The regions comprising and surrounding the T3SS-2 gene clusters of P. syringae pv phaseolicola 1448a, P. syringae pv oryzae str. 1_6, P. syringae pv tabaci ATCC11528, Rhizobium spp. NGR234 and the regions comprising and surrounding the unique T3SS gene clusters of Bradyrhizobium japonicum USDA 110, Rhizobium etli CIAT 652 and R. etli CNF 42 were retrieved from the NCBI Genome database. In the cases of P. syringae pv tabaci ATCC11528 and P. syringae pv aesculi the nucleotide sequence in the region close to the T3SS gene cluster was retrieved (GenBank: N° ACHU01000133 and N° ACXS01000083.1 respectively) after being identified through MegaBLAST searches and found to be present in P. syringae pv phaseolicola 1448a, but absent from P. syringae pv tomato DC3000 and Pseudomonas syringae pv syringae B728A; coding sequences were identified with NCBI’s ORF Finder tool.

Each coding sequence annotated in the T3SS gene clusters of P. syringae pv phaseolicola 1448a, R. etli CIAT 652 and Rhizobium spp. NGR234 was analyzed by Psi-BLAST searches 10 against the NCBI non-redundant database reduced for bacteria using the following parameters: BLOSUM 65 substitution matrix; expected threshold 10; word size 3; gap costs: existence: 11, extension 1; the filter for low complexity regions was set to on. The number of descriptions and alignments to be reported was set to 500 and conditional compositional adjustments were on. The program FoldIndex© was used with default parameters for the prediction of structural disorder propensity from the amino acid sequences 11 . Secondary structure predictions were performed with PSIPRED 12 . Physical and chemical parameters of sequences under study were estimated by ProtParam 13 . Coiled coil predictions and assignment of the heptad repeat positions in proteins were produced in COILS 14 and MATCHER 15 respectively. Sequence threading techniques and fold-recognition algorithms were used to identify distant homologs. 3-D structural profiles for T3SS proteins were predicted from sequence data was performed using the PHYRE pipeline 16 . The program Memstat3 17 was used for the prediction of membrane α-helices in proteins.

The gene synteny of the T3SS-2 clusters of P. syringae pv phaseolicola 1448a, P. syringae pv oryzae str. 1_6, P. syringae pv tabaci ATCC11528, Rhizobium spp. NGR234 and the gene synteny of the unique T3SS gene clusters of B. japonicum USDA 110, R. etli CIAT 652, R. etli CNF 42, were compared to other known T3SS gene clusters of various bacteria using the BLASTN and BLASTP tools of the Genbank. Codon Usage Bias analysis was performed using DnaSP v5 18 .

In the subsequent sections the unified nomenclature for T3SS proteins (Table 1) will be followed 26 . The phylogenetic analysis of various T3SS core proteins (including T3SS-2 proteins), e.g. SctU (RhcU/HrcU/YscU/FlhB and their homologues), SctV (RhcV/HrcV/LcrD/FlhA homolog proteins), SctQ (RhcQ/HrcQ/YscQ/FliN/ and their homologues) and the T3SS ATPases SctN (RhcN/HrcN/YscN/FliI and homologues), confirmed the broad classification of the non-flagellar T3SS into seven families. However, the T3SS-2 proteins were grouped in the same clade with the Rhc T3SS proteins with high bootstrap values, suggesting that these lineages share a more recent common origin than with other T3SS families.

Shaded boxes are indicative of proteins with analog function but no sequence homology to the Ysc T3SS family. Double names are also reported for various cases.

Interestingly, the Rhc T3SS family can be further subdivided into three subgroups: Subgroup I is represented by the well-known T3SSs of Rhizobium sp. NGR234, and B. japonicum USDA 110 while subgroup III is represented by the T3SS present in R. etli. Proteins from the T3SS-2 system of various P. syringae strains are grouped closer to the T3SS-2 of Rhizobium sp. NGR234 (Figure 1, 2, Additional files 1, Additional file 2 & Additional file 3: Figures S1, S2 & S3), forming the subgroup II of the Rhc T3SS family.

