Multiple proteins arising from a single gene: The role of the Spa33 variants in Shigella T3SS regulation

Abstract Shigella invasion and dissemination in intestinal epithelial cells relies on a type 3 secretion system (T3SS), which mediates translocation of virulence proteins into host cells. T3SSs are composed of three major parts: an extracellular needle, a basal body, and a cytoplasmic complex. Three categories of proteins are hierarchically secreted: (a) the needle components, (b) the translocator proteins which form a pore (translocon) inside the host cell membrane and (c) the effectors interfering with the host cell signaling pathways. In the absence of host cell contact, the T3SS is maintained in an “off” state by the presence of a tip complex. Secretion is activated by host cell contact which allows the release of a gatekeeper protein called MxiC. In this work, we have investigated the role of Spa33, a component of the cytoplasmic complex, in the regulation of secretion. The spa33 gene encodes a 33‐kDa protein and a smaller fragment of 12 kDa (Spa33C) which are both essential components of the cytoplasmic complex. We have shown that the spa33 gene gives rise to 5 fragments of various sizes. Among them, three are necessary for T3SS. Interestingly, we have shown that Spa33 is implicated in the regulation of secretion. Indeed, the mutation of a single residue in Spa33 induces an effector mutant phenotype, in which MxiC is sequestered. Moreover, we have shown a direct interaction between Spa33 and MxiC.


| INTRODUC TI ON
Shigellosis was the second leading cause of diarrheal mortality in 2016, accounting for more than 200,000 deaths worldwide (Khalil et al., 2018). Symptoms of shigellosis are mainly due to the invasion of the colon associated with a severe inflammatory reaction and mucosal destruction (Sansonetti, Tran Van Nhieu, & Egile, 1999). The entry of Shigella into the host cell is mediated by the highly conserved type 3 secretion system (T3SS). T3SS spans the whole cell envelope translocating virulence proteins directly into the cytoplasm of the host cells (Cornelis, 2006;Galán & Wolf-Watz, 2006), required for bacterial invasion, intracellular spread, and inhibition of the host immune defenses (Sansonetti, 2006;Schroeder & Hilbi, 2008). The T3SS is divided structurally into three parts: an extracellular needle, a transmembrane basal body, and a cytoplasmic bulb (Blocker et al., 2001;Burkinshaw & Strynadka, 2014;Chatterjee, Chaudhury, McShan, Kaur, & Guzman, 2013). At 37°C, the assembly of the basal body is triggered and the needle subunit MxiH is secreted through the T3SS together with the inner-rod component MxiI (Magdalena et al., 2002). MxiH is a ~9 kDa conserved protein, which forms a ~50 nm long needle structure by polymerization (Blocker et al., 2001;Cordes et al., 2003;Fujii et al., 2012). Secreted proteins are divided into three classes: translocators, early effectors, and late effectors.
Upon activation of T3SS by host cell contact, the translocators IpaB and IpaC, are inserted into the host cell membrane forming a translocation pore (Blocker et al., 1999;Veenendaal et al., 2007), and releasing IpgC in the cytoplasm. Pore insertion triggers a signal, probably transmitted through the needle, to allow MxiC and subsequent effector release, including OspD1 (Kenjale et al., 2005;Veenendaal et al., 2007). IpgC and MxiE together can activate x late effector transcription and subsequent secretion.
Interestingly, deletion of mxiC in these strains restores their ability to secrete effectors (Cherradi et al., 2013;El Hajjami et al., 2018;, suggesting that MxiC is involved in the regulation of effector secretion. A direct interaction between MxiC and MxiI has been shown in this process but no cytoplasmic component receiving the activation signal has been identified to date. In Shigella, the cytoplasmic complex is composed of Spa33, Spa47, MxiK, and MxiN, forming a high molecular weight complex and serves as a sorting platform for T3SS substrates Spa33 is located beneath the basal body and interacts with the cytoplasmic moiety of the basal body proteins MxiG and MxiJ (Morita-Ishihara et al., 2006). It has been proposed that the sorting platform consists of a central hub (Spa47) and six spokes (MxiN), with a pod-like structure (Spa33) at the terminus of each spoke (Hu et al., 2015).
Inactivation of the spa33 gene results in the absence of the cytoplasmic complex, no needles at the surface and consequently lacks protein secretion (Hu et al., 2015;Morita-Ishihara et al., 2006).

