Background

Chlorinated ethene congeners ("chloroethenes") are among the most frequently detected groundwater contaminants in the United States of America and other industrialized countries [1]. Chloroethenes are often incompletely dechlorinated by bacteria in these anoxic environments, leading to an accumulation of vinyl chloride, a Group 1 human carcinogen [2, 3]. Growth-linked reductive dechlorination of vinyl chloride is critical to avoid its accumulation and achieve in situ remediation of chloroethenes [1], but vinyl chloride respiration has only been observed in certain strains of Dehalococcoides[4, 5]. Dehalococcoides is a genus-level phylogenetic group within the Chlorofiexi phylum [6]. Dehalococcoides are strictly anaerobic bacteria that gain metabolic energy exclusively via the oxidation of H2 coupled to the reduction of organohalide compounds [7–9]. This catabolic reductive dehalogenation of organohalide compounds ("organohalide respiration") is catalyzed in Dehalococcoides by heterodimeric, membrane-bound enzymes called "reductive dehalogenases" [10]. Reductive dehalogenases typically contain corrinoid and iron-sulfur clusters as cofactors, and have varied substrate ranges that do not necessarily overlap [10, 11].

The catalytic subunit of reductive dehalogenases is encoded in Dehalococcoides by r eductive d ehalogenase h omologous genes (rdhA). Dehalococcoides possess as many as 36 rdhA per genome [9], but few of the encoded enzymes, RdhA, have been purified and characterized in vitro. Many rdhA are co-expressed [12–16], further confounding a determination of the RdhA responsible for catalysis of an observed reductive dehalogenation activity. The only reductive dehalogenase shown to catabolically reduce vinyl chloride, VcrA, was purified from a highly-enriched vinyl chloride respiring culture dominated by Dehalococcoides strain VS [5]. The operon encoding VcrA, vcrABC, was identified by reverse genetics, and highly-similar vcrA were detected in other vinyl chloride respiring Dehalococcoides cultures [5, 17, 18]. Primers targeting vcrA are now commonly used as an indicator of attenuation potential at vinyl chloride contaminated sites ([5], U.S. Patent Application 20090176210). A putative VC reductase operon, bvcAB, shares only limited similarity with vcrAB and is present in a different VC respiring Dehalococcoides strain, BAV1, which does not contain vcrABC[19].

Although Dehalococcoides are the only known microorganisms capable of vinyl chloride respiration, both vcrA and bvcA appear to be horizontally acquired [9]. Both vcrA and bvcA have a highly unusual, low %(G+C) codon bias that appears maladapted to Dehalococcoides genomes [20], and both are found within a low %(G+C) "genomic island" (GI) [21] that interrupts local gene synteny relative to other Dehalococcoides strains. In strain VS, this vcrABC -containing genomic island (vcr-GI) integrated at the ssrA locus, and as a result is flanked by ssrA and a 20 bp direct repeat of the ssrA 3' end [9]. ssrA is a single-copy gene essential in bacteria [22] encoding t ransfer m essenger RNA (tmRNA), which plays a key role in maintaining the fidelity of protein synthesis [23]. Specific integration of genetic elements at ssrA is also common across many bacterial phyla, and often results in a direct repeat at the genomic island boundary opposite the site of integration [24]. In addition to the vcr-GI, over a dozen ssrA direct repeats were previously detected downstream of ssrA in Dehalococcoides, collocated with many strain-specific rdhA in a region of high genomic variability between Dehalococcoides strains [9]. To further understand the acquisition and dissemination of vcrABC, as well as the impact of ssrA-specific integration on Dehalococcoides genome dynamics, we determined the conserved features of Dehalococcoides ssrA-specific genomic islands (ssrA-GIs) from all publicly available genomes and metagenomes of Dehalococcoides cultures, including the recently-sequenced Dehalococcoides strain GT [17] and the metagenome sequences of the vinyl chloride respiring Dehalococcoides enrichment cultures KB-1 [25, 26] and ANAS [27]. We also amplified and sequenced ssrA-GIs from the vinyl chloride respiring Dehalococcoides enrichment cultures Evanite (EV) [28], PM [28], WBC-2 [29], and WL [30] using primers designed to target either vcr-GIs specifically, or conserved features present in all available Dehalococcoides ssrA-GIs. Previous studies have implicated a subset of rdhA B with horizontal gene acquisition, but evidence for the method of integration, mobilization, replication, and transfer is limited [31, 32]. We describe here a family of putative ssrA-specific integrative and mobilizable elements [33] that share a conserved 'integration module' while also encoding a broad variety of putative and unknown functions, including reductive dehalogenation. The key conserved integrase encoded on these elements is a homolog of the CcrB family of site-specific serine recombinases that specifically integrate/excise the methicillin-resistance element "SCCmec" in Staphylococcus aureus[34]. Using a robust whole-genome phylogeny and several estimates for mutation rate, we estimate the age of the most recent common ancestor of contemporary Dehalococcoides strains, as well as the age of divergence for Dehalococcoides ssrA-GI integration module components.