BLASTP and Psi-BLAST searches revealed the main T3SS components of the novel T3SS-2 gene cluster of P. syringae pv phaseolicola 1448a which are also conserved in P. syringae pv oryzae str. 1_6, P. syringae pv tabaci ATCC11528 (Additional file 4: Table S1) and P. syringae pv aesculi. Similar searches and comparisons were also carried out with the T3SSs of R. etli CNF 42, R. etli CIAT 652 and Rhizobium sp. strain NGR234. In the following, the prefix Hrc _II will be used to specify the conserved T3SS-2 proteins of P. syringae pv phaseolicola 1448a, P. syringae pv oryzae str. 1_6 and P. syringae pv tabaci, while the prefix Rhc _II will be used to distinguish the Rhc proteins of the T3SS-2 gene cluster found in plasmid pNGR234b of Rhizobium sp. NGR234 (see below). The T3SS protein nomenclature when used is indicated by the prefix Sct according to Table 1.

All major T3SS core proteins were found in the T3SS gene clusters mentioned above, including the T3SS ATPase protein SctN (RhcN/HrcN/YscN/FliI homolog), its negative regulator SctL (NolV/HrpE/YscL/FliH homolog), the two T3SS gate proteins SctU and SctV (RhcU/HrcU/YscU/FlhB and RhcV/HrcV/LcrD/FlhA homologs respectively), the protein building the inner ring of the T3SS basal body SctJ (RhcJ/HrcJ/YscJ homolog), the protein building the cytoplasmic ring SctQ (RhcQ/HrcQ/YscQ/FliY homolog) and the three core membrane proteins SctR, SctS, SctT (RhcRST/HrcRST/YscRST/FliPQR homologs) (Additional file 4: Table S1).

It is noteworthy that the promoter regions of the T3SS-2 ORFs/operons of P. syringae pv phaseolicola 1448a, do not appear to harbor "hrp box" elements like those which have been described for the T3SS-1 genes of various P. syringae strains 27 . This, coupled with the low expression level seen in minimal media (Figure 3), leave open the question whether T3SS-2 in this or other P. syringae strains is expressed under in planta conditions and whether it is plays a role in their phytopathogenic potential or in any other aspect of their life cycle.

A distinguishing feature of gene organization in Rhc T3SS clusters is a split gene coding for the outer membrane secretin protein SctC, i.e. a HrcC/YscC homologue 28 . This is also true for the subgroup II Rhc T3SS gene clusters. In the T3SS-2 clusters of the three P. syringae pathovars the secretin gene is split in two ORFs (Figure 4, Additional file 4: Table S1). In P. syringae pv phaseolicola 1448a, loci PSPPH_2524 (hrc _II C1) and PSPPH_2521 (hrc _II C2) code for the N-terminal and the C-terminal part of secretin, respectively, of a HrcC/YscC homolog. Comparisons of Hrc _II C1 and Hrc _II C2 with the RhcC1 and Rhc2 proteins of Rhizobium sp. NGR234 are given in Additional file 5: Figure S4, respectively. A similar situation occurs in P. syringae pv oryzae str. 1_6 while in P. syringae pv tabaci ATCC11528 hrc _II C2 gene is further split into two parts. However in P. syringae pv phaseolicola 1448a and P. syringae pv tabaci ATCC11528 the two hrc _II C1, hrc _II C2 genes are only separated by an opposite facing ORF coding for a TPR-protein, while in the subgroup I Rhc T3SS these two genes are separated even further (Figure 4). Although the functional significance of the split secretin gene is not known, there are reports of constitutive expression of the rhcC1 gene in contrast to the rest of the T3SS operons in rhizobia 29 30 . In subgroup III only the rhcC1 could be identified (RHECIAT_PB0000097 in the R. etli CIAT 652 and RHE_PD00065 in R. etli CNF 42) in Psi-BLAST searches using the Hrc _ΙΙ C1 protein sequence as query (25% identity to RhcC1 of Rhizobium sp. NGR234) (Figure 4).

The PSPPH_2534 locus (designated hrc _II Q) in the T3SS-2 cluster of P. syringae pv phaseolicola 1448A codes for a polypeptide chain of 301 residues, which has sequence similarities with members of the HrcQ/YscQ/FliY family. Members of this family usually consist of two autonomous regions 26 which either are organized as two domains of a single protein or can be split up into two polypeptide chains. The Hrc _II Q is comparable in length with the long proteins of the family. The same is true in the Rhc-T3SS case, where an HrcQ ortholog is found. In agreement with the other HrcQ/YscQ/FliY members the sequence conservation is especially high at the C-terminus 31 32 . In the originally described T3SS-1 (Hrc-Hrp1) of P. syringae strains this gene is split into two adjacent ORFs coding for separate polypeptides (HrcQ_A and HrcQ_B). No splitting occurs however in the T3SS-2 clusters of the P. syringae strains.

