human specificity. Interestingly, most of NUMTs present only in humans are short sequences, and about half of them are less than or equal to 100 bp. In summary, out of 211 NUMTs recognizable in the human genome, at least 27 (or 28, if we also include NUMT 11-541) were specific to humans, and we do not expect this value to increase significantly because 99% of the euchromatic genome of H. sapiens was analyzed, and we assume that most NUMTs with low identity to the mtDNA (less than or equal to 90%) are unlikely to be human-specific. This results in an average of one NUMT integration in the germline for each 180,000 y, in the last 4–6 million years (Myr). Interestingly, one fourth of these NUMTs (6/27, or 7/28 if we include NUMT 11-541) show insertion polymorphism (see Table 1 and above), indicating that they have occurred in more recent times. Figure 3 Distribution of Human-Specific NUMTs in Chromosomes A scale representation of the human chromosomes. The location of human-specific NUMTs is indicated with a red arrow. A green arrow indicates the position of NUMTs showing insertion polymorphism in humans, and a blue arrow indicates a previously described NUMT (11-541). High Frequency of Human-Specific NUMTs in Chromosomes 18 and Y The distribution of human-specific NUMTs in human chromosomes is not proportional either to the chromosome size or to the total number of NUMTs present in the chromosome ( Figure 4 ). In Chromosomes 13 and 20, human-specific NUMTs represent 37% and 50%, respectively, of the NUMT insertions detected in the chromosome. More surprisingly, in Chromosomes 18 and Y there is a proportionally higher number of human-specific NUMTs (2/3 present in each chromosome; see Figure 4 ), whereas at the genome-wide level about 13% (27/211) of NUMTs are specific to humans. The high number of human-specific NUMTs on the Y chromosome is particularly intriguing since this chromosome is 4-fold less present in the human population than the other chromosomes (it is the only haploid chromosome, present only in males). Additionally, the NUMT value for the Y chromosome may be an underestimate, since its large heterochromatic portion has not yet been sequenced. Since no more NUMTs are available to increase sampling size, we cannot formally distinguish between a founder effect and an increased insertion rate in Chromosomes 18 and Y above that in other chromosomes during recent human evolution. Figure 4 Human-Specific NUMTs in Human Chromosomes For each human chromosome, indicated on the x-axis, the number of NUMTs (y-axis, on the left) common to human and chimpanzee (white columns) and specific to humans (black columns) are shown. An open circle indicates the chromosome size in millions of base pairs (Mbp; y-axis on the right). NUMTs Mainly Integrate in Known or Predicted Genes The integration of NUMTs in the human genome takes place, surprisingly, mainly in known or predicted coding or regulatory regions. As indicated in Table 3 and Figure 5 , out of 28 human-specific NUMTs, 22 integrate in a known or predicted intron, one in an exon, and one in a promoter region. Only 4/28 NUMTs are in intergenic regions. This is also the case for older NUMTs, common to humans and chimpanzees, where 10/14 NUMTs are inserted in intron regions, and 4/14 in intergenic regions. All seven of the most recent integrations, those displaying insertion polymorphism, were found in exons or introns. In summary, about 80% of NUMTs are inserted in known or predicted introns/exons, which together should cover about 25% of the human genome ( Venter et al. 2001 ). Analysis of the position of NUMTs inside introns reveals that one NUMT, 12-89, was inserted exactly at the splice-donor site of the last predicted intron ( Figures 2 and 6 ). The other 31 NUMTs appear to be randomly integrated within the introns, although in two cases the insertion generates one or two new exons in the predicted proteins (NUMT 17-653 and NUMT 5-8781, respectively); see Figure 6 . Moreover, NUMT 1-74, which is the only NUMT found inserted within an exon, splits the last exon of the gene Q8N7L5 into two, and the NUMT itself becomes a new intron ( Figures 2 and 6 ). Thus, for at least four NUMTs, out of the 33 that are inserted in genes, the exon/intron pattern looked modified after integration of the sequence of mt origin, essentially by a change in the number of exons. We would expect these to be the most likely candidates to perturb gene function. Figure 5 Insertion Sites of NUMTs in the Human Genome Histogram of the insertion sites of NUMTs in the human genome. Only NUMTs tested in human and in chimpanzee samples are shown. This includes the 27 NUMTs specific to humans and absent from chimpanzees (21 present in all individuals tested and 6 with insertion polymorphism in humans), one additional NUMT with insertion polymorphism, previously described, see text, and 14 NUMTs common to human and chimpanzee, out of 183 found by BLAST search. Colors of the blocks indicate the different target sites. For details see Table 3 . Figure 6 Scheme Representing Some NUMT Insertions in Genes Four known or predicted genes, found in loci with NUMT insertion in humans, have been schematically represented either in the absence (A) or in the presence (B) of the insertion. Boxes represent exons, and thick lines represent introns. Red boxes and lines indicate the sequence corresponding to the NUMT, which has been identified for each case. A dotted line in (A) indicates that, in the absence of insertion, the exon/intron pattern was not identified by gene identification programs. Representation not to scale. Table 3 Insertion Sites of NUMTs in the Human Genome “Gene Reference” indicates the targeted gene, or the transcript code—hypothetical protein when based on prediction programs. Swiss-Prot indicates Swiss-Prot/TrEMBL. Data were obtained using http://www.ensembl.org/Homo_sapiens , http://us.expasy.org/sprot , http://genome.ucsc.edu/cgi-bin/hgBlat , http://genes.mit.edu/GENSCAN.html , and related sites. Detailed coordinates of the predicted genes are shown in Table S2. The last two columns indicate the organ(s) where the corresponding transcript was found and the phenotype associated with mutations in the gene (references in http://us.expasy.org/sprot and in Table S2) CL, colon; FB, fetal brain; PL, placenta; SM, skeletal muscle Twenty-one out of 33 NUMTs are inserted in predicted genes, and the other 12 in known genes, including a heart-specific serine protease and a thiamine transporter ( Table 3 ). Interestingly, three of the genes targeted by NUMTs with insertion polymorphism in humans are MADH2, a tumor-suppressor gene, mutated in colorectal carcinoma ( Eppert et al. 1996 ; NUMT 18-192); a gene coding for a homologue of the thrombospondin gene (an angiogenesis inhibitor that retards tumor growth; Bogdanov et al. 1999 ; NUMT 1-74); and a mt ribosomal precursor (NUMT 13-75; Table 3 ). For NUMT 1-74, which is inserted in an exon, and for NUMTs 18-192 and 13-75, both inserted in an intron, it is not known if individuals carrying the insertion are mutated for these genes. These findings suggest that in humans, the insertion of NUMTs is elevated in gene-containing regions of the genome. Insertions in such regions are potentially mutagenic. Interestingly, at least two cases of NUMT insertions—one in an exon, the other in an intron region—associated with diseases have been recently reported in humans ( Borensztajn et al. 2002 ; Turner et al. 2003 ). Discussion Integration of mt genes into the nuclear genome is a physiologically important process that contributes to the origin and evolution of the eukaryotic cell ( Margulis 1970 ), and the transfer of entire genes from mitochondria to the nucleus appears to be continually active in some plants ( Knoop et al. 1995 ). Although the transfer of entire genes seems to have ended in animals, DNA fragments of mitochondrial origin continue to integrate in the nuclear genome. In the present study, we examined the extent and the consequences of this process in humans and in chimpanzees. NUMTs Are Present in the Human Genome and Display Insertion Polymorphism The direct investigation of samples of different individuals provided in this study clearly demonstrates the presence of DNA fragments of mitochondrial origin in the nuclear genome of humans, as previously suggested by the analysis of the databases ( Mourier et al. 2001 ; Woischnik and Moraes 2002 ) and by a few tests in cells ( Tourmen et al. 2002 ). Although the presence of a single NUMT was previously shown in vivo ( Zischler et al. 1995 ), our study provides direct evidence that the large colonization of the human genome by NUMTs detected in silico, an outcome of the sequencing of the entire human genome, corresponds to the in vivo situation. We can thus exclude that, at least for the tested loci, NUMTs result from contaminations during the sequencing process, a situation that could not be previously ruled out formally (see comments in Venter et al. 2001 , and in Mourier et al. 2001 ). Our investigation of the distribution of NUMTs in human populations, which includes some of the less divergent among the 211 NUMTs, reveals that in most cases NUMTs are present in all the individuals tested, and therefore these sequences have colonized the nuclear genome of all major human populations. However, six NUMTs described here and one described previously ( Zischler et al. 1995 ; Thomas et al. 1996 ) are present only in some individuals. These seven NUMTs, not fixed within the human population, must have been recently acquired. Since they are present in individuals within each ethnic group, their insertion most probably occurred after the origin of modern humans and before the emergence of distinct ethnic groups (see also below). We did not detect NUMTs restricted to only one or a few ethnic groups. Furthermore, these seven NUMTs appear to have colonized the genome of human populations at different rates. Indeed, the frequency of alleles carrying the insertion varies greatly according to the NUMT (from 21% to 98% in our samples). This suggests that each sequence exhibits different colonization dynamics, involving the time of insertion and/or the expansion rate of the founder individual(s). Moreover, the distribution of each NUMT is unequal between ethnic groups, and a larger analysis of human populations will be necessary to reveal distinct population patterns and to perform phylogenetic studies. Nevertheless, most individuals tested had a unique combination of these seven NUMTs, suggesting that the individual pattern of NUMT insertion polymorphism can be useful as genetic fingerprints for familial pedigree studies. We expect that other NUMTs displaying such polymorphism will be discovered when larger population samples are examined. The locus 14-1023, which contains a NUMT according to the genome sequence, does not amplify a NUMT-containing fragment in all of 21 individuals. If this does not represent a sequencing artifact, it may be a further example of insertion polymorphism. Furthermore, we propose that the number of NUMT insertion polymorphisms is currently underestimated, since sequencing of the human genome was done only on a limited number of individuals ( Lander et al. 2001 ). Several independent markers are needed to accurately retrace the phylogeny of human populations ( Rosenberg et al. 2002 ), and insertion polymorphism is particularly interesting because of the low fixation rate and lack of reversion, unlike markers such as single nucleotide polymorphisms. Each of the six insertion polymorphisms described here is a rare event, and if a neutral genetic marker, it provides an important tool for tracing human dispersal. Insertion Rate of NUMTs in H. sapiens An important question concerning the integration of DNA sequences in the nuclear genome is their rate of colonization. For exogenous sequences like NUMTs, this has not been investigated in vivo. To determine the extent of colonization of a given genome, it is necessary to compare the insertions within this genome with those of a closely related species. To date, a comprehensive analysis of the presence of NUMTs has been done for several complete nuclear genomes ( Ricchetti et al. 1999 ; Mourier et al. 2001 ; Tourmen et al. 2002 ; Richly and Leister 2004 ), but the colonization rate of these sequences was not investigated, because of the absence of data in closely related species. In the case of H. sapiens, although a proportion of NUMTs present in its genome appears as ancient insertions (see Results; Tourmen et al. 2002 ; Bensasson et al. 2003 ), it is not clear how many NUMTs were inserted in primate ancestors and how many specifically colonized the human genome. Chimpanzee (P. troglodytes), a species closely related to humans and whose evolutionary relationship with humans has been widely investigated, is an ideal candidate for a comparative analysis. Moreover, the high level of identity (more than 98%) between the two species ( Chen et al. 2001 ; Fujiyama et al. 2002 ) allows the investigation of the respective genomes using molecular tools. No previous analysis in vivo showed the presence of NUMTs in the chimpanzee genome. By direct PCR amplification and sequencing of chimpanzee samples, we found that out of 41 NUMTs, 14 are also integrated in the chimpanzee genome (see above) and were therefore present in the common ancestor of humans and chimpanzees. However 27 NUMTs are absent from the chimpanzee genome, and are therefore recent acquisitions in H. sapiens. For NUMTs fixed in the human genome (not displaying insertion polymorphism), we do not expect this value to increase significantly, since our analysis was made on essentially the entire human genome. Among all NUMTs detected in the human genome, only about 13% (27 out of 211, or 28 if we include NUMT 11-541) are specific to H. sapiens and have integrated in the human genome in the last 4–6 Myr, after the split of the two species from their common ancestor ( Chen and Li 2001 ). This corresponds to an average of one integration in the germline each 180,000 y, or 5.4 insertions per Myr, a value remarkably close to that estimated by a phylogenetic analysis (5.1 insertions per Myr), which assumed a uniform insertion rate over time ( Bensasson et al. 2003 ). However, the rate of integration of the more recent NUMTs may not be consistent with a constant insertion rate. Indeed, out of 28 human-specific NUMTs, seven display insertion polymorphism and are present in all populations; thus they must have appeared early after the origin of modern humans. The date of this origin is still uncertain. If we assume that NUMTs with insertion polymorphism have been inserted at the same rate as the older NUMTs (fixed in the population), then they must have integrated in the genome of the human ancestor not earlier than 1.4 Myr ago, long before the origin of modern humans, and after the spread of Homo erectus out of Africa (1.7 Myr ago). Living humans would still be polymorphic for these NUMTs, as a result of interbreeding of the nonmodern human populations with modern humans ( Templeton 2002 ). On the contrary, if we assume that these insertions are more recent, as suggested by the poor allelic presence of most of them in present populations, then they must have appeared shortly before the expansion of modern humans, estimated at about 100,000 y ago ( Templeton 2002 ). In this case, their integration rate would be significantly higher than that of NUMTs inserted in the human genome in the previous 4–6 Myr. If the latter is the case, this strikingly high difference in the rate of colonization of the human genome may be due to a founder effect (i.e., the sporadic expansion of individuals carrying specific NUMTs) or, alternatively, to a genuine increase in the integration rate in modern humans. A third possibility is that there is no increase in the insertion rate in modern humans and that the number of “recent” NUMTs is overestimated because they include unfixed NUMTs that are destined to be lost eventually. In this latter case, we would need to assume that at least some of the NUMTs with insertion polymorphism are not neutral and are associated with a selectable phenotype. Although this may not be true for NUMTs 2-132 and 13-75, present in more than 95% of alleles tested, we cannot exclude that the low allelic presence (21%) of NUMTs 1-74 and 12-89 (both inserted in the context of an exon) is the result of the progressive counterselection of a defective phenotype; this would have implications for the mutagenic potential of NUMTs (see below). Compared to 28 NUMT insertions in the human nuclear genome in the last 4–6 Myr, it has been calculated that about 5,000 new insertion events of Alu repeats have occurred in the human genome in the same timescale (reviewed in Batzer and Deininger 2002 ). This large difference may be due to the fact that Alu elements are endogenous sequences that can be amplified by reverse transcriptase provided by long interspersed elements and inserted in the genome using L1 endonuclease ( Batzer and Deininger 2002 ), whereas the integration of NUMTs depends only on the availability of DSBs and of the repair machinery ( Ricchetti et al. 1999 ; Yu and Gabriel 1999 ). This suggests that retrotranscription/integration mechanisms increase the insertion efficiency of DNA sequences by two orders of magnitude. Alternatively, the limited number of NUMTs in the human genome may result from the selection process, if NUMTs preferentially integrate in coding regions (see below). Consequences of the Preferential Integration of NUMTs in Genes Contrary to previous findings, which indicated that NUMTs were inserted mostly outside annotated genes ( Woischnik and Moraes 2002 ), we find that NUMTs preferentially integrate in known or predicted genes. The availability of a more powerful database analysis on genome Web sites and the resulting increase in the number of potential new genes may explain this different evaluation. Unlike previous analyses, we investigated more recent insertions, frequently characterized by short sequences, which may have been missed in earlier studies. Moreover, we find that all of the most recent integrations, namely NUMTs with insertion polymorphisms, are integrated in genes. In cases where it was possible to identify the gene, its transcript was detected in one or more tissues ( Table 3 ). Among the targeted genes we found MADH2, a transcriptional modulator with tumor-suppressor properties ( Eppert et al. 1996 ), and a gene involved in microvessel development, and in both cases the NUMTwas present only in some individuals. In these cases it is not known whether the insertion has affected the function of the gene. Recent findings suggest that transcription promotes DNA breaks ( Gonzalez-Barrera et al. 2002 ). Insertion of NUMTs is, at least in yeast, dependent on DSBs ( Ricchetti et al. 1999 ), and in humans it is frequently associated with a hallmark of NHEJ, a DSB repair mechanism (our study). It is therefore possible that highly transcribed genes, perhaps carrying DSBs, are the preferential targets for the insertion of NUMTs. Only eight out of 41 NUMTs were found in intergenic regions, which should represent 75% of the human genome ( Venter et al. 2001 ). The Genscan program, which detected several insertion targets in our analysis, identifies around 20% more genes than previous estimates ( Das et al. 2001 ), but this does not significantly change the proportion of the genome that is noncoding. Approximately 80% of NUMTs are in coding regions, and we consider this to be a statistically significant event. Interestingly, this was not the case for yeast, where NUMTs integrated with 41-fold preference in intergenic regions ( Ricchetti et al. 1999 ). The intronless structure of the yeast genome may explain this difference, since NUMTs inserted in genes would essentially target exons in yeast and would be selected against if deleterious. In humans, the high content of introns would buffer most of the mutagenic potential of these insertions. Nevertheless, we expect that a fraction of insertions would be harmful also in humans. Although most of the analyzed NUMTs are internal to introns, in at least three cases the insertion modified the exon/intron pattern, and this may be mutagenic. Our analysis indeed confirms the recent finding that two NUMTs , occasionally found as new insertions in the human genome, and associated with diseases in humans, are inserted in genes, either in an exon or at the junction between exons and introns ( Borensztajn et al. 2002 ; Turner et al. 2003 ). Therefore, it is likely that future insertion events in the human genome would also preferentially target genes. An intriguing finding is that NUMT 12-89 is located exactly at the splice-donor site of the predicted intron. Insertion in a splice-related site was found also in human factor VII gene, where a 251-bp NUMT integrated a splice-acceptor site in a patient with severe plasma factor VII deficiency ( Borensztajn et al. 2002 ). Taken together, these results account for two insertions at intron-splice sites out of 45 NUMT insertion sites analyzed (42 NUMTs in our study and present in human populations and three more NUMTs found in one or more individuals and correlated with a disease; Willett-Brozick et al. 2001 ; Borensztajn et al. 2002 ; Turner et al. 2003 ). The limited sampling size does not permit us to determine if these finding are significant, although it is tempting to speculate that splice sites can be favored targets for the insertion of NUMTs. Is the rate of de novo insertions in the human germline limited to one each 180,000 y, or even ten times higher? The number of NUMTs detected as very recent insertions in one or a few individuals suggest that the insertion rate of these sequences in humans is currently dramatically underestimated ( Willett-Brozick et al. 2001 ; Borensztajn et al. 2002 ; Turner et al. 2003 ). Three new insertions of NUMTs have been found in living individuals, occasionally detected because of the search for the cause of a mutated phenotype. It seems reasonable to predict that a wider search would reveal many more NUMTs present in single or in small groups of individuals. Since NUMTs preferentially target genes, a fraction of these NUMTs could also be connected with diseases. Moreover, one expects that harmful insertions, whose probability increases as genes become preferential targets, would be subject to negative selection and thus removed from the gene pool. Thus the low fixation rate of NUMTs in the human genome may be a direct consequence of their preference for insertion in genes. NUMTs specific to the genome of H. sapiens and widespread in major human populations may represent only a small fraction of insertions that have occurred continually in the human genome. We propose therefore that the insertion of NUMTs, previously considered as functionless ( Perna et al. 1996 ; Hazkani-Covo et al. 2003 ), at best an evolutionarily important but essentially harmless process, is a potentially mutagenic process, challenging the functional integrity of the human genome. Remarkably, the integration of NUMTs in the nuclear genome can be accelerated under increased induction of DSBs. In the yeast nuclear genome, where only 34 NUMTs were detected, new NUMTs are integrated at an induced DSB site with a high frequency (10 −3 −10 −4 per repair event; Ricchetti et al. 1999 ). In keeping with this notion, a de novo insertion was reported on Chromosome 7 for a patient conceived during the Chernobyl nuclear meltdown ( Turner et al. 2003 ). By analogy to our previous findings in yeast ( Ricchetti et al. 1999 ), it is possible that this novel insertion is the consequence of a de novo DSB in the chromosome resulting from radiation exposure. Consistent with this view, a NUMT has been found inserted at the breakpoint junction of a familial constitutional reciprocal translocation, also associated with the occurrence of a DSB ( Willett-Brozick et al. 2001 ). The fixation rate and the insertion strategy used by NUMTs are probably the prototype for the integration of exogenous sequences in the nuclear genome. Like NUMTs, sequences lacking specific amplification and integration mechanisms would rely on occasional DSBs to integrate in chromosomes. Coding or transcribed sequences could represent the preferred insertion target for these sequences as well. This strategy is in sharp contrast with the integration procedure of retrotranscribed elements, which have successfully colonized the human genome and only rarely target coding regions ( Lander et al. 2001 ). In conclusion, we provide direct evidence that NUMTs are present in the human and in the chimpanzee genomes and that the insertion polymorphisms of six NUMTs reveal new markers for the study of human population genetics. Further, our in vivo analysis reveals that on average one new NUMT is fixed in the human genome each 180,000 y, although during the expansion of modern humans the fixation rate of NUMTs may have increased. The frequency of insertion of NUMTs may represent the genuine fixation rate of exogenous sequences colonizing the human nuclear genome. Strikingly, NUMTs preferentially integrate in introns and in exons, and they are thus potentially mutagenic, and novel NUMT integrations have been shown to be associated with diseases in humans. The recent case of de novo insertion of a NUMT following the Chernobyl accident, if not coincidental, provides a compelling example of how environmental insults can drive NUMTs to colonize the nuclear genome and induce genetic dysfunctions in humans. Materials and Methods BLAST search The human mtDNA sequence ( Anderson et al. 1981 ) was compared to the “ Homo sapiens genomic contig sequences” database version of April 11, 2003, using the National Center for Biotechnology Information (NCBI, Bethesda, Maryland, United States) “BLAST the Human Genome” server ( http://www.ncbi.nlm.nih.gov/genome/seq/page.cgi?F=HsBlast.html&&ORG=Hs ). The blastn program was used with default parameters on April 24, 2003. Only output parameters were changed to 1,000 descriptive lines and to 1,000 segment alignments. In a few cases (see Table 1 ) results of previous searches (July 2001 and January 2003) were also used. BLAST output results were saved locally in a text format and parsed using the readblastn script (see Tekaia et al. 2000 ), so that results were presented in a table format, including the query sequence, its size, the hit sequence, its size, the blastn E-value, the percent identity, the percent similarity, the matching segment size, and its coordinates on the query sequence as well as on the hit sequence. Only scores less than or equal to 10 −15 have been selected. Each selected sequence was further aligned with the mtDNA sequence using http://www.ncbi.nlm.nih.gov/blast/b12seq/b12.html . Amplification and sequencing of NUMTs from human and chimpanzee cells PCR amplification was performed on lysed cells originating from the buccal mucosa of healthy volunteers. Appropriate informed consent was obtained from human subjects. Pygmy samples (two Biakas and two Mbuti pygmies) were obtained as purified DNA from Coriell Institute (Camden, New Jersey, United States). Purified chimpanzee DNA, obtained either from tissues or from fecal material, was a kind gift from J.