Chaperone-mediated secretion switching from early to middle substrates in the type III secretion system encoded by Salmonella pathogenicity island 2

The bacterial type III secretion system (T3SS) delivers virulence proteins, called effectors, into eukaryotic cells. T3SS comprises a transmembrane secretion apparatus and a complex network of specialized chaperones that target protein substrates to this secretion apparatus. However, the regulation of secretion switching from early (needle and inner rod) to middle (tip/filament and translocators) substrates is incompletely understood. Here, we investigated chaperone-mediated secretion switching from early to middle substrates in the T3SS encoded by Salmonella pathogenicity island 2 (SPI2), essential for systemic infection. Our findings revealed that the protein encoded by ssaH regulates the secretion of an inner rod and early substrate, SsaI. Structural modeling revealed that SsaH is structurally similar to class III chaperones, known to associate with proteins in various pathogenic bacteria. The SPI2 protein SsaE was identified as a class V chaperone homolog and partner of SsaH. A pulldown analysis disclosed that SsaH and SsaE form a heterodimer, which interacted with another early substrate, the needle protein SsaG. Moreover, SsaE also helped stabilize SsaH and a middle substrate, SseB. We also found that SsaE regulates cellular SsaH levels to translocate the early substrates SsaG and SsaI and then promotes the translocation of SseB by stabilizing it. In summary, our results indicate that the class III chaperone SsaH facilitates SsaI secretion, and a heterodimer of SsaH and the type V chaperone SsaE then switches secretion to SsaG. This is the first report of a chaperone system that regulates both early and middle substrates during substrate switching for T3SS assembly.

Many Gram-negative pathogens use type III secretion systems (T3SSs) 2 to deliver virulence proteins, termed effectors, directly into the eukaryotic cytoplasm (1). T3SSs comprise several structures, including a sorting platform containing an ATPase complex and a cytoplasmic ring, an inner membrane export apparatus, a basal body with a needle and an inner rod, and a translocation pore. For T3SSs to fully assemble and function correctly, numerous proteins must be secreted in a predefined order. On the basis of the T3SS secretion hierarchy, secreted proteins are divided into early (needle and inner rod), middle (tip/filament and translocators), and late (effectors) substrates.
The components of the T3SS have been uniformly named (1,2). The needle (SctF) and inner rod (SctI) proteins are secreted through the basal body. When the needle reaches a defined length, it signals the completion of the functional type III secretion apparatus, and the system switches to secretion of the middle substrates (3). A needle-length ruler (SctP) is a key regulator responsible for controlling the needle length and for secretion switching. The loss of SctP results in long, unregulated T3SS needles (4,5). The secretion of SctF and SctI is significantly increased upon deletion of SctP (6 -8). SctU, a member of the export apparatus, is also involved in controlling needle length and secretion switching (3,7,9,10).
In some bacterial species, specific chaperones are necessary for SctF secretion. Chaperones in T3SSs bind to newly synthesized substrates and keep them partially unfolded in a secretion-competent state (1,11). Chaperones are classified on the basis of substrate specificity into classes IA/IB, II, III, IV, and V (12). Class III chaperones are known to bind to needle proteins. In Pseudomonas aeruginosa and Yersinia spp., SctF (PscF and YscF) forms a complex between the class III chaperone (PscG and YscG) and a partner protein known as the class V chaperone (PscE and YscE). Structural studies have revealed that class III chaperones bind to the C termini of needle proteins to maintain their soluble state and prevent self-aggregation (13,14). However, the homologs of class III and V chaperones have not been identified in most T3SSs. Thus, it is still unclear whether chaperone-mediated regulation of early substrate secretion contributes to the control of the needle length and substrate switching in other T3SSs.
Salmonella enterica serovar Typhimurium is an enteropathogenic bacterium that causes gastroenteritis in humans and typhoid-like fever in mice. Salmonella produces two T3SSs encoded chromosomally by the Salmonella pathogenicity islands (SPIs) 1 and 2 (15). The SPI1-T3SS plays an important role in gastrointestinal disease but seems to be largely dispensable for systemic infection in mice (16). In contrast, the SPI2-T3SS is required for growth within host cells such as macrophages to establish systemic infection in mice (17,18). After the phagocytosis of Salmonella by macrophages, SPI2 expression is induced in bacteria growing intracellularly in response to phagosomal conditions such as acidic pH and nutrient limitations, whereas SPI1 expression is repressed (19,20). Thirty-one genes in SPI2 are organized into two operons encoding components of the secretion apparatus, ssaBCDE and ssaG-U; one transcriptional unit encodes effectors (sse) and chaperones (ssc), and the other encodes the two-component regulatory system, SsrAB (21) (Fig. S1). To date, two subgroups of substrate proteins transported by the SPI2-T3SS have been identified: translocators and effectors. SseBCD function as translocators of the SPI2-T3SS and show a certain similarity to the translocon proteins EspABD of enteropathogenic Escherichia coli (EPEC) (17,22,23). SseB is a protein with sequence similarity to EspA and composes a filament, whereas SseC and SseD are similar to EspD and EspB, respectively, and compose the pore-forming translocon. In vitro secretion analysis has shown that all three proteins are secreted by the SPI2-T3SS and are loosely attached to the bacterial surface (22,24). Although SseB secretion is independent of SseC and SseD (22), efficient secretion of the latter two proteins occurs only in the presence of SseB (23). After assembly of the translocation pores in phagosomal membranes by SseC and SseD, effectors are translocated into the host cytoplasm upon sensing of the host cytosolic nutrient pH (25). Although many studies have focused on the regulation of secretion and switching from middle (translocators) to late (effectors) substrates, the secretion of early substrates by the SPI2-T3SS is not well understood.
In this study, we aimed to investigate chaperone-mediated secretion switching from early to middle substrates in the SPI2-T3SS. Our findings show that the class V chaperone SsaE regulates the cellular level of the class III chaperone SsaH to translocate early substrates and subsequently promotes secretion of the middle substrate, SseB, through its stabilization. The class V chaperone that regulates both early and middle substrates affecting substrate switching, described here, is expected to provide insight into the mechanism of secretion switching from early to middle substrates in the T3SSs of pathogenic bacteria.