Results

ssrA Genomic Islands in Dehalococcoides

The region downstream of ssrA in available Dehalococcoides (meta)genome sequences contains multiple tandem genomic islands that are primarily distinguished by their boundaries - ssrA or its 20 bp direct repeat - as well as disruption to local gene synteny and in many cases the presence of a characteristic cluster of integration-associated genes adjacent to the left edge (Figure 1). All direct repeats are located within 100 kbp downstream of ssrA, with varying numbers per strain and no duplicate genomic islands within any strain. These findings are consistent with ssrA-specific integration described for other bacteria [21, 24], as well as a class of integrating and mobilizing elements that encode their own specific integration but do not replicate independently from the chromosome nor encode for conjugation [33]. From available Dehalococcoides genomic data (including this study) we have detected a total of 31 ssrA-GIs containing 47 rdhA, 75 hypothetical protein encoding genes, 2 putative complete CRISPR modules and arrays [35], as well as other genes; most of which are not believed to encode a core function and are present in only a subset of Dehalococcoides strains.

Figure 1
figure 1

General structure of ssrA genomic islands ( ssrA -GIs). (A) Generalized structure of Dehalococcoides ssrA-GIs, oriented according to published Dehalococcoides complete genomes. Labels below genes in the integration module indicate the most informative homolog of the protein encoding gene. (B) Alignment of the 84 bp region surrounding the 3' end of ssrA or its d irect r epeat (DR) fragments (shaded black in the consensus) from 40 such positions in available Dehalococcoides genomes. Each sequence is labeled by its strain or enrichment name, underscore, and the order in which it occurs, beginning with the 3' end of ssrA. Positions in the alignment that disagree with the 75% consensus sequence are shaded in darker grey. The alignment is ordered such that sequences corresponding to Dehalococcoides ssrA ("_01") are the top 10 sequences, emphasizing a conserved position of disagreement between ssrA sequences and the direct repeat regions, position 15 in the alignment, 333 in ssrA. The two bases flanking the inferred integration site are marked with a hash. (C) To-scale genomic maps of region downstream of Dehalococcoides ssrA in (meta)genomic datasets. Orientation of genes is indicated with arrows. Key genes are shaded according to the provided legend.

Full size image

Sixteen of the identified Dehalococcoides ssrA-GIs contain an integration module comprised of 6 syntenic protein encoding genes oriented on the reverse strand and located adjacent to attL (in this context, attL and attR are the ssrA direct repeat sequence at the left or right boundary, respectively; Figure 1A).

Beginning from attL, the integration module contains genes that appear to encode (1) a 540 residue serine recombinase family putative site-specific integrase we call D ehalococcoides s srA-specific i ntegrase, DsiB (Figure 2); (2) a smaller (200aa) PinR (COG1961) homolog that also contains a serine recombinase catalytic domain (cd00338), DsiA; (3) a small (150aa) RecF homolog likely involved in DNA recombination or repair [36], (4) a 210 residue protein with ParBc domain, possibly catalyzing single-stranded DNA cleavage, circular element nicking, element segregation ([37, 38], PF02195); (5) a Mom [39] homolog (270 aa), predicted to play a role in restriction endonuclease resistance via methylation [40, 41]; and (6) a large (700 aa) protein containing a DNA-directed RNA polymerase domain in the first 85 residues (GO:0003899). These integration modules also contain a 76 bp conserved tRNA-like locus embedded within the first 150 bp of the fifth protein encoding gene, approximately 4400 bp from attL (Figure 1A). It is usually labeled as 'pseudo-tRNA' by automated annotation pipelines, but alignment-based RNA folding analysis predicts a complete tRNA-Gly-like structure (Additional file 1 Figure S1). The elevated sequence conservation at its 3' end provides an effective target for primers, as do the regions surrounding ssrA direct repeats and a site of locally high nucleotide conservation within dsiB (Figures 1B, Additional file 1 Figure S2).

Figure 2
figure 2

Phylogeny of ssrA -GI integrase, DsiB. (A) Grey-scale similarity 'barcode' representation (black is identical sites) of pairwise global alignment (Needleman-Wunsch, free end-gaps) between a representative DsiB [DhcVS_1282, Genbank: ACZ62382], and CcrB1 of Staphylococcus aureus [Genbank: ADC39978]. Key domains of CcrB1 are annotated below the alignment, and traced in black if they are also detected in DhcVS_1282 by the conserved domain database search [84] incorporated in PSI-BLAST [85]. (B) Maximum Likelihood tree of the putative integrases encoded on Dehalococcoides ssrA-GIs, DsiB, as well as key integrases involved in mobility of SCCmec in Staphylococcus aureus [64] (unless otherwise noted). The clade of integrase sequences found on vcr- GIs are shaded in blue. Nodes with 100% bootstrap support are bolded with a filled circle. CisA of Clostridium acetobutylicum ATCC 824 is rooted as an outgroup, as in [86]. The following abbreviations are used to label CcrA, CcrB, or CcrC from bacteria other than S. aureus: 'Lys spha' - Lysinibacillus sphaericus C3-41; 'Clost perf' - Clostridium perfringens C str. JGS1495; 'Macca' - Macrococcus caseolyticus [86].

Full size image

Alignment of the ~85 bp surrounding each of the 28 ssrA direct repeats reveals additional nucleotide conservation and allows identification of the site of insertion in ssrA (Figure 1B). A 15 bp motif (TTCAGRSMGMRKCCA) occurs adjacent upstream of the direct repeat and does not align well with the corresponding positions in Dehalococcoides ssrA (318-333), indicating that insertion likely occurs between 333 and 334 in ssrA. This location corresponds to the middle of the T-loop of the encoded tmRNA, between the canonical insertion positions called 'Sublocations II and III' [24] (Figure 1B).