Spa33 exhibits sequence similarities with orthologs in other T3SSs
including the flagellar proteins FliM and FliN of Salmonella, SpaO of Salmonella, and YscQ of Yersinia. As shown for SpaO and YscQ, an internal translation start codon is present in spa33, and leads to the expression of a short carboxy-terminal variant, called Spa33 C (12-kDa), which interacts with the full-length protein (Spa33 FL , 33-kDa) (Bzymek, Hamaoka, & Ghosh, 2012;McDowell et al., 2016;Song et al., 2017). Absence of Spa33 C completely abolishes T3S, showing that Spa33 C is crucial for secretion as already shown in other T3SS (Bzymek et al., 2012;McDowell et al., 2016;Song et al., 2017). However, the exact roles of Spa33 FL and Spa33 C in T3SS are still unclear.
In the present study, we strived to characterize the role of Spa33 FL and Spa33 C in the regulation of the T3SS secretion by creating a series of mutants. We presented evidence that multiple proteins result from the spa33 gene, some being required for T3SS function. Moreover, we show that Spa33 plays a significant role in effector secretion and interacts directly with MxiC and MxiI.

| Bacterial strains and cultures
Bacterial strains and plasmids used in this study are listed in Appendix Table A1. Unless stated otherwise, we consistently used Shigella flexneri M90T (serotype 5a) strain as a parental strain during this study. Shigella strains were grown in Tryptic Soy Broth (TSB) at 37°C and phenotypically selected on Congo red (CR) (Meitert et al., 1991). Escherichia coli (E. coli) strains, Top10, and BL21 DE3, were grown in Luria-Bertani (LB) broth. When required, appropriate antibiotics with following final concentrations were added to the bacterial cultures: zeocin 50 μg/ml, kanamycin 50 μg/ml, streptomycin 100 μg/ml, ampicillin 100 µg/ml, and chloramphenicol 25 μg/ml for E. coli strains and 3 μg/ml for Shigella strains.

| Construction of the spa33 and mxiCspa33 mutants
Generation of ∆spa33 and ∆mxiC∆spa33 mutants was achieved by single-step gene inactivation method using the λ Red system as described previously (Datsenko & Wanner, 2000

| Generation of recombinant plasmids and mutagenesis
All the plasmids and primers used in this study are listed in Appendix   Tables A1 and A2, respectively. The plasmid pMK1 (pSU18-spa33) was used to complement the spa33 mutant. The gene encoding Spa33 was amplified using primers tailed with BamHI/HindIII restriction sites. The double digested PCR product with BamHI/HindIII restriction enzymes was ligated into the BamHI/HindIII sites of the low-copy vector pSU18 (Invitrogen). To complement the mxiC spa33 double mutant, we constructed a plasmid, carrying both native MxiC and Spa33 (pMK2) by cloning the spa33 gene with 5′ insertion of the Shine and Dalgarno (SD) sequence into KpnI/PstI restriction sites of the pSU18-MxiC (Cherradi et al., 2013 To combine the expression of Spa33 variants, lacking either Spa33 C or Spa33 CC , and Spa33 C in trans, we constructed pBAD-Sap33 C . The DNA sequence coding for the Sap33 C was amplified by PCR and digested with NcoI/HindIII. The resulting cleaved product was then inserted into the pBAD vector giving rise to pMK11. A series of single amino acid substitutions on pSU18-Spa33 and pSU18-MxiC-Spa33 was generated and co-expressed in trans with Spa33 C from pBADHisA.
All the plasmids expressing GST fusion and His fusion proteins used for the protein-protein interaction assays are listed in Appendix