SsaI is an early substrate of SPI2-T3SS
SsaI is a homolog of the inner rod, commonly known as SctI (1). The SsaI ORF is 246 bp long and encodes a peptide of 82 amino acid residues with a predicted molecular mass of 8.96 kDa. SsaI seems to have low levels of identity with other inner rod SctI proteins (Fig. S2A). SctI proteins are known to be secreted as early T3SS substrates. To determine whether SsaI is also secreted by the SPI2-T3SS, we established a rabbit anti-SsaI antibody (Fig. 1). The antibody showed a band corresponding to ϳ9 kDa in the cytoplasm of the wildtype (WT) strain (Fig.  1A), whereas the protein was not detected in the ssaI-disrupted mutant (⌬ssaI) strain. However, the corresponding protein was detected in the ⌬ssaI strain carrying plasmid pTKY1245, capable of overexpressing ssaI following the addition of isopropyl ␤-D-1-thiogalactopyranoside (IPTG) under the regulation of the P A1lacOϪ1 promoter system ( Fig. 2A), indicating that the anti-SsaI antibody established in this study has the ability to specifically detect SsaI produced in Salmonella cells. The anti-SsaI antibody detected the protein in the supernatant fraction prepared from WT cells (Fig. 1B), demonstrating the secretion of the SsaI protein.
SsaG, SsaP, and SsaU are homologs of the needle protein SctF, the needle-length ruler SctP, and the inner membrane protein SctU, respectively (1). SsaN is an ATPase associated with SPI2 (26,27). SseB is the filament protein secreted by the SPI2-T3SS (23,28). We also detected SseB in the supernatant fraction by immunoblotting with a rabbit anti-SseB antibody used previously (23). To examine whether the disruption of the SPI2 apparatus influenced the production of SsaI and SseB, the cellular levels of SsaI and SseB in the disrupted mutants of ssaG, ssaN, ssaP, ssaU, and sseB were examined (Fig. 1A). Levels of SsaI in ⌬ssaG, ⌬ssaN, ⌬ssaU, and ⌬sseB strains were detectable but slightly lower than that in the WT strain. In contrast, SseB levels in ⌬ssaG, ⌬ssaI, ⌬ssaN, and ⌬ssaU strains were similar to that in the WT strain. In the ⌬ssaP strain, the amount of SsaI was increased, but the production of SseB was not affected. These results suggest that the disruption of the SPI2 apparatus may not reduce the production of SsaI and SseB.
To determine whether SsaI is secreted extracellularly via the SPI2-T3SS, we examined the amounts of SsaI and SseB in the supernatants of WT and SPI2 mutant strains (Fig. 1B). Both SsaI and SseB were detected in the supernatants of the WT strain. The anti-SsaI antibody detected two bands in the supernatant fraction. On the basis of molecular size, the upper band appeared to correspond to the intact SsaI protein, which can be detected in the cell lysate, whereas the lower band was assumed to be truncated SsaI. In the supernatant of the ⌬sseB strain, SsaI was detected at a reduced level compared with that in the WT strain. In contrast, neither SsaI nor SseB was detected in the supernatants of the ⌬ssaN and ⌬ssaU strains (Fig. 1B). These results suggest that SsaI may be secreted via the SPI2-T3SS prior to secretion of the middle substrate SseB.

