Classical and non-classical secretory proteins.

Various proteins that populate the outer surface of Gram-positive bacteria entail some sort of signaling peptide to help facilitate their translocation across the cell envelope barrier. For instance, the route of N-terminal signaled “classical” protein secretion can occur through either the Sec-dependent or twin-arginine translocation (Tat) pathways. Then again, some secreted proteins use “non-classical” pathways and do so without any obvious signal peptides. For our in silico predictions, the SignalP [49], TatP [50], and SecretomeP [51] programs were used to discriminate the number and type of exported proteins utilizing the classical pathways from those that are secreted by non-classical means. Predicted primary structures of loci from all nine L. ruminis genomes were individually processed per genome and the output analyzed afterward. Data from the SignalP and TatP predictions were also sorted into either the core or accessory genomes.

According to the SignalP results (S2 Table), it appears that classically secreted proteins requiring a Sec-signal peptide account for a maximum of only 75 putative loci (ATCC 25644). Among these protein genes, an estimated 33–48 belong to the core genome (equivalent to 2.7 to 3.9%, respectively). Of note, we attribute the range in pooled core genome number as a general inconsistency with in silico prediction programs. For example, with SignalP, this can also partly originate from some of the “secreted” protein loci unexpectedly lacking a deduced primary structure for an intact N-terminal region, which causes them to be missed and not predicted. For the Sec-signaled proteins that are encoded by non-core loci, these number from six to 29 genes and average out to less than one percent of the accessory genome (S2 Table). Prediction estimates using the TatP method indicate that there is a minimum of 15 (DPC 6832) and a maximum of 38 loci (SPM0211) with encoded proteins that would have the sequence potential for a Tat signal peptide (S3 Table). Of these, the designated core genes are between 11 (ATCC 27782 and DPC 6830) and 14 (PEL65 and S23), whereas those loci that make up the accessory genome range from three to 26 in number (DPC 6832 and SPM0211, respectively). By comparison, the SecretomeP prediction data gives an approximation of loci for non-classically secreted proteins that is at least ten-fold greater in number. In total, there are between 526 and 841 putative loci whose encoded products presumably use one or other unconventional routes for exporting proteins that do not have a typical N-terminal signaling peptide (S4 Table). While these numbers of protein loci might appear high, we consider them to be acceptable given that the recommended SecP score (≥ 0.5) for indicating secretion was used in our determinations [51].

Our survey of the corresponding annotations for the putative classical and non-classical secretory proteins indicates that for both these types many of them have some link to a recognizable extracellular function. In the context of cellular adhesion, one of the non-classically secreted proteins is worthy of mention (S4 Table). Here, it might be envisioned that active expression of the core gene annotated as fibronectin-binding protein (Fbp) (GRL1172_498, HMPREF0542_10570, LRC_RS05075, LRN_0851, PEL65_1842, PEL66_466, P869_04425, LRU_00261, and LRP_1613) might contribute to the gut autochthony of the various L. ruminis strains. Fibronectin is an ECM glycoprotein found near the surface of epithelial cells and can serve as a target or attachment site for many types of bacteria, including a variety of Lactobacillus species [63, 64]. For the binding of L. ruminis cells to gut-localized fibronectin, this might be mediated through the Fbp, along with the LrpCBA pilus, which has been shown to have similar substrate specificity [23]. As with other lactobacillar counterparts [35, 65], L. ruminis Fbp is atypical from those in pathogens and is seen to lack any familiar signaling domains at its C- or N-termini, thus explaining why this protein was detected with the SecretomeP method.

Transmembrane proteins.

