Controlled Activity of the Salmonella Invasion-Associated Injectisome Reveals Its Intracellular Role in the Cytosolic Population

ABSTRACT The Salmonella invasion-associated type III secretion system (T3SS1) is an essential virulence factor required for entry into nonphagocytic cells and consequent uptake into a Salmonella-containing vacuole (SCV). While Salmonella is typically regarded as a vacuolar pathogen, a subset of bacteria escape from the SCV in epithelial cells and eventually hyperreplicate in the cytosol. T3SS1 is downregulated following bacterial entry into mammalian cells, but cytosolic Salmonella cells are T3SS1 induced, suggesting prolonged or resurgent activity of T3SS1 in this population. In order to investigate the postinternalization contributions of T3SS1 to the Salmonella infectious cycle in epithelial cells, we bypassed its requirement for bacterial entry by tagging the T3SS1-energizing ATPase InvC at the C terminus with peptides that are recognized by bacterial tail-specific proteases. This caused a dramatic increase in InvC turnover which rendered even assembled injectisomes inactive. Bacterial strains conditionally expressing these unstable InvC variants were proficient for invasion but underwent rapid and sustained intracellular inactivation of T3SS1 activity when InvC expression ceased. This allowed us to directly implicate T3SS1 activity in cytosolic colonization and bacterial egress. We subsequently identified two T3SS1-delivered effectors, SopB and SipA, that are required for efficient colonization of the epithelial cell cytosol. Overall, our findings support a multifaceted, postinvasion role for T3SS1 and its effectors in defining the cytosolic population of intracellular Salmonella.

hampered by its requirement for bacterial internalization, i.e., T3SS1 mutants are invasion defective. Previously, two methods have typically been used to bypass this T3SS1-mediated entry requirement. The first involves rerouting internalization through T3SS1-independent mechanisms, either by using nonphagocytic cell lines that are relatively permissive for T3SS1-independent entry (8,20) or via surface expression of invasin from Yersinia pseudotuberculosis (21,24,25), which forces an S. Typhimurium T3SS1 mutant to utilize a "zipper" mechanism of entry (12). The second method relies on cointernalization of T3SS1 mutants with wild-type S. Typhimurium (25). Collectively, these studies have suggested a role for T3SS1 in both biogenesis and lysis of the nascent SCV (20,21,25). Given our recent finding that the T3SS translocator proteins affect early trafficking events of bacteria-containing vacuoles (8), here we wished to examine the postinvasion contributions of T3SS1 while maintaining the native T3SS1dependent trigger mode of bacterial internalization and biogenesis of the nascent SCV. We targeted a number of T3SS1-associated genes and found that one was particularly amenable to absolute and continual intracellular inactivation: the T3SS1-energizing ATPase InvC. When InvC was genetically engineered to have a C-terminal peptide tag that dramatically shortened its half-life and when its expression was placed under the control of an anhydrotetracycline (ATc)-inducible promoter, T3SS1 activity was robust in the presence of ATc but lacking in the absence of inducer. This allowed us to directly pinpoint new intracellular processes that are dependent on T3SS1 activity, including cytosolic proliferation and subsequent bacterial egress. We further identified two T3SS1 effectors, SopB and SipA, which contribute to bacterial proliferation in the epithelial cell cytosol. Collectively, this work demonstrates that the invasion-associated T3SS also contributes to essential intracellular stages of the S. enterica infectious cycle.

RESULTS
Establishing a system for conditional activation of T3SS1. Our goal was to develop a conditional system whereby extracellular S. Typhimurium would have a functional T3SS1 to allow for bacterial invasion but in which T3SS1 activity would be rapidly and perpetually disabled following mammalian cell entry. We considered the genes for four different T3SS1-associated proteins as potential targets for genetic manipulation: (i) HilE, a negative regulator of SPI1-encoded gene transcription (26,27), which includes T3SS1 structural components and effectors, (ii) SipB and SipC, the two translocator proteins that form the T3SS translocon pore (28), and (iii) InvC, the energy-generating ATPase (29).
