Novel insights into the mechanism of SepL‐mediated control of effector secretion in enteropathogenic Escherichia coli

Abstract Type three secretion systems (T3SSs) are virulence determinants employed by several pathogenic bacteria as molecular syringes to inject effector proteins into host cells. Diarrhea‐producing enteropathogenic Escherichia coli (EPEC) uses a T3SS to colonize the intestinal tract. T3S is a highly coordinated process that ensures hierarchical delivery of three classes of substrates: early (inner rod and needle subunits), middle (translocators), and late (effectors). Translocation of effectors is triggered upon host‐cell contact in response to different environmental cues, such as calcium levels. The T3S substrate specificity switch from middle to late substrates in EPEC is regulated by the SepL and SepD proteins, which interact with each other and form a trimeric complex with the chaperone CesL. In this study, we investigated the link between calcium concentration and secretion regulation by the gatekeeper SepL. We found that calcium depletion promotes late substrate secretion in a translocon‐independent manner. Furthermore, the stability, formation, and subcellular localization of the SepL/SepD/CesL regulatory complex were not affected by the absence of calcium. In addition, we demonstrate that SepL interacts in a calcium‐independent manner with the major export gate component EscV, which in turn interacts with both middle and late secretion substrates, providing a docking site for T3S. These results suggest that EscV serves as a binding platform for both the SepL regulatory protein and secreted substrates during the ordered assembly of the T3SS.

. The effectors required for the formation of the A/E lesion are encoded in a pathogenicity island known as locus of enterocyte effacement (LEE) McDaniel & Kaper, 1997), which also contains all the genes necessary to assemble a functional T3SS Pallen, Beatson, & Bailey, 2005). In addition to the seven effectors encoded in the LEE, there are many others encoded by genes scattered through the genome (Nles: Non-LEE encoded effectors) that are also translocated by the T3SS (Dean & Kenny, 2009;Deng et al., 2012;Iguchi et al., 2009).
The injectisome of EPEC consists of an outer (EscC) and a pair of inner (EscJ and EscD) membrane rings that are interconnected through a periplasmic inner rod (EscI), forming a core structure, the so-called basal body, that spans both bacterial membranes (Ogino et al., 2006; Sal-Man, Spreter et al., 2009;Yip et al., 2005).
These complexes trigger effector secretion by different mechanisms.
In the case of Yersinia, substrate secretion switching occurs upon YopN secretion, whereas in Salmonella it requires the switch protein complex dissociation and the subsequent degradation of its components.
In both cases, the switching event occurs in response to environmental cues detected upon host membrane contact, for example, calcium depletion and a pH change, respectively (Cheng et al., 2001;Day & Plano, 1998;Yu et al., 2010). In EPEC, calcium depletion from the bacterial growth media has an effect on substrate secretion similar to that observed in a sepL or sepD null mutants Ide, Michgehl, Knappstein, Heusipp, & Schmidt, 2003). Yet, the molecular mechanism by which the switch protein complex participates in the regulation of substrate secretion in the absence of calcium is still poorly understood. In this work, we investigated the effect of calcium depletion in the gatekeeper-dependent triggering of effector secretion and uncovered the existence of novel protein interactions.

| Bacterial strains and growth conditions
All bacterial strains and plasmids used in this study are listed in

| Plasmid construction and oligonucleotides
The sequences of the oligonucleotides used in this study are listed in

| Epitope tagging of chromosomal genes
A modification of the λ-Red recombinase system was used for chromosomal epitope tagging as described in (Uzzau, Figueroa-Bossi, Rubino, & Bossi, 2001). The EPEC strains expressing a 3xFLAG-tagged version of sepL (EPEC sepL-3FLAG::km) or a 2xHA version of cesL (EPEC cesL-2HA::km), were constructed by amplifying the kanamycin cassette from either plasmid pSUB11, using the primer pair sepL-3FLAG_Fw and sepL-3FLAG_Rv, or from plasmid pSU315, using the primer pair cesL-2HA_Fw and cesL-2HA_Rv. The resulting PCR products were electroporated into EPEC WT strain carrying plasmid pKD46, which was grown at 30°C in LB medium containing 100 mmol/L of L-arabinose to induce Red recombinase expression. Transformant cells were grown at 37°C to eliminate the pKD46 plasmid. Recombinant EPEC sepL-3FLAG::km and EPEC cesL-2HA::km strains were selected on LB plates supplemented with 300 μg/mL Km and tested for Ap sensitivity. To generate the double-tagged EPEC sepL-3FLAG cesL-2HA::km strain, the kanamycin cassette of the EPEC sepL-3FLAG::km strain was excised using the helper plasmid pFLP2 (Hoang, Karkhoff-Schweizer, Kutchma, & Schweizer, 1998), and the homologous recombination of cesL-2HA::km was performed as described above.