-P. Vartanian at the Pasteur Institute (Paris, France). For both PCR strategies described in the text, primer sequences are available upon request. Cell lysis was performed by incubating fresh cells overnight at 55 °C in a Tris-EDTA buffer (pH 8.5) in the presence of 200 μg/ml of proteinase K. PCR amplification was performed with 30 cycles of denaturation (1′ at 94 °C), annealing (1′ at 68 °C), and DNA synthesis (3′ at 72 °C) using Invitrogen (Carlsbad, California, United States) Taq polymerase. In heterozygous samples, a specific stochiometry of the two bands was found for each couple of primers used. Amplified NUMTs have been sequenced by specialized commercial services, using PCR amplification bands purified by gel extraction. Calculation of the insertion time of NUMTs The age of insertion of NUMTs was estimated using, as reference, the sequence divergence of the NUMT from the mtDNA. We assumed that the NUMT, when inserted into the nuclear genome was identical to the corresponding mt sequence. We also assumed that, once inserted into the nuclear genome, the NUMT mutated at the same rate as the nuclear genome, μN, which corresponds, for noncoding sequences, to 2.5 × 10 −8 mutations per nucleotide per generation, or 1.25 × 10 −9 mutations per nucleotide per year, assuming a generation time of 20 y ( Nachman and Crowell 2000 ). By comparison, the original sequence remaining in DNA is assumed to have undergone mutation at the rate, μM, of 1.7 × 10 −8 substitutions per nucleotide per year, excluding the D-loop ( Ingman et al. 2000 ). Thus, from the date of insertion (in Myr from the present) the sequence divergence between the NUMT and the cognate mitochondrial sequence is expected to be nearly the sum of mutations accumulated in each compartment (the possibility of compensation by two identical mutations is negligible given the limited divergence). It follows that the date of insertion, i, is given by i = d /(μM + μN), where d is the frequency of sequence divergence between the NUMT and present mtDNA sequence. As an example, for a sequence 300 bp long, 94% identity to mtDNA corresponds approximately to an insertion time of 3.3 Myr, and 96% to 2.2 Myr. Supporting Information Table S1 Sequence Analysis of NUMTs in the Human Genome (53 KB DOC). Click here for additional data file. Table S2 Coordinates of the Genes Where NUMTs Are Inserted in the Human Genome (63 KB DOC). Click here for additional data file. Accession Numbers The NCBI ( http://www.ncbi.nlm.nih.gov/genome/seq/page.cgi?F=HsBlast.html&&ORG=Hs ) accession number for the human mtDNA sequence is AB055387.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC515365.xml
387266	“Mosaic” Genes Highlight Forces of Genome Diversity and Adaptation	null	Microbes are arguably the most adaptable organisms on Earth, inhabiting nearly every crevice of nearly every corner of the globe. Some invade the cavities of a wide variety of insects and other invertebrates while others colonize the skin, blood, eyes, and internal organs of animals. Still others thrive in such inhospitable places as the hydrothermal vents of the ocean floor and the Dry Valleys of Antarctica. These “simple” single-celled organisms have evolved unique molecules and strategies over some 3.5 billion years that suit life on the edge. With the sequenced genomes of nearly 140 microbial species in hand, scientists are gaining valuable insights into the nature of this adaptive diversity. Segmentally variable genes Adapting to such radically different niches, it appears, has produced genes with diverse functions that evolve at very different rates. Genes that code for molecules essential for fundamental cellular functions like maintaining cell metabolism and structure tend to evolve rather slowly, while genes that make proteins charged with mediating cellular responses to internal or external changes often evolve relatively quickly. Pathogenic microbes in particular rely on a flexible genome to keep a step ahead of their hosts' similarly evolving defenses in the never-ending struggle to gain adaptive advantage. This adaptability underlies the increasing antibiotic resistance of diseases like tuberculosis, as selective pressures favor the expansion of resistant bacterial populations. Combating such problems requires a molecular understanding of bacterial infections, yet function has been ascribed to only a fraction of the genes found in microbial genomes. One approach to improve functional analyses of genome sequences combines bioinformatics with experimental methods. With such collaborations in mind, Yu Zheng, Richard Roberts, and Simon Kasif have developed a computational approach to help filter out the genetic noise and home in on genomic regions likely to contain clues to gene function. Their method relies on a novel way of classifying genes that flags sequences likely to reward biochemical and genetic efforts to analyze gene function. Many comparative genomic studies have focused on looking for sequence “motifs” that correlate with well-characterized protein sequences (that is, the amino acid sequence) and predicting function based on their similarity to the known protein sequences. Zheng, Roberts, and Kasif took a different approach, classifying genes based on their sequence variation. The researchers analyzed 43 fully sequenced microbial genomes and, after determining the degree of conservation or divergence among similar genes in different species, divided the genes into three broad categories: rapidly evolving genes unique to a particular species; highly conserved genes; and “segmentally variable,” or mosaic, genes. Stipulating that the boundaries between the categories are somewhat blurred, Zheng et al. define segmentally variable genes as regions that show a mosaic pattern of one or more rapidly evolving, variable regions interspersed with conserved regions. Based on evidence suggesting that retained variable regions tend to serve a function, the researchers predicted that these mosaic genes, with their highly variable, fast-evolving regions, would shed light on the forces that shape genome diversity and adaptation. For most of the microbes analyzed, mosaic genes accounted for about 8–20% of their genomes. Selecting several large families of mosaic genes, Zheng et al. explored the relationship between genes with known function and the structure of their variable regions. Noting an overabundance of particular functional categories in different species—such as signaling proteins that come into either direct or indirect contact with the cell's environment—the researchers speculate that the variable regions may constitute an adaptive layer for the microbe, as they not only “play a key role in mediating interactions with other molecules” but also support a microbe's ability to adapt to its particular niche. Several bacteria species, for example, contain roughly 40% more mosaic sensor genes involved in cell motility, which the authors attribute to the microbes' “expanded ability to detect different chemical signals and find favorable environments.” This regional variability appears to reflect the influence of selective pressures that fuel diversity through ongoing interactions with other rapidly evolving molecules in the environment, adding another source of genetic adaptability as cells adjust to new environments and outmaneuver pathogenic threats. While many of the mosaic genes identified encode proteins involved in host-pathogen interactions, defense mechanisms, and intracellular responses to external changes, their function is only broadly understood. While Zheng et al. cannot say to what extent variability affects function—Is extreme variability required for diversity or can modest variation suffice?—they are refining their classification of segmentally variable genes to address such questions. Until then, the authors' “mosaic” approach to understanding gene function promises to improve efforts to annotate the volumes of sequenced genomes on hand, offering biologists a much-needed tool to sift through the mountains of genomic datasets and identify promising targets for further study.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC387266.xml
526181	Computational Mapping of Complex Traits in the Mouse	null	The past few years have seen an explosion of data on the mouse genome, widely hailed as a guidebook to the genetic origins of human disease. Scientists are particularly interested in charting the location of SNPs—single nucleotide polymorphisms—throughout the genome. SNPs are DNA sequence variations, or mutations, that change a single nucleotide in the genome, replacing a cytosine (C) base, say, with thymine (T). Many SNPs, even those found within genes, have no functional effect. Others, however, can increase risk of specific diseases or alter a person's response to pathogens and drugs. Whether or not they are involved directly in a disease, SNPs are attractive markers for population studies aimed at identifying the multiple genes underlying complex diseases like diabetes and cancer. Extensive data exist on multiple inbred strains of mice linking their genetic makeup (genotype) to physical traits (phenotype), and scientists have used these data to guide investigations of gene function and disease. Many data have been gathered by crossing mouse strains and painstakingly analyzing their progeny, to tease out the relative contributions different genes make to pathogenesis. But these efforts take time. Investigators would greatly benefit from high-throughput methods to scan the mouse genome and flag markers for candidate disease genes. In 2001, Andrew Grupe et al. introduced an “in silico” (computational) approach to do that very thing. The method scanned mouse SNP data to home in on chromosomal areas regulating complex traits and reduce the time needed to analyze mouse disease models from “many months” to “milliseconds.” For a number of reasons, however, it wasn't clear whether in silico mapping could deliver on its promise. For one thing, the density of SNP maps was sufficient to provide meaningful markers for only a few mouse strains, and phenotype information was lacking for many strains. In a new study, Tim Wiltshire and colleagues have addressed these limitations by mapping nearly 11,000 SNP probes to 48 mouse strains. They have also been able to use this dataset for in silico mapping to predict genomic regions with functionally important phenotypes. Wiltshire and colleagues first show that their method can predict the genomic location of a Mendelian trait (controlled by a single gene), in this case coat color, which the authors acknowledge is a “minimum requirement for a viable in silico mapping method.” They go on to map complex “quantitative” traits (controlled by differential contributions from multiple genes at what are called quantitative trait loci, or QTLs)—gallstone development and plasma levels of high-density lipoprotein cholesterol—and find that their predictions fall in line with loci identified by traditional mouse disease studies. Noting a high correlation between QTLs predicted in silico and those identified experimentally, the authors argue that loci predicted using this method are “very likely to be biologically relevant.” In silico mapping for mouse genetics Wiltshire and colleagues are careful to point out that in silico mapping is meant to complement, not replace, traditional gene mapping models. After all, computers are no match for living organisms in modeling the subtleties inherent in biological reactions. But this approach is a good starting point for identifying significant genomic areas in a new strain. And as new strains are genotyped and phenotyped, and refinements are made to the SNP database, the robustness of this method should only get better.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526181.xml
387272	A Mechanism for Adding the First Link in a Nascent Actin Filament Chain	null	The capacity for self-generated movement is a defining characteristic of animal life. With the molecular components of cellular locomotion conserved in organisms from protozoa to vertebrates, directed cell motility appears to be an ancient cell process, likely dating back a billion years. Most directed motion relies on the assembly, or polymerization, of actin proteins into filaments. Actin is one of the most abundant proteins in cells; about half of the cellular concentration of actin is bound together in filaments at any given time while the other half floats freely as “monomers” in the cytoplasm. The erection and demolition of actin filaments directs the cell motility that lays down the remarkable million miles of nerve cells that form the nervous system and drives a variety of fundamental biological processes, from effective immune response to embryonic development. Mutations in proteins that regulate actin assembly can lead to the abnormal cell migration associated with metastatic cancer. The actin cytoskeleton also provides the structural support for animal cells that the cell wall provides for plants. Actin addition The molecular mechanisms underlying actin assembly and cell motility remained obscure until 1994, when Thomas Pollard and his colleagues discovered the protein complex that initiates actin polymerization. Called actin-related protein 2/3 (Arp2/3) complex, this molecular machine consists of seven subunits, including the two actin-related proteins. Free actin monomers are primed for rapid polymerization, but polymerization must be initiated by the Arp2/3 complex in a process referred to as nucleation. To nucleate a new filament, the Arp2/3 complex must be activated, a job accomplished by a family of proteins called WASP (after Wiskott Aldrich Syndrome, a genetic disease characterized by defects in platelet development and lymphocyte function). WASP proteins bind to both the Arp2/3 complex and an actin monomer. The Arp2/3 complex also binds two molecules of adenosine triphosphate (ATP) on the Arp2 and Arp3 subunits. ATP releases energy in a process called hydrolysis, which drives most energy-dependent processes, from actin polymerization to muscle contraction. The precise mechanisms governing Arp2/3 activation and nucleation are not known. Now Mark Dayel and Dyche Mullins show where hydrolysis occurs during this crucial first step in polymerization and use this finding to investigate the mechanisms that drive nucleation. In previous experiments, Dayel and Mullins found that Arp2/3 appears to require hydrolysable ATP to effect nucleation. To determine when and if ATP hydrolysis occurs on the Arp2/3 complex, Dayel and Mullins developed a technique that allowed them to analyze the Arp2 and Arp3 subunits separately. Dayel and Mullins discovered that hydrolysis occurs only on the Arp2 subunit of the complex and that it happens during the step when WASP initiates the nucleation of a new filament. The researchers then used ATP hydrolysis on Arp2 to dissect the mechanism by which WASP activates the Arp2/3 complex and develop a model of nucleation. (All previous techniques required actin polymerization to monitor the activity of the Arp2/3 complex, but this technique offers a way to decouple activation from polymerization.) They find that WASP proteins activate the Arp2/3 complex by coordinating its interaction with an actin monomer—the first monomer of the new filament. By developing a novel technique to monitor activation of the Arp2/3 complex, the authors contribute a new tool for further investigations of this central part of the cellular motility machinery. And by showing how Arp2/3 is activated, they offer important insights into the workings of a multiprotein cellular machine and the mechanisms that cells enlist to control their shape and motility—which could suggest potential drug targets to inhibit the abnormal cell movement characteristic of cancer and other diseases.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC387272.xml
529271	Practice patterns of naturopathic physicians: results from a random survey of licensed practitioners in two US States	Background Despite the growing use of complementary and alternative medicine (CAM) by consumers in the U.S., little is known about the practice of CAM providers. The objective of this study was to describe and compare the practice patterns of naturopathic physicians in Washington State and Connecticut. Methods Telephone interviews were conducted with state-wide random samples of licensed naturopathic physicians and data were collected on consecutive patient visits in 1998 and 1999. The main outcome measures were: Sociodemographic, training and practice characteristics of naturopathic physicians; and demographics, reasons for visit, types of treatments, payment source and visit duration for patients. Result One hundred and seventy practitioners were interviewed and 99 recorded data on a total of 1817 patient visits. Naturopathic physicians in Washington and Connecticut had similar demographic and practice characteristics. Both the practitioners and their patients were primarily White and female. Almost 75% of all naturopathic visits were for chronic complaints, most frequently fatigue, headache, and back symptoms. Complete blood counts, serum chemistries, lipids panels and stool analyses were ordered for 4% to 10% of visits. All other diagnostic tests were ordered less frequently. The most commonly prescribed naturopathic therapeutics were: botanical medicines (51% of visits in Connecticut, 43% in Washington), vitamins (41% and 43%), minerals (35% and 39%), homeopathy (29% and 19%) and allergy treatments (11% and 13%). The mean visit length was about 40 minutes. Approximately half the visits were paid directly by the patient. Conclusion This study provides information that will help other health care providers, patients and policy makers better understand the nature of naturopathic care.	Background The number of Americans using complementary and alternative medicine (CAM) rose from 34% to 42% from 1990 to 1997, with annual spending for CAM therapies in excess of $21 billion [ 1 ]. The majority of the research on the use of CAM services has focused on patients or consumers of CAM. Several patient characteristics have been found to be associated with increased CAM use including female gender, child-bearing age, above average income, above average education and diagnosis with a chronic or life-threatening condition [ 1 - 4 ]. Reasons why consumers seek care from CAM practitioners have also been identified. These include: the failure of conventional treatment to alleviate symptoms; psychological congruence between the individual's belief system and the CAM therapy; individuals' expectations that use of CAM will empower them by offering a greater sense of control over personal health care decisions; adverse effects of conventional therapies; and dissatisfaction with the care of conventional practitioners [ 1 - 3 , 5 - 8 ]. Although we now have some understanding of who seeks CAM care and why, relatively little is known about the actual practices of CAM practitioners including the type of patients they treat, the kinds of therapies they use, and the numbers of patients they see each day. In addition, it is unclear if the practice patterns of CAM practitioners vary by state. The objective of this paper is to describe and compare the practice of naturopathic physicians in two US states: Washington and Connecticut. Naturopathic medicine Naturopathic medicine is a system of health care, based on the teachings of Benedict Lust [ 9 ], with a primary goal of enhancing the individual's innate self-healing ability by employing a variety of natural and largely non-invasive healing modalities. In some states, naturopathic physicians are authorized to employ more invasive interventions (i.e., prescription drugs, and minor surgery). Naturopathic medicine is defined by the American Association of Naturopathic Physicians (AANP) as: ... a distinct system of primary health care – an art, science, philosophy and practice of diagnosis, treatment and prevention of illness. Naturopathic medicine is distinguished by the principles upon which its practice is based. These principles are continually reexamined in the light of scientific advances. The techniques of naturopathic medicine include modern and traditional, scientific and empirical methods [ 10 ]. Naturopathic physicians are trained as generalists with expertise in a variety of core treatment methods including nutrition, hydrotherapy, colonic irrigation, physiotherapy, naturopathic manipulation, botanical medicine, homeopathy, pharmacology and minor office surgical procedures. Some licensed naturopathic physicians are also trained in traditional Chinese medicine, acupuncture and Ayurvedic medicine as well as clinical specialties such as natural childbirth [ 9 , 11 - 14 ]. There are currently six North American schools of naturopathic medicine whose graduates are eligible to sit state licensing examinations: Bastyr University (Seattle, WA), the Canadian College of Naturopathic Medicine (Toronto, ON, Canada), National College of Naturopathic Medicine (Portland, OR), Southwest College of Naturopathic Medicine (Tempe, AZ), the University of Bridgeport College of Naturopathic Medicine (Bridgeport, CT) (has applied for accreditation candidacy with the Council on Naturopathic Medical Education (CNME)), and the West Coast Naturopathic Medical College (Vancouver, BC) (not accredited by the CNME) [ 9 ]. Accurate estimates of the number of naturopathic physicians practicing in North America are difficult to obtain. A recent estimate (based on data from licensing bodies) indicated that in 2000 approximately 1300 licensed naturopathic physicians were practicing in the United States and another 500 were practicing in Canada. Naturopathic medicine is a licensed health care profession in twelve US states (Alaska, Arizona, Connecticut, Hawaii, Maine, Montana, New Hampshire, Oregon, Utah, Vermont, Washington, California), Puerto Rico and four Canadian provinces (British Columbia, Manitoba, Ontario and Saskatchewan) [ 9 , 15 ]. In most states and provinces where naturopathic medicine is not regulated, individuals may practice similar therapeutic approaches and/or call themselves naturopaths (whether or not they have been trained at a school for naturopathic medicine) because the term naturopathic medicine is not a restricted term. The number of individuals practicing in unregulated jurisdictions is unknown. In South Carolina and Tennessee, it is illegal to practice naturopathy [ 9 ]. Legal regulation of naturopathic medicine varies by state and province. Connecticut and Washington have relatively similar (and representative of the other ten states) requirements for licensure including: Graduation from an accredited four-year naturopathic medical school, successful completion of licensing examination (normally NPLEX), the submission of completed application forms and paying the required fee. In addition, Washington State requires that individuals be of good moral character with no history of unprofessional conduct [ 9 ]. The legal scopes of naturopathic practice in Connecticut and Washington are similar and relatively broad compared to the other states that license naturopathic physicians. Naturopathic physicians in these two states can diagnose and treat disease, utilizing a wide range of modalities and specialties including: minor surgery such as suturing (limited in Washington); physical therapies including hydrotherapy, naturopathic manipulation, and physiotherapy; colonic irrigation; electrotherapy (for example TENs); diagnostic x-ray; venipuncture; obstetrics (in Washington a midwifery license is required); gynecology; botanical medicine; nutrition; and