We first study a conditional mutant for which transcription of genes encoding structural components of T3SS1 and T3SS1-associated effectors could be modulated. HilE is the major negative regulator of SPI-1 gene expression. It interacts with HilD to inhibit transcription of HilA, a transcriptional activator that has a central role in modulating the expression of T3SS1 apparatus and T3SS1-associated effector genes (26). Overexpression of hilE represses hilA expression and drastically reduces bacterial invasion into nonphagocytic cells, whereas hilE mutants are derepressed for hilA transcription and exhibit a hyperinvasive phenotype (26) (Fig. S2A). We therefore reasoned that modulating hilE expression would allow us to temporally control T3SS1 activity. hilE under the control of the ATc-inducible tetA promoter (P tetA ) (30) was inserted between purA and yjeB-nsrR in the chromosome of an S. Typhimurium ΔhilE mutant (ΔhilE tetRA-hilE). Addition of ATc to the bacterial subculture decreased the invasion efficiency (the percentage of the infective inoculum that was internalized at 1 h p.i.) of the ΔhilE tetRA-hilE strain to levels similar to those of a prgI mutant, indicating an SPI-1 deficiency (Fig. S2A). In the absence of ATc induction, the ΔhilE tetRA-hilE strain also showed decreased invasion compared to ΔhilE bacteria, suggesting leaky expression of hilE. Notably, intracellular induction of hilE expression with 300 ng/ml ATc, a concentration that is sufficient to induce gene expression in vacuolar and cytosolic Salmonella in epithelial cells (20), did not affect intracellular T3SS1 activity (SipC translocation) or bacterial proliferation. Specifically, the proportions of infected cells positive for translocated SipC (Fig. S2B) and vacuolar and cytosolic proliferation ( Fig. S2C) were indistinguishable for cells infected with ΔhilE tetRA-hilE bacteria incubated in the presence or absence of ATc.
Next we attempted to conditionally control T3SS1 by placing sipB or sipC under the control of the araBAD (P BAD ) promoter from Escherichia coli (31). We recently showed that the invasion defect of an S. Typhimurium ΔsipB or ΔsipC mutant could be complemented in trans with sipB or sipC, respectively (32), when under the control of this promoter. A similar arabinose-inducible system has been used previously to investigate the intracellular role of the S. flexneri translocator, IpaB (33), the SipB ortholog. To test the validity of arabinose-controllable expression of the T3SS1 translocator proteins, S. Typhimurium ΔsipB and ΔsipC mutants were transformed with pBAD30 (empty vector) and pBAD30-sipB or pBAD30-sipC constructs and tested for their ability to invade HeLa epithelial cells when grown in the presence or absence of arabinose. Overall, invasion efficiencies indicated that placing expression of sipC, but not of sipB, under the control of the P BAD promoter is a viable option for tightly controlled, transient activation of T3SS1 activity (Fig. S2D). However, when we compared intracellular T3SS1 activity, while no SipCpositive cells were detected for ΔsipC pBAD30-sipC bacteria in the absence of arabinose at 4 h and 8 h p.i. (Fig. S2E), vacuolar and cytosolic replication were unaffected at 8 h p.i. (Fig. S2F). SPI1 gene expression is downregulated inside cells, yet T3SS1 can remain functional for long periods if the apparatus is relatively stable once assembled. Such T3SS1 stability might explain why modulation of hilE and sipC expression did not have any overt effects on intracellular Salmonella (Fig. S2). We therefore tested a system in which T3SS1 activity was conditionally inactivated via manipulation of the stability of the T3SS1-energizing ATPase InvC. invC is essential for Salmonella invasion of nonphagocytic cells (34) and encodes the ATPase which energizes T3SS1 and facilitates unfolding of proteins for export (29,34). Our rationale was that T3SS1 would be nonfunctional without ATP generation, i.e., it would be energetically "dead," even if it remained assembled. We used Tn7 transposon site-specific integration at the attTn7 locus, located downstream of the glmS gene (35), to chromosomally complement an S. Typhimurium ⌬invC mutant with invC under the control of P tetA , thereby allowing ATc-tunable invC expression. A 3ϫ FLAG epitope was also attached to the carboxyl terminus of InvC to facilitate its immunodetection (ΔinvC glmS::tetRA-invC-FLAG). To alter protein stability, we further incorporated one of three peptides recognized by bacterial tail-specific proteases-AANDENYA(ASV), AANDENYA(AAV), and AANDENYA(LVA) (36)-to the C terminus of InvC-FLAG. We tested the ability of these strains to enter HeLa epithelial cells when grown in the absence or presence of ATc. ATc induction fully restored invasion by ΔinvC glmS::tetRA-invC-FLAG bacteria (Fig. 1A), indicating that 3ϫ FLAG tag addition to the C terminus of InvC did not negatively affect its biological function and that ATc addition fully derepressed TetR binding to P tetA . However, these bacteria exhibited a reduced, yet significant, invasion efficiency in the absence of inducer, suggestive of leaky invC transcription under these growth conditions (Fig. 1A). Addition of C-terminal peptides to InvC differentially affected their ability to complement ΔinvC bacteria for invasion: ΔinvC bacteria producing InvC-FLAG(ASV) were the most competent for invasion when grown in the presence of ATc, and InvC-FLAG(LVA) was the least competent (Fig. 1A). Furthermore, under noninducing conditions, ΔinvC glmS::tetRA-invC-FLAG(AAV) and ΔinvC glmS::tetRA-invC-FLAG(LVA) strains exhibited invasion efficiencies equivalent to that of ΔinvC bacteria (Fig. 1A), indicating negligible InvC function. Collectively, these results identified at least two strains for which InvC function is tightly controlled by ATc addition to bacterial subcultures [ΔinvC glmS::tetRA-invC-FLAG(AAV) and ΔinvC glmS:: tetRA-invC-FLAG(LVA)], and these strains therefore are potential candidates for examining the intracellular effects of an inoperative T3SS1.