Specific features of vcr-GIs

vcr-GIs are a distinct subset of Dehalococcoides ssrA-specific genomic islands, present in two Dehalococcoides genomes (VS [GenBank:CP001827], GT [GenBank:NC_013890]) and two metagenomes (KB-1 [JGI:4083612], ANAS [JGI:4085297]). Using primers that target conserved features of all ssrA-GIs or specific features of vcrABC, we amplified and sequenced 4 additional vcr-GIs from independently derived vinyl chloride respiring Dehalococcoides enrichment cultures (WBC-2, PM, EV, WL [GenBank:JN034252-JN034255] see Methods). In all instances the vcr-GI is located immediately adjacent to ssrA, except strain GT where it is the second genomic island downstream of ssrA. Because ssrA is an essential single-copy gene encoding a structural RNA [42], its sequence provides a coarse phylogenetic identity of the chromosome from which it was amplified [43]. This allowed confirmation that the vcr-GIs acquired via metagenomic and targeted sequencing are Dehalococcoides chromosomal segments, even though the source genomic DNA was from a mixed culture (Figure 3, Additional file 1 Figure S3).

Figure 3
figure 3

Organization, alignment, and phylogenetic comparison of 8 vinyl chloride reductase genomic islands ( vcr- GIs). (A) To scale summary plot (0 - 100% ID. 14 bp window) of a multiple alignment of all 8 vcr- GIs. Horizontal axis numbers indicate the distance downstream of ssrA, in nucleotides. Bar heights are shaded darker grey when their value is 100%. Position and orientation of genes are annotated above the plot, shaded according to Figure 1. Regions of categorically different similarity correspond to the integration and vcrABC cargo modules, with region-wide average % IDs of 75.0 and 99.4, respectively. (B) Enlarged view of the multiple alignment at key positions. (Left) The presumed boundary between integration and cargo modules. (Right) The region of atypically high substitutions occurring in the leader sequence of vcrA. Identical sequence is shaded light-grey, nucleotides that disagree with the consensus are indicated with tick marks shaded red, green, yellow or blue representing nucleotides A, T, G, C, respectively. (C) Phylogenetic discontinuity between integration modules (left) and their attached vcrABC cargo (right), represented by separately calculated Maximum Likelihood trees. Middle cartoon summarizes the major phylogenetic separations of the trees, with curves connecting modules if one of the 8 vcr- GIs contains the corresponding combination of module types. vcr- GIs sequences are from Dehalococcoides strain VS [GenBank:CP001827], strain GT [GenBank:NC_013890]) and two metagenomes (KB-1 [JGI:4083612], ANAS [JGI:4085297]), as well as targeted sequencing (this study) from the vinyl chloride-respiring Dehalococcoides enrichment cultures WBC-2, PM, EV, and WL [GenBank:JN034252-JN034255].

Full size image

These vcr-GIs contain integration and cargo (vcrABC ) modules with discordant evolutionary histories. The boundary between integration module and vcrABC is delineated by an unambiguous difference in nucleotide identity, 75.0 and 99.4%, respectively (Figure 3). This boundary reveals that vcr-GI integration modules contain a seventh protein encoding gene oriented in the opposite (forward) direction encoding a rubredoxin domain protein, in addition to the 6 integration module genes described previously (Figure 3). These integration modules (9164 - 11361 bp) are related as three distinct branches with nearly-identical leaves (masking a ~2200 bp deletion in PM, EV), grouped in a topology that is discordant with the corresponding tree of the 3784 bp vcrABC cargo (Figure 3). Relatedness of vcrABC -cargo sequences was estimated based on the 44 variant positions in their alignment, appearing mostly (66%) in the form of substitutions in the leader sequence of vcrA[5] (Figure 3). K a /K s ratios for the vcrA leader sequence (first 129 bp) ranged between approximately 0.05 and 0.2 for different pairwise combinations [44] and phylogenetic nodes [45], suggesting this region is under purifying selection. By contrast, the K a /K s ratio for the remainder of vcrA was incalculable because all 15 variant positions (out of 1431 bp) were non-synonymous substitutions, suggesting recent positive selection on the mature VcrA enzyme. This latter conclusion must be tempered by the limited information available in just 15 variant sites, the complete lack of indels detected in any vcrA (suggesting some purifying selection in the leader sequence), as well as the clear influences of recent horizontal gene transfer and recombination on these vcr-GIs. For example, vcrC is identical across all strains, within a 1650 bp region of perfect identity.