homeopathy. Naturopathic physicians in Washington may prescribe a limited range of drugs; however, this is not part of the scope of practice in Connecticut. The practice of acupuncture requires separate licensure in both states [ 9 ]. Methods Study design The data derive from a parent study of licensed acupuncturists, chiropractors, massage therapists and naturopathic physicians [ 16 ]. The study was conducted in two phases: 1) random samples of licensed naturopathic physicians were interviewed by telephone; 2) a sub-set of those interviewed were recruited to record detailed information on 20 consecutive patient visits. Data were collected for each practitioner group in two states (one in the West and one in the Northeast). The West and Northeast were selected because these are the regions where CAM practitioners are concentrated in the United States [ 14 , 17 ]. The data for naturopathic physicians were collected in Washington and Connecticut. Sample Licensing Boards provided contact information for all licensed naturopathic physicians in Washington (1998) and in Connecticut (1999) with in-state addresses. In total, 142 licensed practitioners were identified in Washington and 63 licensed practitioners were identified in Connecticut. Providers without identifiable telephone numbers, and those not currently practicing were excluded. The proportion of excluded providers was 11% in Connecticut and 29% in Washington. See Table 1 for additional sampling details. Table 1 Selection and participation of naturopathic physicians (NDs) in Washington (1999) and Connecticut (1998) Connecticut Washington NDs licensed in state 71 286 NDs randomly selected for study 71 200 NDs found eligible for study* 63 142 Eligible NDs interviewed 59 111 Interviewed NDs eligible to collect visit data** 55 93 Eligible NDs providing visit data 34 65 Total patient visits reported 631 1186 * NDs confirmed to be practicing in state and to have verifiable and functioning phone number ** NDs who reported seeing 10 or more patients in a typical week To maximize the accuracy of state-wide estimates, data collection efforts were concentrated on high-volume practitioners: those with at least 20 patient visits per week. About 60% to 70% of practitioners had a high-volume practice and accounted for roughly 85% to 90% of all visits to the profession. The remaining practitioners were categorized as low-volume providers. While all high-volume practitioners were asked to collect data on 20 consecutive patient visits, only the first 10 low-volume practitioners were asked to collect data. The rationale was to collect only enough data from low-volume practitioners to ascertain whether their practices differed markedly from those of high-volume practitioners. It was ultimately decided, however, to weight data for high- and low-volume practitioners in a manner that produced annual estimates of visits in each state. Naturopathic physicians who agreed to participate were asked to complete one-page encounter forms for 20 consecutive patient visits. No financial incentives were provided for participation in the study and the protocol was approved by the Group Health Human Subjects Review Committee and the Harvard Pilgrim Health Care "Committee on Clinical Investigations, New Procedures and New Forms of Therapy" prior to the commencement of data collection. Data collection Research assistants conducted telephone interviews with the randomly selected providers in Washington State in 1998 and in all eligible practitioners in Connecticut in 1999. (Given the small number of eligible practitioners in Connecticut, all were contacted). Information about individual demographic characteristics (e.g., age, gender, race/ethnicity); training characteristics (e.g., duration, location); practice characteristics (e.g., length of time in practice, licensure in other health care professions, practice arrangement); and workload characteristics (e.g., number of weeks in practice per year; hours per week of direct patient care) was obtained for each provider who agreed to be interviewed. Among those eligible to be interviewed, the participation rate was 94% in Connecticut and 78% in Washington. Naturopathic physicians who were eligible to collect visit data (i.e., those who saw at least 20 visits per week plus a sample of those who saw at least 10 – 19 visits per week) were asked to complete encounter forms for 20 consecutive patient visits. (Only 2% of all visits were provided by naturopathic doctors with less than 10 visits per week.) Those who agreed were mailed encounter forms that were modeled after those used by the National Ambulatory Medical Care Survey (NAMCS) [ 18 ]. Each naturopathic physician was instructed to return the encounter forms by mail to the study center after completing them. Those who failed to promptly return their forms received a reminder telephone call. Among those eligible to collect visit data, 62% of naturopathic doctors in Connecticut and 70% of naturopathic doctors in Washington provided visit data. Analysis Descriptive statistics (means, standard deviations, frequencies) were used to present the practitioner characteristics in Table 2 . In the analyses of visit characteristics (Tables 4 and 5 ), each visit in the sample was weighted by the inverse of the sampling probability, reflecting both the chance that the particular provider participated and the estimated proportion of that provider's annual visits included in the study. Consequently, our results represent estimates of all visits made to naturopathic physicians in each state except for the roughly 2% of visits made to practitioners with the lowest volumes [ 16 ]. Table 2 Characteristics of naturopathic physicians licensed in Connecticut (1999) and Washington (1998) Connecticut (n = 59) % Washington (n = 111) % Demographics Female 58 57 White 95 94 Mean Age (SD) 43.6 (8.7) years 44.1 (9.2) years Education Institution: Bastyr University 37 80 National College 59 20 Other 3 1 Specialty training # 69 42 Licensed in other Health Profession 17 (acupuncture 10%; all others less than 2%) 33 (chiropractic 8%; nursing 7%; acupuncture 6%; midwifery 5%; all others less than 2%) Practice solo 51 54 significant differences between the two states, chi square, p < 0.05) # significant differences between the two states, t-test, p < 0.05) Table 4 Characteristics of visits to naturopathic physicians in Connecticut (1999) and Washington (1998) Connecticut (n = 631 visits) % Washington (n = 1186 visits) % Gender female 76 74 Age Median 43 years 42 years 0–15 years 12.8 10.9 16–34 years 16.4 20.9 35–64 years 63.0 58.5 65 years+ 7.8 9.7 Race White 97 96 Smoking status: non-smokers 95 94 Type of major problem a Acute problem 22 23 Chronic problem, routine 53 56 Chronic problem, flare up 20 18 Pre/post-surgery/injury 1 1 Non-illness care 5 6 Primary reason for visit b Fatigue 6.1 6.0 Headache 4.4 3.8 Back symptoms 4.4 6.5 Skin rashes 4.2 2.5 Menopausal symptoms 3.7 2.0 Bowel function changes 3.5 2.0 Anxiety or Depression 3.2 5.1 Allergies to food, milk 3.0 3.6 Sinus symptoms 2.8 <2 Upper respiratory symptoms 2.6 <2 Cough 2.4 <2 Neck symptoms 2.3 3.3 Infectious disease 2.2 <2 Ear infection symptoms 2.2 <2 Abdominal pain/cramps 2.2 2.7 Menstrual problems <2 2.1 Diagnosis by naturopathic physician c Menopausal disorders 4.9 3.1 Allergies 4.6 6.3 Back conditions 3.5 5.8 Fibromyalgia 3.0 2.0 Sinusitis 2.9 <2 Fatigue 2.4 2.7 Headache 2.2 3.1 Upper respiratory infections 2.0 <2 Neck conditions <2 3.3 Depression <2 2.6 Asthma <2 2.2 Anxiety <2 2.1 Dermatitis <2 2.0 New patients 22 22 Payment Private insurance 60 43 Worker's compensation 0 3 Personal injury 0 3 Self-pay 37 48 Other 3 3 a Because 3% to 4% indicated multiple reasons, total percentages exceed 100% b Patient report; only reporting reasons totaling 2% or more of visits c Based on ICD 9 criteria; only reporting diagnoses totaling 2% or more of visits Table 5 Characteristics of visits to naturopathic physicians in Connecticut (1998) and Washington (1999) Connecticut (n = 631 visits) % Washington (n = 1186 visits) % Diagnostics/Screen Services : Examinations Vitals (BP, pulse, temp) 28 39 HEENT 18 15 Complete physical 13 9 Mental Status <5 6 Imaging x-ray 1 2 ultrasound 1 1 Blood tests Complete blood count 7 10 Serum chemistry 7 9 Thyroid 3 7 Lipids panel 4 5 Allergy 4 2 Additional Tests Stool analysis 5 4 Urine analysis 4 2 Vitamin/Mineral 3 0 Endocrine 2 3 Allergy Skin test 1 1 TB skin test 1 0 Therapeutics and Preventive Services: Naturopathic Therapeutics Botanical medicine 51 43 Vitamins 41 43 Minerals 35 39 Homeopathy 29 19 Acupuncture 14 4 Allergy treatment 11 13 Glandular therapies 4 13 Physical Therapies Naturopathic manipulation 8 15 Physiotherapy 1 13 Hydrotherapy 4 10 Ultrasound 2 9 Mechanotherapy 2 7 Counseling/Education Therapeutic diet 26 36 Self-care education 17 23 Exercise therapy 9 12 Mental Health 4 6 Visit disposition No follow-up planned 5 4 Return if needed 18 21 Return at specified time 77 74 Referred (% to an MD) 5 (4) 6 (4) Visit Duration Mean, minutes 40 44 <15 minutes 1 3 15–29 minutes 14 20 30 to 44 minutes 49 31 45–59 minutes 15 18 >/=60 minutes 20 28 * Totals add to more than 100% because multiple responses were allowed Given the large sample sizes (631 and 1186), the weighted percentages presented in the tables have small standard errors, generally between 0.5 and 2.5 percentage points and rarely exceeding 3 percentage points. As a result, moderate to large differences between the states are also statistically significant and we do not include standard errors in the tables. Results Naturopathic physician participants The demographic characteristics of the naturopathic physicians in Washington State and Connecticut were remarkably similar (see Table 2 ). Just over half were female, with mean ages of about 44 years. Over 90% of naturopathic physicians identified themselves as White. The majority of practitioners (80%) in Washington State trained at Bastyr University in Seattle, while most practitioners in Connecticut trained at National College of Naturopathic Medicine in Portland, Oregon. Although naturopathic doctors in Connecticut were more likely to report advanced specialty training (such as homeopathy), those in Washington were more likely to be licensed in another health profession. Washington naturopathic physicians were most often also licensed in chiropractic (8%), nursing (7%), acupuncture (6%) and midwifery (5%), while Connecticut naturopathic physicians were most often also licensed in acupuncture (10%). Almost one-quarter (22.5%) of naturopathic doctors licensed in Washington had completed at least one year of post-graduate residency training compared with only 8.5% of those in Connecticut. (Residency training is not required for licensure in either state.) Both groups graduated from 4–5 year accredited full time programs of naturopathic medical education and the resulting naturopathic practice characteristics were reported to be very similar across the two states (Tables 2 and 3 ). Just over half of the practitioners in both states practiced solo, averaged approximately 25 hours per week providing direct patient care and saw an average of just over 30 patient visits per week. Practitioners who reported seeing less than 10 patients per week were excluded from this study. Table 3 Practices of naturopathic physicians licensed in Connecticut (1999) and Washington (1998) Connecticut (n = 59) Washington (n = 110) Median number of years in practice 9 7 Patient care hours in a typical week Mean (S.D.) 25.8 (10.6) 24.5 (12.2) Percentage reporting: 1–14 hours 10.2 20.0 15–24 hours 32.2 30.0 25–34 hours 39.0 26.4 35–44 hours 15.2 17.2 45+ hours 3.4 6.4 Patient visits in a typical week Mean (S.D.) 32.7 (18.9) 30.5 (24.7) Percentage reporting: 1–19 visits 20.7 34.6 20–29 visits 31.0 21.8 30–39 visits 13.8 21.8 40–49 visits 13.8 9.1 50+ visits 20.7 12.7 Characteristics of patients who see naturopathic physicians As was the case for the practitioners, patients of naturopathic physicians in the two states were remarkably similar (Table 4 ). About 75% of patients were female, about 95% were White and non-smokers, and the mean age of the patients was 41 years. Just over 50% of visits were for chronic conditions, followed in frequency by acute problems and flare-ups of chronic problems. There was extensive overlap in the most common presenting complaints (based on ICD 9 classification) in Connecticut and Washington with fatigue, headache, back symptoms among the top four in both states. However, there was a wide variety in the presenting complaints recorded; the most common presenting complaint was identified by less that 7% of patients in both states. More than one of every five visits was made by a new patient. Overall, approximately half the visits were paid for by private insurers and a similar percentage were paid directly by the patient. Visits in Connecticut were slightly more likely to be covered by insurance (60% vs 49% of visits). Characteristics of naturopathic medical visits More than 70% of visits included examinations or the ordering of diagnostic or screening tests (Table 5 ), most often: assessment of vital signs (including blood pressure, pulse and temperature), and Head, Eyes, Ears, Nose, Throat (HEENT) assessment (typically done for upper respiratory symptoms). The most common blood tests ordered in both states were: complete blood counts and serum chemistry. Imaging and other types of diagnostic tests were used for no more than 5% of visits in both states. Therapeutic and preventive modalities were used in most visits (96%) in both states. The most common naturopathic therapeutics used in both states were: botanical medicines, vitamins, minerals, homeopathy and allergy treatments (Table 5 ). In addition, glandular therapies were prevalent in Washington (13% of visits) and acupuncture was prevalent in Connecticut (14% of visits). The only physical therapies that are part of the naturopathic scope of practice used in more than 5% of the visits were naturopathic manipulation, physiotherapy (Washington only) and hydrotherapy (Washington only). Twenty-six per cent of visits in Connecticut included counseling or education compared with 36% of visits in Washington. Most often this included discussion of a therapeutic diet. Three-quarters of all visits in both states resulted in a recommendation that the patient return at a specified time (Table 5 ). The mean length of visit in the two states was similar (40–44 minutes), but there was greater variability in the length in Connecticut. A wide range of generally similar commercially packaged products (including herbs, vitamins and minerals) were used by naturopathic physicians in the two states (Table 6 ). Connecticut providers recommended about 50% more than Washington providers. Almost all of the top 15 commercially packaged products used in both states were vitamins and minerals as single agents or as combinations. The most common commercially packaged products were about 1.5 to 2.0 times more likely to be used in Connecticut compared to Washington. In addition, zinc and vitamin B6 were more likely to be used by NDs in Connecticut, while protomorphogens and vitamin B12 were more commonly used in Washington. (A protomorphogen is thought to be that component of the cell chromosome that is responsible for morphogenic (forming of body and organs) determination of cell characteristics. Theories about what protomorphogens are, as well as their role in health and disease continue to be debated.) Table 6 Most common commercially packaged products used by naturopathic practitionersin Connecticut (1998) and Washington (1999) Connecticut (n = 631 visits) Washington (n = 1186 visits) Number of Products Used per visit Range 0 to 28 0 to 26 Median 3.0 2.0 Mean 4.1 2.7 Most Common Commercially Packaged Products a (weighted percent of visits) Multivitamins 26.7 9.1 Digestive treatments 22.3 13.2 Magnesium 16.3 9.1 Combination vitamin and mineral 16.2 10.0 Calcium 15.8 9.3 Vitamin C 15.1 10.2 Minerals, NOS b 9.4 7.7 Zinc 9.0 <5 Multiminerals 8.4 7.8 Vitamins, NOS b 8.4 10.5 Vitamin E 7.9 5.2 Bioflavonoids 7.1 5.9 Vitamin B6 5.9 <5 Coenzyme Q10 5.8 <5 Selenium 5.2 <5 Protomorphogen <5 9.0 Vitamin B12 <5 7.4 a Excluding diets and foods; reporting only those with a weighted per cent >5% b NOS = Not Otherwise Specified Discussion Given the broad scopes of practice of naturopathic medicine in Washington and Connecticut, the similarities in the practitioner characteristics, patient characteristics and practice characteristics in Washington State and Connecticut are remarkable. The finding that naturopathic medicine is primarily a white and female profession is consistent with other recent studies of naturopathic physicians in the United States and Canada [ 19 , 20 ]. Substantial fractions of naturopathic doctors, especially in Washington, are licensed in other CAM and conventional professions. Practitioners spend about 25 hours per week in patient care and see approximately 32 patients/week, which is slightly more than was reported in a recent Canadian study [ 20 ]. In comparison, according to the American Medical Association, general practitioners in the United States spend approximately 51.3 hours/week providing direct patient care and see approximately 125 patients in a typical week [ 16 ]. Patients who visit naturopathic physicians are primarily female, white, middle aged, non-smokers. In many ways they are demographically similar to the naturopathic physicians themselves. Patients present to naturopathic physicians with a broad range of complaints and are diagnosed by the practitioners with a wide variety of conditions. It should be noted that much of naturopathic practice is focused on chronic problems, particularly those exclusively or primarily presented by women (e.g., fatigue, headache, symptoms associated with menopause, depression). In fact 75% of all visits are made by women. The most common "reasons for visits" identified in this study were very similar to those noted in a Canadian study [ 20 ], suggesting that this sample of provider visits may be representative of naturopathic patient visits across North America. Other than physical examination, blood tests and stool analysis, naturopathic practitioners use few diagnostics tests. However, they commonly recommend commercially packaged natural products such as herbal medicines, vitamins and minerals. Naturopathic practice also involves significant amounts of patient counseling and education. Overall, naturopathic physicians spend more than twice as much time with patients as conventional physicians at each visit (40 minutes vs. 14 minutes) [ 16 ], permitting more time to discuss patients' concerns and counseling/education about lifestyle issues such as diet. Data from this and previous studies [ 19 , 20 ] indicate that patients visit naturopathic physicians for a wide range of conditions and are given a correspondingly wide range of therapies and treatments as part of their naturopathic care. This diversity makes it difficult to describe a "typical" naturopathic medical consultation and to develop standardized research protocols for studies evaluating naturopathic treatments. Our results also have implications both for naturopathic physicians and the American health care system. They indicate that naturopathic physicians treat a broad range of health problems. The results from our study provide baseline data for tracking changes as the practice of naturopathic medicine evolves over time across the US in response to increasing research into the safety, efficacy and cost-effectiveness of naturopathic treatments, as well as to changes in the social, economic and regulatory environment. Strengths and limitations of the study The major strengths of this study are its large sample size, relatively high participation rates, and the ability to compare naturopathic physicians from two geographically distant states. The main limitation is that despite the similarities of the states studied, it is not known if the results are representative of the other ten states where naturopathic medicine is licensed. These two states reflect typical licensing requirements prevalent throughout the US. Washington State has one of highest numbers of licensed naturopathic physicians in the US, while the number of naturopathic physicians in Connecticut is significantly smaller, and more representative of, most other licensed states. Another limitation of the survey was its inability to accurately capture the use of office-compounded tinctures and powders. Our results highlight the use of bottled products as a major part of naturopathic practice; however, the practitioners' use of their own combination preparations was difficult to record on the data collection forms. This aspect of practice is likely under-reported in our data. Conclusion Naturopathic physicians and their practices in Washington and Connecticut, two geographically distinct states, are similar. This study provides baseline data for describing and tracking the practice patterns and scope of practice of naturopathic physicians over time and will help inform researchers and policy makers considering the regulation of this profession in other states. It also helps conventional physicians, other health care providers and patients better understand the nature of naturopathic care. Competing interests The authors declare that they have no competing interests. Authors' contributions HB conceptualized and drafted the manuscript. DC conceptualized the overall project, designed and directed collection of the original data and participated in the conceptualization of this manuscript, and analysis of this data. KS participated in the design of the overall project and this manuscript, and participated in the statistical analyses. JE helped to conceptualize the overall project, directed the data collection, quality control, and participated in the analyses for this paper. BM, JB, EC, and DE helped to conceptualize the overall project including the data collection. MS helped to conceptualize this paper. RD helped to conceptualize the overall project, design data collection procedures, and secure funding. All critically edited drafts of this manuscript and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC529271.xml