A conserved feature in gene organization of T3SS gene clusters and the flagellum is the presence of a small ORF downstream of the gene coding for the ATPase (hrcN/yscN/fliI homologue). These ORFs code for proteins of the HrpO/YscO/FliJ family, a diverse group characterized by low sequence similarity, and heptad repeat motifs suggesting a high tendency for coiled-coil formation and a propensity for structural disorder 33 . Such a gene is also present in the Rhizobium NGR234 T3SS-2 but is absent from the subgroup III Rhc-T3SS where the rhcQ gene is immediately downstream of the rhcN gene (Figure 4). In the P. syringae pathovars included in Figure 4 there is a small ORF (PSPPH_2532 in strain P. syringae pv phaseolicola 1448A, Figure 4) coding for a polypeptide wrongly annotated as Myosin heavy chain B (MHC B) in the NCBI protein database. Sequence analysis of this protein and its homologs in the other two P. syringae strains using BLASTP searches did not reveal any significant similarities to other proteins. However, these small proteins are predicted as unfolded in their entire length, while heptad repeat patterns are recognizable in the largest part of their sequence, thus strongly resembling the properties of members of the HrpO/YscO/FliJ family 33 , (Additional file 6: Figure S5). A potentially important feature in the P. syringae pv phaseolicola 1448a T3SS-2 cluster is a predicted transposase gene between the ORF coding for the above described HrpO/YscO/FliJ family member and the ORF for the Hrc_IIN ATPase (Figure 4); this gene is absent from the P. syringae pv tabaci and P. syringae pv oryzae str.1_6 T3SS-2 clusters. The insertion of the transposase gene does not disrupt genes hrc _II N or hrp _II O as concluded by amino acid sequence comparison with other members of the SctN and SctO protein families respectively (including ORFs from other T3SS-2 P. syringae strains). These genes are capable of producing the respective full-length proteins and no premature termination, due to transposase insertion, is observed.

Another common feature of P. syringae T3SS-2 and the Rhizobium T3SSs excluding subgroup III, is a gene usually positioned upstream of the sctV gene (rhcV/hrcV/lcrD/flhA homolog) and in close proximity to it. Psi-BLAST searches for the PSPPH_2517 encoded protein revealed moderate similarities to the HrpQ/YscD family of T3SS proteins; these were confirmed by sequence threading techniques. For example, a segment of of PSPPH_2517 corresponding to 45% of its amino acid sequence scores an E-value of 2e-05 and a 26% identity with YscD protein from Yersinia enterocolitica (ref|YP_006007912.1); the same segment scores an E-value of 1e-13 with 25% identity to the 90% of its sequence with the equivalent protein from B. japonicum USDA110 (ref| NP_768443.1). The chosen folding templates belong to various forkhead - associated (FHA) protein domains from different origins. FHA cytoplasmic domains characterize the YscD/EscD protein family and may suggest phosphopeptide recognition interactions 34 . A protein with the above characteristics is present in the B. japonicum USDA110 T3SS cluster (encoded by the y4yQ gene) while an ortholog could not be identified in the R. etli T3SS.

A common feature of the R. etli Rhc T3SS (subgroup III) and the T3SS-2 of P. syringae pathovars (but not of the Rhizobium sp. NGR234 T3SS-2) is the presence of an ORF coding for a hypothetical translocator protein: The PSPPH_2540 locus of the P. syringae pv phaseolicola 1448a T3SS-2 codes for a large protein of 1106 residues. The C-terminal part of this protein (residues 421 – 1106) is homologous to the HrpK proteins of the Hrc-Hrp1 T3SS family based on Psi-BLAST searches (25% identity with HrpK of Erwinia amylovora). HrpK shares low similarity with the putative translocator, HrpF, from Xantomonas campestris pv vesicatoria. Furthermore, the C-terminal part of the protein coded by PSPPH_2540 also possesses two predicted transmembrane α-helices comprising residues 879–898 and 1029–1047 (MEMSAT3 analysis). The subgroup I Rhc T3SS lacks a hrpK ortholog. The HrpK protein was initially identified as a component of the Hrc-Hrp1 family of T3S systems 39 . Interestingly, the R. etli T3SS gene cluster possesses two copies of hrpK-like genes, plus an additional hrpW-like gene, coding for an Hrp-secreted protein homologous to class III pectate lyases which is absent from the P. syringae pv phaseolicola 1448a T3SS-2 gene cluster but present in the extremity of the Hrc-Hrp1 gene cluster of P. syringae pv phaseolicola 1448a. These differences possibly suggest variations in the mode of interaction of these bacteria with their hosts.