SsaI and SsaG coregulate SseB secretion
The ⌬ssaI strain was not capable of secreting SseB (Fig. 1B). Moreover, the ⌬ssaG strain did not secrete SsaI or SseB (Fig.  1B). To confirm that the loss of SseB secretion in both ⌬ssaG and ⌬ssaI strains was due to disruption of ssaG and ssaI, respectively, plasmids pTKY1243 and pTKY1245, capable of overexpressing ssaG and ssaI, respectively, under the regulation of the P A1lacOϪ1 promoter system, were introduced into the corresponding strains. The resultant strains were tested for complementation of SseB secretion via detection of SseB protein in cell lysates and supernatants (Fig. 2). The cellular level of SseB in the ⌬ssaI strain containing pTKY1245, which enables inducible expression of ssaI, was increased compared with that in ⌬ssaI strains carrying the empty vector pUHE21-2⌬fd12 and pTKY1243 ( Fig. 2A). In the supernatant of the ⌬ssaI strain carrying pTKY1245, SseB was detected (Fig. 2B), indicating that ssaI gene on pTKY1245 was translated to the functional SsaI. In the ⌬ssaG strain, ssaG expression by the introduced pTKY1243 recovered the secretion of both SseB and SsaI (Fig. 2), indicating that ssaG gene on pTKY1243 was also translated to the functional SsaG. In contrast, the expression of ssaI by pTKY1245 in the ⌬ssaG strain recovered the secretion of neither SseB nor SsaI, whereas the cellular level of SsaI in this strain was increased compared with those in the ⌬ssaG strain carrying pUHE21-2⌬fd12 and pTKY1243. Furthermore, ssaG expression in the ⌬ssaI strain did not recover the secretion of SseB. SsaG is the needle protein (Fig. S2B), which is secreted as the early substrate. Therefore, the results suggest that the secretion of both SsaG and SsaI may be involved in SseB secretion.

SsaI secretion is negatively regulated by SsaP
In the supernatant of the ⌬ssaP strain, an increased level of SsaI was detected compared with that in the WT strain, but no SseB was detected (Fig. 1C). The increased secretion of SsaI was suppressed by the introduction of plasmid pTKY1251, which expresses a functional ssaP gene under the regulation of the lac promoter in a low-copy plasmid. Furthermore, SseB secretion was recovered in the ssaP-complemented strain. The secretion of SPI2 effectors is activated by the exposure of bacteria to pH 7.2 following growth at pH 5.0 (25). SseB was not secreted even when the ⌬ssaP strain was exposed to the medium at pH 7.2 after growth at pH 5.0 (Fig. S3). These results suggest that the repression of SsaI secretion by SsaP may be necessary for substrate switching to SseB in the SPI2-T3SS.

Cytosolic protein SsaH is required for secretion of SsaI
The ssaH gene is located between ssaG and ssaI in the ssaG-U operon in SPI2 (Fig. S1). The ORF of ssaH is 288 bp long and encodes a peptide of 96 amino acid residues with a predicted molecular mass of 10.5 kDa and a predicted pI of 4.83. The homolog of SsaH has not been identified in other T3SSs. To examine the function of SsaH in the SPI2-T3SS, we established a rabbit anti-SsaH antibody and characterized an ssaHdisrupted mutant (Fig. 3A). In the cell lysate prepared from the WT strain, a band corresponding to ϳ10 kDa was detected by immunoblotting using the anti-SsaH antibody. The protein was not detected in the cell lysate of the ⌬ssaH strain. The corresponding protein was also detected in the ⌬ssaH strain carrying plasmid pTKY1246, which overexpressed ssaH under the regulation of the P A1lacOϪ1 promoter upon the addition of IPTG (Fig. 3A), indicating the ability of the established anti-SsaH antibody to detect endogenous SsaH. In the ⌬ssaH strain, SsaI protein, encoded by a gene located downstream of ssaH, was produced at the same level as that in the WT strain (Fig. 3A). Additionally, the amount of SseB was similar to that in the WT strain. However, neither SsaI nor SseB was detected in the supernatant of the ⌬ssaH strain (Fig. 3A). The loss of SPI2 secretion was complemented by ssaH expression in the ⌬ssaH strain. In addition, SsaH was not detected in the supernatants of WT or ssaH-overexpressing strains (data not shown). These results suggest that SsaH in the bacterial cytoplasm is required for the secretion of SsaI via the SPI2-T3SS.