| Secretion tests
The detailed procedure for the preparation of crude extracts, leakage of the Ipa proteins into the culture supernatant and CR-induced protein secretion were described in previous works (Allaoui, Sansonetti, & Parsot, 1992;Botteaux et al., 2009). S. flexneri strains were grown overnight at 37°C in TSB medium with appropriate antibiotics. The overnight cultures were diluted to an optical density at 600 (OD 600 ) of 0.1 in 15 ml of TSB supplemented with appropriate antibiotics and grown at 37°C. In the case of S. flexneri strains carrying pBAD-Spa33 or its derivatives, 0.001% arabinose was added to the culture when they reached an OD 600 of 0.6. Cultures were grown to OD600 of ≅2, and bacteria were collected by centrifugation at 2,800 g for 15 min at 37°C. The supernatants were collected and precipitated with 4.5 g of ammonium sulfate for overnight as described previously (Botteaux et al., 2009). The bacterial pellet was suspended in 1X phosphate buffer saline (PBS) containing 200 μg/ml CR and induced for 20 min at 37°C on a shaker incubator. After incubation, bacteria were centrifuged at 13,000 g for 15 min at RT and supernatants were collected. The CR induced and noninduced samples were mixed with 4× Laemmli sample buffer, resolved on 12% SDS-PAGE, and visualized by Coomassie blue staining or Western blot. All secretion tests were conducted at least three times.

| GST-pulldown assays
The E. coli BL21 DE3 or Top10 strains were used in this study as the host cell for the expression of recombinant (GST and His fused) proteins. All the plasmids expressing GST, GST-MxiC, GST-MxiC F206S , GST-Spa33 C , GST-MxiI, His-tagged Spa33, and its derivatives used for the protein-protein interaction assays are listed in Appendix Table A1. To express the recombinant proteins, the cells were propagated in the LB medium containing appropriate antibiotic at 37°C and 200 rpm. Once the bacterial growth reached 0.6-0.7 at OD 600 , the protein expression was induced by adding isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.1 mmol/L and incubating at 30°C for at least 3 hr. In the case of strains carrying pBAD-Spa33 C , 0.002% L-arabinose was used for the induction. After 3 hr, cells were harvested by centrifugation (8,000 g, 15 min, 4°C).
The harvested cells were resuspended in cold phosphate-buffered saline (PBS) supplemented with 0.1% TritonX-100, 0.15 mM PMSF, and iodoacetamide. Sonication was used to lyse the cell suspensions at the following settings: amplitude, 70; time, 3 min; pulsar, 10 s. Cell lysates were then clarified by centrifugation at 8,000 g for 30 min at 4°C. The cleared lysates were mixed with 200 µl of GST-Bind™ Resin (EMD Millipore Novagen) which had been previously equilibrated with PBS buffer and incubated for one hour at room temperature (RT) while shaking. GST beads were recovered by centrifugation and then washed five times with PBS. Then lysates of target proteins with His-tag expressed from E. coli strains (Rosetta DE3) were applied to the beads, which was followed by overnight incubation at 4°C. Beads were washed again extensively, and the captured proteins were eluted by incubating beads for 20 min at RT with elution buffer (40 mmol/L Tris pH 8.0, 500 mmol/L NaCl, and 50 mmol/L reduced glutathione). Eluted samples were separated by SDS-PAGE and analyzed by Coomassie blue staining and Western blot.
MS1 spectra were collected in the range 400-1,200 m/z for 250 ms.
The 20 most intense precursors with charge state 2-4 were selected for fragmentation, and MS2 spectra were collected in the range 100-2,000 m/z for 100 ms; precursor ions were excluded for reselection for 12 s.