Age of Dehalococcoides

A core-gene phylogenetic tree was constructed to support age estimates based on evolutionary models. The core-gene tree was built from 432 core orthologous protein encoding genes shared between available Dehalococcoides (meta)genomes and Dehalogenimonas lykanthroporepellens BL-DC-9, a Chlorofiexi strain that is a phylogenetic outgroup to Dehalococcoides and its closest completely-sequenced relative [46]. Age estimates depend heavily on the assumed rate of mutation. We iterated our calculations on multiple published mutation rates (see Methods), as well as an empirical observation for mutation rate derived from the known divergence time (16 years, S. Zinder, pers. comm.) between the isolation of Dehalococcoides ethenogenes strain 195 [6] and the generation of a metagenome of its parent culture, DONNA2 (R. E. Richardson, pers. comm.). The latter empirical rate is substantially faster than the published values of faster-growing microbes (Additional file 4 Table S1), possibly because it includes mutations that already existed between strain variants within the DONNA2 culture prior to isolation of strain 195. Although we expect a long-term average mutation rate in the natural environment to be slower, and hence ages based on this rate to be an underestimate, it remains useful as a conservative bound on the 'recentness' of the events in question. Similarly, we used a range of growth rates to estimate the age of Dehalococcoides. For a recent bound we used the fastest reported Dehalococcoides doubling time (0.8 days [6]), as well as a range of slower reported growth rates from anaerobic environmental systems for more realistic estimates (11-14 days [47–49]). The corresponding estimates and lower (recent) bounds are presented in Table 1.

Table 1 Divergence Time Estimates Under Different Rates of Evolution.
Full size table

In relative terms, the divergence of Dehalococcoides and Dehalogenimonas are comparable to the predicted m ost r ecent c ommon a ncestor (MRCA) of available integration modules, approximately an order of magnitude earlier than the MRCA of Dehalococcoides strains. The MRCA of vcr-GI integration modules also significantly precedes the divergence of contemporary Dehalococcoides strains. In contrast, the high similarity among vcrABC sequences results in an estimated age that is at least an order of magnitude younger than Dehalococcoides speciation (Figure 4, Table 1).

Figure 4
figure 4

Date Estimates of Key Events in Dehalococcoides Evolution. Maximum likelihood phylogeny of 432 'core' orthologs. Timing of key evolutionary events are mapped onto the tree. The horizontal line below the tree represents the divergence time to the Dehalococcoides (Dhc) and Dehalogenimonas (Dehly) MRCA, while vertical hash marks indicate the relative divergence times of Dehalococcoides ssrA genomic island (GI) components. These include Dehalococcoides ssrA integration modules, vcrABC-attached integration modules, and vcrAB. Relative divergence times are based on the estimated age of the MRCA of Dehalococcoides and Dehalogenimonas (set to 1). The scale bar represents 5% of the total divergence time. Absolute time scales are from published mutation rate estimates and rate estimates based on the Dehalococcoides strains DONNA2/strain 195 divergence. Black points on the tree are nodes with 100% bootstrap support.

Full size image

Discussion

Age and specific features of vcr ABC acquisition

Nucleotide similarity is significantly higher between vcrABC cargo modules than can be expected if it was an orthologous locus present in the Dehalococcoides common ancestor (Figures 3, 4). In all cases vcrABC is located within a syntenic putatively-mobile element, vcr- GI, that is part of a broader class of ssrA-specific mobile elements that appear to be common among Dehalococcoides. In all vcrABC -containing strains except GT, the vcr- GI is located adjacent to the primary site of integration, ssrA, structural evidence that vcr- GIs are among the most recently integrated of the available Dehalococcoides ssrA-GIs. Within phylogenetic branches, integration modules are perfectly identical, except for a large identical deletion in the EV and PM vcr- GIs. The significantly unusual nucleotide signature of vcr- GIs [9, 20], as well as the discordance between the vcrA tree and the corresponding Dehalococcoides strain phylogeny, indicate that vcrABC has not been stably maintained in Dehalococcoides genomes since their divergence. Taken together, these observations suggest recent horizontal acquisition and dissemination of vcrABC across all Dehalococcoides ecotypes by way of a ssrA-specific mobile element with conserved attachment site and integration module.

Because anthropogenic release of chloroethenes into the environment is a relatively recent phenomenon (~100 years [3]), we are particularly interested in the recent bounds for estimates of the age of the MRCA of these vcrABC sequences as a proxy for their horizontal acquisition by Dehalococcoides. Using our highest estimated rates of mutation and chromosomal replication, the divergence of these vcrABC sequences appears to have occurred 4000 years ago. This value is in flated by the inexplicably high variation within the leader sequence of vcrA. If we remove the vcrA leader sequence from the calculation, the age of divergence decreases to 900 years. However, there is clear signal for positive selection in the remaining vcrAB sequence alignment: all 16 variant positions (15 in vcrA and 1 in vcrB) are predicted to result in amino acid substitutions. If positively selected, these mutations may have accumulated faster than the background rates assumed in our molecular dating calculations. Because the relative increase in substitution rate is unclear and the total information represented by just 16 variant positions is low, we cannot confidently distinguish the divergence of these vcrABC from the first industrial production of chloroethenes. By contrast, our most conservative estimate for the MRCA of contemporary Dehalococcoides strains is 40,000 years ago (ranging as high as 3 Mya, Table 1), long before industrial civilization had a chance to influence the evolution of Dehalococcoides and their streamlined genomes specialized for organohalide respiration.

It is important to note that these molecular dating estimates use the available vcrABC sequences to predict the first horizontal acquisition of vcrABC by Dehalococcoides. This analysis is not meant to predict the age of genesis of the first vinyl chloride reductase. We did not detect partial homology with other rdhA that would suggest vcrA is a chimera resulting from a recent homologous recombination event. Moreover, the existence of an alternate vinyl chloride reductase from strain BAV1, BvcA [19], that shares deeply branching ancestry with VcrA on a tree of available RdhA [9], suggests that vinyl chloride reductases have existed for a considerable period of time, just not within strains of Dehalococcoides for which sequence data is currently available. In fact, naturally occurring vinyl chloride has been detected in soils [50], providing a plausible source of selective pressure to explain the existence of vinyl chloride reductases in nature prior to human pollution. However, we have not identified any candidate lineages as the possible progenitor of vinyl chloride reductases, and we have no way of knowing whether the primary substrate for the ancestral VcrA or BvcA was consistently vinyl chloride, leaving their ancestral history unclear.