The choice of the B. japonicum USDA 110 T3SS as archetypal for subgroup I in the Rhc family (Figure 4) and for synteny comparisons with the subgroup II gene clusters, was based on the DNA segment encompassing rhcV (y4yQ-y4yS). The presence of two small open reading frames upstream of the rhcV gene and downstream of the y4yQ gene of the known Rhizobium T3SS resembled the case of the P. syringae pv phaseolicola 1448a T3SS-2 where loci PSPPH_2518 and PSPPH_2519 are found between the ORF coding for the SctV protein (RhcV/HrcV/LcrD/FlhA homolog) and the ORF coding for the SctD protein (HrpQ/YscD homolog).

The PSPPH_2519 locus, upstream of the hrc _II V gene of P. syringae pv phaseolicola 1448a genome, encodes for a 112 long polypeptide with sequence similarities to the VscY protein of Vibrio parahaemolyticus, according to Psi-BLAST searches (E-value = 0.005). The vscY gene is located upstream of the vcrD gene and this synteny is also conserved in the Ysc T3SS gene cluster family. Proteins YscY, VscY and PSPPH_2519 all possess TPR repeats (Tetratricopeptide Repeats) as predicted by Psi-BLAST searches and fold recognition methods. YscY has been found to directly bind the YscX protein, a secreted component of the Ysc T3SS 40 . The bll1801 locus of B. japonicum USDA110 encodes for a 142 long polypeptide with TPR repeats and sequence similarities to the AscY (Aeromonas salmonicida) and YscY proteins according to Psi-BLAST searches. The position of bll1801 is likewise upstream of the rhcV gene in B. japonicum USDA110 T3SS gene cluster. A protein with the above characteristics could not be identified for the R. etli T3SS (subgroup III), however it is present in the T3SS-2 of Rhizobium NGR234.

The Hrc-Hrp2 and the Rhc T3S (subgroup I) systems possess transcription regulators that belong to the AraC/XylS in contrast to the Hrc-Hrp1 T3SS that depends on the alternative sigma factor HrpL. The known transcription factors are related to the T3SS regulation of AraC and LuxR/UhaP families of transcription regulators and characterized by two α-helix-turn-α-helix (HTH) motifs in a tetrahelical bundle.

However, the PSPPH_2539 locus of P. syringae T3SS-2 codes for a hypothetical transcription regulator with different characteristics. The N-terminal part of the hypothetical protein (Figure 5, blue-purple area) is predicted to adopt a structure similar to the DNA-binding domains of the PhoB transcription factor. The characteristic HTH motif is a common feature of transcription factors. Although the PSPPH_2539 ORF is annotated in the NCBI as a LuxR-type of transcription regulator, the choice of the DNA-binding domain of PhoB as a structural template indicates that PSPPH_2539 probably has an α-/β- doubly wound fold (distinguished by the presence of a C-terminal β-strand hairpin unit that packs against the shallow cleft of the partially open tri-helical HTH core) motif. Transcription factors are usually multidomain proteins, thus the assignment of PSPPH_2539 as a LuxR-type transcription regulator in the NCBI is probably due to full-length inadequate Psi-BLAST searches biased by the presence of Tetratricopeptide Repeats (TPR) in the large carboxyterminal domain.

The middle part of the protein (Figure 5, yellow area) was found homologous to the AAA⁺ ATPases (COG3903) based on fold-recognition algorithms and Psi-BLAST searches. These ATPases are associated with diverse cellular activities and are able to induce conformational changes in their targets 41 . In the context of the transcription process, AAA⁺ ATPase domains are involved in the remodeling of σ⁵⁴ RNA polymerases. Especially the residues 195 to 300 probably possess the receiver or ligand binding domain of the hypothetical transcription factor (green area, Figure 5).

Apart from the PSPPH_2539 C-terminal domain, there are two more ORFs, PSPPH_2519 and PSPPH_2523, from the P. syringae pv phaseolicola 1448a T3SS-2 that are predicted to code for proteins that possess TPR domains. TPR domains are typically found in class II chaperones of T3S systems - chaperones of the translocators - as well as in transcriptional regulators of the T3S systems, e.g. the HrpB protein of Ralstonia solanacearum, HilA of Salmonella enterica 42 and SicA, of Salmonella typhimurium involved in the activations of T3SS virulence genes 43 . Proteins with TPR repeats also exist in the Hrc-Hrp2 T3S system of X. campestris (HrpB2 protein) and in the T3S system of Rhizobia (e.g. the 182 residue long Y4yS protein). On the other hand, the Hrc-Hrp1 system of P. syringae does not possess proteins with TPR repeats.