| Immunoblotting
Samples subjected to SDS-PAGE were transferred onto nitrocellu-

| Protein overproduction and pull-down assays
Salmonella SJW1368 strain, which lacks the flagellar master operon and has been extensively used for expression of pTrc99A-based plas-  To purify the SepL/SepD/CesL protein complex, Salmonella SJW1368 cells were cotransformed with the bicistronic plasmid pMTBISpDcL and the plasmid pMATpL or pMATpL ∆C11 . Protein production was induced by the addition of 1 mmol/L IPTG. The cleared lysate was prepared as described above, incubated with 100 μL of Ni-NTA for 2 hr at 4°C and then loaded onto a polypropylene column. After extensive washing, proteins were eluted with 500 mmol/L imidazole. To assess the effect of calcium on the EscVc/SepL and SepL/SepD/CesL interactions, 2 mmol/L CaCl 2 was added to all buffers used in the pulldown assays.
To evaluate the interaction of EscVc and EscD N with T3-secreted proteins, a modified version of the method employed by Thomas et al. (2005) to identify cognate substrates of the CesT chaperone was used.
Briefly, the cleared lysate of His-EscVc or His-EscD N was incubated

| Subcellular fractionation
Subcellular fractionation was carried out as previously reported (Gauthier, Puente, & Finlay, 2003) with slight modifications. The EPEC sepL-3FLAG cesL-2HA::km strain was grown under T3S inducing conditions in regular DMEM or DMEM without calcium until an OD 600 of 1. Cells were harvested by centrifugation (15,500g for 10 min), the cell pellet was washed once with phosphate-buffered saline and then resuspended in 20 mmol/L Tris-HCl pH 7.5 containing 1 mmol/L PMSF.
Cells were disrupted by sonication and unbroken cells were removed by centrifugation at 19,800g for 10 min. The cleared lysate was subjected to ultracentrifugation at 90,000g for 1 hr. The supernatant (containing the cytoplasmic proteins) was collected into a clean tube, and the membrane-containing pellet was washed with 20 mmol/L Tris-HCl pH 7.5. Both fractions were ultracentrifuged once again as previously described, and the cytoplasmic and membrane fractions were collected. Total protein concentration of the samples was quantified using the DC protein assay (Bio-Rad), and proteins were analyzed by immunoblotting as described above.