The phylogenetic discord between integration modules and their attached vcrABC indicates that homologous recombination - or perhaps a more directed form of 'module swapping' - has recently occurred between vcr- GIs (Figure 3). This additional inter-element recombination may be independent of ssrA-specific integration, but it would still require horizontal transfer so that 2 or more vcr- GIs are collocated within the same cell. Multiple vcr- GI variants have not been detected in the same complete genome. However, we did detect a low-coverage variant in the KB-1 metagenome assembly with 3 corroborating reads that perfectly match a different vcr- GI integration module found in VS, WL, GT, and WBC-2 cultures, providing preliminary evidence of the physical collocation of two vcr- GIs within the KB-1 culture (Additional file 5 Figure S4).

ssrA-GIs appear to be integrative and mobilizable elements

A subset of Dehalococcoides rdhA B were previously implicated in horizontal transfer [31, 32], including the trichloroethene reductase gene, tceAB[32]. Although the selective conditions in chloroethene-contaminated environments favors maintenance of tceAB and vcrABC, the genes implicated in tceAB transfer [32] share no detectable homology with the ssrA-specific system described in detail here. We hypothesize that these Dehalococcoides ssrA-GIs behave as integrative and mobilizable elements ("IMEs") because they do not appear to encode conjugation, although they share many other features of the broadly defined class of integrative and conjugative elements ("ICEs") [33]. It may be possible that conjugation is encoded by a surprisingly minimal gene set within the integration modules [33], similar to the small (10.9 kbp) integrating and conjugating element 'pSAM2' of Streptomyces ambofaciens, which requires only a single gene, traSA, for inter-mycelial (conjugal) transfer [51]. Dehalococcoides core genes do include putative pil genes, the functions of which are unclear but may play a role in conjugation. Some strains of Dehalococcoides contain unambiguous prophages, providing an alternative hypothesis for the mechanism of ssrA-GI transfer, via illegitimate packaging of the excised ssrA-GI into a phage capsule. The length of Dehalococcoides ssrA-GIs is within the range of typical phage genomes. However, evidence for a complete prophage is not as ubiquitous among Dehalococcoides as the presence of ssrA-GIs, and there have been no descriptions to date of Dehalococcoides phage that also encode an rdhA, leaving the influence of phage on rdhA evolution unclear. Based on currently available evidence, we hypothesize that Dehalococcoides ssrA-GIs are mobilizable but not conjugating elements that sometimes mobilize adjacent tandem islands but in all cases rely on a host- or phage-encoded system for cell-cell transfer of a transient, presumably circular, intermediate.

Dehalococcoides also contains comEA, and it is unknown if Dehalococcoides is transiently competent for uptake of exogenous DNA. However, transfer via stochastic competence is an unsatisfying explanation, mainly because Dehalococcoides ssrA-GIs appear to lack genes encoding independent replication, and stable non-phage extrachromosomal elements have not been observed in Dehalococcoides[7–9].

Occasionally integrating and conjugating elements do have replicative forms [33], as in the case of rolling circle replication of pSAM2 in the donor cell [52]. Maphosa et al. recently described a field site in which there were 1 to 2 orders of magnitude more vcrA copies detected than copies of tceA, bvcA, or Dehalococcoides 16S rRNA genes [53]. vcrA was also more abundant than Dehalococcoides 16S rRNA genes in a dechlorinating bioreactor inoculated from the site [53], suggesting either (1) there exists a vcr- IME that can replicate independently or has integrated within an element that can replicate independently, or (2) they detected a non-Dehalococcoides population that also possesses vcrA, coexisting with a Dehalococcoides population.

It is important to note that, while a conspicuous and common feature, not all Dehalococcoides ssrA-GIs contain an integration module. We identified 15 ssrA-GIs without integration modules, containing a total of 38 rdhA as well as other genes. These might be 'cis-mobilizable elements' that encode neither integration nor transfer, but retain functional attL/attR sites [33] and are occasionally or constitutively mobilized with adjacent genomic islands through a process known as accretion [54]. In some cases these tandem ssrA-GIs may have been previously mobile but are now fixed in the chromosome. For example, there is a region immediately downstream of the direct repeats furthest from ssrA that is similarly dense in rdhA while also syntenic across Dehalococcoides strains, phylogenetically coherent with whole genome estimates, and devoid of ssrA-GI signatures (Additional file 6 Figure S5); suggesting this region was present in the MRCA of available Dehalococcoides[9]. Some or all of this region may have been acquired originally as an ssrA-GI, but deletion and amelioration has erased evidence of horizontal gene transfer.