InvC is destabilized by peptide tail addition. The destabilizing effect of adding peptide tail variants has been assessed for green fluorescent protein (GFPmut3*) tagging in E. coli (36). Addition of these peptides to GFPmut3* dramatically shortens its half-life from Ͼ24 h to 110 min (ASV), 60 min (AAV), or 40 min (LVA) (36). Since our invC is under the control of P tetA and was integrated by using a Tn7 transposon at the site-specific attTn7, located downstream of the glmS gene (SL1344_3828). Invasion efficiency of S. Typhimurium wild type (WT), ΔinvC, ΔinvC tetRA-invC-FLAG (InvC), ΔinvC tetRA-invC-FLAG(ASV), ΔinvC tetRA-invC-FLAG(AAV) and ΔinvC tetRA-invC-FLAG(LVA). Where indicated, subcultures were grown in the absence or presence of 100 ng/ml ATc. Invasion efficiency (the percentage of inoculum internalized) was determined in a gentamicin protection assay in HeLa cells at 1 h p.i.; results are expressed as a percentage of the invasion efficiency for wild-type bacteria (set to 100%). Data are from Ն3 independent experiments (means Ϯ SD). Asterisks indicate data significantly different from WT. (B, C, and D) InvC stability in broth. ΔinvC tetRA-invC-FLAG (InvC), ΔinvC tetRA-invC-FLAG(ASV), ΔinvC tetRA-invC-FLAG(AAV), and ΔinvC tetRA-invC-FLAG(LVA) bacterial subcultures were grown in the presence of 100 ng/ml ATc for 1 h to induce the synthesis of InvC-FLAG. ATc was washed out, and chloramphenicol (30 g/ml) was added to stop de novo protein synthesis (t0). Aliquots were removed at the indicated times and processed for immunoblotting with anti-FLAG antibodies (␣-FLAG). Similar exposure times were used for chemiluminescent detection. Graphs depict densitometric analysis results (means Ϯ SD; 3 independent experiments). (E) InvC stability in HeLa epithelial cells. Bacterial strains (those shown in panel A) were grown in the presence of 100 ng/ml ATc for 1.5 h prior to infection to induce InvC-FLAG production. MOIs were adjusted so that equivalent numbers of bacteria were internalized into (Continued on next page) Controlled Intracellular Activity of Salmonella T3SS1 ® invasion data suggested that addition of each of these peptide tags differentially affected InvC turnover in Salmonella (Fig. 1A), we used a modified pulse-chase experiment to directly compare the half-lives of InvC-FLAG, InvC-FLAG(ASV), InvC-FLAG(AAV), and InvC-FLAG(LVA). For quantification of InvC variant half-lives, invC expression was induced by the addition of ATc to bacterial subcultures for 1 h. Bacteria were then washed to remove ATc and treated with chloramphenicol (time 0 [t0]) to prevent de novo protein synthesis. Bacterial lysates were collected over a time course and assessed for InvC-FLAG levels by immunoblotting (Fig. 1B, C, and D). We calculated the half-life of InvC-FLAG to be 55 min in S. Typhimurium; the addition of AANDENYA(ASV), AANDENYA(AAV), or AANDENYA(LVA) peptides to InvC reduced its half-life to 20 min, 3 min, and 1 min, respectively (Fig. 1B, C, D, and F). Therefore, these peptide sequences have a destabilizing effect on S. Typhimurium proteins, as observed for E. coli. Because we were aiming to rapidly inactivate T3SS1 energetics after bacterial internalization, we also compared the turnover of InvC and destabilized variants of InvC inside mammalian cells. HeLa epithelial cells were infected with ATc-induced bacterial subcultures, and at 30 min p.i., chloramphenicol was added (t0). Whole-cell lysates were collected at various time points, and InvC-FLAG levels were analyzed by immunoblotting (Fig. 1D). Interestingly, InvC-FLAG and its variants were more stable in intracellular S. Typhimurium than in liquid broth. InvC-FLAG and InvC-FLAG(ASV) had comparable half-lives for intracellular S. Typhimurium, 140 min and 160 min, respectively ( Fig. 1E and F), whereas the half-life of InvC-FLAG(AAV) was considerably shorter (40 min) ( Fig. 1E and F). We were unable to detect a band corresponding to the predicted size of InvC-FLAG(LVA) at t0 (30 min p.i.) (Fig. 1E), presumably due to its extremely rapid turnover, which precluded us from calculating its half-life. Our conservative estimate of the half-life of InvC-FLAG(LVA) is Ͻ10 min inside mammalian cells. Due to their accelerated turnover compared to InvC-FLAG, both InvC-FLAG(AAV) and InvC-FLAG(LVA) were further considered to effect inactivation of T3SS1 inside mammalian cells, a requirement for assessing its intracellular role. We also considered it advantageous to use multiple destabilized variants, since their differential stabilities allowed us to study the effects of "tuning" InvC production, akin to a dose-response curve for T3SS1 activity.
