Phylogenetic comparison between Type IX Secretion System (T9SS) protein components suggests evidence of horizontal gene transfer

Discussion

The 19 Bayesian Inference (BI) trees of T9SS protein components exhibit monophyletic clades for all major classes under Bacteroidetes with strong support for the monophyletic clades or its subclades (Figs. 2–6). Similar to the 19 BI trees of T9SS protein components, the BI tree of 16S rRNA also exhibits monophyletic clades for all major classes under Bacteroidetes with strong support (Fig. 7). 16S rRNA has been extensively used in phylogenetic analysis for the purpose of evolutionary comparison and classification. The reliability of this approach lies on the assumption that the 16S rRNA gene undergoes hierarchical and unidirectional evolution and no gene transfer of 16S rRNA occurs between species (Karlsson et al., 2011). Due to the advantages that the 16S rRNA gene has such as ubiquity in bacterial genomes, being easily sequenced, and widely available in public sequence databases, the current universal tree of life is based on the phylogeny of this gene (Winker & Woese, 1991; Coutinho et al., 1999; Pylro et al., 2012). That assumption has been challenged due to the presence of multiple copies of 16S rRNA in a bacterial genome and the 16S rRNA genes from operons in the same genome are rather distinct which might suggest that such genes might have undergone horizontal gene transfer (Pei et al., 2010; Karlsson et al., 2011). However, the extent of 16S rRNA evolution remains considerably less compared to the other genes in the bacterial genome (Espejo & Plaza, 2018). Thus, 16S rRNA remains relevant for the purpose of evolutionary comparison and classification.

It is expected that the BI trees of T9SS protein components would exhibit similar phylogeny with the BI tree of 16S rRNA. However, the BI trees of T9SS protein components exhibit inconsistent positions of terminal nodes from minor classes among themselves and the phylogeny for the minor classes deviate from the phylogeny exhibited by the BI tree of 16S rRNA (Figs. 2–7). Hence, the minor classes under Bacteroidetes are excluded from the comparison between the 20 BI trees of T9SS protein components. This might arise due to insufficient taxa from minor classes provided to construct those BI trees. Hence, the information that is provided is insufficient to fully resolve the phylogeny of minor classes. As more T9SS-acquiring species from minor classes are sequenced later on, the phylogeny of T9SS protein components will be more resolved (Alvizu et al., 2018).

Different from the other 19 BI trees of T9SS protein components, the BI tree of PorR does not exhibit monophyletic clades for all major classes under Bacteroidetes (Figs. 2–6).

The presence of strong support (posterior probability value >0.95) as denoted by the black branch leading to the top half of the BI tree of PorR (Fig. 3C) suggests that there is strong support that the phylogeny exhibited by the BI tree of PorR deviates from the phylogeny based on the 16S rRNA sequence (Fig. 7). Thus, there is a possibility that the porR gene is subjected to horizontal transfer hence causing deviation from the expected phylogeny (Pylro et al., 2012). Hirt, Schlievert & Dunny have demonstrated that virulence factors and antibiotic resistance genes could be horizontally transferred (Hirt, Schlievert & Dunny, 2002). Hence, this suggests the possibility that the porR gene that encodes one of the virulence factors produced by P. gingivalis can be horizontally transferred (Shoji et al., 2002; Shoji et al., 2014).

The arrangement of porR and its neighbouring genes in the P. gingivalis ATCC 33277 genome was identified in order to support the possibility that porR is horizontally transferred. P. gingivalis ATCC 33277 genome was chosen because many gene orthologs that are involved in A-LPS biosynthesis have been identified in this genome (Shoji et al., 2018). porR and its neighbouring genes are found to be flanked by insertion sequences (IS5 family transposons) (Fig. 8). The IS5 family transposons (cyan rectangles) contain a single open reading frame that encodes for IS5 family transposase that cleaves the 12 bp inverted repeats (purple triangles) that flank the insertion sequences (Fig. 8). The 12 bp inverted repeats show imperfect homology to each other with the consensus sequence: GAGACCTTTG[CA]A. Both of the IS5 family transposons are ∼1300 bp in length. These features are typical of IS5 family transposons (Mahillon & Chandler, 1998; Naito et al., 2008). The intervening DNA segment and both IS5 family transposons that flank it might form a composite transposon where the cleaving action of IS5 family transposases on inverted repeats can mobilise the intervening DNA segment that contains the porR gene and possibly subject it to conjugative transfer (Thomas & Nielsen, 2005; Brochet et al., 2009). The length of the composite transposon is ∼70 kbp. However, it is also possible for IS5 family transposase to cleave the inverted repeat directly downstream of rfa (PGN_1255) (Fig. 8) which will reduce the length of the composite transposon to ∼47 kbp. It has been reported that a transposon of ∼47 kbp in length is able to undergo both transposition and conjugation processes (Brochet et al., 2009). Hence, it might be possible for composite transposons of such length to undergo transposition and subsequently be horizontally transferred via bacterial conjugation.