Likely Roles within ssrA-GI Integration Modules

The first identified Dehalococcoides ssrA-specific integrase gene (dsiB) (DhcVS_1292) was sequenced following the original identification and characterization of VcrA, and noted for its proximity to vcrA on the chromosome [5]. It is now clear that DhcVS_1292 is part of an integration module in an adjacent downstream ssrA-GI (GI 02 in VS, Figure 1), one of 16 dsiB homologs detected in Dehalococcoides genome sequences. The closest relative to dsiB in the public database is present on a fully-sequenced metagenomic fosmid from a deep (4000 m) ocean subsurface sample (EU016565, Figure 2), within an apparent integration module that also includes homologs to dsiA, parB, mom, and a putative tRNA embedded in mom, as well as an unambiguous ssrA-direct repeat at the homologous attL position embedded in dsiB (Additional file 7 Figure S6). This is especially intriguing in light of the recent sequencing of 32 novel rdhA amplified from various marine subsurface sediments [55], many of which appear phylogenetically within a major rdhA branch (Cluster I [9]) that is otherwise populated only by rdhA from Dehalococcoides or Dehalogenimonas. Given this indirect evidence and the large diversity of organohalogens detected in marine systems [56], it is tempting to speculate that Dehalococcoides plays a role in these settings. However, in the absence of direct observation of Dehalococcoides-like microorganisms in marine (subsurface) settings, this role remains unclear.

A more sensitive database search indicated that DsiB is a structurally similar homolog of CcrB, containing the serine-recombinase-catalytic domain at the N terminus, as well as similar motifs along its ~500 residue length (mean 22% ID, Figure 2A). CcrB specifically integrates/excises the so-called 'Staphylococcus Cassette Chromosome' (SCC [57]) family of mobile elements that are a vector of antimicrobial resistance (among other phenotypes [58, 59]), with major consequences for hospitals and the greater community [60–63]. CcrB was shown to have DNA-binding and recombination activity for attS of SCC [64], but SCC integration [57] and attB-specific excision both required the product of a smaller, co-transcribed serine recombinase gene, ccrA, that does not encode a DNA-binding domain [64]. Similarly, Dehalococcoides integration modules encode on a putative operon a second, smaller serine recombinase, DsiA, that also lacks a detectable DNA-binding motif. Dehalococcoides ssrA-GIs and SCC also share overlapping size ranges and specifically integrate at a non-tRNA, single-copy essential gene. We hypothesize that integration/excision of Dehalococcoides ssrA-GIs occurs in a homologous mechanism to SCC, via DsiB in concert with DsiA, with other integration module elements likely playing a role in regulation of integrase/excisionase activity or modification of the excised element to facilitate transfer or maintenance. Unfortunately, the mode of SCC transfer among Staphylococcus is unclear [65], and so does not provide additional clues regarding a likely transfer mechanism.

Interestingly, dsiB is always found overlapping attL at its 3' end. A stop codon occurs only upstream of the genomic island, even if that means overlapping substantially with an adjacent genomic island or ssrA itself. Complimentary overlap of ssrA with small open reading frames has been detected in some bacteria with ambiguous implications [66]. It seems unlikely in this instance that the 3' terminal ~70 bp of ssrA also encode a functional region of dsiB on its complementary strand. Accordingly, alignments of DsiB are divergent at this portion of their sequence, both in length and amino acid similarity. The majority of dsiB is upstream of ssrA or its direct repeat, and already comprises the expected length for homologs of ccrB (1600 bp). In addition to a trivial explanation in which dsiB undergoes low-efficiency translation that is variable at the C-terminus, it may be that dsiB is only fully functional when encoded on the circularized element, or alternatively when encoded on the chromosome downstream of an adjacent genomic island containing the requisite 3' gene fragment. In any case, the overlap of dsiB with attP/attL leaves the stop codon of dsiB unclear, and may have functional relevance or affect regulation of dsiB.

Conclusions

Structural comparison of new (meta)genomic data, as well as targeted sequencing from unsequenced vinyl chloride respiring enrichment cultures, resulted in identification of 8 homologous mobile elements containing the vinyl chloride reductase operon, vcrABC. These vcr- GIs are a subset of mobile genetic elements in Dehalococcoides that specifically integrate at the single-copy gene, ssrA. A detailed comparison of these ssrA-GIs allowed identification of the precise position of insertion, the direct repeat created by the insertion event, as well as a conserved module of syntenic integration-associated genes that includes the likely ssrA-specific integrase, which we named dsiB. ssrA-GIs are most likely 'integrating and mobilizable elements' (IMEs) that do not encode their own mechanism of cellular transfer. Core (meta)genome phylogenetic analysis allowed an estimation of timing of divergence of Dehalococcoides strains, between 40,000 and 400,000 years ago, suggesting that the specialization by Dehalococcoides for respiration of organohalide compounds far precedes industrial synthesis by humans. By contrast, time estimates for the first horizontal acquisition of vcrABC sequences by Dehalococcoides are not confidently distinguishable from the first industrial synthesis of chloroethenes ~100 years ago. Vinyl chloride reductases may be ancient, nevertheless, and the donor(s) of recent genetic diversity to Dehalococcoides remain undetermined.