SsaH forms a heteromolecular complex with SsaE
Structural modeling of SsaH was performed with the SWISS-MODEL expert system for protein structure modeling (https:// swissmodel.expasy.org) to evaluate the biological function of SsaH (29). The structure of SsaH was based on the class III chaperone PscG in P. aeruginosa (SWISS-MODEL Template Library (SMTL) ID 2uwj.1.C) (Fig. 3B), although the sequence identity between SsaH and PscG was only 12.5% (Fig. 3C). AscG, in Aeromonas hydrophila, and YscG, in Yersinia spp., are also known as class III chaperones. Sequence alignment of SsaH

Chaperone-mediated secretion switching in SPI2-T3SS
showed low identities of 10.4 and 14.9% with AscG and YscG, respectively ( Fig. 3C). PscG interacts with PscE, a class V chaperone, to form the heteromonomeric chaperone PscGE (14). AscG and YscG also interact with AscE and YscE, respectively, which are class V chaperones (30,31). Another homolog of a class V chaperone in the SPI2-T3SS is SsaE (28), encoded by a gene located within the ssaBCDE operon in the SPI2 locus (Fig. S1). Recently, functionality of the SWISS-MODEL system has been extended to modeling heteromeric complexes (32). Thus, a structural model of the complex of SsaH with SsaE was predicted using SWISS-MODEL. Three heteromolecular complexes of SsaH and SsaE were feasible based on a PscGE hetero-2-2-mer, a YscGE heterodimer, and an AscGE heterodimer (Fig. S4, A-C). This suggested that SsaH interacts with SsaE. To test this possibility, GSH S-transferase (GST)-tagged SsaE (GST-SsaE) and SsaH were coexpressed in E. coli and subsequently used to assess the interaction between SsaH and SsaE via a GST pulldown assay (Fig. 4A). When GST-SsaE and SsaH were expressed together in E. coli, both bound to the GSH particles. Following cleavage of the GST tag from SsaE by the addition of PreScission protease, SsaH could be coeluted with SsaE but not with GST, indicating that SsaH directly may interact with SsaE. The gel-filtration profile of the SsaH-SsaE complex showed that the complex was eluted at a volume corresponding to its expected molecular weight as a heterodimer (Fig. 4B). As shown in Fig. 4A, when proteins in the SsaH-SsaE complex were eluted by gel filtration and separated by a Tricine gel system, the resultant band intensity corresponding to the SsaH protein was similar to that of SsaE. These results indicated that SsaH forms a 1:1 heteromolecular complex with SsaE. The SsaH-SsaE heterodimer model predicted that the N-terminal region of SsaH may be involved in the interaction with SsaE (Fig. S4, A-C). Coexpression of SsaH lacking the 21 N-terminal residues and GST-SsaE demonstrated that SsaH lacking the 21 N-terminal residues could not bind to GST-SsaE (Fig. S4D). This result suggests that the N terminus of SsaH is involved in the direct interaction with SsaE in the complex.
Additionally, the gel-filtration profile of SsaE showed that SsaE exists as a homodimer in solution (Fig. 4B). The crystal structures of YscE from Yersinia pestis and AscE from A. hydrophila have been reported as that of dimeric helical proteins (30,33). These findings suggested that SsaE may exist as a homodimer as well as form a heterodimer with SsaH.

Chaperone-mediated secretion switching in SPI2-T3SS SsaE regulates the cellular levels of SsaH and SseB in the bacterial cytoplasm
Although SsaE has been shown to recognize SseB as a middle substrate to regulate its secretion via the SPI2-T3SS (28), the involvement of SsaE in the secretion of early substrates, including SsaI, is not clear. Thus, we constructed an ssaE-disrupted mutant and assessed its ability to secrete SsaI (Fig. 5A). Immunoblotting of the supernatant indicated that neither SsaI nor SseB was secreted by the ⌬ssaE strain. To confirm that the loss of secretion of these proteins in the ⌬ssaE strain was due to the ssaE disruption, plasmid pTKY1261, expressing ssaE under the regulation of the ssaB promoter, was introduced into the ⌬ssaE strain. The amount of SsaI secreted by the ⌬ssaE strain carrying pTKY1261 was significantly increased compared with that in the ⌬ssaE strain, whereas the amount of SseB was only slightly increased. This result indicated that SsaE may also be involved in the secretion of SsaI. The cellular level of SsaI was not affected by the ssaE disruption (Fig. 5A). Interestingly, cellular levels of SsaH and SseB in the ⌬ssaE strain were significantly reduced compared with that in the WT strain. This reduction in the ⌬ssaE strain was complemented by introducing the functional ssaE gene into pTKY1261, suggesting that SsaE may contribute to the stabilization of SsaH and SseB. In contrast, ssaH disruption did not affect the cellular level of SseB (Fig. 3A). These results suggest that SsaE may regulate the production of SseB in an SsaH-independent manner.
To examine the possibility that impaired SsaH production by the ssaE-disrupted strain could result in the loss of SsaI secretion, ssaH was expressed in the ⌬ssaE strain by introducing pTKY1246 (Fig. 5B). The enhanced production of SsaH protein in the ssaE-disrupted background did not recover SsaI secre-

Chaperone-mediated secretion switching in SPI2-T3SS
tion. These results suggest that both SsaH and SsaE may function as a complex to regulate SsaI secretion.