The intervening DNA segment contains seven genes that are involved in the biosynthesis of A-LPS (Fig. 8). Both porR (PGN_1236) and ugdA (PGN_1243) genes are involved in the Wbp pathway that is important for the biosynthesis of di-acetylated glucuronic acid which is the structural sugar of A-LPS (Shoji et al., 2002; Shoji et al., 2014). The porS gene (PGN_1235), which is an O-antigen flippase, and wzy gene (PGN_1242), which is an O-antigen polymerase, are involved in the assembly of A-LPS on the periplasmic side of bacterial IM (Shoji et al., 2013). gtfB (PGN_1251) and gtfE (PGN_1240) glycosyltransferase genes are involved in the biosynthesis of the sugar moiety of A-LPS. rfa (PGN_1255) glycosyltransferase gene is involved in the biosynthesis of the lipid A-core moiety of A-LPS (Shoji et al., 2018). However, there are other genes that are involved in the biosynthesis of A-LPS and they are spread out throughout the genome (Shoji et al., 2018). Usually, genes that are co-regulated and involved in a similar pathway are clustered in a single operon (Yanofsky & Lennox, 1959; Osbourn & Field, 2009). Thus, it is possible that the other genes do not form a cluster with the seven genes that are identified to be flanked by insertion sequences (IS5 family transposons) because they are not co-regulated.

It is possible that those seven genes might be co-transferred via horizontal gene transfer. Thus, phylogenetic analysis was performed for the protein alignment of UgdA that is encoded by ugdA which, together with porR, are involved in the Wbp pathway and are co-localised in the intervening DNA segment flanked by IS5 family transposons (Fig. 8). The BI tree of UgdA (Fig. 9B) was constructed to be compared with the BI tree of PorR (Fig. 9A). Different to the 19 BI trees of T9SS protein components, both BI trees do not exhibit monophyletic clades for all major classes under Bacteroidetes. They also exhibit similar topology where four similar clusters (I, II, III, and IV) with strong support (denoted by a black branch leading to the cluster) have been identified in both BI trees. Cluster I consists of terminal nodes from Flavobacteriia and a few terminal nodes from other classes. Cluster II consists of terminal nodes from Porphyromonas, Tannerella, and Parabacteroides genera. Cluster III consists of terminal nodes from Rufibacter and Hymenobacter genera. Cluster IV consists of terminal nodes from Prevotella, Bacteroides, Proteiniphilum, and other genera. These four clusters exhibit similar relative positions to each other in both BI trees (e.g., cluster I is closer to cluster II than the other clusters and cluster III is closer to cluster II than the other clusters). However, due to the differences in branch lengths between both BI trees, they look slightly different as the upper part of the UgdA tree (Fig. 9B) appears more elongated than the upper part of the PorR tree (Fig. 9A), while the lower part of the UgdA tree (cluster I) appears more shortened than the lower part of the PorR tree (cluster I).

Other than the BI tree of PorR, the BI trees of the other 19 T9SS protein components also exhibit evidence of horizontal gene transfer perhaps between classes under Bacteroidetes. As listed in Table S1, there are terminal nodes that are out of their expected monophyletic clades in the BI trees of those components that suggests the genes that encode them might be horizontally transferred. In theory, the common ancestral species of a monophyletic clade for a class under Bacteroidetes passes the gene that encodes T9SS protein components to its descendant species. Thus, the descendant species that are out of their expected monophyletic clades most likely acquired that gene from the common ancestral species of a monophyletic clade from another class that could be interpreted as a horizontal gene transfer between classes under Bacteroidetes (Thomas & Nielsen, 2005; Brochet et al., 2009). It is interesting to highlight that there are species that frequently have their corresponding terminal nodes in those 19 BI trees out of their expected monophyletic clades (Figs. 2–6) such as F. taffensis DSM 16823, bacterium L21-Spi-D4, O. hongkongensis DSM 17368, and D. orientale. Thus, it is likely that those bacteria frequently acquire their T9SS components through horizontal gene transfer. However, for the genes that encode those 19 components, they might undergo horizontal gene transfer less frequently compared to porR that causes most of the terminal nodes of BI trees of those components to cluster according to their respective classes. It might be because the intervening DNA segment that contains the porR gene is easily exchanged between bacteria under Bacteroidetes due to the presence of insertion sequences (IS5 family transposons) that flank it (Fig. 8).

