Earlier studies using heterologous systems have led to the identification of several bona-fide or putative C. trachomatis T3S effectors [13–15, 21, 22, 24–27]. While these and other analyses covered a significant portion of all C. trachomatis proteins, we hypothesized that there could be previously unidentified T3S substrates. By combining basic bioinformatics searches, exhaustive T3S assays, translocation assays, and analyses of chlamydial gene expression in infected cells, we revealed 10 C. trachomatis proteins (CT053, CT105, CT142, CT143, CT144, CT161, CT338, CT429, CT656, and CT849) as likely T3S substrates and possible effectors. In particular, CT053, CT105, CT142, CT143, CT338, and CT429 were type III secreted by Y. enterocolitica, could be translocated into host cells, and their encoding genes were clearly expressed in C. trachomatis strain L2/434. Therefore, these 6 proteins have a high likelihood of being effectors. However, additional future studies are required to show that all of these 10 proteins are indeed translocated by C. trachomatis into host cells and to show that they are bona-fide effectors, i.e., that they interfere with host cell processes.
Among the likely T3S effectors of C. trachomatis that we identified, CT105 and CT142 have been previously singled out as possible modulators of host cell functions, based on the phenotypic consequences of their ectopic expression in yeast S. cerevisiae. In addition, the genes encoding CT142, CT143, and CT144 have been shown to be markedly transcriptionally regulated by a protein (Pgp4) encoded by the Chlamydia virulence plasmid . This plasmid is present in almost all C. trachomatis clinical isolates , and studies in animal models of infection showed that it is a virulence factor in vivo[67, 68]. Additional studies are needed to understand if the putative effector function of CT142, CT143, and CT144 can partially explain the virulence role of the chlamydial plasmid. Furthermore, the predicted amino acid sequence of CT849 reveals a domain of unknown function (DUF720) that can only be found in Chlamydia proteins. In C. trachomatis, besides CT849, a DUF720 domain is found in CT847, a T3S effector that interacts with human Grap2 cyclin D-interacting protein (GCIP) , and in CT848, which has been indicated as a T3S substrate using S. flexneri as a heterologous system . Therefore, this further supports a possible role of CT849 as an effector. In contrast with CT105, CT142, CT143, CT144 or CT849, no significant information is available or could be retrieved about CT053, CT338, CT429, or CT656.
CT161 is a possible T3S substrate and effector, but we could not detect significant levels of ct161 mRNA during the developmental cycle of strain L2/434. The ct161 gene is localized within the “plasticity zone”, a chromosomal region of rare high genetic diversity among C. trachomatis strains. In fact, although C. trachomatis includes strains showing remarkably different tropisms (strains that can spread into lymph nodes and cause lymphogranuloma venereum [LGV], such as L2/434, and strains causing infections usually restricted to the mucosa of the conjunctiva and genitals), their genomes are all highly similar . Preliminary data indicate that, contrarily to what is seen in LGV strains, the ct161 seems to be more expressed in some ocular and urogenital isolates (data not shown). We are currently investigating the possibility that ct161 is a pseudogene in LGV strains, perhaps inactivated by a mutation in its promoter region. Interestingly, CT161 has been shown by yeast two-hybrid to bind CT274 (a possible chlamydial T3S chaperone) . Another feature of this protein is that part of its amino acid sequence (residues 40–224, out of 246) shows 28% of identity to a region of Lda2/CT163 (residues 167–361, out of 548), a known C. trachomatis translocated protein .
Among the proteins for which we found a secretion signal but could not demonstrate their T3S as full-length proteins, we highlight CT153 and GrgA/CT504. Regarding CT153, this protein possesses a membrane attack complex/perforin (MACPF) domain , and there is previous evidence that it may be translocated by C. trachomatis, which is consistent with our data. The ct504 gene has been recently shown to encode a transcriptional activator, GrgA . Therefore, T3S of CT50420-TEM-1 could be false a positive. However, if GrgA is a T3S substrate, as our data suggests, it could have a function within the host cell or, more likely and similarly to what has been described in the T3SSs of Yersinia or Shigella[74, 75], it could be discarded by secretion once its intra-bacterial regulatory activity needs to be shut down.
We found T3S signals in 56% proteins analyzed (26 out of 46, including controls). This high percentage of proteins showing a T3S signal suggests that some should be false positives. It is conceivable that within a single bacterium non-secreted proteins possess T3S signals but are not targeted to the T3SS machinery because they also carry signals (e.g. DNA-, membrane-, or protein-binding) that preferentially direct them to other location within the bacterial cell. To help differentiating between true or false positives among chlamydial proteins carrying a T3S signal we analyzed their secretion as full-length proteins. This is because, as explained above in the Results section, not all proteins have folding characteristics compatible with T3S [59–62]. However, we cannot exclude that some of the C. trachomatis full-length proteins that were not type III secreted by Yersinia have a T3S chaperone that maintains them in a secretion-competent state  and enables their secretion during infection by C. trachomatis. Intriguingly, CT082 or CT694 have dedicated T3S chaperones, CT584 and Slc1, respectively , and, in agreement with what we previously observed , they were both secreted as full-length proteins in the absence of the chaperones. Considering that T3S chaperones have various functions [76, 77], the chaperone role of CT584 or Slc1 should be different from maintaining their substrates in a secretion-competent state.
Eleven of the Chlamydia proteins that we analyzed have been previously studied for T3S using S. flexneri has a heterologous system . In the majority of the cases the outcome of the experiments was identical; however, differently from what was shown in Shigella, we detected a T3S signal in the N-terminal of CT429 (which was also secreted as a full-length protein, and could be translocated into HeLa cells), GrgA/CT504, and CT779 and we did not detect a T3S signal in CT577. Evidence for a T3S signal in only one of the heterologous systems may suggest a false positive. However, there is a myriad of possible explanations for these discrepancies, when considering that different heterologous systems (Shigella and Yersinia) and reporter proteins (Cya and TEM-1) were used, and that the N-terminal regions in the hybrid proteins consisted in different lengths of amino acids and were in some cases from different Chlamydia species.
We compared the data from our T3S assays (including the controls, CT082, CT694, and RplJ) with predictions of T3S substrates by in silico methods (Effective T3S , SIEVE , Modlab , and T3_MM ) using resources available in the Web (Effective T3S, Modlab and T3_MM) and Table three in reference  (SIEVE), as detailed in Additional file 3: Table S3. When considering the analysis of T3S signals in TEM-1 hybrids, the vast majority of proteins (60%; 12 out of 20) in which we did not find a T3S signal were also predicted not to be secreted by each of the in silico methods. In contrast, the vast majority of proteins (58%; 15 out of 26) in which we detected a T3S signal were also predicted to be secreted by at least one of the in silico methods. The correlation between our experimental data and the in silico predictions was more striking when considering the T3S of full-length proteins. Among the 16 full-length proteins for which we could not find definitive evidence of T3S, 10 (i.e., 62.5%) were also predicted not to be secreted by each of the in silico methods, but among the 11 proteins that we showed or confirmed to be T3S substrates, 10 (i.e., 83%) were also predicted to be secreted by at least one of the in silico methods. Overall, this indicates some correlation between our experimental data and the in silico methods that predict T3S substrates. However, for many proteins, each of these in silico methods generates different predictions (see Additional file 3: Table S3). It is possible that the quantitative data on T3S such as the one we generated in this and in a previous study , can be used to normalize and improve the predictive value of such methods.