| The spa33 gene encodes 5 fragments
To better understand the role of Spa33 in T3S, we cloned the spa33 gene in an expression vector (pET30a), allowing fusion of a hexa-histidine peptide to both ends of the protein (His-Spa33-His). By analyzing whole cell extracts with anti-his antibodies, we observed 5 fragments of different sizes ( Figure 1a). Three of these peptides were well expressed: ~33 kDa corresponding to the Spa33 full-length (Spa33 FL ), ~12 kDa, which was previously discovered by McDowell et al. (2016), named Spa33 C , and a smaller peptide of ~7 kDa (called hereafter Spa33 CC ) ( Figure 1a and Table 1).
The two other peptide fragments were barely detectable; one of ~11 kDa (called Spa33 N ) and another of ~10 kDa (called Spa33 X )   Table A3 and Figure A2). Based on the apparent size of Spa33 CC (7-kDa) and our MS data, we searched for potential alternative start codons in spa33.
We mutated an ATG codon, encoding the M237 residue in Spa33, by replacing it with an alanine (Spa33 M237A ). Study of the protein expression showed that M237A mutation totally abolished expression of Spa33 CC (Figure 2a) suggesting that M237 acts as an internal translation start codon for Spa33 CC . Insertion of a stop codon upstream of the start codon of the Spa33 CC (M237) fragment within the spa33 gene still allows Spa33 CC production, confirming the presence of an alternative start codon for Spa33 CC (Appendix Figure A4).
We generated the same mutation on a low copy plasmid (pSU18) carrying the spa33 gene (pSpa33), which can restore proteins secretion in a spa33 knockout mutant (Δspa33) (Figure 2b). To analyze the secretion profile in the absence of Spa33 CC , we induced secretion by adding Congo Red dye (CR), a small amphipathic molecule, which mimics host cell contact (Meitert et al., 1991). in a medium copy plasmid with a stronger promoter (pBAD) also allows perfect complementation of the spa33 mutant (Appendix Figure A3). The Spa33 M237A mutant generated on this plasmid presents the same phenotype as pSU18-spa33 (data not shown).
To understand if the absence of secretion was only due to the absence of Spa33 CC , we restored expression of Spa33 CC by  Spa33 CC transforming a second plasmid (pBAD-Spa33 C -His) that encodes Spa33 C and Spa33 CC in the spa33 mutant expressing Spa33 M237A (trans-complementation). This plasmid allows the expression of Spa33 C but also of Spa33 CC (Appendix Figure A4). We observed that trans-complementation in the Δspa33/pSpa33 M237A background restored the secretion to the WT level (Figure 2b).

| Spa33 N is a slippage product from the spa33 gene but is not required for T3SS secretion
Work published by Penno et al. demonstrates that at the RNA level, a string of 9 alanines at position 180-189 bp of spa33 allows lowlevel transcriptional slippages (Penno et al., 2006). According to our in silico analysis, the molecular weight of Spa33 N (around 11 kDa), potentially corresponds to a protein produced during a +1 slippage (Appendix Figure A5). To check this hypothesis, we mutated the slippage site to prevent frameshifting and analyzed the expression profile of the resulting His-Spa33 Slipp* . This plasmid allows a better detection of both Spa33 N and Spa33 X than His-Spa33-His.
Our results showed that the Spa33 N is totally absent in this mutant ( Figure 3a) while Spa33 X is still produced although at a lower level.
However, the mutation of the slippage site was not associated with any detectable change in T3SS secretion under induced condition ( Figure 3b).

