Identification of protein secretion systems in bacterial genomes

Abstract

Bacteria with two cell membranes (diderms) have evolved complex systems for protein secretion. These systems were extensively studied in some model bacteria, but the characterisation of their diversity has lagged behind due to lack of standard annotation tools. We built online and standalone computational tools to accurately predict protein secretion systems and related appendages in bacteria with LPS-containing outer membranes. They consist of models describing the systems' components and genetic organization to be used with MacSyFinder to search for T1SS-T6SS, T9SS, flagella, Type IV pili and Tad pili. We identified ~10,000 candidate systems in bacterial genomes, where T1SS and T5SS were by far the most abundant and widespread. All these data are made available in a public database. The recently described T6SS(iii) and T9SS were restricted to Bacteroidetes, and T6SS(ii) to Francisella. The T2SS, T3SS, and T4SS were frequently encoded in single-copy in one locus, whereas most T1SS were encoded in two loci. The secretion systems of diderm Firmicutes were similar to those found in other diderms. Novel systems may remain to be discovered, since some clades of environmental bacteria lacked all known protein secretion systems. Our models can be fully customized, which should facilitate the identification of novel systems.

Figure 1. Model and results for the T1SS.

(a) Schema of the structure (left panel) and model of the genetic organisation (right panel) of T1SS. We built protein profiles for the three components and modelled the two possible genetic architectures of the T1SS: one with the three components encoded in a single locus (inter_gene_max_space parameter in MacSyFinder: d ≤ 5 genes), another with the ABC transporter and the MFP encoded in a locus while the OMF is further away (loner attribute). A single OMF can also be used by different T1SS and this is noted by the attribute multi_system. (b) Distribution of hits for the protein profiles of the T1SS components, separated in two groups: hits effectively part of a T1SS main locus (i.e., containing at least ABC and MFP, blue) and hits found elsewhere (“Other”, grey). Even if encoded outside of “main loci” (grey area of the bar), OMF might be involved in T1SS (loner property), whereas it is not the case for ABC and MFP. (c) T1SS encoded in one single locus (ABC, MFP and OMF co-localise) (3) or in two (OMF encoded away from the other components) (2 + 1).

Figure 2. Model of the T2SS for detection and discrimination from the T4 and Tad pili.

(a) Schema of the structure (left panel) of T2SS, and model of its genetic organisation (right panel), indicating components with homologies with T4P and Tad pilus. We built protein profiles for all these components (Tables S4 and S5). Protein families represented by the same colour are homologous, and their profiles often match proteins from the other systems (except for the Flp and TadE/F families that are less similar). Some prepilin peptidases of T2SS and T4P are defined as functionally interchangeable (curved double-headed arrow, exchangeable attribute). Boxes represent components: mandatory (plain), accessory (dashed) and forbidden (red crosses). (b) Scores of proteins matched with the profiles of T2SS and T4P. The components of actual T2SS (dark blue) and actual T4P (in red) are well separated, indicating that in each case the best match corresponds to the profile of the correct model system. The exceptions (blue points surrounded by a black ellipse) concern the prepilin peptidases (light blue squares, circled in blue), which are effectively inter-changeable. (c) Representation similar to (b), but for the comparison between T2SS (blue) and Tad (grey) systems. In this case, the separation is perfect: the proteins always match better the protein profile of the correct system. (d) Number of detected systems per genome among the 1,528 genomes of diderm bacteria.

Figure 3. Model of T3SS and T4SS.

(a) The models of T3SS and flagellum were built based on a previous study (representation conventions as in Fig. 2). Of the nine mandatory components for the T3SS only the secretin is forbidden in the model of the flagellum. Conversely, three flagellum-specific components are forbidden in the T3SS model. Three different types of secretins are found in T3SS derived from different appendages, which are thus defined as exchangeable in the model. (b) Models of the T4SS were built based on a previous study. Two different proteins have been described as type 4 coupling proteins (T4CP: VirD4 and TcpA) and two as the major ATPases (VirB4 and TraU, which are homologous). Some pT4SS lack a T4CP and secrete proteins from the periplasm. The relaxase (rel), is necessary for conjugation but not for protein secretion, although some relaxase-encoding T4SS are found in both cT4SS and pT4SS. Only two MPF types are associated with protein secretion - pT4SS_I and pT4SS_T, corresponding to MPF_I and MPF_T types. The specificity of type-specific profiles is assessed in Fig. S2.