516042	Binding of estrogen receptor with estrogen conjugated to bovine serum albumin (BSA)	Background The classic model of estrogen action requires that the estrogen receptor (ER) activates gene expression by binding directly or indirectly to DNA. Recent studies, however, strongly suggest that ER can act through nongenomic signal transduction pathways and may be mediated by a membrane bound form of the ER. Estradiol covalently linked to membrane impermeable BSA (E 2 -BSA) has been widely used as an agent to study these novel membrane-associated ER events. However, a recent report suggests that E 2 -BSA does not compete for E 2 binding to purified ER in vitro . To resolve this apparent discrepancy, we performed competition studies examining the binding of E 2 and E 2 -BSA to both purified ER preparations and ER within intact cells. To eliminate potential artifacts due to contamination of commercially available E 2 -BSA preparations with unconjugated E 2 (usually between 3–5%), the latter was carefully removed by ultrafiltration. Results As previously reported, a 10-to 1000-fold molar excess of E 2 -BSA was unable to compete with 3 H-E 2 binding to ER when added simultaneously. However, when ER was pre-incubated with the same concentrations of E 2 -BSA, the binding of 3 H-E 2 was significantly reduced. E 2 -BSA binding to a putative membrane-associated ER was directly visualized using fluorescein labeled E 2 -BSA (E 2 -BSA-FITC). Staining was restricted to the cell membrane when E 2 -BSA-FITC was incubated with stable transfectants of the murine ERα within ER-negative HeLa cells and with MC7 cells that endogenously produce ERα. This staining appeared highly specific since it was competed by pre-incubation with E 2 in a dose dependent manner and with the competitor ICI-182,780. Conclusions These results demonstrate that E 2 -BSA does bind to purified ER in vitro and to ER in intact cells. It seems likely that the size and structure of E 2 -BSA requires more energy for it to bind to the ER and consequently binds more slowly than E 2 . More importantly, these findings demonstrate that in intact cells that express ER, E 2 -BSA binding is localized to the cell membrane, strongly suggesting a membrane bound form of the ER.	Background For many years, estrogen actions have been presumed to be mediated almost exclusively through the regulation of target gene transcription by a chromosomal bound estrogen receptor. These genomic estrogen effects are the well described interactions between the estrogen receptor and adapter transcription factors that result in activation or inhibition of the basal transcription protein machinery. However, there is a growing body of evidence that several rapid estrogen effects are non-transcriptional in nature. These rapid estrogen effects include changes of calcium flux in several cell types [ 1 - 3 ], MAPK activation [ 4 , 5 ], cAMP levels [ 6 , 7 ], and nitric oxide release [ 8 ]. That many of these effects are mediated by a membrane-localized estrogen receptor has been postulated for some time [ 9 , 10 ], but the majority of evidence supporting this hypothesis is indirect, relying on the induction of these non-genomic effects using estrogen covalently conjugated to BSA by a 6 atom hydrocarbon tether (E 2 -BSA) [ 11 , 12 ]. However, the relative binding efficiency of these conjugates is low and concern has been raised regarding the use of these conjugates as direct surrogates for estrogen [ 13 ]. A recent report added to this controversy by showing that commercially available E 2 -BSA is contaminated by unconjugated free E 2 and a series of binding experiments demonstrated that E 2 -BSA was unable to bind to ER after the contaminant E 2 was removed. [ 14 ]. These findings contradict studies where fluorescein-labeled E 2 -BSA (E 2 -BSA-FITC) specifically bound to a putative ER on the cell membrane [ 15 - 17 ]. Elucidation of novel membrane-associated ER effects is crucial to our understanding of the non-genomic signaling pathways of ER and other hormone receptors. Hormone-conjugated BSA is an important tool in this pursuit. We believe the contradictory results are explained by differences in the rates of binding of the bulky E 2 -BSA and E 2 with the ER. We show that pre-incubation of E 2 -BSA with ERα results in a highly significant decrease in the binding of 3 H-E 2 . The binding of 3 H-E 2 with ERα is unaffected by the simultaneous addition of E 2 -BSA. We also demonstrate that fluorescein conjugated E 2 -BSA binds to the membrane of cells that endogenously produce ERα and to HeLa cell lines stably expressing mERα. Results E 2 -BSA binding to purified estrogen receptor Although E 2 is covalently attached to BSA using a relatively long six atom hydrocarbon tether, the bulky BSA moiety of E 2 -BSA still may be interfering with the binding between the estrogen molecule and the estrogen receptor. This would result in an increase in the energy of activation required for E 2 -BSA binding. If so, increasing the reaction time would allow for the establishment of an equilibrium between bound and free forms of E 2 -BSA, maximizing the amount of E 2 -BSA bound to the receptor. To test this hypothesis, E 2 -BSA free of contaminant E 2 was prepared by ultrafiltration. Competition between the purified E 2 -BSA and labeled E 2 for binding to purified ERα was determined after E 2 -BSA was pre-incubated with ERα and also when added at the same time as labeled E 2 . As shown in figure 1 , concurrent addition of labeled E 2 and E 2 -BSA had no effect on labeled E 2 binding. However, a four-hour pre-incubation of E 2 -BSA with ER significantly decreased E 2 binding. These results suggest that the large BSA molecule retards, but does not prevent binding of E 2 -BSA. Figure 1 Pre-incubation of purified hERα with E 2 -BSA competes for estradiol binding. Purified ERα was incubated with E 2 (solid line) or E 2 -BSA (dotted line) for four hours before (a) or concurrently with (b) the addition of labeled E 2 . Incubation was continued for another 2 h at room temperature, and at the end of this period, specific binding was determined by adsorption, removal, and counting of free labeled E 2 . E 2 -BSA binding to ER in intact cells Non-genomic actions of the estrogen receptor are now well established. Several investigators have demonstrated that fluorescein labeled E 2 -BSA (E 2 -BSA-FITC) binds to the cell membrane, suggesting that a form of the estrogen receptor is present within the cell membrane and capable of binding to extracellular E 2 . Specific binding of E 2 -BSA-FITC to this membrane-localized form of the ER would further establish that E 2 conjugated to the BSA molecule is capable of binding to the ER. To examine this possibility, E 2 -BSA-FITC binding studies were performed with MC7 cells that contain endogenous ERα and with ER-deficient HeLa cells stably transfected with the ERα (HeLa-ERα). Expression of ERα within the HeLa cells was established by demonstrating specific binding of labeled E 2 to HeLa-ERα, but not native HeLa cells (figure 2 ). Scatchard analysis of the binding of E 2 to HeLa-ERα cells showed that although weakly expressed, the Kd for the expressed ERα was 7.04 nM, similar to published values (figure 3 ). HeLa-ERα cells, but not native HeLa cells, exhibited fluorescent staining of the cell membrane after incubation with E 2 -BSA-FITC (figure 4 ). The heterogeneous staining pattern reflected the low level of ERα expression. This fluorescence was not seen when HeLa-ERα cells were incubated with BSA conjugated to fluorescein alone (data not shown). Figure 2 Whole cell binding of estradiol to MC7 cells and HeLa cells stably transfected with ERα. 2 × 10 6 cells were incubated with 3 H-17β-estradiol (10 -8 M) in the absence (solid) and presence (white) of 100 fold excess of unlabeled estradiol for 15 minutes at room temperature, washed, and placed on ice for 30 minutes. Cells were then pelleted, lysed and counted. Results are expressed as the mean +/- the SEM of 3 experiments (* p < 0.01) Figure 3 Estradiol binding to HeLa cells stably transfected with ERα. Subconfluent HeLa-ERα cells were trypsinized and aliquots (2 × 10 6 cells) incubated with several concentrations of 3 H-17β-estradiol in the presence and absence of a 200-fold excess of cold 17β-estradiol for 30 min at 37°. Cells were then incubated on ice for 15 min, washed three times with 2 ml of ice cold 0.2% BSA-saline and pelleted by centrifugation at 1,5000 rpm for 10 min at 4°C. Cells were lysed by the addition of 100 ul of lysis buffer, vortexed and counted. a) Representative binding results of 3 independent experiments with total binding (solid box), non-specific binding (open box), and specific binding (triangle). b) Scatchard analysis of binding results. Figure 4 Membrane localization of ERα. HeLa cells stably transfected with ERα(a) and native HeLa cells (b) were incubated with fluorescein labeled membrane impermeable BSA conjugated to estradiol (E 2 -BSA-FITC) and visualized under phase contrast bright field and with UV light with an excitation filter for FITC. To establish the specificity of E 2 -BSA binding, MC7 cells and HeLa-ERα cells were incubated with E 2 -BSA-FITC after pre-incubation with various concentrations of E 2 and the anti-estrogen ICI-182,780. As shown in figure 5 , fluorescence was lost in both cell types in a dose dependent manner with increasing concentrations of E 2 . Fluorescence was almost completely eliminated by pre-incubation with the specific competitor ICI-182,780. BSA conjugated to FITC alone did not bind. These results suggest that estrogen covalently bound to BSA can bind to ER in a biologically significant manner. Figure 5 E 2 competition with E 2 -BSA-FITC binding. MC7 cells (black bars) or HeLa-ERα (white bars) were incubated with vehicle, various concentrations of E 2 , or ICI 182,780 (10 -8 M) for 30 minutes and then incubated an additional 30 minutes with E 2 -BSA-FITC or BSA-FITC alone (grey bar). Cells were fixed, and visualized by confocal microscopy. Digitized images were inverted to black on white and pixel density for each cell was determined by averaging the density across the cell membrane at four orthogonal points. Each bar represents >20 cells counted +/- SEM. (* p < 0.05). The possibility that E 2 -BSA-FITC could be degraded during incubation with intact cells was examined using HPLC. E 2 -BSA-FITC was incubated in empty wells or wells containing MC7 cells under the same conditions employed for the binding studies described above. Media was removed from the cells and HPLC performed with a reverse phase column. Peaks were visualized using a scanning fluorescence detector. Aqueous solutions of E 2 -BSA-FITC produced a single peak with a retention time of 5.5 minutes using a methanol-water gradient from 80% to 50% over 30 minutes at 1 ml/minute. E 2 and E 2 -BSA did not fluoresce at the excitation and emission wavelengths used (data not shown). Spectra obtained from media containing E 2 -BSA-FITC alone and media containing E 2 -BSA-FITC incubated with MC7 cells are shown in figure 6 . The average area under the curve for E 2 -BSA-FITC was the same (p < 0.05) for solutions incubated in the presence (44,556 +/- 432) and absence (43,436 +/- 289) of MC7 cells (p, 0.05). These results demonstrate that E 2 -BSA-FITC is stable under the culture conditions employed for the binding experiments. Figure 6 Stability of E 2 -BSA-FITC. E 2 -BSA-FITC (10 -8 M in estrogen) was placed in empty wells or wells containing MC7 cells and incubated for 30 minutes at 4°C. E 2 -BSA-FITC was detected by reverse phase HPLC using a methanol-water gradient from 80% methanol to 50% methanol over 30 minutes at 1.0 ml/min. The assay was run in triplicate. Representative spectra are shown for E 2 -BSA-FITC alone (A) and E 2 -BSA-FITC incubated with MC7 cells (B). Discussion The cellular effects elicited by estrogen [ 11 , 12 , 18 - 20 , 3 ] testosterone [ 3 , 21 , 22 ] and progesterone [ 23 - 25 ] covalently conjugated to membrane impermeable BSA have been attributed to non-genomic actions mediated by membrane associated hormone receptors. The use of these reagents for this purpose remains controversial for several reasons. A recent report demonstrated that E 2 -BSA does not bind to purified ER in competition assays with labeled E 2 [ 14 ]. The studies were performed when E 2 -BSA or cold E 2 were added concurrently with labeled E 2 . We obtained similar results under these conditions. However, pre-incubation of E 2 -BSA with purified ER results in significant competition with labeled E 2 . These conflicting results may be explained by differences in the rate of binding between E 2 and E 2 -BSA. E 2 -BSA is a large, bulky molecule similar in size to the ER and is probably spherical in general structure as is the parent BSA molecule. The BSA protein conformation immediately adjacent to the covalently bound estrogen undoubtedly provides substantial steric hindrance to the proper presentation of conjugated E 2 to the binding pocket of ER. The increased size of the E 2 -BSA molecule would also reduce the rate of its diffusion compared with the smaller E 2 . Correct orientation of E2 in the ER binding pocket is also impeded by the restraint on three-dimensional movement imposed by the six atom spacer used to connect BSA and E 2 . Lastly, the use of E 2 -BSA solutions that are formulated in terms of the molarity of total bound-E 2 probably overestimates the amount of E 2 available for binding. The rate of binding between E 2 and ER can be expressed using the second order rate equation: rate = k [E 2 ] [ER], where [E 2 ] is the concentration of estradiol, [ER] is the concentration of the ER, and k the rate constant. Commercially available E 2 -BSA is commonly composed of approximately 10 molecules of E2 attached to every BSA molecule. An E 2 -BSA solution equimolar in estradiol to a solution of estradiol alone would contain one-tenth the molarity of E 2 -BSA with respect to the concentration of estradiol alone. However the rate of E 2 -BSA binding is dependent upon the concentration of E 2 -BSA (rate = k [E 2 -BSA] [ER]). Even if every collision between E 2 -BSA and ER produced binding as successful as collisions between E2 alone and the ER, an E2-BSA solution equimolar in E2 would have approximately one-tenth the rate. Taken together, these factors reduce the binding efficiency of E 2 -BSA to ER compared with free E 2 . However, once binding has occurred, the stability of the E 2 molecule in the ER binding pocket may be only modestly impaired. This may explain how pre-incubation with E 2 -BSA results in successful binding, whereas immediate addition of E 2 -BSA does not have sufficient time to establish successfully bound forms. A similar rationale may explain our results and those of other investigators [ 15 ] that demonstrate specific cell surface binding of E 2 -BSA-FITC only to cells that express ERα. These studies typically employ at least a 30-minute incubation time with E 2 -BSA-FITC, which may be sufficient to result in significant binding. These factors strongly suggest that the rate of binding is an important consideration in experiments assessing potential interactions between E 2 -BSA and ER. Although an estrogen receptor has not been directly isolated and characterized from the cell membrane, evidence other than E 2 -BSA activation of non-genomic effects has recently been reported that strongly supports the existence of a membrane ER. Immunocytochemistry using antibodies specific to several domains of the ERα stained only on the membrane of GH3 cells [ 26 ]. Membrane specific staining was prevented by treatment with antisense ERα mRNA or peptides that interfere with antibody binding. E 2 conjugated to peroxidase also bound only to the membrane of pancreatic islet cells and this binding was competed by E 2 [ 27 ]. The membrane impermeable E 2 -BSA-FITC was shown to stain only the membrane of ER deficient CHO cells transiently transfected with ERα and ERβ [ 15 ]. Moreover, ERα and ERβ interact directly with the membrane associated Src complex to trigger prostate cancer cell proliferation through the RAF-1/Erk-2 signal transduction pathway [ 5 ]. Lastly, we demonstrate that E 2 -BSA-FITC membrane staining is absent with ER deficient HeLa cells and present only on the membrane of cells that endogenously produce ER or HeLa cells that stably express mERα. Taken together these data strongly suggest that non-genomic effects of E 2 are at least partially mediated by a membrane associated ER. However, whether the receptor is the classical nuclear ER translocated to the membrane or an ER unique to the membrane remains unanswered. Conclusions The results presented here suggest that E 2 -BSA can bind to the estrogen receptor but the rate of binding is impeded due to steric and other considerations. Commercially available forms of the reagent are contaminated with dissociable E 2 and should be purified prior to studies designed to demonstrate effects mediated through a membrane ER. Although we demonstrate that classical nuclear ERs can be translocated to the membrane, the conclusive identity of the endogenous membrane receptor awaits purification and sequencing of the putative membrane ER protein. Materials and Methods Establishment of ER stable transfectants Full-length cDNA encoding the mouse ERα was cloned into a vector containing the CMV promoter driving the neomycin resistance gene (pcERα). HeLa cells maintained in MEM containing 10% fetal bovine serum under 5% CO2 were transfected with pcERα and successful transfectants (HeLa-ERα) were selected by survival in media containing the neomycin analog, G418 (400 ug/ml). Preparation of E 2 -BSA free of E 2 400 ul of E 2 -BSA (10 -5 M in estrogen dissolved in 50 mM tris, pH 8.5, Sigma) was added to a centrifugal filter unit with a MW cut-off of 3,000 (Millipore) and centrifuged at 14,000 × g until 50 ul of retentate remained. The retentate was washed 3 times with 350 ul of buffer, recovered and volume adjusted to 400 ul. Binding of estradiol to purified estrogen receptor 3 H-labeled E 2 (NEN, specific activity 48 Ci/mmol, 10 -8 M) was incubated with recombinant ERα (.035 pM, Alexis Corp) for four hours at room temperature in binding buffer (10 mM tris, 10% glycerol, 2 mM DTT, and 1 mg/ml BSA). The binding of labeled E 2 to ERα was competed by various concentrations of ultrafiltered E 2 -BSA or E 2 (10 -9 to 3.5 × 10 -6 M in E 2 ) added four hours prior to or concurrently with the addition of labeled E 2 . ERα was precipitated by the addition of a hydroxyapatite slurry (50% v/v in TE) and centrifugation at 10,000 × g. The pellet was washed three times with wash buffer (40 mM tris, 100 mM KCl, 1 mM EDTA, and 1 mM EGTA) and 3 H-E 2 binding determined by liquid scintillation counting. E 2 -BSA-FITC binding to cell membranes of ER producing cells HeLa-ERα cells or mammary tumor cells (MC7, ATTC) were plated on glass cover slips and incubated with 500 ul of 10 -8 M (in estrogen) E 2 -BSA conjugated to FITC (E 2 -BSA-FITC, Sigma, 10 moles E2 and 3.5 moles FITC per mole BSA) or BSA-FITC (Sigma, equimolar to E 2 -BSA-FITC with respect to BSA) for 30 minutes at 4°C. Binding of E 2 -BSA-FITC to MC7 cells was competed by a 30 minute pre-incubation with E 2 , ICI-182,780, or E 2 -BSA (Sigma, 10 -7 to 10 -9 M). Cells were fixed and FITC staining visualized by confocal microscopy. Images were digitized, inverted to black on white, and pixel density for each cell determined by averaging the density across the cell membrane at four orthogonal points (Scion Image, Scion Corp). The stability of E 2 -BSA-FITC during the incubation with MC7 cells was assessed by HPLC. E 2 -BSA-FITC (500 ul, 10 -8 M in estrogen) was added to empty wells and to wells containing MC7 cells prepared as above for 30 minutes at 4°C. 10 ul of supernatant was resolved using a C-18 reverse phase column (Xterra C-18 RP, 5 um, 4.6 mm × 250 mm, Waters). A multiple solvent deliver system (BIO CM 4000, Milton Roy) provided a methanol-water gradient from 80% methanol to 50% methanol over 30 minutes at a flow rate of one ml/minute. Peaks were detected by a scanning fluorescence detector (model 747, Waters) at an excitation wavelength of 495 nm and emission wavelength of 519 nm. Area under the curve was calculated using standard algorithms (Millenium Software). Assays were performed in triplicate. Estradiol binding studies Subconfluent HeLa-ERα or native HeLa cells were trypsinized and aliquots (2 × 10 6 cells) incubated with several concentrations of 3 H 17β-estradiol in the presence and absence of a 200-fold excess of cold 17β-estradiol for 30 min at 37°. Cells were then incubated on ice for 15 min, washed three times with 2 ml of ice cold 0.2% BSA-saline and pelleted by centrifugation at 1,5000 rpm for 10 min at 4°C. Cells were lysed by the addition of 100 ul of lysis buffer, vortexed and counted. Data was analyzed by Scatchard analysis. Competing interests None declared. Authors' contributions DB wrote the manuscript and performed binding assays. MK generated the stable cell lines. YT performed binding assays.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC516042.xml
549546	A quantitative survey of intern's knowledge of communication skills: an Iranian exploration	Background It is a high priority that health care providers have effective communication skills. It has been well documented that the doctor-patient relationship is central to the delivery of high quality medical care, and it has been shown to affect patient satisfaction, to decrease the use of pain killers, to shorten hospital stays, to improve recovery from surgery and a variety of other biological, psychological and social outcomes. This study sought to quantify the current knowledge of interns in Iran about communication skills. Methods A cross-sectional study using a self-report questionnaire was conducted among interns. Data analysis was based on 223 questionnaires. The internal consistency of the items was 0.8979. Results Overall, knowledge levels were unsatisfactory. Results indicated that interns had a limited knowledge of communication skills, including identification of communication skills. In addition, there was a significant difference between the mean scores of interns on breaking bad news and sex education. The confidence of males about their communication skills was significantly higher than for females. Analysis of the total scores by age and sex showed that there was a statistically significant main effect for sex and the interaction with age was statistically significant. Free response comments of the interns are also discussed. Conclusions It is argued that there is a real need for integrating a communication skills course, which is linked to the various different ethnic and religious backgrounds of interns, into Iranian medical curricula. Some recommendations are made and the limitations of the study are discussed.	Background The expectations of the public have been dramatically increased and the majority of them are familiar with their rights in the health care system. As a consequence, it is a high priority that health care providers have effective communication skills. It has been well documented that the doctor-patient relationship is central to the delivery of high quality medical care. It has been shown to affect patient satisfaction, to decrease the use of pain killers, to shorten hospital stays, to improve recovery from surgery and a variety of other biological, psychological and social outcomes [ 1 - 4 ]. Lack of knowledge of communication skills, or an inability to use them effectively, can be distressing and is potentially hazardous for patients. It may also be a cause of stress for medical students arriving on the ward for the first time [ 5 ]. There is a large body of evidence indicating the importance of students' knowledge of communication skills and [ 6 , 7 ] how behaviours learned from communication skills training transfer into the clinical setting and such training is known to have long term effects on students behaviour [ 8 - 11 ]. However, little is known about the importance of communication skills in the practice and training of doctors in Iran, where the culture differs greatly from that of the West. Sensitivity to religious matters is particularly important in Iranian doctor-patient relationships where Islam is more than a religion; it is a way of life. It controls politics, local laws, behaviour and many other aspects of daily life. It gives guidance in all spheres of human activity from birth to death. Therefore doctors coming into contact with religious patients need to be aware that there are numerous potential barriers to good communication [ 12 ]. A major criticism of current medical training in Iran is that communication skills have not been embedded in the curriculum of Iranian medical students, despite the richness and variety of evidence from elsewhere concerning the importance of communication skills. Concerns over poor doctor-patient communication amongst Iranian doctors led to an exploration of the current situation [ 13 ]. In this paper we investigate the knowledge level of interns about communication skills to gain a clearer picture of some challenges relating to health care promotion, especially patient satisfaction and adherence to treatment. Two questions guided the study: (a) How do interns assess their knowledge about communication skills? (b) Is there a significant difference between the level of knowledge among male and female interns? Methods A quantitative survey was performed at Tehran University of Medical Science (TUMS). A cross-sectional study was conducted using a questionnaire administered to 235 interns. Anonymity was maintained throughout. The subjects received the self-administered questionnaire with a covering letter explaining the project and the subject's rights. 12 subjects did not return the questionnaire and an additional 7 subjects did not give their age and one person did not give his/her sex. Therefore data analysis was based on 223 questionnaires, but covariate-based analysis on fewer. The subjects were asked to complete the questionnaire without referring to source books. The questionnaire consisted of three sections. The first section asked students to give personal details including the demographic items age and gender (summarised in Table 1 ). The second section is related to the educational items: subjects studied or attended in a specific course about communication skills (Table 2 ). The third section asked students to rate their knowledge of communication skills and, if they rated themselves higher than 5, discuss the item briefly in the space provided in order to assess their real knowledge with regard to that communication skill. In addition, they were encouraged to provide additional written comments on the questionnaire. The communication skills knowledge scale (CSKS) developed here consists of 10 items about communication skills. Each item is measured on a 10-point scale, ranging from 1 (low) to 10 (high). Table 1 Distribution of background characteristics Variables Number Percentages Sex Male 132 59.5 Female 90 40.5 Total 222 100 Age Less than 25 114 52.8 25–30 96 44.4 More than 30 6 2.8 Total 216 100 Table 2 Percentage response to educational items by interns Educational item Yes No % % 1. Have you studied a paper in relation to communication skills? (n = 219) 21.9 78.1 2. Have you formally attended communication skills courses? (n = 221) 8.6 91.4 The choice of items was based on the communication skills an intern will need. All items were verified and subjected to content validation by three major experts in communication skills. These experts were given copies of the CSKS and the purpose and objectives of the study. They then evaluated the CSKS on an individual basis. Comparisons were made between these evaluations and the authors then made some minor changes within the CSKS. The CSKS had a high internal consistency ( Cronbach alpha = 0.8979). The validity of the CSKS can only be examined through logical rather than empirical means. Since the CSKS was not compared to a standardised test, it was impossible to obtain a numerical estimate of the validity of the test. However, based on logical means, i.e., a respectable Cronbach alpha and high inter-rater agreements on each item, the authors believe that the test is valid. The questions and responses have been translated from Persian into English for this paper. Results The potential score range from the 10-item CSKS (by summing all 10 item scores) is 10 to 100, with 10 indicating low knowledge. Analysis of the total scores produced a mean score of 51.30 [95 per cent confidence interval (CI) 49.05–53.55]. The subjects' performance on the CSKS suggests a knowledge deficit in communication skills. The mean scores for males and females were respectively 53.6 and 48.2 (P = 0.02). The vast majority of interns (78.1%) had not studied a paper on communication skills. When asked whether they had formally attended communication skills courses, 91.4% of interns reported "no". Of the few interns who reported "yes", these interns specified courses such as CPR, injections and semiology (Table 3 ), which are not formal communication skills courses. Table 3 Courses of communication skills training reported by interns Courses Number CPR 2 EBM 1 Ethics 4 Health 2 Injection 1 Skills lab 3 Semiology 3 Workshop 1 Total 17 The analysis of the scores by topic is shown in table 4 . The possible range of scores for each item was 1 to 10. Mean scores for topics ranged from 2.8 to 6.1. Interns were most confident on "giving and receiving information", and the least confident on "sex education". Table 4 Analysis of results by communication skill Topics Mean P Value Male (SD) Female (SD) Breaking bad news 4.6 (2.0) 3.7 (2.0) 0.02 Dealing with anger/difficult patient 5.0 (2.4) 4.4 (2.2) 0.81 Demonstration of empathy 5.3 (2.3) 5.5 (2.4) 0.56 Giving and receiving information 6.1 (2.3) 5.9 (2.2) 0.46 Non-verbal communication skills 5.3 (2.5) 5.0 (2.0) 0.37 Dealing with patient perception 5.8 (2.4) 5.4 (2.1) 0.22 Shared decision making 5.7 (2.2) 5.4 (2.3) 0.28 Patient-oriented interviewing 5.4 (2.3) 5.3 (2.3) 0.17 Sex education 4.6 (2.7) 2.8 (1.9) 0.00 Closing skills 5.4 (2.4) 4.8 (2.4) 0.08 Total 53.6 (17.4) 48.2 (15.8) 0.02 A two-way between-groups analysis was conducted to explore the impact of sex and age on levels of knowledge, as measured by the CSKS. Subjects were divided into two groups according to their age (less than 25 years, or 25 years and above). There was a statistically significant main effect for sex [F (1, 212) = 4.90, p = 0.02] and the interaction effect [F (1, 212) = 4.06, p = 0.04) did reach statistical significance. However the effect size was small (eta squared = 0.02). The young male interns were more confident than average, while the young female interns were less confident. Free responses included the following comments: ' Nobody has trained us about communication skills. Our knowledge in respect of communication skills is very poor. Your items show that we are very far behind other countries. Our universities are not as advanced as other universities' . 