| Role of Spa33 C in T3SS
Spa33 C arises from an alternative translation start codon (GTG) which encodes a valine at position 192 in the full-length protein (McDowell et al., 2016). To further investigate the role of Spa33 C in T3SS, we constructed 3 mutants by changing the alternative start codon by a synonymous mutation (Spa33 GTC ) or two nonsynonymous mutations, where V192 was replaced with an alanine (Spa33 V192A ) or with an aspartic acid (Spa33 V192D ). These mutations were first introduced in the His-Spa33-His plasmid to allow detection of all the Spa33 fragments. As expected, the three constructs allow expression of Spa33 FL but lack the Spa33 C product (Figure 4a).
In the absence of Spa33 C , we still detect Spa33 N , though scarcely, but do not detect Spa33 X anymore.
We then generated the same mutations in pSpa33 and introduced the resulting plasmids (pSpa33 GTC , pSpa33 V192A , and pSpa33 V192D ) into the spa33 mutant. Analysis of CR-induced culture supernatants showed that, as expected, all these variants (McDowell et al., 2016), were unable to restore protein secretion compared to the complemented strain (Figure 4b). Same mutations were generated in pBAD-Spa33-His and transformed in the spa33 mutant. They all present the same phenotype as the corresponding mutations in pSU18-spa33 (data not shown).
To understand if the absence of secretion was only due to the absence of Spa33 C , we restored expression of Spa33 C by transforming a second plasmid expressing Spa33 C and Spa33 CC in the spa33 mutant (trans-complementation; Appendix Figure A6). Supernatants from CR induced cells ( Figure 4) and whole cell extracts (Appendix Figure A6) were analyzed by Western blot using antibodies targeting different classes of secreted proteins. We observed that expression of Spa33 C with pSpa33 GTC or pSpa33 V192A allows secretion of proteins at wild-type levels ( Figure 4b). Interestingly, trans-complemen- Spa33 N Spa33 X ∆spa33 late effectors (IpaH) were detected after CR induction (Figure 4b).
This "effector mutant" phenotype, as previously described (Cherradi et al., 2013;El Hajjami et al., 2018;Kenjale et al., 2005), suggests a role of Spa33 in the regulation of the secretion hierarchy. Expression of Spa33 C alone in the spa33 mutant does not allow T3 secretion (Appendix Figure A3). co-expressed them with Spa33 C in the double mxiC spa33 mutant. We analyzed the ability of the mxiC spa33 mutant and its derivatives to secrete virulence proteins under both constitutive (leakage) and induced conditions (CR). We observed that, when MxiC F206S , Spa33 V192D , and Spa33 C were simultaneously expressed, effector secretion was restored, as observed in the absence of MxiC ( Figure 5b). Interestingly, analysis of noninduced culture supernatants established that this mutant did not show an increased leakage as usually seen with MxiC F206S (Figure 5c). Our results suggest that these two mutations, one on MxiC and one on Spa33, rescue each other's phenotypes and support that Spa33 is involved in T3SS regulation. Whole cell extracts analysis showed that the lack of secretion in some mutants is not due to impaired proteins production (Appendix Figures A7 and A8).

As our previous results suggest a (direct or indirect) link between
Spa33 and MxiC, we tested the potential interaction between these two proteins. GST-MxiC was immobilized on glutathionesepharose beads and incubated with E. coli lysates expressing His-Spa33-His ( Figure 6a). In lysates, only two forms of Spa33 were detectable: Spa33 FL and Spa33 C , suggesting that the other fragments are not soluble. Subsequently, fractions eluted were analyzed by Western blot, using anti-His antibody. As shown in    (Figure 6a). Our results showed that Spa33 C and Spa33 CC co-eluted with GST-MxiC, in the absence of Spa33 FL (Figure 6b).

The interaction between Spa33 and the gatekeeper protein
MxiC prompted us to investigate whether Spa33 interacts with other proteins, implicated in the regulation of effectors secretion.
In Shigella, the inner-rod component, MxiI, interacts with MxiC to prevent effectors secretion (Cherradi et al., 2013;El Hajjami et al., 2018). Cleared lysates prepared from E. coli producing His-Spa33-His and Spa33 C -His were incubated with GST-MxiI or GST alone, which had been preincubated with glutathione-sepharose beads.
We found that Spa33 C -His, produced either with Spa33 FL or with Spa33 CC , interacts with GST-MxiI ( Figure 6c).
As Spa33

| Spa33 V192D mutation did not abolish interaction with Spa33 C
As the expression of Spa33 V192D and Spa33 C allows translocator secretion but not that of effector, we wanted to test if the mutation had any effect with respect to binding of Spa33 C , MxiC, or MxiI. We cloned the COOH-terminal part of spa33 in a plasmid allowing its NH2-terminal fusion with a GST-tag (pGEX4T1-Spa33 C ) and tested for the interaction with His-Spa33-His, His-Spa33 GTC -His, or His-Spa33 V192D -His.
We know that Spa33 FL is less stable in the absence of Spa33 C (Figure 4a and McDowell et al., 2016). We failed to detect Spa33 FL in the lysate in the absence of Spa33 C suggesting that Spa33 C also has a role in solubilization of Spa33 FL . To overcome this problem, we expressed all recombinant proteins separately and then mixed cell suspensions before sonication. The premixes of clarified cell lysates ( Figure 7a) were then immobilized on glutathione-sepharose beads and washed. Analysis of eluted fractions revealed that the Spa33 V192D mutation did not affect the ability of Spa33 FL to bind to Spa33 C (Figure 7b). We could observe Spa33 CC in all the eluted fractions but not in the lysates, as previously observed (Figure 6a).
Since the previously observed deficiency in effectors secretion could be due to changes in the interaction of Spa33 V192D with MxiC and/or MxiI, we tested the potential interaction between GST-MxiC, GST-MxiI, and His-Sap33 V192D -His. Unfortunately, the expression of Spa33 FL (GTC or V192D) in the absence of Spa33 C was barely detectable in the cell lysates and not sufficient to work with, even when premixed with GST-MxiI or GST-MxiC (Appendix Figure A9 and data not shown).