| Protein stability assay
The EPEC sepL-3FLAG cesL-2HA::km strain was grown in calciumfree DMEM added with 1.8 mmol/L CaCl 2 under T3S inducing conditions. When an OD 600 of 1 was reached, cells were harvested by centrifugation at 4,600g for 10 min, washed with and resuspended in either calcium-free DMEM or calcium-free DMEM added with 1.8 mmol/L CaCl 2 , containing 50 μg/ml Cm in order to stop protein synthesis. Bacterial growth was continued and samples were taken every 30 min during 3 hr. Samples were normalized according to the OD 600 and proteins were analyzed by immunoblotting as described above. we did not observe a significant decrease in translocator secretion ( Figure 1a). Therefore, to provide a more quantitative measure of the F I G U R E 1 Calcium depletion from the bacterial growth media enhances effector secretion. (a) T3 secretion profiles of the EPEC wildtype strain (WT) grown in regular DMEM containing 1.8 mmol/L of CaCl 2 (Regular), calcium-free DMEM (Ca 2+ -free) or calcium-free DMEM supplemented with 1.8 mmol/L of CaCl 2 (Ca 2+ -free + Ca 2+ ), and the ∆sepL and ∆escU mutant strains, visualized by SDS-PAGE stained with Coomassie brilliant blue (CBB) (upper panel). The presence of DnaK, Tir, Map, and EspA in the supernatants (SN) and whole-cell lysates (WC) was examined by immunoblotting, using anti-DnaK, anti-Tir, anti-Map, and anti-EspA antibodies (lower panels). (b) Protein secretion profiles of EPEC WT strain overproducing HA-tagged T3 substrates, grown in the presence or absence of calcium as described in (a). Immunodetection of LEE and non-LEE-encoded effectors (NleH2, EspH, and NleC) and the inner rod protein EscI, was performed in the supernatants (SN) and whole-cell lysates (WC) using specific antibodies against the HA tag. The results shown are representative of three independent experiments. (c) Relative abundance of EspA and Tir proteins in the supernatant of EPEC wild-type strain grown in the presence or absence of calcium as described in (a). CBB-stained protein bands were quantified from six independent secretion assays by gel densitometry using the ImageJ software (Schneider, Rasband, & Eliceiri, 2012). The secretion level of EspA and Tir proteins was normalized relative to the secretion level of the EspC autotransporter band. The average and the standard deviation of normalized data are displayed. Significant statistical differences compared with the regular DMEM condition are denoted by asterisks. A p value < .05 was considered statistically significant. **p = .002, Wilcoxon-test overall change in the secretion of effectors and translocators, we performed a statistical analysis on the secretion data of several independent experiments by quantifying the Coomassie-stained protein band intensities of the translocator EspA and the effector Tir. The values were normalized relative to the corresponding band intensity of the autotransporter EspC, which is secreted through the type V secretion system and thus serves as a loading control (Mellies et al., 2001;Stein, Kenny, Stein, & Finlay, 1996). This analysis showed no significant differences in EspA translocator secretion in the different media used in the assay, whereas there was a clear 11-fold increase in Tir effector secretion in the absence of calcium (Figure 1c). Moreover, we investigated the effect of calcium on the secretion of an early substrate, namely, the inner rod component EscI. As shown in Figure 1b, secretion of EscI-2HA was not altered by different calcium levels in the growth media. It is worth mentioning that although overproduction of EscI-2HA caused a decrease in T3 secretion, EPEC was still responsive to calcium depletion as demonstrated by an increased secretion of Tir ( Figure S1A). Besides, overproduction of an untagged version of EscI, which did not interfere with T3 secretion, further demonstrated that calcium levels had no effect on the secretion of this early substrate ( Figure S1B). Overall, our results suggest that only late substrate secretion is significantly affected by calcium depletion from the growth media. Thus, we next decided to investigate the mechanism that triggers effector secretion under this condition.

| Calcium sensing is translocon-independent
It has been previously shown that effector secretion is elicited upon host cell contact, and that different environmental cues are involved in the activation of this secretion stage, for example, a pH shift in Salmonella and low-calcium in Yersinia, Pseudomonas, and A/E pathogens (Beuzon, Banks, Deiwick, Hensel, & Holden, 1999;Deng et al., 2005;Kim et al., 2005;Lee, Mazmanian, & Schneewind, 2001;Mills, Baruch, Charpentier, Kobi, & Rosenshine, 2008;Rosqvist, Magnusson, & Wolf-Watz, 1994;Yu et al., 2010). These in vitro conditions are proposed to mimic those found during the bacterial infection process. Therefore, we reasoned that the translocon proteins in EPEC, which establish contact with host cells, could be involved in sensing calcium-level changes. This hypothesis was evaluated by comparing the secretion profile of the EPEC wild-type strain with that of the isogenic translocator null mutants espA, espB, and espD, when grown in the presence or absence of calcium. Translocator null mutants were still responsive to calcium depletion, as shown by an increased secretion of the effector Tir in calcium-free medium (Figure 2).
Tir production was not affected in the different culture media or in the mutant strains (Figure 2 lower panel). Thus, although translocators are the proteins that make direct contact with host cells, they appear not to be required for in vitro calcium sensing.

