In this study we have applied a metagenomic approach to isolate and characterise novel plasmids from the human gut microbiome. These plasmids ranged between 3.7 and 10.8 kb, and at present it is unclear if this size range represents a limitation of the TRACA system used to isolate them, a pre-dominance of smaller plasmids in the gut microbiome or a combination of these factors [12
]. The range of G+C contents of plasmids isolated corresponds with the overall range in genomic G+C content observed among the major bacterial divisions comprising the human gut microbiota. The Actinobacteria
constitute a major phylogenetic division within the human gut comprising high G+C Gram positive species [2
]. In contrast other major phylogenetic groups, Firmicutes
, are comprised of low G+C Gram-positives and Gram-negatives respectively [2
]. As such plasmids of a high G+C content (pTRACA10, pTRACA18) are likely to originate from bacteria belonging to the Actinobacteria
Of the six complete plasmid sequences utilised in this study, no significant homology to pTRACA17, pTRACA20 or pTRACA30 was identified in any of the metagenome data sets examined. This may simply be due to a low abundance of these plasmids in the human gut microbiota, the generally low coverage of bacterial communities offered by current metagenomic data sets, or may reflect intra-individual variation in plasmids associated with the gut community. Alternatively, the apparent absence of these or closely related plasmids in the human gut metagenomes examined, may be attributed to a general incompatibility of plasmid DNA with the standard metagenomic approach used to generate many of the currently available datasets, which is focused on acquisition and sequencing of bacterial genomic DNA [7
In contrast sequences homologous to plasmids pTRACA10 and pTRACA22 were identified in multiple human gut metagenomes. In particular, pTRACA22 appeared well represented in the combined human gut metagenomes analysed, and the absence of nucleotide sequences homologous to this plasmid in the non-human datasets, coupled with its presence in human gut datasets from geographically isolated individuals with a broad global distribution, suggests that pTRACA22 may be unique to the human gut microbiota. This has also been observed for bacteriophage present in the human gut community and recently Ebdon et al
. (2007) reported that bacteriophages isolated from human faecal material were specific to the human gut, and absent in the general environment, as well as faecal samples from horses, sheep, pigs, cattle, rabbits and poultry [38
]. Furthermore the general enrichment of certain MGE in the human gut microbiome has also been reported in recent metagenomic studies [10
Investigation of the relative abundance of functions encoded by pTRACA10 and pTRACA22 revealed that several show a higher abundance in the human gut microbiome (Fig ). In particular the increased prevalence of the pTRACA22 TA addiction module in the human gut microbiome was unexpected, and the significance of this observation is currently unclear. TA modules have been shown to contribute to plasmid stability and maintenance through a post segregational killing mechanism, and consist of a stable toxin component which is inactivated by an unstable antitoxin [29
]. Loss of the plasmid from daughter cells ultimately results in a loss of the antitoxin component and the killing of plasmid free cells [29
]. In the case of the RelBE system, RelE encodes an RNA interferase which cleaves mRNA in a site specific manner at ribosomes, blocking translation [40
]. Plasmid encoded TA modules are also involved in plasmid-plasmid competition [42
], and plasmids harbouring a TA module have been shown to outcompete counterparts from the same incompatibility group lacking a TA module [42
]. These attributes of TA modules may facilitate the maintenance of plasmids, or other DNA molecules harbouring them, in the human gut microbiome, and contribute to the observed prevalence of RelBE modules in this ecosystem (Fig ). However, this explanation does not account for the lower prevalence observed for non-human metagenomes.
TA modules that are active and functional in surrogate cloning hosts may also generate bias in metagenomic libraries by enhancing stabilisation of constructs (such as metagenomic clones) harbouring them. Any such effects of TA modules could also account for the increased prevalence of the pTRACA22 RelBE and homologous modules in the human gut microbiome, as well as the differences between the relative abundance of toxin and antitoxin homologs (Fig ). However, considering that orphan RelE toxin genes were identified on sequences retrieved from human gut metagenomes, and no increase in relative abundance was observed for the other TA modules analysed in this study, this explanation seems unlikely (Table S2).