'I feel we are not familiar with the ABC of communication skills' . 'A good guide to communication skills needed' . 'I feel communication skills would be an excellent course since it gives us an idea of how we can handle bad news' . 'Attending doctors are not totally familiar with the aims and use of communication skills in the clinical setting' . 'All our courses only focus on biological issues rather than psychosocial issues' Limitations There were a number of limitations to this study. 1. The CSKS has not been normed for a population of interns. 2. Criterion-related validity of the CSKS was not determined, although content validity was established on the instrument. 3. Since it is a self-assessed questionnaire, these may be problems with bias, such as prestige bias. Discussion The very high response rate (95%) of this questionnaire may have reflected general interest, or may have resulted from the advantages of self-assessment which itself may improve performance. The results on the CSKS show that basic knowledge of interns in Iran about communication skills is limited. Researchers have reported similar findings in other countries which reveal a deficit in the knowledge of doctors about communication skills [ 14 ]. The importance of communication skills has long been acknowledged in general practice training [ 15 ] and the need to teach communication skills formally, as part of British undergraduate medical education, has also been recognised [ 16 ]. In Iran, interns' knowledge deficiency may be attributed to the fact that interns have never been trained to consult in the general practice setting, and their skills are limited to making value judgements, often using the only available criterion, comparison with their own style [ 13 ]. This approach to a patient is not cost effective and may lead to negative health outcomes such as patient dissatisfaction, poor adherence to treatment and medical errors [ 17 ]. A few students reported their attendance at courses such as EBM, semiology, skills lab or CPR, which have no relation with communication skills training. This indicates that students are not familiar with the tasks of communication skills [ 18 ]. The vast majority of research studies have been conducted on the outcome of communication skills in the practice and training of doctors in western countries. Even here, despite doctors trained in communication skills and the advocacy of the use of a patient-oriented approach, some evidence suggests that there are difficulties in practice [ 19 , 20 ]. However, research has demonstrated that communication skills training intervention using behavioural, cognitive and affective domains can increase not only potentially beneficial and effective interviewing styles, but also alter attitudes and confer other benefits [ 9 , 21 ]. The results of the study show that there were significant differences between males and females with regard to their reported knowledge of the main communication skills. Women were less confident of their skills. The deficit may partly be an artefact of an inadvertent prestige bias of the male students. The deficit is particularly notable in sex education. There are three possible explanations for this. Firstly, in general, Iranian female interns are very shy to ask patients about sexual issues. Therefore they may feel that sex education skills have no implications for their practice and hence pay less attention to sex education training. Secondly, it may be a systematic error in female respondents, i.e. they may be shy to discuss their knowledge about sex education skills rather than lack knowledge. Thirdly, in the past, sex education was regarded as a taboo in Iran and was not available in schools, especially for girls [ 22 ]. This perhaps acts as an inhibitory factor on the basic knowledge of sex education. Within this context, there is no evidence that shows similar results for gender difference on the knowledge of sex education in the practice and training of doctors. The results on the CSKS suggest that there are areas of weakness in the communication skills confidence of interns, particularly in breaking bad news. While it is well recognised that delivering bad news is a difficult task that requires skills and sensitivity [ 23 ], both female interns and male interns reported that their confidence in breaking bad news is low, especially the female interns. While the interns commented on the need to improve medical students' communication skills, it seems that guidelines on delivering bad news to patients and patients' family members have not been seriously taken into consideration in the practice and training of doctors in Iran. This could be due to interns possessing deep fears regarding delivering bad news to patients' family members, or because they are unaware of the general guidelines about delivering bad news [ 24 ]. Three studies which have attempted to address residents' perception of delivering bad news indicate that residents had experienced discomfort with psychosocial issues related to the conveyance of bad news, such as personal fears and different perceptions of bad news [ 25 - 27 ]. There is a significant difference between the mean score of the interns on breaking bad news. The female interns have reported lower confidence than the male interns. The deficit could be an inadvertent prestige bias of the male students. However, to our knowledge, there is no evidence that underpin such finding. Although Orlander et al's work [ 28 ] demonstrated there were no significant differences between males and females with regard to the type of bad news, residents' knowledge with regard to breaking bad news was not reported by the authors. Therefore, some empirical research is essential. Given the poor levels of confidence about communication skills, particularly sex education skills, revealed in this study, it is concluded that educational programmes are necessary. In sex education skills training, given the complex interplay of cultural and religious beliefs in Iran, particular attention must be paid to multicultural and religious issues. Therefore, further work is needed on gender education and stereotypes in sex education; learning styles; the 'hidden curriculum'; and how far medical schools make organisational and administrative arrangements on the basis of gender and the implication for female and male interns. The enthusiastic response to the questionnaires may suggest that medicine is accepting the need for developing communication skills within the medical curriculum. Medical education in Iran must respond to this challenge. Finally, our findings may be somewhat limited in generalisability because they are derived from only one medical school in Iran. Self-assessment data may suffer from biases such as prestige bias. Despite these caveats, the authors believe the data to be an accurate reflection of current practice in Iran, based on the Iranian authors training experiences, and consistency with previous accounts. Conclusions Whilst the approach to this research has been shaped by a government-recognised health need, the authors recognise the need for, and welcome, further examination of these findings from multiple perspectives, especially with regards to ethnicity and social issues. Since not enough attention has been focused on individuals as makers of health as a service rather than customers of health care services, it is strongly recommended, therefore, that medical students be trained in the context of psychosocial issues that may influence health behaviour, as has been indicated by one of the participants. It is particularly important that this type of approach be incorporated into the curricula of medical training. This may assist in transferring from the disease-oriented to the patient-oriented approach and ultimately lead to patients understanding more and taking greater responsibility for their own health. Competing interests The author(s) declare that they have no competing interests. Authors' contributions MT and ST carried out the conception, design, initial analysis and interpretation of the data. MT drafted the paper. ODL was involved in revising the draft critically, revising the statistical analysis and gave final approval of the version to be published. AAZ contributed to the collection of data and the reviewing of the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC549546.xml
529259	BIOKID: Randomized controlled trial comparing bicarbonate and lactate buffer in biocompatible peritoneal dialysis solutions in children [ISRCTN81137991]	Background Peritoneal dialysis (PD) is the preferred dialysis modality in children. Its major drawback is the limited technique survival due to infections and progressive ultrafiltration failure. Conventional PD solutions exert marked acute and chronic toxicity to local tissues. Prolonged exposure is associated with severe histopathological alterations including vasculopathy, neoangiogenesis, submesothelial fibrosis and a gradual loss of the mesothelial cell layer. Recently, more biocompatible PD solutions containing reduced amounts of toxic glucose degradation products (GDPs) and buffered at neutral pH have been introduced into clinical practice. These solutions contain lactate, bicarbonate or a combination of both as buffer substance. Increasing evidence from clinical trials in adults and children suggests that the new PD fluids may allow for better long-term preservation of peritoneal morphology and function. However, the relative importance of the buffer in neutral-pH, low-GDP fluids is still unclear. In vitro, lactate is cytotoxic and vasoactive at the concentrations used in PD fluids. The BIOKID trial is designed to clarify the clinical significance of the buffer choice in biocompatible PD fluids. Methods/design The objective of the study is to test the hypothesis that bicarbonate based PD solutions may allow for a better preservation of peritoneal transport characteristics in children than solutions containing lactate buffer. Secondary objectives are to assess any impact of the buffer system on acid-base status, peritoneal tissue integrity and the incidence and severity of peritonitis. After a run-in period of 2 months during which a targeted cohort of 60 patients is treated with a conventional, lactate buffered, acidic, GDP containing PD fluid, patients will be stratified according to residual renal function and type of phosphate binding medication and randomized to receive either the lactate-containing Balance solution or the bicarbonate-buffered Bicavera ® solution for a period of 10 months. Patients will be monitored by monthly physical and laboratory examinations. Peritoneal equilibration tests, 24-h dialysate and urine collections will be performed 4 times. Peritoneal biopsies will be obtained on occasion of intraabdominal surgery. Changes in small solute transport rates, markers of peritoneal tissue turnover in the effluent, acid-base status and peritonitis rates and severity will be analyzed.	Background Peritoneal dialysis (PD) is the preferred dialysis modality in children. Advantages of PD over hemodialysis relevant to pediatric patients include its compatibility with a normal lifestyle and full psychosocial integration, the continuous mode of blood purification without dysequilibrium conditions, the absence of vascular access issues and the avoidance of puncture pain. However, the major drawback of PD is its limited technique survival. Almost fifty percent of adult as well as pediatric PD patients must switch to hemodialysis within 4 to 5 years of treatment [ 1 , 2 ]. While the incidence of PD failure due to infectious complications is steadily decreasing, loss of ultrafiltration due to degenerative changes of the peritoneal tissue is becoming the leading cause of non-elective termination of PD [ 1 ]. Histopathological alterations induced by exposure to PD solutions include a severe vasculopathy, neoangiogenesis, submesothelial fibrosis and a progressive loss of the mesothelial cell layer [ 3 - 5 ]. Acute and chronic toxicity of standard PD fluids to mesothelial cells, affecting cell turnover and the pattern of growth factor and cytokine release, is considered a key mechanism underlying the progressive transformation of the peritoneum. Conventional PD fluids contain large doses of glucose, are lactate-buffered at acidic pH and contaminated with toxic glucose degradation products (GDP) formed during heat sterilization. Low pH, lactate and hyperosmolar glucose independently impair mesothelial cell functions [ 6 - 9 ]. GDPs impair the viability and functional integrity of mesothelial cells upon extended exposure [ 10 ], and stimulate VEGF and TGF-β release by mesothelial cells [ 11 , 12 ]. In recent years, a new generation of more biocompatible PD fluids has been introduced into clinical practice. The separation of alkaline and acidic fluid compartments in pluri-chamber bags permits to sterilize glucose at very low pH with greatly reduced GDP formation and yet produce pH-neutral final dialysis solutions, using lactate and/or bicarbonate as a buffer. We recently compared the safety and efficacy of Bicavera ® (Fresenius), a purely bicarbonate-buffered biocompatible PD solution, with that of a conventional acidic, lactate buffered solution by a three-month crossover trial in children on automated PD [ 13 ]. We observed a marked increase of the mesothelial cell marker CA125 in the effluent during Bicavera ® treatment, which was readily reversible when patients returned to conventional solution. This effect was also observed with lactate- or lactate/bicarbonate-buffered biocompatible PD solutions [ 14 , 15 ], and is interpreted as a functional and/or numeric recovery of mesothelial cells exposed to these fluids. Moreover, in agreement with previous studies [ 7 , 16 ] we observed a trend towards increasing small solute permeability with the standard solution; this trend was absent when Bicavera ® was used. Two studies observed a slightly lower initial increase of the functional peritoneal surface area during a single PD dwell with pH-neutral compared to acidic solutions compatible with reduced peritoneal capillary recruitment; this trend was significant in one study [ 17 - 19 ]. Fischbach et al. also demonstrated lower intraperitoneal pressure and less inflow pain in children receiving a low-GDP, neutral-pH PD solution [ 19 ]. Finally, we noted a more effective compensation of metabolic acidosis with Bicavera ® than with lactate buffered conventional fluid despite identical content of base equivalents. While these results are encouraging with respect to the long term preservation of the peritoneal membrane and strongly favour the primary use of low-GDP, neutral-pH biocompatible PD fluids, the relative importance of the buffer system is still unclear. In vitro data suggest that lactate per se may compromise local cell functions independently of pH by affecting the cellular redox state and reducing cellular energy sources [ 6 , 20 - 22 ]. By intravital microcopy of rat peritoneum, lactate-based neutral-pH PD solution caused mesenteric vasodilation whereas bicarbonate buffered PD fluid had no hemodynamic effects [ 23 ]. All previous clinical trials comparing conventional and biocompatible PD fluids were unsuitable by design to identify any role of the buffer for peritoneal tissue integrity, perfusion and the acute or chronic regulation of peritoneal solute transport, since the solutions tested differed not only by the buffer used, but also by pH and GDP content. To clarify the role of the buffer the BIOKID trial has been designed. Patients participating in this trial will be exposed to two solutions which are both pH neutral and of low GDP content, but contain either pure bicarbonate or pure lactate as buffer compound. Methods/design Objectives of the study The European Pediatric Peritoneal Dialysis Study Group (EPPS) plans a prospective, randomized study with administration of pH neutral, low-GDP PD solutions containing either lactate or bicarbonate buffer over a period of 10 months. The primary objective is to evaluate the effect of lactate vs. bicarbonate buffer on peritoneal transport capacity in children. The hypothesis to be tested is that bicarbonate based PD solutions may allow for a significantly better preservation of peritoneal transport characteristics (D/P Crea ) in children compared to a solution containing lactate buffer. Secondary objectives will be to assess differential effects of lactate and bicarbonate buffered PD fluids on acid-base status, surrogate parameters of peritoneal biocompatibility and local and systemic carbonyl stress, peritoneal morphology, the incidence and severity of peritonitis, statural growth and nutritional status. Moreover, this study will be used to assess genetic determinants of the peritoneal transporter status and the evolution of peritoneal morphology over time. Study design This is a multicenter open-labelled, controlled, randomized clinical trial, designed to test the effects of the buffer substance in biocompatible PD fluids on peritoneal small solute transport capacity. All subjects will undergo a 2-month run-in period, in which they receive conventional lactate buffered, acidic, GDP containing PD fluid. During this period, patient eligibility for the trial will be verified and the dialysis dose will be optimized if necessary to ensure appropriate PD adequacy. At the end of this period, the patients will be stratified according to residual renal function (greater or less than 100 ml urine output/day/1.73 m 2 ) and the type of phosphate binder therapy (Sevelamer vs. calcium-containing phosphate binders), since these variables may affect the overall efficacy of metabolic acidosis control. Following stratification, subjects will be randomized centrally to receive either the lactate-containing Balance solution or the bicarbonate buffered Bicavera ® solution for a period of 10 months. Both during the run-in period and during the intervention phase, patients will be monitored by monthly clinical and laboratory examinations, including capillary blood gas analyses. In addition, peritoneal equilibration tests (PETs), 24-hour dialysate and urine collections and intraabdominal pressure assessments will be performed at time of randomization (with conventional PD fluid) and after 3, 6 and 10 months of treatment (with the study solutions). Also, peritoneal biopsies will be performed on occasion of intraabdominal sugery or laparoscopy prior to start and after termination of the study (usually at time of catheter insertion and renal transplantation). Primary outcome measure The primary outcome measure will be the longitudinal change in 4h-D/P creatinine in the sequential PET examinations. Differential changes in this parameter will indicate differences in the development of the peritoneal solute transport status over time. Secondary outcome measures Secondary outcome measures will be surrogate parameters of mesothelial cell viability (CA-125), peritoneal neoangiogenesis (VEGF), fibrotic activity (TGF-β) and local inflammation (IL-6). With the same intention, the evolution of peritoneal histomorphology will be assessed in all patients available for sequential biopsies. Moreover, possible differential effects of lactate and bicarbonate buffer on the control of metabolic acidosis will be assessed by monthly blood gas analyses. Finally, the incidence and clinical course of peritonitis will be recorded as a possible indirect marker of local peritoneal macrophage function. Inclusion criteria Criteria for inclusion in the study are 1) patients above 1 month and less than 19 years of age, 2) end-stage renal disease with manual or automated continuous peritoneal dialysis as maintenance treatment modality, 3) a fill volume approximately of 1100 ml/1.73 m 2 body surface area, 4) the most recent episode of PD-associated peritonitis, if any, occurred more than 3 weeks ago, 5) signed informed consent by parent/guardian, with a subject aged > 7 years also signing an age-appropriate assent form. Exclusion criteria Criteria for exclusion from the study are 1) reduced efficiency of peritoneal dialysis due to anatomic anomalies or intraperitoneal adhesions, 2) uncontrolled hyperphosphatemia, 3) severe pulmonary, cardiac, hepatic or systemic disease including any kind of malignancy, and 4) current or recent (within 30 days) exposure to any investigational drug. Exit criteria Reasons for permanently discontinuing the study medication are 1) renal transplantation, 2) switch to hemodialysis due to PD technique failure, 3) patient/parent withdrawal of consent of participate, 4) patient moving out of the area to a location with no participating center within reasonable distance, and 5) a severe adverse event. Study medications The composition of the study fluids is given in Table 1 . Both Bicavera ® and Balance will be available in three different glucose concentrations to meet individual ultrafiltration requirements. The fluids will be administered at a dose of approximately 1,100 ml/m 2 body surface area per dwell. Both in patients on CAPD and CCPD, the number of cycles and dwell times can be varied according to clinical needs. The dose of dialysis will be tailored individually in order to ascertain a minimum total weekly Kt/V Urea of ≥ 2.0. Table 1 Balance 1.5% Balance 2.3% Balance 4.25% Bicavera ® 1.5% Bicavera ® 2.3% Bicavera ® 4.25% Sodium (mmol/l) 134 134 134 134 134 134 Calcium (mmol/l) 1.75 1.75 1.75 1.75 1.75 1.75 Magnesium (mmol/l) 0.5 0.5 0.5 0.5 0.5 0.5 Chloride (mmol/l) 101.5 101.5 101.5 104.5 104.5 104.5 L-Lactate (mmol/l) 35 35 35 - - - Bicarbonate (mmol/l) - - - 34 34 34 Glucose-monohydrate (g/l) 15 23 42.5 15 23 42.5 Osmolarity (mosmol/l) 358 401 511 358 399 509 PH 7.4 7.4 7.4 7.4 7.4 7.4 Any kind of concomitant medication during the run-in period and the study period will be documented in the case report form with respect to type, dosage and mode of delivery. In case of peritonitis (defined by cloudy effluent, white blood cell count greater than 100/mm 3 with more than 50% polymorphonuclear leukocytes), treatment will be given intraperitoneally according to international pediatric guidelines [ 24 ] using cefazoline and ceftazidime in patients with mild peritonitis and a glycopeptide/ceftazidime combination in patients with defined risk factors for severe course and poor outcome. Bicarbonate supplementation will be discontinued at start of the run-in period and only re-instituted if blood bicarbonate levels drop below 17 mmol/l despite sufficient dialysis efficacy. The recommended dosage is 0.5 mmol/kg/day divided into 3 doses. Clinical safety monitoring An adverse event is defined as any untoward medical occurence in a patient who takes the study medication. It does not necessarily have a causal relationship with this treatment. This may be an unfavourable and unintended sign, symptom or disease, which is observed after exposure to the study medication, whether or not considered related to the treatment. Moreover, the participating investigators will report all treatment-emergent adverse events that are observed on the online adverse event form. This applies regardless of the clinical significance or the assessment of study drug causality. In this trial, such adverse events may include inflow pain, severe changes of the state of hydration, abnormal electrolyte and glucose blood levels, peritonitis, abdominal hernia, and allergic reactions. A severe adverse event or reaction is any untoward medical occurrence that a) results in death, b) is life-threatening, c) requires inpatient hospitalization or prolonges an existing hospitalization, or d) results in persistent or significant disability/incapacity. Any serious adverse event, whether or not considered related to the study medication, and any unexpected drug reactions with significant hazard to the patient population will be reported to the responsible safety assessor at Fresenius Medical Care by phone or by fax within 24 hours following first knowledge of the event. Alternatively the clinical monitor may be informed. This information will be forwarded to and evaluated by the safety monitoring committee, and reports of serious adverse reactions will be disseminated to all participating centers for submission to their respective institutional review boards. Fresenius Medical Care is responsible for passing on the information to relevant supervisory authorities. Data management Data acquisition will be entirely through the internet. The case report form menus will be used to record the following: 1) baseline clinical patient information, 2) physical and biochemical examination variables, 3) study and concurrent medications, and 4) adverse events. Periodic computerized audit reports will be run to monitor data quality and completeness. The data base is stored on a server drive that is backed up to tape daily by Tel-A-Vision, Media Networking GmbH. In order to insure confidentiality, data from each patient will be recorded in the computer data base with a unique contributing center code, study code, sequence number and patient initials. Patient names are never entered online or forwarded in any other form to the coordinating office. Sample size estimation The primary study outcome is the change in 4h-D/P Cr from the time of randomization to the conclusion of the 10-month study period. In a previous trial we demonstrated a 6% increase of D/P Cr in children on lactate-buffered PD fluid in contrast to a 4% decrease in D/P Cr using bicarbonate buffered fluid within 3 months of exposure, resulting in a statistically significant 10% difference in the evolution of peritoneal creatinine transport rate [ 13 ]. Assuming that this effect was at least in part due to the different buffer substances applied, a difference in D/P Cr at least 7 ± 10% can be expected when Balance and Bicavera ® are applied for 10 months. To detect this difference with a sensitivity of 80% and an error probability of 5%, at least 15 patients per randomization group will be required. Assuming a 50% drop-out rate due to renal transplantation and other reasons in the course of the study, 60 patients will have to be enrolled. Statistical approach A computer based randomization protocol will be generated and applied centrally at the end of the run-in period. Baseline comparability between the two treatment groups will be evaluated with respect to entry criteria. Chi-square and t-tests will be used to assess differences between the two groups on the baseline variables. Any variables that are found to be discrepant between the two groups and that are related to the outcome variables will be treated as co-variates in later analyses. In order to evaluate the patients' change in D/P Cr ratios, two strategies will be employed: 1. Repeated measure ANOVA will be performed on those patients who complete the 10-month observation period on the study medications. 