| D ISCUSS I ON
In this study, we showed that Spa33, a C-ring component, plays a significant role in the regulation of secretion and encodes multiple proteins from a single gene, namely Spa33 FL , Spa33 C , Spa33 CC , Spa33 N , and Spa33 X (Table 1).
For the first time with Spa33 homologous proteins, we have identified an internal start codon, encoding a methionine at position 237 and leading to a small ~7 kDa fragment. The absence of Spa33 CC results in the lack of T3SS substrate secretion indicating F I G U R E 7 Spa33 V192D mutation did not abolish interaction with Spa33 C . Cell suspensions of E. coli (BL21) producing His-Spa33-His, His-Spa33 GTC-His, and His-Spa33 V192D -His were mixed with cell suspensions of GST-Spa33 C or GST alone followed by sonication and centrifugation. Clarified lysates were incubated with glutathione-sepharose beads and proteins were eluted as described in experimental procedure. Mw that this fragment is critical for T3SS function. It is noteworthy that lack of either Spa33 C or Spa33 CC mimics the knockout mutant phenotype in terms of protein secretion. Therefore, we cannot exclude the possibility that Spa33 CC and Spa33 C also contribute to T3SS assembly. Spa33 CC is not expressed significantly and we failed to find a ribosomal binding site (Shine-Dalgarno sequence) upstream of the M237. Spa33 FL and Spa33 C were shown to interact with each other to form a complex in a 1:2 ratio, and further oligomerize in a defined order to form a functional cytoplasmic complex structure (McDowell et al., 2016). However, our results showed that Spa33 CC also copurified with Spa33 FL and Spa33 C (Appendix Figure A10), indicating that Spa33 CC is also part of Spa33 FL -Spa33 C complex. Spa33 X , the only fragment we failed to find an origin for, seems linked, directly or indirectly, to Spa33 CC expression as the fragment totally disappeared in the absence of Spa33 CC . A more detailed study of compositional and conformational changes of these protein assemblies under various conditions may provide a better insight to understand the complex regulatory dynamics involved.
Spa33 N was identified as a slippage product of the spa33 gene.
This fragment is barely expressed and therefore barely detected in whole cell extracts and not detected at all in soluble fractions. The rate of slippage events from spa33, measured by Penno et al., was <14% of total RNA, which can explain the difficulties encountered to detect and purify this fragment. We failed to find a role for Spa33 N in T3SS under our experimental conditions. However, it is well-known that transcription of spa33 is repressed by fumarate and nitrate reductase (FNR) binding in the absence of oxygen (O 2 ), leading to T3SS functional impairment (Marteyn et al., 2010). Hence, Spa33 N could have a role in regulation under different conditions that bacteria could encounter during the infection process and not represented in our experimental design. Future work to identify the exact role of Spa33 N is needed as it appears to be a specific feature of Shigella's T3SS regulation, as the homologous genes (yscQ, spaO) do not harbor any slippage sites in their genes (data not shown).
Spa33 FL and Spa33 C have been previously shown to be important for T3SS (McDowell et al., 2016) as the absence of Spa33 C lead to a total deficiency in secretion. Nevertheless, the mechanism by which Spa33 C plays a role in secretion remains unclear. We have shown that expression of Spa33 C in trans could restore secretion in the Spa33 GTG variant. This result is not surprising as the homologous proteins of Spa33 FL and Spa33 C are encoded by separate genes in other systems (i.e., fliM and fliN genes in Salmonella flagella). More surprisingly, trans-complementation with Spa33 C in the Spa33 V192D background, also lacking Spa33 C but harboring a point mutation on the full-length form, restored secretion of translocators, but not of effectors. In several mutational studies on MxiH and MxiI, both implicated in signal transmission, this phenotype has already been described (Cherradi et al., 2013;Kenjale et al., 2005;Roehrich, Guillossou, Blocker, & Martinez-Argudo, 2013). In the mutants MxiH K69A and MxiI Q67A , the lack of effector secretion was shown to be dependent on the presence of MxiC as mxiC inactivation in these strains restores effector secretion to wildtype levels (Cherradi et al., 2013;El Hajjami et al., 2018). Moreover, a mutation in MxiC that inhibits its interaction with MxiI (MxiC F206S ), presents the same rescue effect. It was hypothesized that as this variant is secreted too early (before induction), it opens the way for effector secretion. Moreover, MxiI-MxiC complexes could be physically implicated in the inhibition of effector secretion before host cell contact (Cherradi et al., 2013;El Hajjami et al., 2018).
Inactivation of mxiC or expression of MxiC F206S in our "effector mutant" (Spa33 V192D /Spa33 C ) also leads to effector secretion like in the wild-type strain. These results suggest that Spa33 implicated as well in signal transmission/reception. We showed that Spa33 is able to interact directly with MxiC and MxiI, at least by its C-terminal fragment. The interaction domain of Spa33 with MxiI and MxiC is probably different as the 3 proteins can form a complex in vitro.
The interaction between MxiI with Spa33 was expected as Spa33-MxiN-MxiK complex acts as a "sorting platform" that determines the recognition, timing, and sorting of specific substrates exported in a defined order to form a functional T3SS (Hu et al., 2015). Moreover, most of the virulence proteins intended for secretion are produced by the bacteria during T3SS assembly and are believed to be predocked at the base of the injectisome until favorable conditions are available (Spaeth, Chen, & Valdivia, 2009). MxiC is also a secreted effector but its interaction with Spa33 could be more than a T3SS substrate-secretion machinery interaction. Indeed, MxiC F206S is able to rescue the phenotype of the effector mutant (Spa33 V192D /Spa33 C ).
Interestingly, the timing of MxiC F206S secretion is restored to wildtype level in the Spa33 V192D /Spa33 C background that allows a wildtype secretion profile. In the case of MxiC F206S , which maintained an interaction with Spa33, but failed to interact with MxiI (Cherradi et al., 2013), we observed a lack of secretion control before induction while secretion is normal following CR induction (Figure 5b,c). In the case of Spa33 V192D /Spa33 C , we can postulate that interaction between Spa33 and MxiC is stronger and leads to a defect of MxiC and effector secretion, in response to pore insertion.