Methods

Identification of Dehalococcoides sequences in metagenome data

For ANAS and KB-1 enrichment cultures, complete genomes have not been assembled. However, identification of Dehalococcoides contigs was performed by comparison with closely-related complete genomes of strains 195 and CBDB1, respectively. For KB-1, gap closure was performed to create a closed draft genome of the dominant Dehalococcoides strain in the metagenome, including primer-walking of gap-spanning fosmid inserts, as well as gap-spanning PCR amplification from an aliquot of the originally-submitted KB-1 genomic DNA. For ANAS, structural genomic information comes from a single contig (contig 2014738012; 119815 bp) that contains a Dehalococcoides ssrA on one end and a series of tandem ssrA-GIs downstream (Figure 1). A second contig containing a Dehalococcoides ssrA was also detected (2014739670), but it contained no detectable ssrA-GIs. Dehalococcoides orthologs present in the ANAS dataset were detected using reciprocal best-hit BLASTp criteria described previously [8, 9], treating the collection of all Dehalococcoides protein encoding genes in ANAS as if it were one genome. Although ANAS contains more than one Dehalococcoides strain, this approach had little effect on the number of orthologous groups considered 'core' between all Dehalococcoides, mainly because sequencing was sufficiently deep and because the Dehalococcoides strains in ANAS are closely-related to the Cornell lineage from which ethenogenes 195 is the only completely-sequenced representative [67].

DNA Source, Primer Design, Amplification Optimization

Enrichment cultures were derived from samples from geographically distinct contaminated sites (Additional file 8 Figure S7): WL from Western Louisiana [30]; EV from the Evanite site in Corvallis, Oregon [28]; PM from the Point Mugu Naval Weapon Facility, California [28]; and WBC-2 from the West Branch Canal Creek, Aberdeen Proving Ground, Maryland [29]. Culture, culture pellets, or genomic DNA were provided by L. Semprini (EV, PM) or E. Edwards (WL, WBC-2).

Nucleotide positions strongly conserved at ssrA, its direct repeat, and a few locations within integration modules were used for primer design (Additional file 2 Figure S2). Amplification was successful with a variety of combinations of predicted melting temperature and degeneracy. We described only the best-performing primer pairs, especially those that contributed to vcr- GI amplification and sequencing. All PCR amplifications from mixed cultures were performed using Phusion polymerase under standard conditions using 'HF' buffer. Primer pairs were optimized toward amplification of regions of ssrA-GIs in mixed genomic DNA template by DMSO and annealing temperature gradients. For most target amplicons the optimal DMSO concentration was at or near 4%, with optimal annealing temperature depending on the primer, and summarized in Additional file 2 Figure S2. In particular, a 20 bp forward primer beginning at Dehalococcoides ssrA position 8 (CGTGGTTTCGACAGGGAAGG - 'ssrA _03F'), successfully amplified ~90% of all 4 novel vcr- GIs when paired with a reverse primer upstream of vcrA (GTTCCTGACCATGCCGTACC - 'vcrA_05R'). The resulting (8.3 - 10.5 kbp) amplicons were purified in agarose gel electrophoresis and sequenced directly by the Sanger method (MCLAB, ELIM) and primer walking. No single primer-pair was determined that could amplify a complete ssrA-GI in one reaction from attL to attR, as these would be reverse complements of one another and produce primer dimers. Instead, combinations of PCR reactions were amplified and sequenced separately, and their resulting sequence data was assembled in silico and verified manually. For example, forward primers targeting a conserved position in the integration module (TGGAGCGCCGCCGTNGG - 'REC_003eF') amplify a portion of the integration module and all of the genetic cargo (~7 kbp) when coupled with a reverse primer that targets the ssrA-direct repeat (TGGTGGAGACGGGGGAGGG - 'REC_001eR'). Two-fold or greater coverage and perfect agreement between different amplicons from the same sample was required in assembly. In some instances ssrA-GI-derived amplicons were cloned in Escherichia coli following agarose gel purification. Efficient ligation to a vector was achieved with Enzymatic Assembly [68] and pSMART-LC-Kan (AF532106; Lucigen Corp.) or by blunt ligation into the pJAZZ-OK linear vector (FJ160465; Lucigen). Transformation was achieved chemically in E. coli DH5α or electrically in E. coli BigEasy-TSA (Lucigen) cells, respectively.

Core Genome and Genomic Island Phylogenies

The reciprocal BLASTp procedure for identifying orthologous groups among Dehalococcoides was also applied to Dehalogenimonas lykanthroporepellens BL-DC-9 in comparison to Dehalococcoides ANAS, KB-1, 195, DONNA2, CBDB1, BAV1, GT, and VS; resulting in 432 core orthologous groups that were also free of paralogs. Global alignments of each orthologous group were performed by Muscle (version 3.8.31) [69]. Single gene trees were generated using RAxMLHPC (version 7.0.3) [70] under the GTR + γ model [71] with Dehalogenimonas lykanthroporepellens BL-DC-9 constrained as the outgroup to prevent long-branch artifacts. The resulting trees were entered into Splitstree4 [72] and a consensus network was generated. The single gene global alignments were concatenated to generate a single large alignment for the 9 organisms. A core-genome phylogeny was generated using RAxMLHPC as described above, with 10 initial random starting tree iterations and 100 bootstrap replications. The tree with the highest likelihood is presented in Figure 4 and used for evolutionary analysis. Alignments for components of genomic islands were generated using Muscle and refined with hmmer (version 2.3.2) [73], then masked manually. Phylogenies were generated in RAxMLHPC under the GTR + γ model with 10 random starting trees and 100 bootstrap replications. In each case, the appropriate sequence was constrained as an outgroup. The trees with the best likelihood were identified and used for further age estimate calculations.