| The SepL/SepD/CesL complex is insensitive to calcium concentration changes
It has been reported that the SsaL/SpiC/SsaM complex of the S. enterica SPI-2 T3SS prevents premature effector secretion at pH 5.0, a condition found within host cell vacuoles. However, upon T3SS vacuolar membrane contact, the detection of a pH increase to 7.2 in the cytoplasm leads to dissociation of the protein complex and degradation of its components, resulting in effector secretion (Yu et al., 2010).
Since the proteins forming the SsaL/SpiC/SsaM complex in Salmonella are homologous to the ones forming the SepL/SepD/CesL complex in EPEC, and both protein complexes participate in the substrate specificity switch from translocators to effectors without secretion of the gatekeeper protein (Coombes et al., 2004;Younis et al., 2010;Yu et al., 2010), we argued that the mechanism by which these complexes respond to environmental cues that induce effector secretion could be similar. Therefore, we examined the effect of calcium depletion on the stability, formation, and subcellular localization of the SepL/SepD/ CesL complex. To this end, we constructed a sepL-3FLAG cesL-2HA chromosomally tagged strain and showed that its T3 secretion profile is comparable to that of the wild-type strain ( Figure S2). However, we were unable to obtain a functional SepD chromosomally tagged strain, because the C-terminal tag affected the protein secretion profile (data not shown). Then, we evaluated the effect of calcium depletion on the protein stability of SepL-3FLAG and CesL-2HA. Stability assays were performed as described under Materials and Methods, inhibiting de novo protein synthesis with chloramphenicol in cells grown in the presence or absence of calcium. As shown in Figure 3a, the stability Moreover, it has been suggested that environmental signals can diffuse through the needle channel to a sensor protein at the base of the secretion apparatus (Notti & Stebbins, 2016;Portaliou et al., 2016;Shaulov et al., 2017;Yu et al., 2010). Therefore, we decided to investigate whether calcium directly affected the formation of the SepL/ SepD/CesL trimeric complex. To this end, in vitro pull-down assays were performed in the presence or absence of calcium as described in Materials and Methods. The results showed no differences in the copurification assays with or without calcium ( Figure S3). This is consistent with a previous report showing that calcium did not affect the SepL-SepD interaction . Overall, these findings indicate that the mechanism by which effector secretion increases upon calcium removal from the growth media does not involve changes in the stability, formation, or localization of the SepL/SepD/CesL secretion regulatory complex.

| SepL interacts with EscV, a T3SS component engaged in substrate recognition
The gatekeeper protein SepL and its homologs in different bacteria have been shown to interact with diverse T3SS components (Archuleta & Spiller, 2014;Botteaux et al., 2009;Cherradi et al., 2013;Kubori & Galan, 2002;Lee et al., 2014;Roehrich et al., 2017;Shen & Blocker, 2016;Wang et al., 2008). On the basis of our previous results, we reasoned that T3SS proteins involved in substrate recognition or targeting might be prone to calcium-mediated control.