It is also possible that bias between the metagenomic data sets due to variation in methods for construction of metagenomic libraries, such as efficiency of DNA extraction protocols, may account for the observed differences in prevalence of plasmid encoded functions between human and non-human metagenomes. However, the data sets utilised in this study were generated using comparable approaches, and in the case of the 15 human gut metagenomes utilised, which were generated by 2 distinct research teams [9
], no demarcation between data sets from different groups was identified during our analysis. This is despite the observed paucity of sequences from Bacteriodes
sp. in the metagenomes generated by Gill et al
. (2006) [9
]. Furthermore, analysis of other TA modules revealed a similar prevalence among the all metagenomic data sets, with few significant differences between human and non-human data sets (Fig ).
Plasmids and other MGE are also potentially involved in the maintenance of ecosystem functions, and retention of genes within this community in the absence of any selective pressure for the functions they encode. Many plasmids are highly stable even when providing no obvious survival advantage to the host cell, and numerous explanations have been suggested for the apparent stability of plasmids encoding traits such as antibiotic or heavy metal resistance, in the absence of direct selective pressure. These include the existence of unidentified selective pressures acting directly on the resistance or other genes encoded by a plasmid, or provision of a survival advantage sufficient to overcome the metabolic burden imposed on plasmid carrying cells, when selective pressure is removed [43
However, TA modules may also constitute a mechanism by which plasmids and the genes they encode are retained in the absence of direct selective pressure [12
]. It has also been proposed that chromosomally encoded TA modules may stabilise regions of adjacent DNA and maintain their vertical transmission [32
]. Thus both plasmid and chromosomally encoded TA modules may be important in maintaining functional stability of microbial ecosystems such as the human gut microbiome.
The wide distribution and high prevalence of chromosomally encoded TA modules in both bacteria and archaea (particularly in free living species) has also lead to the hypothesis that these systems are advantageous to the host bacterium [32
]. Since the majority of sequence data in the metagenomic libraries utilised will be derived from chromosomal DNA, most of the RelBE TA modules identified in the human gut microbiomes will be chromosomally encoded. Therefore it is also plausible that the increased abundance of these TA modules in the human gut microbiome is due to effects on the fitness of the bacterial host. Interestingly, while assessment of other TA modules did not reveal the prevalence observed for the pTRACA22 RelBE, the MazF toxin, which also functions as a ribonuclease and blocks translation by the same mechanism as RelE [32
], did exhibit an increased prevalence in the human gut microbiome as well as other environments, albeit to a lesser extent (Fig ). Diverse functions have been proposed for chromosomally encoded TA modules in bacteria and archaea, and it is possible that these functions are also important in gut associated bacterial species.
RelBE and MazEF TA modules have been shown to modulate gene expression, undertake quality control of this process, and facilitate bacterial survival under nutrient limiting conditions or other environmental stresses [32
]. The involvement of TA modules in gene expression may also be relevant to adaptation of prokaryotes to new environments, and TA modules have been implicated in the formation of persister cells which can survive exposure to antimicrobials and other stresses that are otherwise fatal [50
]. It has also been proposed that chromosomally encoded RelE and ParE toxins are exploited as cellular regulators, and that the acquisition of these elements through HGT mediates the rapid development of novel regulatory pathways which facilitates adaptation to new environments through modulation of gene expression [47
]. The formation of persister cells and the development of new regulatory pathways via TA have important implications for colonization of the gut, as well as survival of bacterial cells in the external environment, or during transmission to new hosts. However, the exact function of RelBE TA modules in bacteria inhabiting the human intestinal tract is currently unknown, and will require further detailed study to elucidate.