SsaH-SsaE complex interacts with SsaG but not SsaI
Most T3SS chaperones directly interact with their substrates in the bacterial cytoplasm to stabilize them and recruit them to the sorting platform, resulting in the translocation of the substrates into target host cells (34). Although SsaH and SsaE were involved in the secretion of SsaI, neither appeared to be involved in the stabilization of SsaI (Figs. 3A and 5A). We examined the binding of SsaI to SsaH-SsaE in E. coli using the GST pulldown assay (Fig. 6). When GST-SsaE, SsaH, and SsaI were coexpressed in E. coli and proteins bound to SsaE were collected, SsaH but not SsaI was detected in the eluted fraction of SsaE by immunoblotting analysis (Fig. 6). We also performed the GST pulldown assay in the ssaE-disrupted strain of Salmonella (Fig. S5). Cellular SsaH and SseB were detected in the GST-SsaE-expressing strain but not in the GST-expressing strain. Moreover, the GST pulldown assay demonstrated that GST-SsaE interacted with SsaH and SseB but not SsaI on a chromosomal level, suggesting that GST-SsaE may bind to both SsaH and SseB (Fig. S5). These results indicated that the SsaH-SsaE complex recognizes other substrates involved in SsaI secretion.
The PscGE, YscGE, and AscGE complexes form ternary structures with needle proteins PscF, YscF, and AscF, respectively, to avoid premature assembly in the cytoplasm (14,30,35). Furthermore, the needle proteins PscF and YscF are necessary for the secretion of inner rod proteins PscI and YscI, respectively (10,36). In the SPI2-T3SS, the needle protein SsaG is also involved in SsaI secretion (Figs. 1 and 2). Therefore, we hypothesized that SsaH-SsaE complex may act as a chaperone of SsaG to regulate SsaI secretion. Because the establishment of an anti-SsaG antibody was unsuccessful in the present study, the GST pulldown assay in E. coli was performed to investigate the possibility of interaction between the SsaH-SsaE complex and SsaG (Fig. 6). When SsaG was coexpressed with GST-SsaE and SsaH in E. coli, a band corresponding to ϳ8 kDa was observed after collecting proteins bound to SsaE. This protein was also detected upon coexpression of SsaG, SsaH, and SsaI. In contrast, it was not detected in the absence of ssaH expression. The ORF for SsaG is 213 bp long and encodes a peptide of 71 amino acid residues with a predicted molecular mass of 7.9 kDa. Thus, the observed protein corresponds to the predicted size of SsaG. These findings indicated that SsaG may be a substrate of the SsaH-SsaE complex in the SPI2-T3SS.

Discussion
Class III chaperones, which form heterodimeric complexes with partner proteins known as class V chaperones, bind and stabilize the monomeric needle protein to prevent self-aggregation in the bacterial cytosol. Complexes of class III and V chaperones and needle proteins have been identified in P. aeruginosa (PscG-PscE-PscF) (14), Yersinia spp. (YscG-YscE-YscF) (35), A. hydrophila (AscG-AscE-AscF) (30), EPEC (EscG-EscE-EscF) (37), and Chlamydia trachomatis (CdsG-CdsE-CdsF) (38). In this study, we found that SsaH functions as a class III chaperone in S. enterica Typhimurium SPI2-T3SS. S. enterica Typhimurium strains with disruptions in ssaG, ssaH, or ssaE showed no secretion of the inner rod, SsaI, suggesting that SsaG, SsaH, and SsaE may be required for secretion of SsaI, an early substrate of SPI2-T3SS. Prediction of the structures of SsaH and SsaE revealed that their tertiary structures could be aligned and superimposed over the known structures of PscG-PscE, YscG-YscE, and AscG-AscE (Fig.  S3). Biological analysis showed that SsaH forms a heterodimer with the class V chaperone SsaE (Fig. 4). The binding of the needle protein SsaG to SsaE could be detected only in the presence of SsaH (Fig. 6). Therefore, it is suggested that SsaH may interact directly with SsaG, promoting the secretion of both SsaG and SsaI. Previous studies of PscG-PscE-PscF, YscG-YscE-YscF, and AscG-AscE-AscF complexes have shown that the chaperone-binding region of the needle protein is located in the C-terminal region (13,30,39). In the PscG-PscE-PscF complex, the C terminus of PscF is contained within the hydrophobic groove of PscG (39). Site-directed mutagenesis revealed that the hydrophobic residues Ile-83 and Leu-84, located within the last five C-terminal residues ("ILQKI") of AscF, are essential for AscG binding (30). The C-terminal sequence "IIAKI" of SsaG is also hydrophobic, and the Ile residue is conserved (Fig.  S2B). Another study demonstrated that the last five C-terminal residues of MxiH and PrgI, which are needle proteins in Shigella and SPI1, respectively, are essential for oligomerization. When these five residues were removed, these proteins could be expressed as soluble proteins but were unable to assemble (40). Therefore, the hydrophobic C-terminal region of SsaG is likely protected by the binding of SsaH to prevent oligomerization of the premature SsaG inside the cell, contributing to SPI2-T3SS assembly.
The amount of SsaH in presence of SsaG was decreased compared with that in the absence of SsaG (Fig. 6). The structure of