Homologous proteins of the EPEC export apparatus component
EscV have been implicated in the recruitment of T3 secretion substrates and chaperones, in both the flagellar and virulence T3SSs. In Xanthomonas spp., the cytoplasmic domain of HrcV was shown to bind early and late substrates, as well as chaperones (Alegria et al., 2004;Buttner, Lorenz, Weber, & Bonas, 2006;Hartmann & Buttner, 2013), whereas the cytoplasmic domain of the flagellar export gate protein FlhA, binds different chaperone/substrate complexes (Bange et al., 2010;Kinoshita, Hara, Imada, Namba, & Minamino, 2013). To investigate whether EscV might play a role in substrate recognition, we tested the ability of the C-terminal cytoplasmic domain of EscV (EscVc) to interact with secreted proteins. To this aim, His-EscVc was coupled to Ni-NTA agarose beads and incubated with the combined supernatants of the secretion assays of a ∆sepL mutant, which hypersecretes effectors, and a ∆grlR mutant, which produces and secretes more effectors and translocators (Deng et al., 2004Wang et al., 2008). EscK (data not shown), as previously observed in a yeast two hybrid F I G U R E 4 EscVc directly interacts with secreted proteins. Pull-down assay of His-EscVc (left panels) and His-EscD N (middle panels) with secreted proteins, performed by nickel affinity chromatography. The cleared lysate containing His-EscVc or His-EscD N (Lys), was incubated with Ni-NTA beads and loaded into a column. The proteins secreted into the supernatant by the ∆sepL and ∆grlR null mutant strains (SN) were passed through the His-EscVc or His-EscD N -coupled resin in the column and the flow through (FT) was collected. After extensive washing (W), proteins were eluted (E1 and E2). Samples were visualized by SDS-PAGE stained with Coomassie brilliant blue (CBB) (upper panels). Detection of copurified proteins was performed by immunoblotting with specific antibodies against Tir, EspD, and EspA (lower panels). As an additional control, the nonspecific binding of Tir, EspD, and EspA to the Ni-NTA beads was analyzed by SDS-PAGE stained with CBB and immunoblotting (right panels) screen (Soto et al., 2017), corroborating the specificity of the EscV-SepL interaction.
To further dissect the SepL interacting region with EscVc, we generated SepL truncated protein versions. In P. aeruginosa, the EscV homolog PcrD interacts with the Pcr1 protein, which is homologous to the C-terminal domain of SepL (Lee et al., 2014). Therefore, we generated two C-terminal truncated versions of SepL (SepL ∆C75 and SepL ∆C11 ) and assessed their ability to interact with His-EscVc by in vitro pull-down assays. We found that in contrast to full-length SepL, neither SepL ∆C75 nor SepL ∆C11 interacted with His-EscVc (Figure 5b). In addition, we demonstrated that the structural integrity of the SepL ∆C11 truncated protein is not affected, since it is able to form the SepL ∆C11 / SepD/CesL ternary complex ( Figure S4). These results indicate that the last 11 amino acids of SepL are crucial for its interaction with EscVc.
Moreover, we further evaluated the ability of the SepL truncated versions to complement the phenotype of the ∆sepL null strain. Neither SepL ∆C75 nor SepL ∆C11 could restore the secretion of translocators, and only partially reduced the hypersecretion of effectors of the sepL mutant ( Figure 5c). Therefore, our results highlight the importance of the extreme C-terminal region of SepL for substrate secretion regulation.
Secretion of translocators prior to effectors ensures that the latter are translocated directly into the host cell cytoplasm. In EPEC, this secretion hierarchy is proposed to be controlled by the gatekeeper SepL (Deng et al., 2004Wang et al., 2008), and to respond to environmental cues such as calcium concentration Shaulov et al., 2017). However, the exact mechanism underlying this specificity-switching process is still poorly understood. In this work, F I G U R E 5 SepL interacts with the export gate component EscV. (a) Pull-down assays of His-EscVc and SepL (His-EscVc + SepL) performed by nickel affinity chromatography in the presence or absence of 2 mmol/L CaCl 2 . Cleared lysates containing His-EscVc or untagged SepL were mixed and incubated for 2 hr (In) in the absence or presence of calcium. Samples were loaded into columns packed with Ni-NTA resin, washed (W) and eluted (E1, E2, and E3). His-EscVc was visualized by SDS-PAGE stained with Coomassie brilliant blue (CBB). The molecular mass markers (M) of 37 and 50 kDa are shown. Detection of copurified SepL was performed by immunoblotting with polyclonal antibodies against SepL (middle panel). As a negative control, SepL was loaded into a Ni-NTA column (Ni-NTA + SepL) and treated under the same conditions described above (lower panel). (b) Pull-down assays of His-EscVc and C-terminal truncated versions of SepL (SepL ∆C75 and SepL ∆C11 ) performed by nickel affinity chromatography. Cleared lysates containing His-EscVc and full-length SepL (SepL FL ), SepL ∆C75 , or SepL ∆C11 were mixed and incubated for 2 hr (Lys). Samples were loaded into columns packed with Ni-NTA resin, washed (W), and eluted (E). His-EscV was visualized by SDS-PAGE stained with CBB. Copurified proteins were detected by immunoblotting using polyclonal anti-SepL antibodies (middle panels). As negative controls, the nonspecific binding of SepL FL , SepL ∆C75 , and SepL ∆C11 to the Ni-NTA resin was analyzed by immunoblotting (lower panels).