In Gram-positive bacteria, many of the proteins that are being exported through the cell envelope also possess hydrophobic regions that facilitate membrane insertion, but as well, where other regions extend beyond the cell wall to the outer surface. These transmembrane proteins (as they are classified) contain one or more membrane-spanning helical regions whose hydrophobicity allows them to be embedded in the cytoplasmic membrane. Functionally, Gram-positive transmembrane proteins have varied purposes, these ranging from the metabolism of macromolecules (polysaccharides, nucleic acids, proteins, and lipids) to different cellular signaling processes and gene regulation activities [60]. By using the TMHMM algorithm [52], which predicts the presence of transmembrane helices (TMHs), we were able to gauge which of the L. ruminis genes from the pan-genome are representing putative transmembrane proteins (S5 Table). In total, we found there were between 308–480 hits (DPC 6832 and ATCC 25644, respectively) for protein loci that are predicted to encode transmembrane helical regions, but where 222–341 of these predictions are for proteins containing two or more TMHs. While not unexpected, functional annotations for these transmembrane proteins were in line with other Gram-positive bacteria. Interestingly, the maximum number of predicted TMHs per encoded protein was found to be 14, and of these integral membrane proteins they accounted for five to eight loci, most of which were annotated as various members of the transporter protein family.

Among the transmembrane protein hits, we identified an ORF in two human strains (HMPREF0542_11925 from ATCC 25644 and P869_03355 from S23) that encodes for a possible fucose permease (FucP) [66]. This transporter protein is responsible for the cellular uptake of fucose, a hexose sugar commonly found as a glycan component of the mucin proteins forming the mucus layer of the gut epithelium, and which for certain bacteria can serve as an energy substrate instead of glucose should nutrients become limiting [67]. Noticeably, none of the two genomes contains the gene encoding for an α-fucosidase, the enzyme that would hydrolyze the release of fucose from the mucus glycan chain as found with Bifidobacterium bifidum [68]. This can explain the findings of another study where fucosidase activity was not detected in the ATCC 25644 and S23 strains [19]. Even so, fucose could instead be available as scavenged remnants in the gut, and subsequently absorbed into the cells via the putative fucose permease transporter. Thus, with the possibility of having the transport machinery that enables cells to feed on spent intestinal mucus, this can be regarded as a habitat-specific fitness advantage for these two human strains of L. ruminis. Interestingly, despite the fact that mucus-lined intestines are a universal feature of mammals, the fucose permease-encoding gene is not shared by the genomes of the other host-derived L. ruminis strains, and thus it belongs to the accessory genome.

Lipoproteins.

Lipoproteins in Gram-positive bacteria represent a functionally diverse group of membrane proteins (transporters, enzymes, adhesins, and receptors) that protrude outwardly from their covalent anchoring at the interface region between the cytoplasmic membrane and the peptidoglycan layer of the outer cell wall [60]. Bioinformatically, Gram-positive lipoproteins can be readily recognized based on their so-called lipobox, a pattern of residues ([LVI]-[AST]-[GA]-C) in the C-region of its N-terminal secretion signal sequence that serves as the cleavage recognition site for signal peptidase II [60]. In our pan-genomic survey of L. ruminis, loci encoding putative lipoproteins were identified for the presence of a lipobox using the PRED-LIPO prediction program [53], wherein a total of between 16 and 30 genes had been uncovered (S6 Table). After sorting into the core and accessory genomes, 12–19 core and four to 13 non-core genes were found to exist. Aside from a few hypothetical protein annotations, the majority of the predicted lipoprotein genes are annotated with functions as ABC transporter proteins (or otherwise related), these including a variety of potential substrate binding abilities. Given their essential role in the exchange of metabolic products, such putative lipoproteins can be regarded as a necessary feature of the various L. ruminis strains and their evolved capacity to utilize the nutrient richness of the gut environment.

LPXTG-anchored surface proteins.

Often exposed on the cell surface of many different Gram-positive genera and species are various types of proteins whose covalent cell wall anchoring relies on the action of transpeptidyl enzymes called sortases. Such sortase-dependent proteins (SDPs) require the presence of a C-terminal signaling region that includes a five-residue recognition and cleavage motif (LPXTG), which is itself followed by a stretch of aliphatic amino acids and a tail end of positive-charged residues (for detailed reviews, see [60, 62, 64]). To identify any potential loci for SDPs within the nine L. ruminis genomes, we made use of the CW-PRED algorithm, a Hidden Markov Model (HMM)-based tool that predicts the occurrence of LPXTG (or LPXTG-like) domains [54]. The output of these predictions revealed that per genome there is a minimum of three and maximum of seven hits for SDP loci. For these, each of the corresponding primary structures was visually inspected for an intact LPXTG domain region, whereupon one hit was judged an artifact and subsequently rejected. In total, nine types of SDPs were found, of which three loci are part of the core genome, with the six others being associated with the accessory genome (S7 Table). Surprisingly, although it is common for SDPs to also have a Sec-signal peptide at the N-terminus, not all of these nine protein loci were recognized within our SignalP prediction data (S2 Table). On the other hand, largely due to the predicted presence of an intact LPXTG domain, each of these genes was identified by using the TMHMM method (S5 Table).