Chaperone-mediated secretion switching in SPI2-T3SS
the YscG-YscE-YscF complex (13) revealed that the N terminus of YscG is involved in the interaction between both YscG/ YscE and YscG/YscF. The 21 N-terminal residues of SsaH may be involved in the interaction with SsaE (Fig. S4D). The hydrophobic C-terminal region of SsaG likely binds to the N-terminal region of SsaH. The binding activity of SsaH may change in the presence of SsaG because a portion of the region of SsaH that binds to SsaG and SsaE is overlapped. The present results suggest that SsaE may be involved in the stabilization of SsaH via a direct interaction (Figs. 4 and 5A). In P. aeruginosa, PscE and PscG appear to costabilize each other (14). Reportedly, SsaE interacts directly with SseB and regulates its secretion via SPI2-T3SS (28). The secretion of SsaI and SseB by the ssaP-disrupted mutant strain indicates that SsaI secretion is negatively regulated by SsaP prior to SseB secretion (Fig.  1). This suggests that SsaE may function only as a chaperone for the early substrate SsaG through SsaH stabilization but not as a chaperone for the middle substrate SseB. Interestingly, overexpression of GST-SsaE in the ssaE-disrupted mutant restored the production of both SsaH and SseB (Fig. S5), but neither SsaI nor SseB was secreted (data not shown). The chaperonesubstrate complex is recruited to the sorting platform (34). Therefore, SsaE may recruit SsaG and SseB to the secretion apparatus of SPI2-T3SS, facilitating the secretion of early and middle substrates.
Recently, Souza et al. (41) showed that YscE of Y. pestis may play a direct role in the secretion of YscF by mediating contact with the T3SS apparatus. SsaE interacts with SsaK, SsaN, and SsaQ in SPI2, and binding of SseB to SsaE is abrogated by the ATPase SsaN (27,28). SsaK and SsaQ are homologs of OrgB and SpaO, respectively, in SPI1-T3SS (1), and SpaO, OrgB, and OrgA comprise the sorting platform in SPI1-T3SS. The SpaO-OrgA-OrgB complex ensures secretion of translocators before effectors (34). This platform may also contribute to substrate selection during needle assembly as well as to the substructure of the needle complex (42). In this process, the binding of the chaperone to its substrate is required for targeting of type III secreted proteins to the SpaO-OrgA-OrgB sorting platform (34). The hierarchy of secretion may reflect different affinities of the different substrate-chaperone complexes for the sorting platform. Thus, this suggests a possible hierarchy of secretion from the early substrates SsaG and SsaI to the middle substrate SseB in the SPI2-T3SS as the affinity of SsaE in the SsaE-SsaH-SsaG and SsaE-SseB complexes to the sorting platform differed even though the common chaperone SsaE interacted directly with the sorting platform in both complexes.
In addition to the sorting platform, substrate switching from early to middle substrates in the T3SS assembly depends on the termination of the needle complex assembly, regulated by the needle-length ruler SctP (5,9,(43)(44)(45). SsaP, the homolog of SctP, is also required for substrate switching from SsaI to SseB secretion (Fig. 1). Although SctP secretion in other T3SSs is known to be involved in regulating needle length, no SsaP was detected in the fraction of secreted proteins from the S. enterica Typhimurium WT strain (data not shown). Interestingly, SsaP interacted with GST-SsaE (Fig. S5). SsaH production and SsaI secretion were complemented in the ssaE-disrupted mutant when ssaE was expressed from a plasmid, although only some SseB was secreted (Fig. 5). These results raise the possibility of SsaE involvement in the titration of SsaP in bacterial cells to regulate substrate switching. EscP in EPEC has been reported to interact with the multieffector chaperone CesT, suggesting that EscP may prevent effector secretion until the translocation pore has been formed in the host cell membrane (6).
Considering our findings and previous studies together, we propose a model for chaperone function in the substrate switching of SPI2 (Fig. 7). Until the assembly of outer and inner membrane rings and export apparatus, including the sorting platform (46 -48), SsaE binds to the SsaG-SsaH complex to stabilize SsaG and SsaH in the bacterial cytosol. SsaE also forms a complex with SseB for stabilization. Following the assembly of the sorting platform, SsaG is recognized by the sorting platform via an interaction with SsaE in the SsaE-SsaG-SsaH complex, and secretion is initiated (Fig. 7A). Simultaneously, SsaI is also recognized and secreted regardless of binding to SsaE and SsaH. During the assembly of the needle and the inner rod, the binding of SsaP to SsaE likely occurs to regulate recognition of the SsaE-SseB complex by the sorting platform (Fig. 7B). Once SsaG and SsaI polymerization is complete, SsaP promotes substrate switching to the translocators (Fig. 7C). This initiates interaction between SsaE in the SsaE-SseB complex and the sorting platform, resulting in the secretion of translocators (Fig.  7D). Finally, following the formation of the translocation pore, the membrane-bound regulatory complex consisting of SsaL, SsaM, and SsaB is degraded upon sensing change in host cytosolic pH, which initiates the switch from translocators to effectors (Fig. 7E) (25). To our knowledge, this is the first report of a chaperone that regulates both early and middle substrates affecting substrate switching in the T3SS assembly. Class V chaperones such as SsaE have been identified in other T3SSs of pathogenic bacteria. Thus, the regulation of both early and middle substrates by class V chaperones affecting substrate switching demonstrated here may provide insight into the mechanism of secretion switching from early to middle substrates in other T3SSs of pathogenic bacteria. However, it is unclear how the inner rod protein is secreted at the same time as the needle protein. The inner rod controls switching from early to middle substrates and needle length (3,9,49). Cao et al. (36) revealed that interaction between the inner rod protein YscI and the needle protein YscF is required to assemble the needle structure of the Yersinia T3SS. YscI is necessary for the needle tip protein LcrV. They demonstrated that direct interaction between YscF and YscI is critical for this process. Similarly, it is assumed that the secretion of SseB may require an interaction between SsaG and SsaI. Additional studies regarding the function and regulation of the inner rod are felt to be required for a better understanding of substrate switching.