Figure 4. Model of the T5aSS, T5bSS and T5cSS.

The left panel shows simplified schemas of the T5SS, and the right panel displays the respective genetic model (only one component that is classed as loner). The translocator, pore-forming domains were searched using PFAM domains for T5aSS and T5cSS (resp. PF03797 and PF03895), and a profile built for this work for the T5bSS (Tables S1, S4 and S5).

Figure 5. Model and results for the detection of T6SS.

(a) The left panel shows the schema of the structure of T6SSⁱ, and the right panel displays the genetic model of the three sub-types of T6SS (representation conventions as in Fig. 2). For T6SSⁱ, we built profiles for the 14 mandatory components, which were clustered if at a distance of d ≤ 20 (see Fig. S4). For T6SSⁱⁱ and T6SSⁱⁱⁱ, we built 17 and 13 profiles respectively. All components were set as mandatory, except for TssQ, which is found in half of the T6SSⁱⁱ. Homologies between components that are displayed by the mean of the same colours of boxes between the different sub-types are based on previous studies. *Putative type-specific genes are displayed in grey boxes that do not represent homologies. However, several putative homologies were retrieved using Hhsearch (e-value < 1 and p-value < 0.05) on T6SSⁱⁱ components: iglC (tssG), iglG (tssF), iglH (tssE), iglJ (tssH) and pdpD (tssH). (b) Number of different components per cluster of T6SSⁱ. Following this analysis, we set the quorum parameter of T6SSⁱ to 11. (c) Frequency of hits for each type of T6SSⁱ components in the genomes. Hits matching a single-locus T6SSⁱ are in blue. The other hits match outside the T6SSⁱ loci. (d) Frequency of each component within single-locus T6SSⁱ. The components EvpJ and TssA were detected in less than 45% of the T6SSⁱ, while the other components were found in most T6SSⁱ loci (>89%).

Figure 6. Genetic model of the T9SS.

The representation follows the conventions of Fig. 2. The model includes 11 components for which 13 protein profiles were obtained from PFAM (SprA and SprA-2), TIGRFAM (GldJ, GldK, GldL, GldM, GldN and SprA-3) or designed for this study (PorU, PorV, PorQ, SprE, SprT). Four components were declared as loners. The co-localisation distance for the others was set at d ≤ 5 (see Fig. S5). As several profiles were available for SprA, we included them all in the models, and declared them as exchangeable homologs in the model. GldJ is not part of the secretion system, but of the gliding motility system. It was included in the model as it facilitates the detection of T9SS components that co-localise with it.

Figure 7. Phylogenetic distribution of protein secretion systems in bacteria.

Within each clade, the proportion of genomes harbouring each system is indicated in boxes whose colours follow a gradient from full red (100%) to white (0%) (see legend). Clades were classed as monoderms (grey or “M” symbol), diderms with Lipopolysaccharide-containing outer membranes (Diderm-LPS in bold, no symbol), diderms with homologs of LPS pathway that putatively have LPS (D^?) and diderms with no LPS (D−). The peculiar envelope of the Thermotogae is indicated (T). The Firmicutes are typically monoderms, but some of their members are diderms (the Negativicutes, some Clostridia, Mycobacteria). The bar plot shows the number of detected systems. Bars are split in two categories to separate on one side Alpha- Beta- and Gamma-proteobacteria, and on the other genomes from other bacteria. We display the number of occurrences of systems occurring rarely in our dataset on top of the bars. Clades with less than 4 genomes and/or with unreported phylogenetic position are not shown (i.e., Chrysiogenetes, Gemmatimonadetes, Nitrospirae and Thermodesulfobacteria). This sketch tree was drawn from the compilation of different published phylogenetic analyses.