2. The Kaplan-Meier method will be used to estimate the time in which patients are likely to display a ≥ 7.5 % increase in the D/P Crea ratio during the 10-month period. All patients started on study medication will be available for this analytical approach. The same strategies will be used to analyze secondary outcome variables. Publication of study results All publications will be authored by members of the scientific advisory committee. Co-authors will have contributed to the design, analysis, execution and actual reporting of the study. Study investigators Steering committee : C.P. Schmitt, F. Schaefer, K.E. Bonzel, K. Rönnholm Scientific advisory committee : M. Almeida, K. Arbeiter, G. Ardissino, K.E. Bonzel, A. Edefonti, M. Fischbach, K. Haluany, J. Misselwitz, Markus J. Kemper, K. Rönnholm, F. Schaefer, C.P. Schmitt, S. Wygoda Safety monitoring committee : V. Schwenger, U. Querfeld, G. Offner Discussion We report the protocol of a randomized clinical trial designed to test the effect of the buffer type on the evolution of peritoneal tissue integrity and transport function in children treated with 'biocompatible', i.e. neutral-pH, low-GDP PD solutions. The study utilizes the availability of two novel biocompatible PD solutions manufactured by the same company which differ selectively in the buffer employed, namely either pure lactate or pure bicarbonate. The comparative administration of these solutions will provide information on differences in peritoneal cell viability, tissue morphology, local host defense and solute transfer capacity potentially inferred by cytotoxic and/or vasocative effects of lactate administered to the abdominal cavity in unphysiological concentrations. Competing interests This investigator-initiated trial was designed exclusively by the members of the EPPS scientific advisory committee. None of the investigators have any financial relationship with the manufacturer of the study medication. Financial support is received from Fresenius Medical Care to cover coordination costs and investigator meetings. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC529259.xml
526195	Day-to-day variations during clinical drug monitoring of morphine, morphine-3-glucuronide and morphine-6-glucuronide serum concentrations in cancer patients. A prospective observational study	Background The feasibility of drug monitoring of serum concentrations of morphine, morphine-6-glucuronide (M6G) and morphine-3-glucuronide (M3G) during chronic morphine therapy is not established. One important factor relevant to drug monitoring is to what extent morphine, M6G and M3G serum concentrations fluctuate during stable morphine treatment. Methods We included twenty-nine patients admitted to a palliative care unit receiving oral morphine (n = 19) or continuous subcutaneous (sc) morphine infusions (n = 10). Serum concentrations of morphine, M6G and M3G were obtained at the same time on four consecutive days. If readmitted, the patients were followed for another trial period. Day-to-day variations in serum concentrations and ratios were determined by estimating the percent coefficient of variation (CV = (mean/SD) ×100). Results The patients' median morphine doses were 90 (range; 20–1460) mg/24 h and 135 (range; 30–440) mg/24 h during oral and sc administration, respectively. Intraindividual fluctuations of serum concentrations estimated by median coefficients of day-to-day variation were in the oral group for morphine 46%, for M6G 25% and for M3G 18%. The median coefficients of variation were lower in patients receiving continuous sc morphine infusions (morphine 10%, M6G 13%, M3G 9%). Conclusion These findings indicate that serum concentrations of morphine and morphine metabolites fluctuate. The fluctuations found in our study are not explained by changes in morphine doses, administration of other drugs or by time for collection of blood samples. As expected the day-to-day variation was lower in patients receiving continuous sc morphine infusions compared with patients receiving oral morphine.	Background Morphine is degraded in the liver to several metabolites of which morphine-6-glucuronide (M6G) and morphine-3-glucuronide (M3G) are biological active [ 1 ]. M6G is shown to contribute to the analgesia produced by morphine and may cause opioid related adverse effects such as sedation or nausea [ 2 - 5 ]. Due to first pass metabolism and slow accumulation of M6G in the brain the analgesic activity of M6G is most prominent during oral long-term treatment with morphine while single dose studies show less contribution from M6G to the analgesic effects from morphine [ 2 , 3 , 6 ]. M3G may in exceptional cases cause excitatory adverse effects such as delirium, myoclonus or allodynia [ 7 ]. Animal studies observed that M3G have an anti-nociceptive effect [ 8 , 9 ], but this effect was not reproduced in a study administering M3G to volunteers exposed for human experimental pain [ 10 ]. The most obvious determinants for serum concentrations of morphine, M6G and M3G are morphine doses, route of administration and renal function. However, a considerable variation of serum concentrations between patients remains after correcting for dose and route of administration [ 3 , 11 - 13 ]. Measurements of morphine, M3G and M6G serum concentrations can explain individual responses in patients where morphine treatment turns out to have unexpected effects and help physicians to determine changes in pain treatment. Physicians tend to believe that samples obtained for therapeutic drug monitoring during steady state conditions will be representative irrespective of which day the sample is collected. However, morphine, M6G and M3G serum concentrations may also have fluctuations not caused by changes in morphine doses, administration of other drugs or by time for collection of blood samples. This variation represents the day-to-day variability. In order to evaluate the clinical implications from morphine and metabolites serum concentrations measurements it is necessary to know if these serum concentrations have fluctuations not related to changes in drug administrations. The day-to-day variability in serum concentrations of morphine, M6G and M3G are previously reported in a study of 8 cancer patients treated with subcutaneous (sc) morphine infusions. This study observed coefficients of variation (CV) ranging from 26%–56% for morphine, 20% to 51% for M6G and 20%–49% for M3G [ 12 ]. To our knowledge, the day-to-day variations of morphine, M6G and M3G serum concentrations obtained from consecutive days or during chronic oral administration of morphine are not previously reported. Thus, the aim of this study is to investigate the day-to-day variations of morphine, M6G and M3G serum concentrations during stable chronic oral and sc morphine administration to cancer patients. Methods Patients We included twenty-nine patients admitted during a nine-month period to the Palliative Care Unit at the University Hospital in Trondheim. The inclusion criteria were; verified malignant disease, expected survival time less than 6 months, scheduled morphine treatment started at least three days prior to inclusion, stable scheduled doses of morphine for a minimum of three days and age more than eighteen years. The exclusion criteria were; planned hospitalisation less than three days and lack of ability to communicate (e.g. dementia, deafness). All patients gave their written informed consent before inclusion. The study was conducted according to the guidelines of the Helsinki declaration. The Regional Committee for Medical Research Ethics, Health Region IV, Norway, approved the study. Study design Inclusion The patients were included in the study within three days after admission to the Palliative Care Unit. Each patient was followed for four days. Patients readmitted to the Palliative Care Unit were allowed to a new trial period identical to the first trial period. No patients were included in more than three trial periods. The patients' age, gender, primary malignant diagnosis, presence of metastasis, and other medications were registered. Morphine treatment during the last 24 hours was registered with respect to route of administration, morphine formulation, scheduled dose and consumption of rescue morphine for breakthrough pain. The patients' functional status was assessed using the Karnofsky performance status score [ 15 ]. Blood samples Blood samples were obtained each day during the trial period. The samples were obtained at the same time each day during the routine morning round for collecting blood samples. Observations In order to observe if the patients were studied during stable treatment conditions the scheduled morphine dose, rescue morphine consumption and route of administration were registered each study day. The use of other medications was also registered daily. Pain, nausea and sedation were assessed at day study two, three and four during the trial period using a 5 category verbal rating scale (VRS) score ranging from no to very severe. All symptoms were assessed for the last 24 h. Analyses The blood samples were placed in EDTA tubes until separated by centrifugation (3000 rpm, ten minutes) and stored at -85°C until analysed. All samples were analysed for serum concentrations of morphine, M6G and M3G applying liquid chromatography mass spectrometry [ 16 ]. The limits of detection were for morphine 0,35 nmol/l and for M6G and M3G 2,2 nmol/l. The analytical coefficients of variation obtained in quality control samples (CV Analytical ) were for morphine 3,0%, for M6G 5,5% and for M3G 7,0%. The analytical coefficients of variation were determined at 100 nmol/l for morphine and 1000 nmol/l for M6G and M3G. Serum values of creatinine concentrations, alanin aminotransferase activities (ALAT), aspartat aminotransferase activities (ASAT) and albumin concentrations were determined using standard analytical methods. Statistical evaluation Total use of morphine for each trial day was calculated by adding scheduled morphine doses and rescue morphine consumption. Samples obtained less than two hour after the administration of a morphine rescue dose were excluded from the analyses. Day-to-day variations of morphine and its metabolites are presented as biological coeffecients of variation. This biological variation (CV Biological ), expressed in terms of percent coefficient of variation, was calculated for each patient in each trial period using the equation [ 15 , 16 ]: CV Biological = CV Observed - CV Analytical The observed coefficients of variation (CV Observed ) for morphine, M6G and M3G, which represent the variation in serum concentrations for each patient during each trial period, were calculated using the equation [ 17 , 18 ]: At least three observations were needed in order to calculate an observed coefficient of variation. Statistical comparisons between the trial days and trial periods were performed using one-way analysis of variance tests. Due to multiple comparisons statistical significance was defined as p < 0,01. The statistical software SPSS version 9.0 for Windows was used throughout the analyses. Results Patient characteristics The patients (16 males and 13 women) median age at inclusion was 68 years (range; 39–89). The patients' Karnofsky performance status, primary tumor diagnoses and presence of metastases are shown in Table 1 . The median serum creatinine concentrations at inclusion were 72 μmol/l (range; 45–121). The median values at inclusion of ASAT and ALAT were 31 IU/l (range; 7–154) and 17 IU/l (range; 5–65), respectively. No patient had clinical significant liver failure. The median serum albumin concentrations at inclusion were 32 g/l (range; 23–42). Table 1 Patient characteristics Karnofsky performance status (median (range)) 60 (40–80) Cancer Diagnoses Prostate 11 Colorectal 5 Kidney 3 Breast 3 Pancreatic 2 Lung 2 Gastric 1 Malignant melanoma 1 Leiomyosarcoma 1 Metastases Liver 7 Bone 16 Other 16 Antidepressants 4 Neuroleptics 1 Benzodiazepines 6 Corticosteroids 13 Antiemetics 7 Ten patients used non-opioid analgesics (nine paracetamol, one acetylsalicylic acid). The patients used a median number of 5 (range; 1–10) non-pain medications. The numbers of patients using psychotropic drugs, antiemetics or corticosteroids are given in Table 1 . All except one patients received laxatives, lactulose and bisacodyl, during the study period. All medications were stable during the study period. Similar pain, nausea and sedation scores were observed throughout the trial periods (Table 2 ). Twenty-seven patients had died at the time of manuscript preparation. The median survival time from inclusion was three months. Table 2 Symptom scores All scores were obtained using a 5 category verbal rating scale score (scores; 1–5). All results are given as mean (SD). No significant differences in scores were observed between trial days. Trial day 2 Trial day 3 Trial day 4 Pain 2.8 (0.7) 2.2 (1.1) 2.2 (1,1) Nausea 1.7 (1.0) 1.7 (1.1) 1.8 (0.9) Sedation 3.4 (1.2) 3.4 (1.0) 3.2 (1.1) Of the nineteen patients receiving oral morphine sixteen patients completed one trial period, two patients completed two trial periods and one patient completed three trial periods. The corresponding numbers for the ten patients receiving sc morphine were four, five and one, respectively. Six patients were excluded during a study period. The reasons were; discharge from hospital (n = 2), opioid treatment changed to fentanyl patch (n = 1) and fatigue (n = 3). Opioid induced adverse effects caused no exclusions. Sixteen blood samples were not obtained due to circumstances related to the patients' or relatives' needs (e.g. visits from relatives at the time of a planned blood sample). Morphine treatment The median duration of morphine treatment before entering the study was 7 months (range; 0–29). The median morphine dose at inclusion for the patients receiving oral treatment (controlled-release morphine) was 90 mg/24 h (range; 20–1460). The median morphine dose for the patients receiving sc morphine infusions was 135 mg/24 h (range; 30–340). The morphine doses varied between the study days because the patients were allowed to use rescue morphine. This variation, however, was minor (Table 3 and 4 ). Table 3 Serum concentrations for morphine and metabolites for the oral route The morphine doses (mg/24 h) vary because of variable doses of rescue morphine. A total of 23 trial periods in 19 patients were studied. All data are given as median and range. Day 1 Day 2 Day 3 Day 4 Morphine dose (mg/24 h) 90 (20–1460) 80 (20 – 1700) 95 (30 – 1520) 90 (20 – 1580) Serum morphine (nmol/l) 255 (46–2520) 59 (17–1437) 94 (12–1429) 77 (9–2296) Serum M6G (nmol/l) 1156 (149–7874) 568 (66–7874) 516 (66–9678) 620 (80–8026) Serum M3G (nmol/l) 6341 (1734–31997) 3696 (404–36887) 3226 (595–41452) 3778 (526–43043) Table 4 Serum concentrations for morphine and metabolites for the subcutaneous route The morphine doses (mg/24 h) vary because of variable doses of rescue morphine. A total of 17 trial periods in 10 patients were studied. All data are given as median and range. Day 1 Day 2 Day 3 Day 4 Morphine dose (mg/24 h) 135 (30–340) 163 (30 – 335) 164 (50 – 440) 150 (84 – 440) Serum morphine (nmol/l) 240 (42–741) 254 (62–1297) 305 (106–1045) 373 (103–1222) Serum M6G (nmol/l) 723 (78–1811) 674 (88–2867) 1009 (374–2023) 1225 (400–2339) Serum M3G (nmol/l) 5350 (578–11784) 4490 (779–16312) 5631 (3028–8342) 6119 (2777–13715) The diurnal distributions of rescue morphine administrations were recorded in order to assess the possible influence from rescue morphine on the serum concentration observations. Three blood samples were obtained during the two-hour interval following an administration of rescue morphine. The results from these samples were excluded from the analyses. Morphine, M6G and M3G serum concentrations The median serum concentrations of morphine during oral morphine treatment ranged from 59 to 255 nmol/l during the four study days. The median serum concentrations for M6G and M3G on each study day for patients receiving oral morphine are given in Table 3 . The median serum concentrations were more stable during sc morphine treatment compared with oral treatment. The median serum concentrations of morphine during sc morphine treatment ranged from 240 to 373 nmol/l during the study days. The median serum concentrations for M6G and M3G on each study day for patients receiving sc morphine are given in Table 4 . Day-to-day variation The median biological coefficient of variation (CV) for morphine serum concentrations was 46% (range; 13–103) during oral morphine therapy and 10% (range; 0–36) during sc morphine infusions (Table 5 ). The median biological CV values for M6G were 25% (range; 1–72) during oral therapy and 13% (range; 2–40) during s.c. therapy. The corresponding results for M3G serum concentrations were 18% (range; 0–57) and 9% (range; 0–34), respectively (Table 5 ). Table 5 Biological coefficients of variation (CV) of morphine, morphine-6-glucuronide (M6G) and morphine-3-glucuronide (M3G) serum concentrations during four consecutive days of oral or subcutaneous morphine treatment. All values are given as median and range. Biological coefficient of variation % Oral morphine Subcutaneous morphine Morphine 46 (13–103) 10 (0–36) M6G 25 (1–72) 13 (2–40) M3G 18 (0–57) 9 (0–34) Discussion Intraindividual fluctuation of drug serum concentrations not explained by changes in doses, administration of other drugs or by time for collection of blood samples, is the day-to-day variation. Routine measurements of serum concentrations of morphine and metabolites are of questionable value because of the large variability of minimum effective serum concentration and the lack of a direct relationship between serum concentrations and adverse effects [ 19 ]. However, measurements of serum concentrations of morphine and metabolites are of importance in patients displaying unexpected opioid toxity [ 4 , 7 ] Physicians assessing results from serum drug concentrations determinations should be aware to what extent serum concentrations of drugs fluctuate during stable treatment conditions. Without this knowledge differences and changes in serum concentrations observations may be unduly interpreted. The available data on day-to-day variability during chronic morphine treatment is sparse. Vermeire et al. reported day-to-day variations during morphine treatment in eight cancer patients receiving continuous sc morphine infusion for 1 to 23 weeks. The individual CV values observed in their study varied between 26% to 56% for morphine, 20% to 51% for M6G and 20% to 49% for M3G [ 14 ]. We observed less day-to-day variations of morphine and metabolites concentrations during sc morphine treatment (morphine 10%, M6G 13%, M3G 9%) compared to the fluctuations reported by Vermeire et al. . One explanation for this discrepancy is that Vermeire et al. obtained blood samples during treatment periods up to 23 weeks. This study design may overestimate day-to-day variability since patient related factors will vary more during long time intervals than between consecutive days. To our knowledge this is the first study to assess the day-to-day variation of morphine and morphine metabolites serum concentrations during oral morphine therapy. The median observed CV values for serum concentrations of morphine, M6G and M3G during oral morphine therapy (morphine 46%, M6G 25%, M3G 18%) were higher than the CV values observed in patients receiving sc morphine treatment (morphine 10%, M6G 13%, M3G 9%). This observation was expected due to a more stable delivery rate and since absorption during sc administration is not influenced by food intake, gastric retention, malabsorption, vomiting or variable first-pass metabolism. The results in our study, as in the study by Vermeire et al. , represent day-to-day variability in cancer patients admitted to a palliative care unit. In this patient population pharmacological observations will be influenced by variations in food intake, gastric retention, malabsorption, effects from other drugs on gastric emptying, vomiting and drug interactions. In order to perform a study on day-to-day variation not suspect to these confounding factors patients or volunteers must be recruited into a controlled experimental environment. We believe that studies in controlled experimental environments and studies in patients with advanced cancer disease are complementary to each other. The first targets the pharmacokinetic phenomenon of day-to-day fluctuations, the second targets the fluctuations met during clinical real-life conditions. We recognise some limitations in our study. First, blood samples were collected during four trial days. An extended trial period in order to obtain a larger number of samples from each patient gives a more precise estimate of day-to-day variation. However, due to ethical considerations, taking into account the strain on each patient from serial blood sampling, we chose to not extend the trial periods beyond four days. A second potential confounding factor is absorption peaks in serum concentrations caused by rescue doses of morphine. We chose to allow for rescue morphine because we wanted to observe the variability of serum concentrations as observed in a normal clinical setting in patients considered to be clinical stable in respect to pain treatment. We belive that the variability caused by serum concentration peaks is limited since samples obtained within a time interval of two hours after administration of a morphine rescue dose were excluded. However, it is important to recognize that in order to observe the exact pharmacological day-to-day variability of serum concentrations of morphine and morphine metabolites a design with a stable baseline morphine dose and a non-morphine alternative for breakthrough pain should be applied. Third, the use of rescue morphine implies that the daily morphine doses were not constant. However, the small changes in daily morphine doses can not explain the observed day-to-day variability. In this study we assessed clinical symptoms related to opioid treatment in order to verify the stable intensities of symptoms during the study period. We did not attempt to explore the relationships between serum concentrations and clinical outcome measures. As a rule of thumb 25 patients are required per independent variable in order to give valid results in studies exploring the effects from factors predicting clinical observations [ 20 ]. Consequently, the size of this study was not sufficient to investigate the relationships between opioid serum concentrations and clinical symptoms. Conclusions Morphine, M6G and M3G serum concentrations vary considerably in samples obtained on consecutive days. Such variability is present during stable morphine doses and stable clinical symptoms. The day-to-day variability was lower in patients receiving continuous sc morphine infusions compared with patients receiving oral morphine. These findings indicate that results from blood samples taken in order to assess a patient's pharmacological morphine status should be interpreted with the understanding of that variability is partly caused by day-to-day variation. Competing interests The authors declare that they have no competing interests. Authors' contributions PK, PH and JM participated in the design of the study, running of the study and preparation of the manuscript. SK, PCB and OD participated in the design of the study and preparation of the manuscript. KZ was responsible for measurements of morphine and morphine metabolite serum concentrations. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526195.xml