Interestingly, no RelE sequences homologous to the pTRACA22 RelE were identified in gut associated archaeal species. Since TA modules such as RelBE have been observed as highly prevalent and widely distributed in free living archaea, and were identified in bacterial species from all major phylogenetic divisions in the gut microbial community, the lack of sequences homologus to the pTRACA22 RelE in dominant gut archaeal species (M. smithii and Ms. stadmanae) was surprising. This may reflect differences in selective pressures imposed on bacteria and archaea colonising the human gut, or the adoption of alternate strategies to cope with the same environmental stress. While the exact nature and source of the observed difference between gut archaea and bacteria remain unclear, the identification of functions shared or distinct to each lineage will be important in developing a full understanding of the gut microbiota, and the contribution of constituent species to the development and output of this community.
The putative RelBE TA addiction module identified on pTRACA22 also has potential to impact on eukaryotic cells present in the human gut through activity of the RelE toxin component. While no eukaryotic homologues of RelE have been identified, this toxin has been shown to be active against both prokaryotic and eukaryotic organisms. The RelE toxin is functional in eukaryotic cells and has been shown to induce apoptosis in cultured human cell lines [51
], inhibit the growth of yeast [52
], and to cleave eukaryotic mRNA in the A site of eukaryotic ribosomes in a similar fashion to that described in bacteria [53
]. However, studies demonstrating the activity of RelE in eukaryotic cells are currently limited to artificial expression of the toxin using a specialised vector [51
]. As such further study is required to identify the in vivo
effect, if any, of RelE toxins on eukaryotic cells, particularly in light of the increased prevalence of this TA module in the gut human gut microbiome.
As well as a wide distribution among human metagenomic data sets, pTRACA22 was also found to illuminate the transfer of genetic material between diverse members of the gut microbiota (Fig ). The high identity of pTRACA22 to the E. coli
pARS3 IS26-like transposon, indicates that this region of pARS3 is derived from a plasmid closely related to pTRACA22 (Fig ), raising several intriguing possibilities. In particular the presence of the pTRACA22 putative replication protein on the pARS3 IS26Tn may allow pARS3 to function as a natural shuttle vector and replicate in disparate bacterial hosts. Considering the high identity of pTRACA22 encoded genes to B. hydrogenotrophica
, this scenario would likely encompass replication in both Gram -ve and Gram +ve species. Furthermore, the presence of a RelBE addiction module (with high identity to that encoded by pTRACA22) in the pARS3 IS26Tn could stabilise pARS3 during transition between bacterial hosts, and provide this plasmid with a competitive advantage over other related elements in the same incompatibility group [29
While the site of gene exchange that resulted in the formation of the pARS3 IS26Tn cannot be determined with complete certainty, several factors indicate the human or mammalian GI tract: i) Both species involved in the genetic exchange (E. coli
and a gut associated Firmicute
sp.) naturally inhabit the human gut; ii) The high oxygen sensitivity of many cultivated Firmicutes
comprising the gut microbiota points to an anaerobic environment such as the human gut [54
]; iii) The paucity of sequences homologous to pTRACA22 in the non-human gut metagenomes analysed here. Overall the high identity of pTRACA22 with the pARS3 IS26Tn supports the hypothesis that the human gut is a potential hotspot for HGT between disparate bacterial species [12
], and highlights the potential for gene transfer from commensal organisms to clinically relevant pathogenic speices.
In light of the observed prevalence of RelBE TA modules in the human gut microbiome, the reported functionality of the RelE toxins in eukaryotic cells, and the potential role of this system in mediating bacterial survival in the gut environment, the RelBE module encoded by pTRACA22 merits further investigation, and these studies are currently underway in our laboratory. Furthermore, continued analysis of the mobile metagenomes associated with microbial communities will be important in understanding many aspects of microbial community function and evolution. A greater knowledge of the MGE associated with bacterial communities will also facilitate the development of novel strategies and tools to dissect and manipulate these complex ecosystems.