(c) Protein secretion profiles of EPEC wild-type (WT), ∆escN and ∆sepL mutant strains, and the ∆sepL strain carrying the empty vector pTrc99A_FF4 (pTrc99A), or the pTrc99A_FF4-based plasmids expressing sepL or the C-terminal truncated versions sepL ∆C75 and sepL ∆C11 . Secreted proteins were visualized by SDS-PAGE stained with CBB (upper panel) and detected by immunoblotting in the supernatants (SN) with specific antibodies against Tir and EspA. The production of SepL and its truncated versions in whole-cell lysates (WC) was examined by immunoblotting with polyclonal antibodies against SepL we evaluated the link between calcium sensing and SepL hierarchical regulation.
Previous reports showed that calcium depletion from the growth medium in A/E pathogens has a differential effect on T3 substrate secretion, that is, reduced secretion of translocators and increased secretion of effectors Ide et al., 2003). In accordance, we showed that the absence of calcium in the EPEC growth medium considerably increased the secretion of several LEE and Nle effectors. However, although the secretion of translocators appears to be slightly reduced in the absence of calcium, our statistical analysis demonstrated that the secretion of middle substrates is not significantly affected under this condition (Figure 1). The dissimilar results on the effect of calcium on translocator secretion could be attributed to either differences in the experimental setups for the secretion assay or to the use of chelators to eliminate calcium Ide et al., 2003), given that EGTA and BAPTA bind in addition to calcium other multivalent cations, which in turn could also affect T3-dependent secretion (Gode-Potratz, Chodur, & McCarter, 2010;Kenny et al., 1997;Sarty et al., 2012). In agreement with our results, a previous report in P. aeruginosa showed that in vitro calcium levels influence effector but not translocator secretion (Cisz, Lee, & Rietsch, 2008). Hence, it is possible that secretion of translocators and effectors can be promoted by different environmental signals as recently suggested (Roehrich et al., 2017). Additionally, we extended our analysis to investigate the effect of calcium absence on the secretion of an early substrate, the inner rod protein EscI, and demonstrated that its secretion is not affected under this condition. This result supports our proposal that the calcium signal exclusively regulates the substrate specificity switch promoting late substrate secretion.
Moreover, it has been proposed that physical contact between the bacteria and host cells along with the assembly of the translocation pore, allows for detection of lowered calcium levels in the eukaryotic cell cytoplasm, triggering effector secretion Lee et al., 2001). Accordingly, the translocon components would be expected to be involved in signal sensing as has been previously reported (Armentrout & Rietsch, 2016;Cisz et al., 2008;Roehrich, Guillossou, Blocker, & Martinez-Argudo, 2013;Urbanowski, Brutinel, & Yahr, 2007). Here we demonstrated that the translocator proteins EspA, EspB, and EspD are not implicated in transducing the signal that triggers effector secretion upon calcium depletion in vitro ( Figure 2). This finding is in agreement with previous reports in S. enterica and P. aeruginosa, where mutants lacking the translocators are still able to trigger effector secretion upon changes in chemical cues (Cisz et al., 2008;Yu et al., 2010). The apparent discrepancy regarding the requirement of translocators could be explained by the nature of the triggering signal. While translocators have been shown to be essential for effector translocation upon host cell contact (Armentrout & Rietsch, 2016;Cisz et al., 2008;Urbanowski et al., 2007), they are not required for induction of effector secretion under calcium depletion in vitro (Cisz et al., 2008;Lee, Stopford, Svenson, & Rietsch, 2010). Thus, under in vitro calcium-depleted conditions in EPEC, either the needle plays an active role in detecting and conveying the environmental signal as has been demonstrated for S. flexneri and Y. pestis (Deane et al., 2006;Kenjale et al., 2005;Torruellas, Jackson, Pennock, & Plano, 2005); or as recently suggested, diffusion of chemical signals through the needle channel might alter the local concentration of different ions at the base of the secretion apparatus (Notti & Stebbins, 2016;Portaliou et al., 2016;Shaulov et al., 2017;Yu et al., 2010), resulting in modifications at the T3SS basal machinery that trigger effector secretion.
Nonetheless, further studies are required to elucidate the mechanistic basis of physiological signal sensing and transmission.
The deregulation of effector secretion upon calcium depletion somewhat mimics the hypersecretion phenotype seen in the absence of the SepL and SepD proteins . SepL and SepD form a ternary complex with CesL that is homologous to the S. enterica SsaL/SpiC/SsaM complex (Younis et al., 2010). Besides, in contrast to other gatekeeper proteins, SepL and SsaL are not secreted, share more than 40% sequence similarity and SepL partially complements an ssaL null mutant (Coombes et al., 2004;Younis et al., 2010). For these reasons, we hypothesized that the SepL/SepD/CesL complex regulates the switch from middle to late substrates in response to environmental calcium in the same manner as the SsaL/SpiC/SsaM complex does in response to pH (Yu et al., 2010). Nevertheless, our results demonstrated that the stability, localization, and association of the SepL/SepD/CesL complex are not affected by calcium changes (Figure 3 and S3), indicating that this gatekeeper complex is not prone to direct or indirect calcium regulation. These data prompted us to investigate whether the interaction with other T3SS components might be involved in the calcium-signaling pathway.