544937	Reticulate sympatric speciation in Cameroonian crater lake cichlids	Background Traditionally the rapid origin of megadiverse species flocks of extremely closely related species is explained by the combinatory action of three factors: Disruptive natural selection, disruptive sexual selection and partial isolation by distance. However, recent empirical data and theoretical advances suggest that the diversity of complex species assemblages is based at least partially on the hybridization of numerous ancestral allopatric lineages that formed hybrids upon invasion of new environments. That reticulate speciation within species flocks may occur under sympatric conditions after the primary formation of species has been proposed but not been tested critically. Results We reconstructed the phylogeny of a complex cichlid species flock confined to the tiny Cameroonian crater lake Barombi Mbo using both mitochondrial and nuclear (AFLP) data. The nuclear phylogeny confirms previous findings which suggested the monophyly and sympatric origin of the flock. However, discordant intra-flock phylogenies reconstructed from mitochondrial and nuclear data suggest strongly that secondary hybridization among lineages that primarily diverged under sympatric conditions had occurred. Using canonical phylogenetic ordination and tree-based tests we infer that hybridization of two ancient lineages resulted in the formation of a new and ecologically highly distinct species, Pungu maclareni . Conclusions Our findings show that sympatric hybrid speciation is able to contribute significantly to the evolution of complex species assemblages even without the prior formation of hybrids derived from allopatrically differentiated lineages.	Background Recent empirical data and theoretical advances suggest that the diversity of complex species assemblages is based at least partially on the hybridization of numerous ancestral allopatric lineages that formed hybrids upon invasion of new environments [ 1 - 5 ]. A growing amount of studies show that cytoplasmatic (mitochondrial or chloroplast) gene phylogenies of recent diverse species radiations often conflict with phylogenies based on numerous nuclear genes [ 3 , 5 - 9 ]. Theoretical arguments as well as empirical evidence from hybrid zones predict that in newly colonized habitats the effect of transgressive segregation, i.e. the generation of extreme traits in hybrid populations, may lead to a drastically increased phenotypic variation. This effect may in turn serve as a substrate for evolution of novel adaptive traits [ 2 , 10 - 12 ]. These arguments in combination with an increasing body of evidence showing that species resulting from interspecific hybridization are common in plants [ 13 ] and highly probable in animals [ 5 , 7 , 9 - 14 ] gave rise to hypotheses about a prominent role of hybridization for the evolution of adaptive radiations [ 6 ]. Both, the initial formation of hybrid swarms of originally allopatric populations meeting in a newly colonized habitat ("hybrid swarm origin hypothesis") as well as secondary hybridization of in situ diverged lineages ("syngameon hypothesis") could possibly explain the rapid formation of megadiverse species flocks. The scenario may either involve secondary localized hybridization, i.e. hybridization of parapatric ("microallopatric") lineages within the geographical range of the primary radiation, or alternatively hybridization of sympatrically diverged lineages. Distinguishing between these two alternatives is central for the understanding of the processes that lead to the evolution of megadiversity. The first alternative predicts that increased species richness due to hybridization is dependent primarily on the spatial scale and the accompanied possibility to establish localized metapopulations. The second alternative predicts that hybridisation can aid the build-up of diversity even under fully sympatric conditions. Several studies, some published, some in preparation support the hybrid swarm origin hypothesis for some species assemblages endemic to comparatively large areas [ 3 - 5 , 8 ] but a critical evaluation of the syngameon hypotheses rests on the ability to test for sympatric hybrid speciation. However, despite increasing evidence for sympatric speciation, uncontested examples remain rare and rarely go beyond the formation of single species-pairs [ 15 , 16 ]. In addition, evidence for sympatric speciation of complex species-assemblages is often based on mitochondrial phylogenies of limited taxon-sampling. This is problematic as mitochondrial phylogenies or those based on few nuclear loci may obscure true species phylogenies either due to introgressive hybridization among already established species or due to incomplete lineage sorting during rapid speciation. Hence it is not surprising, that studies applying several nuclear markers occasionally yield phylogenetic hypotheses about the origin and pattern of sympatric species assemblages which contrast with mitochondrial hypotheses. As a consequence of the uncertainty about phylogenetic relationships among members of complex species flocks questions about the processes that contribute to sympatric speciation remain difficult to test due to the lack of appropriate model systems. Until recently the phylogeny of mitochondrial lineages of the cichlid species flock of crater lake Barombi Mbo (Cameroon) [ 17 ] was considered as one of the best examples for sympatric speciation [ 15 ]. According to this phylogeny which was based on haplotypes of single specimens, the monophyly of the 11 endemic species suggested strongly that they had formed after a single colonisation by a riverine founder species, Sarotherodon galilaeus . Because the lake's conical basin is only 2.15 km in diameter, because there are no migration-barriers along the shore, and because the lake is isolated from nearby river systems by cataracts of its outflow, allopatric scenarios for the origin and diversification of the flock were ruled out. However, in the light of the aforementioned methodological drawbacks of mitochondrial phylogenies, both the monophyly-hypothesis for the Barombi flock and the relationships among its 11 endemic species are worth to be reevaluated. The ability to score thousands of amplified fragment-length polymorphisms (AFLPs) has created a powerful possibility for the phylogenetic reconstruction of rapidly originated species flocks. It has been successfully applied to a limited number of taxa belonging to the Lake Malawi and Lake Victoria haplochromine species flocks and revealed previously undetectable phylogenetic patterns including those supporting the hybrid swarm origin hypothesis [ 4 , 18 , 19 ]. In this study, we tested hypotheses about sympatric speciation with a focus on hybridization by applying a combination of mitochondrial DNA-sequencing and AFLP-genotyping as well as a set of recently proposed analytical tools [ 20 ] to the phylogenetic analysis of a complete and complex species flock. Results Mitochondrial phylogenetic inferences We obtained a DNA sequence-alignment with 2553 bp including two complete mitochondrial genes, NADH dehydrogenase subunit 2 (ND2) and cytochrome b (cytb), partial proline tRNA as well as from part of the control region from all Barombi species (two samples per species) and relevant S. galilaeus populations (one to two samples per population). 2191 sites of the alignment were constant, 198 variable characters were parsimony-uninformative and the number of parsimony-informative characters was 164. Empirical base frequencies in this data set were A = 0.2721; C = 0.3271; G = 0.1268; T = 0.2740. Bootstrapped Maximum Parsimony (MP), Maximum Likelihood (ML) and Neighbour Joining (NJ) trees all recovered identical 50%-majority rule consensus-trees (figure 1 ). As the sistergroup to the monophyletic Barombi flock a Sarotherodon galilaeus clade was recovered which includes all west African populations except S. galilaeus sanagaensis which emerged as the sistergroup to all other ingroup taxa. Within the Barombi Mbo flock four lineages were recovered with high bootstrap support, one containing the predators of genus Stomatepia , one combining the fine-particel feeders of the genus Sarotherodon , one consisting only of the dwarf zooplanctivore Myaka and one containing the macro-invertebrate or eggfeeding sistertaxa of the genus Konia plus the highly specialized spongivore Pungu . For taxa represented by more than one sample, all conspecific samples grouped together except those of the morphologically merely distinguishable S. caroli and S. linnellii . A rough time estimate as deduced from the ultrametric tree (chronogram) [ 21 ] derived from non parametric rate smoothening (NPRS) of bootstrapped ML-distances suggests that all four lineages almost simultaneously came into existence, which must have taken place approx. 1 myr years ago. Soon after this primary radiation, the divergence of the Pungu haplotypes from Konia took place, while all other clades radiated into several species much later. A 94 sample data set with only cytochrome b and partial proline tRNA sequences (3 to 7 samples for the Barombi taxa, and 1 to 7 for all other; 1212 bp with 1003 constant characters, 59 parsimony-uninformative and 150 parsimony-informative characters) confirmed the previous findings for the Barombi flock and suggests that lineage sorting between the four large clades is complete (for a Neighbour Joining Tree see Additional File 1 ). However, between species within the clades lineage sorting was only complete for a subset of taxa ( Pungu , Myaka , Konia ssp , S. lohbergeri and S. steinbachi ), but not within Stomatepia ssp ., S. caroli and S. linnellii . AFLP based phylogenetic inferences The phylogenetic reconstruction based on the same individuals and using 22 restrictive primer combinations with 3489 AFLP size fragments (3004 variable) confirmed the monophyletic origin of the Barombi Mbo flock and the monophyly of the Stomatepia and Konia clades (figure 1 ). However, several other phylogenetic groupings were recovered with high bootstrap support, which contrasted conspicuously with the mitochondrial phylogenetic hypothesis. S. gal. multifasciatus and S. gal. "Niger" were recovered as sistergroups to the clade containing all other ingroups including S. gal. sanagaensis . Within the Barombi flock Pungu is now sistergroup to a Sarotherodon subclade ( S. steinbachi and S. lohbergeri ), Myaka is sistergroup to the other Sarotherodon subclade consisting of the pelagic species S. caroli and S. linnellii . The Konia clade is resolved as the sistergroup to the rest of the flock. In contrast to the unresolved basal topology in the mitochondrial tree, the two species from Lake Ejagham are resolved as a monophylum which is the sistergroup to the geographically closest S. galilaeus population from the river Cross. An additional data set with 3 selective amplifications but 80 samples recovered only three supraspecific nodes with moderately high bootstrap support, the monophyly of the Barombi flock, Konia and the node containing the rest of the Barombi flock without Pungu . The latter was placed intermediate between the two well supported nodes. Testing for sympatric reticulate speciation Both the Shimodaira-Hasegawa and Templeton's test confirmed significantly the difference for the alternative tree topologies for each sequence and AFLP data sets, respectively (figure 1 ). These discordant phylogenies suggested strongly that hybridization among previously evolved lineages had taken place and that at least one taxon of the Barombi Mbo flock, Pungu maclareni , is the result of speciation by hybridization. By identifying the clades which contain taxa with discordant phylogenies we hypothesized that traces of three ancient hybridization events are still detectable in the multilocus AFLP data. To test for the presence of the respective phylogenetic signal for these three hypothetical ancient syngameons in the large AFLP data set, we used the recently developed method of Canonical Phylogenetic Ordination (CPO) [ 20 ]. In addition, this method is useful for differentiating between contributions to variation in the observed AFLP character pattern that were generated by the segregation of ancestral polymorphisms inherited from a common ancestor due to incomplete lineage sorting rather than by the contribution of derived characters of hybridizing lineages. This, as the contribution to the variation that is assignable to the phylogenetic group uniting the common ancestor of the hybridizing lineages (coded as phylogenetic variables) is partialled out in the CPO separately from the contribution of the phylogenetic groups characterizing the derived lineages that may have hybridized (see also Methods section). All phylogenetic groups detected either by the mitochondrial or the AFLP data set, or formulated according to the three hypothetical syngameons were coded as phylogenetic variables and their explanatory value for the variance in the AFLP data set was tested (table 1 ). As a result, almost all supraspecific clades which were not conflicting among the two data sets were recovered as contributing significantly to the variation in the data set. Conflicting clades, which were either found in the mitochondrial or AFLP phylogenetic hypothesis, yielded non-significant results except for the AFLP based nodes uniting (1) all S. galilaeus and lake taxa without S. gal multifasciatus and S. gal "Niger" and (2) the node of the AFLP based hypothesis of the monophyly of the Lake Ejagham sister pair. Of the three hypothetical syngameons, the one uniting Pungu with its potential ancestor clades represented now by the Konia ssp and S. steinbachi and S. lohbergeri contributed significantly to the variation in the AFLP data set, whereas the other two did not. As a final test for the putative hybrid origin of Pungu , a tree-based method as outlined in Seehausen [ 6 ] was performed in order to test for homoplasy excess introduced by this potential hybrid taxon in the AFLP-data set (figure 2 ). Only the removal of Pungu resulted in far outlying higher support values for the respective nodes, whereas all other removals resulted in much lower values (Fig. 2 ). In addition, the other nodes tested were not affected by the removal of Pungu . Interestingly, the removal of several other taxa produced far outlier values in other nodes. However, these removals concerned part of the two hypothetical syngameons that were not supported significantly in the CPO, which were therefore not explicitly tested. The removal of Konia dikume resulted in an increased bootstrap support for the clade uniting Konia and Pungu in the analysis of the 530 loci data set and an accompanying though weaker signal in the S. caroli / S.linnellii clade. The removal of S. caroli resulted in a distinct although weak increase of the bootstrap support for the Myaka / S. linnellii node, the exclusion of both Myaka and Konia eisentrauti increased the bootstrap support for the S. linnellii/S. caroli node. Finally, the removal of S. mariae increased the support for the S. pindu/S. mongo split strongly. Among the riverine populations of S. galilaeus , the removal of S. gal. "Meme" increased strongly the value for the clade uniting the S. gal. "Cross" specimen with the two endemic species from Lake Ejagham (Details in Additional File 2 ). These additional findings suggest that the comparative phylogenetic analysis of mtDNA and AFLP data alone was not sufficient to uncover all possible hybridization events. Figure 2 Box-plots of the distribution of %-bootstrap support values for the nodes uniting the two Konia species (Figs. 2a) or Sarotherodon lohbergeri and S. steinbachi (Figs. 2b) after iterative removal of single species or taxon groups. Values are based on 2000 bootstrap replicates in the AFLP-based tree reconstruction using the Link et al algorithm [48]; they are based on either the 32-sample/2355 loci dataset for the S. lohbergeri / S. steinbachi split or on the 80 sample / 530 loci dataset for the Konia split. Outside () and far outside values (°) are plotted as asterisks and circles, respectively. Arrows denote far outside values resulting in a distinctly higher bootstrap support for the two clades after exclusion of Pungu maclareni . n refers to the number out of 18 maximum possible removal experiments. Removal of Pungu did not result in outside or far outside bootstrap support values for any other node out of 30 nodes tested (Additional information concerning far outside values yielded for other nodes see Additional File 2). Discussion Our results demonstrate that the sympatric origin of a diverse and complex species flock was aided substantially by reticulate evolution among lineages that emerged in a much smaller primary radiation. Our data suggest that at least one out of 11 taxa of the species flock in Lake Barombi Mbo, Pungu maclareni , is the result of speciation by hybridization. On the other hand, conflicts among mitochondrial and nuclear data sets as well as results of the homoplasy excess tests suggest that in the course of the evolution of the flock hybridization must have taken place among several additional Barombi taxa, too. According to the ultrametric time-calibrated tree ("chronogram") based on the well supported phylogeny of mitochondrial haplotypes in the lake, the primary radiation in Barombi Mbo resulted into the almost instantaneous split into four distinct lineages approximately one million years ago. The accumulation of numerous apomorphic characters that support these mitochondrial lineages suggests strongly that they represented reproductively isolated species at that time. Only in an advanced stage of species flock formation and after considerable time had elapsed, their cohesion was broken partially by hybridization events between these lineages. However, according to the chronogram the ancestral mitochondrial clades that contributed, for example, to the hybrid origin of Pungu continued to accumulate apomorphic characters well after the origin of Pungu . This suggests that the species status of the ancient hybridizing lineages in terms of sufficient reproductive isolation must have allowed for their ongoing genetic cohesion and accompanied coalescence of haplotypes before additional speciation events took place. Traditionally the rapid origin of megadiverse species flocks of extremely closely related species was explained by the combinatory action of three factors: Disruptive natural selection, disruptive sexual selection and partial isolation by distance [ 22 - 24 ]. Although introgression among species is known for many fish species [ 7 , 9 , 25 - 27 ] and although reticulate evolution and hybrid origins of species are well documented in plants [ 14 ], it is only of recent that hybridization has been proposed to play a major role in generating diversity in animals in general [ 27 - 30 ] and in "explosive" speciation in species flocks in particular [ 6 ]. Especially in newly colonized habitats with increased ecological opportunities, secondary hybridization of primarily diverged lineages may provide rapidly sources of heritable advantageous variation by producing additional adaptive diversity through recombination of functional genotypes. Interestingly, the species with the most likely hybrid origin in Lake Barombi Mbo, Pungu maclareni , represents an ecologically highly specialized ecotype. Both its peculiar dentition and the accompanying hypertrophic jaw-muscles are unique not only in Barombi Mbo but in cichlids in general [ 31 ]. Accordingly, one putative second species with a hybrid genome in the lake, Konia dikume , ranks among the most unusual cichlids as it is the single species which is able to exploit chironimid larvae in the almost oxygen-free deep water due to its extremely high haemoglobin concentration in its blood [ 32 ]. In the light of our findings we hypothesize that hybridization produced these extreme phenotypes by transgressive segregation which allowed the exploitation of extreme niches. This supports the notion that speciation by hybridization is not only able to produce additional random variation but may significantly increase the ecological complexity in a rapidly evolving species community by providing extraordinary genetic opportunities. If indeed transgressive segregation of hybrid genotypes plays a major role in the evolution of evolutionary novelties, members of complex species-assemblages with unusual ecological adaptations should predictably turn out to be of hybrid origin more often than species with common adaptations. Methods Taxon sampling, collection of samples and deposition of vouchers We obtained genetic data from relevant Sarotherodon galilaeus populations and from all species endemic to crater lakes Barombi Mbo and Ejagham, which are related to S. galilaeus . Oreochromis niloticus and Sarotherodon melanotheron were used as outgroups based on published information and pilot-study data confirming their outgroup status with respect to S. galilaeus and the investigated species flocks. Adult specimens from Lake Barombi Mbo were collected by UKS during field visits to the lake in 2001 and 2002 and from Lake Ejagham in 1994. Identification to species is according to Trewavas et al. [ 33 ] and was straightforward except for Sarotherodon caroli and S. linnellii . Black adult breeding males of the S. caroli/S. linnellii -phenotype from the deepwater were identified as S. caroli , whereas golden males from the shallow inshore area were identified as S. linnellii . Fin-clips were taken from the right pectoral fin in the field directly after collection and preserved in 96% Ethanol p.A.. Samples from additional specimens belonging to different populations of Sarotherodon galilaeus and outgroups were either collected by UKS in the field in Cameroon or donated by others. Vouchers are deposited in the ichthyological collection of Bavarian State Collection of Zoology (ZSM). Of few S. galilaeus -samples, only photographs of the specimens are available and are deposited in the ZSM, too. Informations about specimens and their species identifications, geographic origin, accession numbers and information about which specimens were sequenced and AFLP typed in different data sets are provided in Additional file 3 . Molecular Methods DNA Preparation DNA samples were isolated from approx 10 mm 2 fin tissue with the DNeasy™ Tissue Kit (Qiagen). DNA-quality was visually inspected under UV-light on a 0.8% agarose-gel stained with ethidium bromide. For subsequent AFLP-analysis only samples with a clearly visible high-molecular band were used. DNA-concentration was determined using the VersaFluor-Fluorometer-System (BioRad) using the stain Picogreen ® dsDNA Quantitation Kit (Molecular Probes). All samples were adjusted to 60 ng/μl. Mitochondrial DNA Amplification and Sequencing Two different data sets were assembled, one long data set with approx. 