T9SS is made up of various protein components that form the regulation, translocation, energetic, and modification components. Currently, the secretion system is primarily found in bacteria under the Bacteroidetes phylum (Abby et al., 2016). Bacteria from classes under Bacteroidetes (Bacteroidia, Flavobacteriia, Cytophagia, Chitinophagia, Sphingobacteriia, Saprospiria, Incertae sedis, and unclassified) are found to acquire T9SS protein components (Figs. 2–6). However, not all of them acquire the 20 components that have been reported (Sato et al., 2010; Lasica et al., 2017). As shown in the 20 BI trees of T9SS protein components (Figs. 2–6), only bacteria under Bacteroidia, Flavobacteriia, and Chitinophagia acquired the 20 components investigated. The bacteria under Cytophagia only acquired 19 components (except PorN). The bacteria under Saprospiria only acquired 18 components (except PorL and PorG). The bacteria under Sphingobacteriia only acquired 17 components (except PorQ, PorU, and PorZ). The unclassified bacteria only acquired 17 components (except PorN, PorU, and PorG). The bacteria under Incertae sedis only acquired 11 components (PorQ, PorR, Sov, PorU, PorV, PorX, PorY, PorZ, SigP, Omp17, and PorF). It is interesting to note that PorU, PorZ, and PorQ form the modification components of T9SS. Thus, Sphingobacteriia does not acquire the components that perform post-translational modifications on T9SS cargo proteins such as cleavage of CTD and A-LPS glycosylation. Perhaps, T9SS acquired by Sphingobacteriia does not cleave the CTD of cargo protein and glycosylate it with A-LPS, but leaves the cargo protein bounded to PorV after it is translocated to bacterial cell surface by Sov. Another possible explanation is that Sphingobacteriia does have proteins that perform the functions of missing protein components. However, those proteins exhibit limited sequence similarity with any currently known T9SS protein component. Thus, they could not be detected by the homology searching method used in this work. This explanation could also be applied for other species of bacteria under Bacteroidetes that do not acquire the homologs of the 20 T9SS components.

This work has found other species under Chitinophagia, Saprospiria, and those that are unclassified that acquired homologs of T9SS components that, to our knowledge, might not have been reported (McBride & Zhu, 2013). Those other species and the homologs of T9SS components they acquired are shown in Fig. 10. This identification might be due to the analysis that was performed which might cover more bacterial species than previous works as more bacterial genomes have been completely sequenced in the past few years.

Conclusions

The objective of this work was to investigate the phylogenetic relationship among putative members of 20 T9SS component protein families (Emrizal & Muhammad, 2018). The Bayesian Inference (BI) trees for 19 T9SS protein components exhibit monophyletic clades for all major classes under Bacteroidetes with strong support for the monophyletic clades or its subclades, which is consistent with the phylogeny exhibited by the constructed BI tree of 16S rRNA. However, the BI tree of PorR is different from the other 19 BI trees of T9SS protein components as it does not exhibit monophyletic clades for all major classes under Bacteroidetes. There is strong support for the phylogeny exhibited by the BI tree of PorR which deviates from the phylogeny based on the 16S rRNA sequence. Thus, there is a possibility that the porR gene is subjected to horizontal transfer as it is known that virulence factor genes could be horizontally transferred. Seven genes that are involved in the biosynthesis of A-LPS that includes porR are found to be flanked by insertion sequences (IS5 family transposons). This suggests that the intervening DNA segment that contains the porR gene can be transposed and subjected to conjugative transfer. Thus, the seven genes might be co-transferred via horizontal gene transfer. Similar to the BI tree of PorR, the BI tree of UgdA does not exhibit monophyletic clades for all major classes under Bacteroidetes (both are a part of the seven genes). Both BI trees also exhibit similar topology where the four identified clusters with strong support have similar relative positions to each other in both BI trees. Other than the BI tree of PorR, the BI trees of the other 19 components also exhibit evidence of horizontal gene transfer. However, for the genes that encode those 19 components, they might undergo horizontal gene transfer less frequently compared to porR because the intervening DNA segment that contains porR is easily exchanged between bacteria under Bacteroidetes due to the presence of insertion sequences (IS5 family transposons) that flank it. This work also found other species under Chitinophagia, Saprospiria, and those that are unclassified that acquired T9SS component protein homologs that, to our knowledge, might not have been reported.

Supplemental Information