Amongst the designated core genes (S7 Table), we had identified those encoding for the backbone (LrpA) and tip (LrpC) pilin subunits of the LrpCBA pilus [23]. As our earlier work revealed the LrpCBA pilus is a macromolecular structure and assembled from three different protein subunits [23], it was unexpected that the gene for the basal LrpB pilin would not be found in the core genome, particularly since the three LrpCBA pilin subunits are encoded by loci that are clustered and expressed together as a fimbrial operon (called lrpCBA) [23]. Of the nine L. ruminis genomes encompassing the core genome, the genome from the ATCC 27782 strain was identified as being the source of this discrepancy. After manually checking the NCBI database where the ATCC 27782 genome sequence is deposited, we found the corresponding LrpB locus tag (LRC_RS00315) had its ORF hidden as a pseudogene. Previously, under an old locus tag (LRC_00610), amino acid sequence for this ORF was attainable and upon inspection we found it represented a truncated version that was missing about 40 residues at the N-terminus, including those for a secretion signal peptide. Further, we noticed that this shortened LrpB primary structure results from an insertion of two adenines along the lrpB coding sequence, thus producing a reading-frameshift change. Relatedly, it is worth mentioning we had previously observed that for the DPC 6832-sourced LrpB pilin, there is a shift in the reading-frame of its deduced primary structure, which we found is due to a missing cytosine in a serine codon [23]. Consequently, during the annotation process the LrpB pilin was detected as two ORFs and had been assigned separate locus tags (LRN_0080 and LRN_0081). Despite the fact that LRN_0081 was confirmed as having the entire LPXTG domain region, it was the LRN_0080 ORF that became aligned with the other-sourced LrpB loci in the L. ruminis pan-genome, which for us led to another inconsistency during our interpretation of these results. However, for both types of detected nucleotide alterations, it is unclear if they arose from a genuine indel mutation or as an unintentional mistake in DNA sequencing. In all likelihood, there is the possibility that the lrpB gene would instead be better placed within the core genome of L. ruminis.

As for the third core gene whose ORF was predicted as having a LPXTG domain, it had been given various annotations (S7 Table), but none offered any sort of functional description other than being a cell wall-anchored protein. Since the amino acid identity between the various-sourced versions of this LPXTG surface protein is quite high (92.7–100%), we examined whether any homology occurs with recognized binding domains. Here, no hits were obtained from a Pfam search (at http://pfam.xfam.org; [69]) when using the HMPREF0542_11626 sequence of ATCC 25644. However, when performing a BlastP search of the NCBI database with the same primary structure, it was revealed that some similarity to the MucBP (Mucin-Binding Protein; pfam06458) domain exists, suggesting this cell wall-anchored protein might adhere to mucosal substrates. This possibility comes as somewhat unexpected, as previously we found that L. ruminis cells show little binding affinity for mucus [23].

Regarding the LPXTG-domained protein loci found within the accessory genome (S7 Table), three of them appear to be strain-specific (LRP_348 in porcine DPC 6830, and P869_01660 and P869_04430 in human S23). As far as any Pfam-predicted functions, no noteworthy matches were obtained from a database search, which is expected considering that all three genes are annotated as hypothetical protein. Also of note, two additional accessory genes are each common to two different strains. However, whereas the LRP_1521 (from DPC 6830) and PEL65_242 (from PEL65) ORFs are assigned with a starch-debranching pullulanase function, the LRU_00897 and P869_09985 ORFs (from the human SPM0211 and S23 strains, respectively) have less revealing hypothetical protein annotations. Likewise, annotations for the remaining non-core loci (P869_01660 and P869_04430 from the S23 strain) are also functionally uninformative, as both corresponding products have been designated as a hypothetical protein. Based on the annotations of these various accessory genes, nothing can be adequately deduced about their prospective roles in the gut-niche specialization of L. ruminis.