2550 bp including the complete NADH dehydrogenase subunit 2 (ND2), the complete cytochrome b (cytb) gene and part of the proline tRNA, and finally, one part of the control region. For this long data set only two samples per species were used, but a short data set with more individuals but only the cytochrome b and the partial proline tRNA genes was generated, too. ND2 was PCR-amplified using the primers "ND2Met" 5'-CATACCCCAAACATGTTGGT-3'"ND2Trp" 5'-GTSGSTTTTCACTCCCGCTTA-3'; a second fragment containing the cytb, proline and threonine tRNAs and the 5'-end of the control region was amplified with the primers "L14725" 5'-TGACTTGAAAAACCATCGTTG and "H16498" 5'-CCTGAAGTAGGAACCAGATG [ 34 ] ; internal sequencing primers were the newly designed "cytL640" 5'-CACGAAACCGGATCAAAC-3' for cytochrome b and "L71" 5'-TACCCCTAGCTCCCAAAGCT-3' 7 for the 5'-end of the control region. PCR was performed by using a PTC 220 DYAD thermocycler (MJ Research) in a 25 μl reaction volume using the Expand PCR system (Roche Diagnostics) with 25 pmol of each primer, 20 pmol of dNTPs, 12.5 pmol MgCl 2 and 0.88 units of Taq polymerase. PCR parameters were 94°C for 4 min, 35 cycles with 94°C for 1.5 min, 55°C for 1 min, 72°C for 1.5 min, followed by a final elongation at 72°C for 3 min. PCR products were cleaned by using MinElute PCR purification kit (QIAGEN) and their DNA-concentration adjusted to 100 ng/μl. PCR-Products were then used as templates for cycle-sequencing reaction using the "Ready Reaction DyeDeoxy Terminator Cycle Sequencing Kit" (Applied Biosystems) with each of the PCR primers or internal primers. Cycle parameters were the following: 94°C, 2 min; 25 cycles of 94°C, 20 s; 52°C, 10 s; 60°C, 4 min. The sequenced product was filtered through Sephadex-G50 fine (Fluka) packed spin columns (Amersham) to remove unincorporated dye terminators, primers, and salts, and finally dried in a speed-vac. These products were resuspended, electrophoresed and analysed with an ABI PRISM™ 377XL-96 automated sequencer using a 4.25 % polyacrylamid gel (BioRad). Electrophoretic information was transcribed to sequence data using the program Genescan (PE Applied Biosystems). Individual sequence files were edited and contigs assembled using Sequence Navigator™ (PE Applied Biosystems). Homologous protein-coding regions (ND2, cytb) were aligned manually and confirmed by translating DNA data into amino acid sequences in BioEdit [ 35 ]. The short fragment of the control-region was first aligned with default settings in Clustal W as implemented in Sequence Navigator™. No indels larger than 1 basepair (bp) were detected and alignment therefore was straightforward. All sequences were tested for an anti-G bias characteristic of the mitochondrial DNA to confirm that we have collected genuine mitochondrial DNA data [ 36 ]. Sequence data have been deposited in GenBank (for accession numbers se Additional File 3 ). Amplified Fragment-Length Polymorphisms (AFLPs) We followed the original protocol of the AFLP-method [ 37 ] using the AFLP™ Plant Mapping Kit (Applied Biosystems) with slight modifications of the accompanied protocol: Restriction and ligation were carried out in a single step under standardized conditions in a thermocycler (2 h at 37°C and 8 h at 16°C). 1,5 μl of the preselective amplification product were used in only 10 μl total reaction volume of the selective amplifications. The restriction enzymes used were EcoRI and MseI. Primer sequences for preselective PCR were GACTGCGTACCAATTCA and GATGAGTCCTGAGTAAC. An additional two bases were added to the 3' end for selective PCR. Analogous to the mitochondrial data set we assembled two datasets. For the long one in total 22 primer pairs were used in the following combinations and fluorescent dye-labelling (MseI-primer/ EcoRI DYE ): TC-CA FAM , AT-CA FAM , AC-CC NED , TA-CA FAM , AA-CT FAM ; AA-GG JOE ; AC-CA FAM ; AA-CA FAM ; AG-CA FAM ; AA-CG JOE ; AC-CT FAM ; AG-CT FAM ; AT-CT FAM ; AG-GG JOE ; TC-CT FAM ; TA-GG JOE ; AG-AC NED ; AT-AC NED ; AG-CC NED ; AC-AC NED ; AT-CC NED ; TA-CG NED . For the short data only the five primer pairs TC-CA FAM , TA-GG JOE , AC-CA FAM , AC-CC NED , AT-CA FAM were used but typed for 80 individuals (see Additional File 3 ). Selectively amplified fragments were separated on 6% LongRanger polyacrylamid gels (FMC BioProducts) with an ABI PRISM™ 377XL-96 sequencer. Fluorescent signals were detected using the GENESCAN software (Applied Biosystems) with internal size standard (GS-500 ROX; Applied Biosystems). The fluorescent threshold was set to 50 units and the correct identification of ROX-marker bands by GENESCAN was checked for all electropherograms. Bands between 100.5 bp and 499.5 bp were scored in a first step for presence or absence using the software Binthere (developed by N. Garnhart and available through the T. Kocher laboratory . The program generates aligned spreadsheets from GENESCAN-sized AFLP-data by assigning each sized fragment to a size-class of user-defined distance to the next size-class. Using a spreadsheet routine, fragments were inferred in a second step to be homologous if they differed by no more than 1.00 bp, and if the scoring procedure identified the same size-classes whether scored from small to large size-fragments (forward) or vice versa (backward). Size-classes with inconsistent allocation of fragments according to forward and backward scoring were excluded, as well as adjacent size-classes differing by less than 0.35 bp, which corresponds to the double standard deviation of 0.15 bp of the sequencer [ 38 ]. As a result of this procedure a final 0/1 data-matrix for all scored individuals was prepared. All samples of the same primer combination were run on same gel for the 33 specimen dataset and on two gels for the 80 taxon dataset. Single unsuccessful amplifications were repeated and fitted to the data matrix using the size-class assignment criteria as outlined above. Phylogenetic Analyses Mitochondrial Sequence Data MrModeltest 1.1b [ 39 ], a simplified version of David Posada's "Modeltest 3.06" [ 40 ] was used to perform hierarchical likelihood ratio tests (HLRT) and to calculate approximate Akaike Information criteria (AIC) to determine the optimal nucleotide substitution models for the dataset. If the two tests did not select the same model, we chose AIC over HLRT, as AIC is a useful measure that rewards models for good fit but imposes a penalty for unnecessary parameters [ 41 , 42 ], which may cause erroneous phylogenetic conclusions especially in Bayesean phylogenetic analyses [ 43 ]. For the 33 sequence dataset the HLRT selected the HK+I+Γ model (α = 0.2781; proportion of invariable sites = 0) a transition/transversion (Ti/tv) ratio of 8.5662 was calculated. The AIC selected the HKY+I+ Γ model (α = 0.9468; proportion of invariable sites = 0.4264) with a transition/transversion ratio of 8.6616. For the 94 sequence dataset, the HLRT selected the GTR+ Γ model (α = 0.4014; proportion of invariable sites = 0), whereas AIC selected the GTR+I model (proportion of invariable sites = 0.5817). Empirical base frequencies in the data 2553/1212 data sets were A = 0.2443/0.2440; C = 0.3264/0.3261; G = 0.1443/0.1446; T = 0.2581/0.2852. The AIC settings were subsequently used for Maximum Likelihood (ML) analyses and to estimate ML distances for minimum evolution (ME) analyses in the program PAUP 4.0b1.0 (PPC/Altivec) [ 44 ] and for the 94 sequence dataset in Treefinder [ 45 ]. Maximum Parsimony (MP) analyses were conducted with heuristic searches (TBR branch swapping and MULTREES option effective; 10 random stepwise additions of taxa for 33 sequence set and simple addition for the 94 sequence set; gaps in the control region treated as a 5 th character). Non-parametric bootstrapping with 1000 (ME or MP analyses) or 100 (ML) pseudoreplicates was used for testing the robustness of the inferred trees. Tree topologies using the HLRT settings under ME were not different from topologies gained with AIC settings (data not shown). A LRT [ 46 ] as implemented in PAUP* was performed with the respective 33 sequence ME tree under the AIC model assumptions with (-ln L = 6770.47061) and without (-ln L = 7120.92833) molecular clock enforced. Overall constancy of rates of evolution was rejected (chi 2 = 701.0154, df = 32, p = 0.001). To date cladogenetic events in the absence of rate constancy, the nonparametric rate smoothing (NPRS) method [ 21 ] as implemented in Treefinder [ 45 ] was used to construct an ultrametric tree ("chronogram") using the bootstrapped 33 sequence ML derived tree-topology and associated bootstrapped branch lengths as input. AFLP Data We used PAUP* [ 44 ] to calculate the skewness parameter g1 [ 47 ] to test for adequacy of phylogenetic signal in the 0/1-data set. g1 calculated from 1000000 random trees revealed significant non-random structure under the parsimony optimality criterion in the 22 restrictive amplifications for the complete data set: g1 -0.805, 33 samples, 3004 variable sites out of 3489 scored); g1 values were lower in the 3 restrictive amplifications data set used for some homoplasy excess tests (see below): g1 -0.298, 80 samples, 717 variable sites out of 859 scored. 2355 and 530 loci respectively were parsimony informative within Sarotherodon galilaeus sensu lato (excluding Oreochromis niloticus and Sarotherodon melanotheron ). "Pruned" data matrices using only those parsimony informative sites were constructed for Principal Canonical Ordination and homoplasy excess tests (see below) in order to account for noise in the data potentially introduced by the distant outgroup. Pairwise genetic distances were calculated from the binary data-matrix with two different algorithms: One developed by Link et al. [ 48 ] as implemented in TREECON v.1.3b [ 49 ], which is based on shared and unique characters and ignores shared absence. This algorithm is adequate for AFLP data, since noise in the data may often be created by weak signal intensities and hence absence of band-detection despite a possible weak presence of signal. Alternatively, we calculated pair-wise distance matrices with the restriction-site program RESTDIST within the PHYLIP 3.6A2 package [ 50 ]. Trees were constructed from the Link et al.-distances with the neighbour joining (NJ) algorithm as implemented within TREECON or from the RESTDIST-distances using the Fitch-and-Margoliash-algorithm [ 51 ] with unconstrained branch length as implemented in the program FITCH within the PHYLIP 3.6A2 package [ 50 ]. Non-parametric bootstrapping was performed with 100 bootstrapped data sets analyzed 10 times with random input orders, and with local and global optimization. Hypothesis testing Alternative phylogenetic hypotheses produced as described by tree topologies based on mtDNA and AFLP data were compared with each other and statistically evaluated using either the Shimodaira-Hasegawa LRT [ 52 ] for mitochondrial data or the Templeton's Wilcoxon signed-rank test [ 53 ] for AFLP-data (both as implemented in PAUP). In order to test for the presence of a phylogenetic signal that possibly reflects reticulate events in the AFLP-data, two methods were applied. First, a canonical correspondence analysis (CCA) was performed using CANOCO 4.0 [ 54 ]. This method has previously been used successfully for testing the effect of tree-like hydrogeographic data and supplementary ecological data on microsatellite allele-frequencies in freshwater fishes [ 55 ], as well as an alternative to traditional phylogenetic comparative methods [ 20 ]. A presence/absence matrix with the 2355 AFLP-characters which were parsimony-informative within the Sarotherodon galilaeus -clade ( S. galilaeus sensu lato and lake endemics) provided the data-matrix to be tested. Phylogenetic hypotheses derived from mtDNA- and AFLP-analyses as well as hypothetical syngameons as derived from the conflict between the two data-sets were translated into a phylogenetic matrix by assigning binary indicator variables, each coding for the membership of investigated samples to phylogenetic groups (e.g. nodes in phylogenetic trees or hypothetical syngameons) [ 20 , 54 ]. 9999 full model Monte Carlo (MC) permutations were used to test whether a given phylogenetic group as coded by the indicator variables and identified by automatic forward selection of variables was significantly related to the AFLP-data pattern. Second, a tree-based method as outlined in Seehausen [ 6 ] was performed in order to test for homoplasy excess introduced by potential hybrid taxa in the AFLP-data as suggested by the mtDNA-AFLP-phylogeny conflict and the CCA. Theoretically, hybrid taxa are overall intermediate to the parental taxa because they carry a mosaic of parental characters. Consequently, the inclusion of a hybrid taxon into a multilocus based phylogeny estimate introduces an excess of homoplasies and therefore conflict in the subset of clades that contributed to hybridization. Removal of the putative hybrid taxon should therefore decrease the amount of homoplasies and hence increase support for those nodes that unite descendants from taxa which gave rise to a hybrid taxon. In contrast, removal of a non-hybrid taxon should not affect support for the respective nodes. We computed Link et al bootstrap-supports (2000 replicates) for the nodes uniting Sarotherodon steinbachi and S. lohbergeri in the 2355 loci data with n = 16 experiments (each taxon removed once). Analogous support values for the node uniting the Konia eisentrauti and K. dikume were computed with the 530 loci data set with n = 14 experiments, because bootstrap support in the larger data set was always larger than 98.85% and identical runs yielded values differing by more than 1.15 % (data not shown). By reducing the number of loci but increasing the number of samples we obtained a meaningful distribution of bootstrap support values for that node. Authors' contributions UKS designed the study, carried out the field work, part of the lab work, the major part of the data analysis and wrote the manuscript. BK carried out the major part of the lab work and carried out parts of the data analysis. Both authors read and approved the final manuscript. Table 1 Results of canonical phylogenetic ordination of AFLP data Variation explained¶ Node ¥ Marginal effects† λ 1 Conditional effects‡ λ a P-Value# Phylogenetic groups according to mtDNA-based phylogeny S. galilaeus sensu lato incl. Barombi taxa 1 0.15 0.10 0.000 S. galilaeus sensu lato w/o S. g. sanagaensis incl. Barombi taxa 2 0.12 0.05 n.s. S. galilaeus w/o S. g. sanagaensis excluding Barombi taxa 3 0.15 - n.s. S. g. multifasciatus * 4 0.18 0.13 0.000 S. galilaeus w/o S. g. sanagaensis and S. g. multifasciatus 5 0.12 - n.s. S. galilaeus "Meme" * 6 0.10 0.08 0.000 "Cross-clade" + S. g. " Niger " 7 0.10 - n.s. "Cross-clade" * 8 0.09 0.07 0.001 "Barombi Mbo clade" * 9 0.18 0.11 0.000 Stomatepia ssp.* 10 0.08 0.06 0.000 St. mongo * 11 0.07 0.05 0.054 St. mariae * 12 0.06 0.05 0.008 St. pindu * 13 0.05 - n.s. St. mariae – St. pindu 14 0.06 - n.s. Pungu maclareni – Konia ssp. 15 0.07 0.05 0.025 Konia ssp. * 16 0.06 0.06 0.001 Pungu maclareni * 17 0.06 - n.s. Konia eisentrauti * 18 0.05 0.04 n.s. Konia dikume * 19 0.05 - n.s. Myaka myaka * 20 0.05 0.04 n.s. Sarotherodon lohbergeri * 21 0.06 0.04 n.s. Sarotherodon steinbachi * 22 0.05 - n.s. "Barombi Sarotherodon clade" 23 0.08 - n.s. Phylogenetic groups according to AFLP-phylogeny Myaka + S. caroli / S.linnellii 27 0.07 - n.s. Pungu + S. lohbergeri / S. steinbachi 28 0.07 - n.s. St. mongo + St. pindu 29 0.07 - n.s. S. galilaeus s. l. incl. Barombi taxa w/o S.g.multifasciatus and S.g. "Niger" 30 0.21 0.21 0.000 S. lohbergeri + S. steinbachi 31 0.06 - n.s. S. linnellii + S. caroli 32 0.06 - n.s. S. sp. "mudfeeder" + S. sp. "bighead" 33 0.08 0.05 0.013 Hypothetical ancient syngameons according to conflict between mtDNA-based and AFLP-based phylogenetic hypotheses groups ¥¥ P. maclareni + Konia ssp. + S. lohbergeri + S. steinbachi red (24) 0.08 0.06 0.002 M. myaka + S. caroli + S. linnellii + S. lohbergeri + S. steinbachi green (25) 0.09 - n.s. S. g. sanagaensis + S. galilaeus w/o S. g. multifasciatus. S. g. "Niger" blue (26) 0.15 - n.s. Each phylogenetic group as identified by nodes in phylogenetic analyses of mtDNA-, or AFLP-data or by the conflict between the data sets (hypothetical syngameons) was tested for their explanatory value in explaining variance in the AFLP data set. ¥ Node numbers correspond to encircled node numbers for mtDNA-clades and AFLP-clades supported in Fig. 1. ¥¥ Groups refer to hypothetical syngameons as indicated by coloured blocks in Fig. 1a and b. ¶ "Variation explained" refers to eigenvalues of each phylogenetic group that explain part of the variation in the pruned AFLP-data (2355 loci / 33 samples identical to those in Fig. 1a. Sum of all unconstrained eigenvalues was 1.667, sum of all canonical eigenvalues 1.179. † Marginal effects refer to eigenvalues ("fit") of each phylogenetic group if taken singly as the only variable on the pruned AFLP data set. ‡ Conditional effects refer to the increase in eigenvalues ("additional fit") of each phylogenetic group as selected by automatic forward selection. # P-values refer to the significance level of the conditional effects as obtained with a Monte Carlo permutation test under the full model with 9999 random permutations. * mtDNA-clades as indicated in Fig. 1b that are compatible with clades as identified by the AFLP phylogenetic tree in Fig. 1a. Figure 1 Phylogenetic tree (a) based on AFLP-data and chronogram (b) based mtDNA data of all Barombi Mbo cichlids (inside blue box), reference specimens of the closest riverine ancestor of the Barombi flock, Sarotherodon galilaeus , and two undescribed Sarotherodon from Lake Ejagham ( S. sp. "bighead" and S. sp. "mudfeeder" ). Photographs refer to the taxon to the left, different populations of S. galilaeus are depicted by two identical photographs only. (a) Numbers at nodes in the AFLP-tree are bootstrap-values (%) of tree reconstructions using the pruned (above) and unpruned (below) AFLP data set. Topologies were identical except for the position of S. g. "Niger" , which was sistergroup to S. g. multifasciatus when using the unpruned data (bootstrap value: 86). Long terminal branches in the original phylogram were cut to identical lengths for graphical reasons; interior branch lengths are as in the original phylogram. (b) Numbers at nodes in the mtDNA-chronogram refer to bootstrap-values of the ML (above) and MP (below) tree reconstructions. The absolute time scale above the tree is based on the maximum age of the Barombi Mbo crater formation of 1.0 mya [56]. Encircled numbers mark nodes referring to phylogenetic groups tested with PCO (see Table 1). Red, blue and green shaded boxes unite hypothetical ancient syngameons as deduced from the conflict between mtDNA- and AFLP-based phylogenetic hypotheses. Further details of tree reconstruction see Methods . Supplementary Material Additional File 1 Neighbour Joining phylogram of 94 specimens Neighbour. Joining phylogram calculated from complete sequences of the mitochondrial cytochrome b gene (1141 bp) and the partial proline tRNA gene (71 bp) sequences for 94 specimens based on the GTR+ Γ model and rooted with Sarotherodon melanotheron (not shown) Click here for file Additional File 2 Additional box-plots of bootstrap supports in taxon removal experiments. Box-plots of the distribution of %-bootstrap support values for nodes that yielded additional higher outside or far outside values in taxon removal experiments (methodology compare with legend of Fig. 2 of the main text). a. The removal of Konia dikume resulted in an increased bootstrap support for the clade uniting Konia and Pungu in the analysis of the 530 loci data; set. b. The removal of S. caroli resulted in increased outside values for the bootstrap support of the b. the Myaka / S. linnellii node and the c. the clade that unites all Barombi taxa with the exclusion of Konia ; d. the exclusion of both Myaka and K. eisentrauti increased the bootstrap support for the S. linnellii/S. caroli node; e. Finally, the removal of S. mariae increased the support for the S. pindu/S. mongo split strongly. f. Among the riverine populations of S. galilaeus , the removal of S. gal. "Meme" increased strongly the value for the clade uniting the S. gal. "Cross" specimen with the two endemic species from Lake Ejagham ("Cross clade"); in addition, the exclusion of S. lohbergeri increased the bootstrap value for the same node, albeit weakly. Click here for file Additional File 3 Specimens genotyped, Genbank Accession Numbers and vouchers. Specimens included in the study with information on DNA-reference-numbers, voucher deposition in the Zoological State Collection Munich (ZSM), Genbank-Accession numbers and representation of specimens in two AFLP-data sets. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544937.xml
514494	Policing Relative Conflicts of Interest in Social Insects	Workers of social insects prevent other workers laying eggs to increase colony efficiency and not -- as traditionally thought - purely because workers are more related to the queen of the colony	The order and harmony that appears to bless the lives of many social insects, from ants to bees, has long fascinated naturalists. That individual workers seem to routinely sacrifice their own (reproductive) interests for the good of the colony has also piqued the interest of philosophers and kings, for obvious reasons. But scratch the surface and that blissful harmony reveals a complex feat of social engineering that is both exquisitely organized and potentially ruthless. One of the altruistic behaviors that social insects are famous for is that one or a few queens perform most or all of the reproduction in a colony, while workers are, for the most part, non-reproductive. The evolution of this social structure partly stems from the unusual sex determination system of social insects, in which unfertilized eggs (of either workers or queens) develop into males and fertilized eggs (produced only by queens) develop into female queens and workers. This creates unusual relationships between family members that affect how W. D. Hamilton's theory of “kin selection” operates in these species. Kin selection, as elegantly summarized by “Hamilton's Rule,” predicts that the altruistic behavior of workers—that is, investing in the reproduction of others in the colony rather than in their own reproduction—can evolve if the indirect reproductive payoff to workers (i.e., via reproduction by relatives) is higher than the cost of the missed opportunity for direct reproduction. Kin selection revolves around relatedness because relatedness determines the magnitude of indirect reproductive payoffs. However, based on a survey of 50 species of ants, wasps, and bees, Rob Hammond and Laurent Keller now demonstrate that the behavior in the colony cannot be accounted for simply based on relatedness patterns, but that it is necessary to consider how colony efficiency influences behavior. In some social insect colonies, workers do lay eggs, in a sense “cheating” on the other workers who are investing in the queen's reproduction rather than in their own. Such action can be severely penalized by other workers, who aggressively police the colony for the illicit offspring—or behavior—of their guilty colleagues. In honeybees, where this behavior was first shown, workers remove worker-laid eggs within hours by eating them, and, in some ants, more draconian methods lead to the mutilation of the culprit caught in the act of laying. Conflict between ants (Photo: Christian König, www.konig-photo.com ) Why workers police worker reproduction in some colonies and not in others can also be influenced by relatedness. If a queen is monogamous and mates only once, then each worker will actually be more related to her nephew (produced by a sister worker) than to her brother (produced by the queen); in this case, workers should tolerate other workers' male offspring. But if the queen mates more than twice (as in honeybees) or if there are multiple queens heading a colony, then the relationship between workers becomes diluted (they do not all have the same father), and workers are more closely related to brothers than to nephews. In this case, workers should clamp down hard on any worker breeding and raise only the queen's sons (in addition to her daughters). But workers policing the reproduction of their fellow workers could also be advantageous if the energy invested by workers into laying eggs—which would otherwise be used in foraging and legitimate brood rearing—detracts from the overall efficiency and growth of the colony. Although there is some evidence for this “efficiency hypothesis,” it is widely accepted that the driving force behind policing is primarily explained by patterns of relatedness. By doing a detailed comparative phylogenetic analysis of different species, Hammond and Keller put the “relatedness hypothesis” to the test and—contrary to expectations—found evidence that this genetic incentive for workers to police the reproduction of other workers cannot account for its widespread prevalence among social insects. One prediction from the relatedness hypothesis is that the extent to which workers produce male offspring is determined by the relatedness of the workers. By contrast, the efficiency hypothesis predicts no such relationship. In line with this, Hammond and Keller's survey reveals that no matter how related workers are to each other, most males across this broad range of species are produced by queens. In other words, worker-policing does not depend on relatedness, so other factors—such as colony efficiency—must act as an important constraint on worker reproduction. This, Hammond and Keller emphasize, does not amount to showing that kin selection is unimportant—but it does mean that the harmony and regulation of reproduction in social insects is much more complex than expected from simple theoretical expectations based solely on relatedness.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514494.xml

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.