The SepL region involved in the interaction with EscV was mapped to the last 11 amino acids (Figure 5b), and we showed that this protein version, SepL ∆C11 , is unable to rescue the phenotype of a sepL null mutant (Figure 5c). This result is similar to previous observations in enterohemorrhagic E. coli, where complementation of a sepL mutant with a SepL ∆C11 protein version did not recover the secretion of translocators to wild-type levels, and only marginally reduced the boosted secretion of effectors (Wang et al., 2008). These results are also in agreement with recent observations in Shigella showing that the C-terminal region of the gatekeeper MxiC is essential for substrate secretion regulation (Roehrich et al., 2017). Remarkably, the last 11 residues of SepL are also involved in EscD and Tir binding (Wang et al., 2008), suggesting a plausible model for SepL control of substrate secretion that could involve its dynamic interactions with EscD, Tir, and EscV. The interaction of SepL with Tir was proposed to delay effector secretion (since Tir secretion is required for the secretion of other effectors (Thomas, Deng, Baker, Puente, & Finlay, 2007)), while its interaction with EscD was suggested to release Tir and promote effector secretion (Wang et al., 2008). This is supported by the finding that overexpression of escD leads to hypersecretion of Tir (Ogino et al., 2006). In addition, since SepL resembles a T3 effector but without being secreted (Younis et al., 2010), it could function as a Trojan horse that, after its recognition as a T3 substrate, blocks a substrate acceptor site on EscV, or modifies the affinity of this export gate component for different substrates. Nonetheless, whether the SepL-EscV interaction participates in secretion regulation remains to be investigated. Furthermore, it was recently shown that SepL interacts with the molecular ruler protein EscP, and that the SepL/EscP complex is stabilized by calcium, preventing effector secretion. The model proposed that once the translocation pore is assembled in the host cell membrane, the calcium flux to the T3SS base is reduced, resulting in dissociation of the SepL/EscP complex and effector secretion (Shaulov et al., 2017). Thus, the EscP switch protein, which has been shown to play a role in effector secretion control (Monjaras Feria et al., 2012), emerges as an important player in the in vitro calcium-dependent SepL secretion regulatory pathway (Shaulov et al., 2017). Besides, EscP was previously shown to bind to the cognate Tir chaperone CesT, and suggested to act, together with Tir and the SepL/SepD complex, in the restriction of effector secretion (Monjaras Feria et al., 2012). Overall, the intricate interaction network in which SepL participates may reflect the different regulation layers required to fine-tune the timing of substrate secretion ( Figure 6). However, based on previous observations and our results (i.e., neither the translocon senses the absence of calcium nor the SepL/SepD/CesL switch complex is affected by calcium changes), we suggest that the molecular mechanism triggering F I G U R E 6 SepL protein-protein interaction network involved in middle and late substrate secretion regulation. (a) The SepL protein binds directly to five different T3SS components: SepD, EscP, Tir, EscD, and EscV. Although SepL forms a ternary complex with SepD and CesL (Younis et al., 2010), no direct interaction between CesL (dashed circle) and SepL or SepD has been reported. The effector Tir binds to its cognate chaperone CesT (Elliott et al., 1999), which in turn has been demonstrated to interact with EscP (Monjaras Feria et al., 2012). Finally, the interaction between EscV and different T3S substrates is also displayed.  (Wang et al., 2008), and EscV [red] (this work). In addition, SepL was reported to form two mutually exclusive complexes: the SepL-EscP [blue] complex, which was proposed to regulate substrate secretion in response to calcium changes in vitro (Shaulov et al., 2017); and the SepL-SepD [yellow]-CesL [orange] complex (Younis et al., 2010), whose role in substrate secretion regulation remains to be determined effector secretion upon in vitro calcium removal could be different from the one triggering effector translocation upon host cell contact, as outlined below.
However, by increasing host intracellular calcium levels, Cisz et al. (2008) demonstrated that the in vivo translocation of effectors is not influenced by calcium concentration changes. Therefore, a distinction should be made between induction of effector secretion in vitro and upon host cell contact. While the former can promote effector secretion in a translocator-independent mechanism ( Figure 2) that involves the SepL/EscP protein complex in EPEC (Shaulov et al., 2017), the mechanism participating in effector translocation in vivo is still largely unknown. Finally, our data revealed EscV as a new player that serves as a docking site for both SepL and T3 substrates.
Nevertheless, whether EscV acts in concert with the SepL/SepD/ CesL regulatory switch complex to ensure the proper timing of substrate secretion remains to be determined.