Of parenthetical interest, our finding that L. ruminis ATCC 27782 encodes for a maximum number of three SDPs seems to be in conflict with an earlier genomics study that had predicted the occurrence of ten such proteins in this bovine strain [21]. The loci in common with our study are those for the LrpA and LrpC pilus proteins (conceivably as well for LrpB) and LRC_RS00370, but no others were detected. Concerning the additional six loci identified in the other study [21], our examination of their predicted primary structures revealed that none have the C-terminal characteristics for a complete LPXTG domain region (S8 Table). Moreover, we noticed that for four of them (LRC_01690, LRC_16530, LRC_16780, and LRC_16790), they no longer exist in the NCBI database after the ATCC 27782-derived genome had been provided with new locus tags.

With the above in mind, it would seem that the potential variety of SDPs is not too extensive on the outer surface of the L. ruminis species. However, of the modest number of predicted SDPs, those that assemble into the LrpCBA pilus are likely to be advantageous and important for the adhesion of L. ruminis cells to the gut epithelium, and presumably for their inherent autochthonous character as well. Yet, one cannot exclude the possibility that among the other LPXTG-domained proteins whose function is not known, there still might be a useful adhesive role during host colonization. Further, given the host niche environment of the L. ruminis species, its nutritional sustenance weighs heavily on the ability to degrade and utilize various complex carbohydrates, particularly as found in the gut of herbivorous livestock [2, 56]. Thus, it comes as no surprise that two of the L. ruminis genomes (from bovine PEL65 and porcine DPC 6830) encode for a SDP whose putative function is a starch-acting pullulanase enzyme, though still somewhat puzzling that it is not more widespread among the genomes of the other animal-derived strains.

Flagellar and chemotaxis proteins.

Often found on bacterial cell surfaces are long flagellar structures, and while these peripheral features primarily provide the driving force for cellular movement, they are also known to contribute to adhesion, biofilm assembly, and immuno-responsiveness (for detailed reviews, see [70, 71]). As mentioned beforehand, certain L. ruminis strains display a motility function (swimming, swarming, or tumbling) where one or more flagella are used. Presumably, the type of movement would be controlled as a response to external stimuli, which characteristically consists of numerous interactions between so-called chemotaxis proteins [70, 71]. As flagellation and chemotaxis are both complex processes that involve a large number of different proteins with a variety of functions, this would be reflected by the gene content of the L. ruminis pan-genome.

Among the pan-genome loci that are clearly annotated as having a flagellar-related function, there are an estimated 39 core genes, along with an additional nine accessory genes (S9 Table). From this it is evident that the genetics of L. ruminis flagellation is not open to much variation and that there is a highly conserved commonality in the genes needed for encoding the structural constituents and biosynthesis of a flagellum. Here, a majority of L. ruminis genomes (GRL1172, ATCC 25644, ATCC 27782, DPC 6832, PEL65, and S23) have only a single non-core flagellar gene, with one of the genomes (DPC 6830) having none at all. At a molecular level, this suggests that from strain-to-strain the flagella are structured and assembled quite similarly, and that any fundamental divergence from this in form or function is unlikely to be supported by the varying gene content within the accessory genome.