Bacterial strains, plasmids, and DNA oligonucleotides
All Salmonella strains were derivatives of S. enterica serovar Typhimurium 3306 (50). Bacteria were routinely grown in L
Prior to construction of the plasmids pTKY1251 and pTKY1261 encoding ssaP and ssaE, respectively, pTKY1231 encoding ssaB and its endogenous promoter was constructed by PCR amplification of the KpnI-SphI fragment carrying ssaB, including the 359 bp upstream, with the primer set pMW119-ssaB-KpnI-F and pMW119-ssaB-SphI-R. The fragment was then cloned into pMW119. Using the resultant plasmid, pTKY1231, as a template, a vector including ssaB promoter was amplified with the primer set pMW119-ssaB-XbaI-F and pMW119-ssaB-BamHI-R. To construct pTKY1251, the BamHI-XbaI fragment of ssaP, amplified with the primer set ssaP-BamHI-F and ssaP-XbaI-R, was ligated into the vector. The EcoRI-HindIII fragment of ssaP with ssaB promoter was digested from the resultant plasmid and then ligated into pMW118. To construct pTKY1261, the BglII-SphI fragment of ssaE, amplified with the primer set BglII-SD-ssaE-F and ssaE-SphI-R, was ligated into the vector. To construct plasmid pTKY1263 for expression of SsaE protein fused to GST, ssaE was amplified via PCR with the primer set GST-SsaE-BglII-F and GST-SsaE-EcoRI-R, and the BglII-EcoRI fragment was cloned into BamHI-EcoRI-digested pGEX6p-1.