Date Estimations

Estimates of the age of the Dehalococcoides/Dehalogenimonas divergence, the Dehalococcoides clade, as well as the various components of the genomic islands were determined under three different estimates for the rate of Dehalococcoides evolution. Two mutation rates from published values were used: one from a universal estimate of bacterial mutation rates in natural environments [74], and one from an empirical analysis of E. coli in lab cultures [75] (Additional file 3 Table S1). A third rate was based on a known divergence time of approximately 16 years between the separation of Dehalococcoides ethenogenes strain 195 ("strain 195") [6] from its mother culture - the "TCE/MeOH" culture (Prof. S. Zinder, pers. comm.) - in 1992, and the 2008 metagenome sequencing of the "DONNA2" enrichment culture. DONNA2 was also derived from the TCE/MeOH culture and maintained in parallel from strain 195 until its subsequent metagenome sequencing (R. E. Richardson, pers. comm., see DONNA2 Mutation Detection, below). Branch lengths between strain 195/DONNA2 were calculated from single-gene trees of the 387 core protein encoding genes, after excluding 45 trees that did not have strain 195 and DONNA2 as a monophyletic group, most likely resulting from frame-shift mutations. The mean branch length of the 387 protein encoding gene trees, the core-gene concatenated ML tree, as well as the splitstree [72] network average branch length were all approximately 3(10) -5. For a minimum separation of 16 years, this corresponds to 2(10) -6 branch length per year of Dehalococcoides divergence. It is important to note that some unknown fraction of the observed mutations could have already existed within the mother culture prior to isolation of strain 195 if parents of the two contemporary strain variants coexisted at that time. Combined with the imposed pressures for rapid growth inherent to a laboratory culture, we expect that the Dehalococcoides mutation rates observed by this approach represent an unrealistic upper bound to what is likely to occur in Dehalococcoides in nature. This value is still useful, however, for creating lower bounds in molecular dating estimates that are compared with relatively recent events (e.g. human civilization, anthropogenic chloroethene pollution, etc.).

DONNA2 Dehalococcoides Mutation Detection

Because the dominant bacterium in the DONNA2 culture was our target variant of strain 195, the DONNA2 metagenome data included a high coverage of this variant. A comparative assembly of the DONNA2 shotgun reads on the strain 195 genome allowed identification of reliable mutations between these two strains, using the V ariant A scertainment Al gorithm (VAAL) under default settings [76]). The DONNA2 metagenome project has gone through successive rounds of sequencing, and the mutation analysis described here is based on the raw 454 GS FLX Titanium shotgun reads available on 06 November 2009, which were subsequently filtered by alignment to the genome sequence of Dehalococcoides strain 195. The resulting 455,062 Dehalococcoides-derived reads had a mean length of 365 ± 142 nucleotides, and %(G+C) of 48.8. Our version of VAAL did not produce assembly statistics, but a separate comparative assembly using Geneious Pro v5.4 (medium-sensitivity default parameters) successfully aligned 454,342 reads to the strain 195 genome, for a coverage of 115.2 ± 41.2. The consensus sequence of the comparative assembly produced by VAAL formed the basis for the subsequent strain-level mutation analysis. Gene annotations from strain 195 were mapped onto the DONNA2-variant genome sequence and the protein-encoding genes among these were extracted and included as a separate whole-genome collection in the genome-wide core gene phylogenetic analysis (above). The cumulative length of the protein encoding genes shared between 195 and DONNA2 was 1,301,665 bp; and among these genes we detected a total of 192 mutations, with adjacent SNPs considered part of a single mutation. Of these 192 mutations, 39% were deletions, 28% were insertions, 28% were transitions, and 4% were transversions. With respect to the predicted effects relative to the encoded protein in strain 195, 40% were frame-shift mutations, 40% were synonymous (probably no change), 19% were non-synonymous substitutions, and 1% were predicted to cause a truncation due to an early stop codon. It should be noted that frame-shift and truncation mutations would probably not directly affect our subsequent tree calculations because those genes would likely fail our orthology criteria (above) and thus would not be included in the set of "core" genes.

Ka/Ks ratios

K a /K s ratios are an intrinsically pairwise calculation that was performed on a subset of the most different pairs of vcrA (full-length, and leader sequence only) using the 'kaks' function in the SeqinR package [44] of R [77]. K a /K s ratios were also calculated for all adjacent branches in a phylogenetic tree of the 8 sequences, using the Ka/Ks Calculation tool [45].

Integration Module tRNA Secondary Structure

The putative tRNA sequence was originally detected by ARAGORN [78] and annotated previously in publicly available annotations of Dehalococcoides strains CBDB1 and VS. Secondary structure was predicted from the alignment of all 16 detected tRNAs in available ssrA-GI integration modules, submitted to the RNAalifold [79, 80], Pfold [81], and PETfold [82] web servers for independent calculations. The resulting structures were compared manually, including a comparison to classical tRNA secondary structure for identification of the conserved "DCC" anti-codon within a 5 nt anti-codon loop (Additional file 1 Figure S1).

Description of additional data files

Additional data file 1 is a PDF format file containing the supplemental figures and associated legends. Additional data file 2 is a Microsoft excel (.xls) file containing tables of growth rates and rates of evolution, as well as other parameters and example calculations used in the molecular dating analyses.