For the loci of the pan-genome that encode the chemotaxis proteins, there are a total of 16 core genes whose annotation predicts an involvement with flagellar chemotactic responses (S10 Table). The most conspicuous of these loci are those having functional annotations for regulating the frequency of tumbling, a phenomenon manifested when the rotation of the flagellum switches from an anti-clockwise to clockwise direction [70, 71]. Correspondingly, the L. ruminis core genome includes variously annotated loci for the cytoplasmic CheA, CheB, CheC, CheD, CheR, CheW, and CheY proteins. Found also prominent among the core genes are those encoding for a number of transmembrane methyl-accepting chemotaxis proteins (MCPs). These sensory receptor proteins are localized on the outer surface of bacterial cells and operate through the detection and binding of chemoattractants or chemorepellents in the surrounding environment [70, 71]. Ordinarily, this leads to a signaled activation of the Che proteins involving cyclical methylation and demethylation, and which ultimately controls tumbling behavior [70, 71]. It is thus clear that MCPs would be obligatory for the active functioning of the L. ruminis flagella. Interestingly, of the three to six non-core loci predicted to encode chemotaxis proteins most are having MCP annotations. Thus, for each of the various L. ruminis strains encompassing the pan-genome, there could be the functional potential for flagella to respond to differing stimuli that are representative of the intestinal milieu. Indeed, this might reflect a phenotype evolvability that tailors the adaptation of a L. ruminis strain to the specific environmental conditions in a particular host gut. Incidentally, it is tempting to speculate that the oddly resuscitated tumbling motility observed with the ATCC 25644 human strain in vivo [22] might rest with one or more anomalous or possibly defective chemotaxis proteins. This in itself can be seen as an alternative explanation to an earlier proposed model involving the transcriptive regulation of a flagellin gene [22].

Successful bacterial colonization in the host GI tract can be construed as the logical precondition of cellular persistence, and which can lead to an indigenous lifestyle. For the L. ruminis species, its potential for flagellar chemotactic movement is advantageous and can play a key adaptive role in occupying the gut autochthonously. For instance, through a response to certain attractants, flagella would likely be able to propel cells toward a favorable niche location containing abundant nutrients, reduced oxygen levels, and optimal attachment sites, such as what can be found throughout the folded crevices of the inner intestinal walls. Presumably, in reaching these deep and concealed locales, flagellum-directed motility will have helped prevent L. ruminis cells from becoming washed out along with the transiting luminal contents. Further, because chemotactic flagella are also able to sense and recognize repulsive stimuli, L. ruminis would be equipped to avoid the less ecologically hospitable aspects of the host intestine (e.g., oxygenated regions). By being flagellated on the outer cell surface, the various L. ruminis strains will have evolved an effective molecular strategy that supports an ecological adaptation to a very specific gut microenvironment, which can be viewed as also promoting a characteristic autochthonous presence within a given host.

Fermentation.

Enough experimental evidence already exists to indicate that L. ruminis maintains an obligately homofermentative metabolism [17]. For this, cells will undergo lactic acid fermentation to generate cellular energy in a process that breaks down one molecule of glucose to yield two molecules of lactate [72]. Genomically, obligate homolactic bacteria are distinguished by the presence of fructose-1,6-bisphosphate (FBP) aldolase, a key glycolytic enzyme of the Embden-Meyerhof-Parnas (EMP) pathway through which hexose sugars are metabolized [72]. Our analysis of the L. ruminis pan-genome indicates that each of the nine strains has an ORF whose predicted product is annotated as a FBP aldolase (Fig 5A), which makes it part of the core genome (S1 Table). Further inspection of the pan-genome revealed that the annotated loci encoding all other EMP pathway enzymes, including lactate dehydrogenase for catalyzing the conversion of pyruvate to lactate, are as well found in the core-gene pool (Fig 5A). With such pathway-related genes in the core genome, this not only highlights an essential physiological role, but it also provides the genetic basis for the homolactic fermentation profile characteristically observed in L. ruminis cells.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Schematic outline of the putative homolactic and heterolactic fermentative pathways in L. ruminis.

Depicted is the glycolytic metabolism that can occur during homolactic fermentation via the Embden-Meyerhof-Parnas (EMP) pathway (A) or heterolactic fermentation via the phosphoketolase pathway (PK) (B). Included are the various pathway reactions with their precursor metabolites and enzymes. Shown in the insets are the names of the glycolytic enzymes, either numbered (1–11) or alphabetized (a-e) along with the locus tags of their putative genes in the nine L. ruminis genomes. Enzymes that overlap in both pathways are numbered similarly.