| CON CLUS IONS
In conclusion, we present evidences that spa33 encodes multiple proteins that are required for T3SS function. For the first time, we show that Spa33 is involved in secretion hierarchy, regulating effector secretion upon host cell sensing. These results clearly indicate that Spa33 is involved in the timing of MxiC secretion. This study therefore provides comprehensive and critical insights into the complex regulation of T3SS and opens new avenues for future research endeavors.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

E TH I C A L A PPROVA L
None required.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data are provided in full in the results section of this paper.

Anne Botteaux
https://orcid.org/0000-0001-9208-515X   F I G U R E A 4 (a) Insertion of stop codon after the start codon for Spa33 C (V192) did not abolish the production of Spa33 CC . Whole cell extracts (WCE) of the E. coli strains (Top10) harboring plasmid expressing Spa33-His or its mutated derivative (Spa33 stop-His) were analyzed by Western blot using anti-his monoclonal antibodies. (b) Absence of Spa33 CC did not impair virulence proteins production. Analysis of whole cell extracts (WCE) of the wild-type strain (WT), the spa33 mutant (∆spa33), the spa33 mutant complemented by pSU18-Spa33 (pSpa33) or its mutated derivative (pSpa33 M237A ), with or without Spa33 C-CC , resolved by SDS-PAGE and analyzed by Western blot using anti-IpaC, IpaB, IpaD, MxiC, IpaA, IscB, and His antibodies. Mw, molecular weight in kDa. *Nonspecific band