Purification of N-terminally His-tagged proteins and generation of anti-SsaH and anti-SsaI antibodies
For purification of N-terminally His-tagged SsaH⌬N and SsaI, 3-liter cultures of E. coli strains CS6234 and CS6235 were incubated at 37°C until the cell density reached an A 600 of 1.0. Next, IPTG was added to a final concentration of 1 mM for 3 h followed by collection of cells via centrifugation. Wet cell pastes were resuspended in buffer (100 mM sodium phosphate buffer, pH 8.0, 0.1 M NaCl, 8 M urea) and sonicated. After further centrifugation, the supernatant was loaded onto a Ni 2ϩ -nitrilotriacetic acid Superflow column (10 ml; Millipore). After washing with buffer (100 mM sodium phosphate buffer, pH 8.0, 0.1 M NaCl, 8 M urea, 20 mM imidazole), the protein was eluted with 50, 75, 100, 150, 250, and 500 mM imidazole. A 5.0-ml portion of the peak fraction was electrophoresed by 15% SDS-PAGE, and the eluent corresponding to SsaH⌬N and SsaI bands was used to immunize a rabbit. The resultant anti-SsaH and -SsaI antibodies were purified from antisera using a Melon TM Gel IgG spin purification kit (Thermo Fisher Scientific).

Preparation of cell lysates and proteins secreted into the medium
To prepare cell lysates, bacterial cells were harvested by centrifuging 1 ml of the culture and suspended in SDS sample buffer. The suspension was incubated at 95°C for 5 min, sonicated, and recentrifuged. A portion of the supernatant was used as cell lysates. To prepare the proteins secreted into the medium, the same culture was centrifuged to remove cells. The filtrated supernatant was mixed with prechilled TCA (final concentration, 10%), chilled on ice, and centrifuged. The resultant pellet was washed once with acetone and then solubilized in SDS sample buffer.

GST pulldown assay
GST pulldown assay was performed using the MagneGST TM Protein Purification System (Promega) according to the manufacturer's instructions.

Preparation of SsaE-SsaH complex and SsaE protein
E. coli strains CS7141 and CS7145 were grown at 37°C to an A 600 of 0.5 in L broth containing 50 g/ml ampicillin and 0.5% glucose before adding IPTG to 1 mM to induce GST-SsaE and SsaH. Following a 3-h incubation at 37°C, cells were pelleted and lysed in B-PER Bacterial Protein Extraction Reagent buffer (Thermo Fisher Scientific) containing DNase I. Following centrifugation, the supernatant was added to MagneGST particles (Promega) equilibrated with reaction buffer (50 mM Tris-HCl, pH7.0, 150 mM NaCl, 1 mM EDTA, 1 mM DTT) and then rotated at 4°C for 1 h. After washing with reaction buffer, the particles were incubated for 16 h at 4°C in reaction buffer containing PreScission protease (GE Healthcare) to cleave the GST tag from the SsaE protein. The proteins were separated by gel chromatography (Superose TM 12 10/300, GE Healthcare) with reaction buffer containing 10% glycerol. The eluents were analyzed by gel electrophoresis (Fig. S6). The peak fractions containing SsaE at 14.5 ml and SsaE/SsaH at 14 ml, respectively, were concentrated using VivaSpin 500 with 3000 molecular weight cutoff (GE Healthcare). For stoichiometric analysis, a total of 12 g of each of SsaE/SsaH and SsaE was subjected to size-exclusion chromatography.

Separation of SPI2 proteins by gel electrophoresis and immunoblotting
Separation of low-molecular-weight SPI2 proteins was carried out using the Invitrogen Novex TM Tricine gel systems (Thermo Fisher Scientific). Protein samples were mixed with Tricine SDS sample buffer (Thermo Fisher Scientific), and proteins were electrophoresed using Novex 16% Tricine protein gels (Thermo Fisher Scientific). Precision Plus Protein TM WesternC TM Standards (Bio-Rad) or PageRuler Unstained Low Range Protein Ladder (Thermo Fisher Scientific) was used as protein standards. Separated proteins were stained with Oriole TM fluorescent gel stain (Bio-Rad) and visualized using a Gel Doc TM Plus system (Bio-Rad). For immunoblotting, separated proteins were transferred onto Immun-Blot polyvinylidene difluoride membranes (Bio-Rad) and incubated with rabbit anti-SsaI, anti-SsaH, anti-SsaP (22), and anti-SseB (22) antibodies followed by HRP-conjugated anti-rabbit IgG (Jackson ImmunoResearch Laboratories). Precision Protein TM Streptactin-HRP complex (Bio-Rad) was used for detection of protein standards. Enzymatic reactions were carried out in the presence of ECL Prime (GE Healthcare), and the resultant reaction was visualized via LAS4000mini

Chaperone-mediated secretion switching in SPI2-T3SS
(GE Healthcare). All results were based on at least three independent experiments.
Author contributions-A. T. and T. Y. conceptualization; A. T. and H. T. resources; A. T., H. T., and S. T. data curation; A. T., H. T., and S. T. formal analysis; A. T. funding acquisition; A. T. and T. Y. validation; A. T. investigation; A. T. writing-original draft; A. T. and T. Y. project administration; H. K. and T. Y. writing-review and editing; T. Y. supervision.