https://doi.org/10.1371/journal.pone.0175541.g005

Other Lactobacillus species are classified as either facultative or obligate heterofermenters and will instead metabolize glucose using the phosphoketolase (PK) pathway, but whereby lactate, ethanol, and CO₂ are the end products [72–74]. In fact, the PK pathway is the same means that some lactobacilli use for fermenting pentose sugars [72]. One example is Lactobacillus sakei, which can grow on ribose as the sole source of carbon and energy [75]. In order for this to occur, ribose must be transported into the cytoplasm and then become phosphorylated. In L. sakei, these processes necessitate the involvement of an operon (rbsUDK) encoding the ribose transporter (RbsU), D-ribose pyranase (RbsD), and ribokinase (RbsK) proteins, and whose expression is regulated by the RbsR repressor protein [75]. In the case of L. ruminis, its own ability to catabolize ribose is the subject of conflicting studies, one reporting that it occurs in the bovine strain NRIC 1689 (also called ATCC 27780 or PEL65) [76], but another concluding that the same isolate along with several other strains shows no growth on this pentose sugar [18]. In this latter study [18], it was suggested that because the gene for a putative transaldolase (one of the enzymes in the pentose phosphate pathway) is not found in the genome of two strains (ATCC 27782 and ATCC 25644), this can be inferred as the possible reason why ribose is non-fermentable by L. ruminis. However, given that the foremost function of the pentose phosphate pathway is an anabolic one, a missing transaldolase enzyme might not necessarily represent an adequate explanation, particularly since pentose fermentation by heterolactic bacteria is mainly through the PK pathway [72].

To tackle this discrepancy, we examined our pan-genome data (S1 Table) to determine whether L. ruminis contains the genes that encode the necessary enzymes for a PK pathway. Loci for the full repertoire of PK pathway enzymes were detected in the L. ruminis core genome (Fig 5B), but with the exception of the all-important pentose-cleaving phosphoketolase, which was not found in the genome of the DPC 6832 strain. Of related interest, following the completion of this article, we became aware that the phosphoketolase gene is present in the L. ruminis DPC 6832 genome after its locus tags in the NCBI database were reassigned. Nonetheless, while it remains uncertain that these gene products are actively expressed and participate in a functional pathway that ferments pentose sugars, one might begin to speculate and reconsider whether L. ruminis is suitably placed in a group that contains obligately homofermentative lactobacilli. Yet, on the other hand, an examination of the L. ruminis pan-genome for the genes encoding the proteins involved with mediating active ribose transport had revealed that while annotated genes for the ribokinase (GRL1172_644, HMPREF0542_10159, LRC_RS07720, LRN_1311, PEL65_950, PEL66_2068, P869_06985, LRU_01270, and LRP_932) and ribose operon repressor (GRL1172_1281, HMPREF0542_12070, LRC_RS08440, LRN_1438, PEL65_1929, PEL66_741, P869_08795, LRU_01112, and LRP_465) proteins were identified within the core genome, those for the ribose transporter and D-ribose pyranase proteins were not present. Though other factors can be involved, one can justifiably argue the lack of growth on ribose by L. ruminis strains that others had observed [18] might be more attributed to the missing components of the ribose transport machinery. This in turn would serve to lessen any legitimacy that the bovine isolate of L. ruminis (NRIC 1689) is a ribose-fermenting strain [76]. Thus, if it is the situation that ribose (or xylose [18]) cannot be utilized by L. ruminis, it raises the question as to which type of pentose sugar would be transported into the cell and processed through a prospective PK pathway. However, were it otherwise to be shown that a particular L. ruminis strain can as well undergo a heterofermentative metabolism via the PK pathway using various pentose and hexose sugar substrates, this can be seen as an adaptive gain or benefit when competing in the gut microcosm, as it would offer a wider access to fermentable carbon and energy sources, particularly during periods of fluctuating nutrient availability. Presumably, the potential capacity for cellular growth using a more mixed fermentative metabolism would further help support the perpetuity of L. ruminis colonization within the host gut.

Anaerobic respiration.

Contrary to their namesake, certain members of the LAB also have the genetic means of obtaining energy by respiration, whereupon lactate is no longer the main end product, but then provided that an exogenous source of both heme and menaquinone is available beforehand to cells [77]. In an anoxic environment, the absence of oxygen precludes its use as the terminal electron acceptor in the respiratory chain, and thus other oxyanionic compounds must be used instead, and for which a corresponding oxidoreductase enzyme is considered indispensable [78, 79]. Thus far, among the LAB only a few have the minimal number of necessary genes for an anaerobic respiratory metabolism, such as when either nitrate (Lactobacillus plantarum, Lactobacillus fermentum, and Lactobacillus reuteri [80]) or fumarate (Enterococcus faecalis [81]) is used as the final electron acceptor.

Generally, for bacteria that respire anaerobically using nitrate, this ability can be identified at the genomic level by the clustering of genes (narGHI operon) for a nitrate reductase complex, which itself consists of three subunits (NarG, NarH, and NarI) and is membrane-bound [78]. Still, for other bacteria, a single locus that encodes a monomeric soluble nitrate reductase can as well be found, and presumably, this form would represent the smallest structural and functional unit that mediates nitrate reduction [78]. Gene prediction within the L. ruminis pan-genome shows no evidence of what resembles the narGHI operon, but there is one ORF with a nitrate reductase annotation (LRC_RS08915 from ATCC 27782) (S1 Table). However, a closer inspection of its primary structure had revealed this ORF is likely mis-annotated. Based on the domain assignments from the Pfam database [69] (data not shown), the putative protein product of this ORF actually belongs to the amidinotransferase family, which as well includes the arginine deiminase enzyme. It should be noted that orthologous genes from other genomes in the pan-genome have these two types of annotations. Evidently, there is no genetic rationale in L. ruminis for carrying out anaerobic nitrate respiratory growth.

In contrast, our analysis of the pan-genome data did seem to indicate the possibility of fumarate respiration in L. ruminis (S1 Table). Here, we found ORFs in the core genome (GRL1172_1261, HMPREF0542_10777, LRC_RS01705, LRN_0304, PEL65_229, PEL66_1400, P869_08075, LRU_02203, and LRP_1424) whose annotation suggests that they encode for a fumarate reductase (Fig 6), the key enzyme allowing the use of fumarate as the terminal oxidant during anaerobic respiratory growth [79]. Following a BlastP search, we were more convinced of this identity, as its primary structure shows strong similarity with the flavocytochrome c fumarate reductases from other types of bacteria (data not shown). Since bacterial fumarate reductases are able to exist as both a membrane-bound or soluble enzyme, this is also differentiated in the way they are encoded in the genome. For instance, fumarate reductase in Escherichia coli is located on the inner surface of the cytoplasmic membrane and consists of four different subunits that form the catalytic (FrdA and FrdB) and membrane-embedded (FrdC and FrdD) domains, with their corresponding genes encoded within the frdABCD operon [79]. Alternatively, for the genus Shewanella, fumarate reductase is a soluble periplasmic enzyme, existing as a monomer and thus encoded by a single gene [79]. Likewise, in the LAB E. faecalis, it is recognized that only one gene encodes its fumarate reductase [81]. As for the putative fumarate reductase we uncovered in L. ruminis, its locus is not part of a gene cluster, thus suggesting that this enzyme is monomeric. Since the L. ruminis fumarate reductase gene was not found amongst our SignalP or TMHMM prediction data (S1 and S5 Tables, respectively), but was identified when using the SecretomeP program (S3 Table), it is reasonable to assume that this enzyme is not membrane-bound, and perhaps might even be non-secretable and confined to the cytoplasm. Nonetheless, even though anaerobic respiration is regarded as less energetically efficient than aerobic respiration, it still yields more energy than when bacterial cells use fermentative processes and, by comparison, will lead to increased cellular biomass and long-term survival [82]. In terms of metabolic versatility, any putative capacity of L. ruminis to respire anaerobically on fumarate instead of undergoing lactic acid fermentation would likely represent a competitive edge over the other non-respiring intestinal bacteria and a potential fitness benefit for autochthonous growth in the gut.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Schematic representation of the fumarate reductase-catalyzed reaction.

Shown is the reaction catalyzed by fumarate reductase, where the reduction of fumarate to succinate is coupled with the oxidation of menaquinol to menaquinone. Included are the nine locus tags for the putative fumarate reductase gene found present in the L. ruminis core genome.

https://doi.org/10.1371/journal.pone.0175541.g006