T3SS1 activity is diminished for strains producing InvC-FLAG(AAV) and InvC-FLAG(LVA). To examine the effect of InvC destabilization on T3SS1, we screened for the temporal abundance of T3SS1 components in liquid broth and infected cells. Whole-cell lysates from bacterial subcultures (induced with ATc where appropriate) were subject to immunoblotting with anti-FLAG, anti-SipB, anti-SipC, anti-SipD, anti-SipA, and anti-DnaK (bacterial loading control) antibodies. In agreement with the calculated half-lives of the InvC variants ( Fig. 1F), the steady-state levels of InvC-FLAG decreased as the stability decreased ( Fig. 2A). Levels of SipB, SipC, SipD, and SipA were comparable for all tested strains ( Fig. 2A), indicating that there was no feedback response on the stability of other T3SS1 components or T3SS1-associated effectors for broth-grown bacteria.

FIG 1 Legend (Continued)
HeLa epithelial cells for all strains. At 30 min p.i., chloramphenicol (60 g/ml) was added, then lysates were collected at the indicated times and processed for immunoblotting with anti-FLAG antibodies. Similar exposure times were used for chemiluminescence detection. The graph depicts densitometric analysis results (means Ϯ SD; 3 independent experiments). (F) Summary of estimated half-lives for InvC-FLAG and its destabilized variants. ND, not determined.
anti-SipB antibodies. Loading was normalized for equivalent CFU per lane for each time point. There was a progressive reduction in intracellular InvC-FLAG levels at both 1 h and 8 h p.i., as protein stability decreased (Fig. 2C), i.e., the temporal profiles for InvC-FLAG and InvC-FLAG(ASV) were equivalent, whereas InvC-FLAG(AAV) and InvC-FLAG(LVA) were moderately and drastically reduced, respectively. Intracellular SipB levels for all strains were equivalent to those in wild-type bacteria at 1 h p.i. and 8 h p.i., except for ΔinvC glmS::tetRA-invC-FLAG(LVA) bacteria, which had considerably lower levels at 8 h p.i. than observed for all other strains (Fig. 2C). Collectively, these data indicate that, at later times during infection, InvC abundance is affected for the InvC-FLAG(ASV) strain and for both InvC and SipB abundance levels for the InvC-FLAG(LVA) strain.
We then compared intracellular T3SS1 activity for these strains via immunostaining and adenylate cyclase (CyaA) assays. For immunostaining, bacteria harbored pFPV-mCherry, a plasmid that carries mCherry under the control of the constitutive S. Typhimurium rpsM promoter (37), to facilitate fluorescence detection. Subcultures were induced with ATc where appropriate, and infected HeLa cells were processed for SipC immunodetection at 30 min, 4 h, and 8 h p.i. SipC is a T3SS1 translocator protein and HeLa epithelial cells were infected as described for panel B with strains harboring a SopB-CyaA plasmid. Lysates were collected at 0.5 h and 8 h p.i. and subjected to an ELISA for cAMP quantification (total cAMP/well). For reference, cAMP levels in WT bacteria-infected lysates (i.e., no SopB-CyaA plasmid) were 0.24 Ϯ 0.15 pmol/well and 0.10 Ϯ 0.044 pmol/well at 0.5 h and 8 h p.i., respectively. An asterisk denotes a result significantly different from that of WT SopB-CyaA bacteria.
Identification of intracellular activities associated with T3SS1 activity. We used these strains with tunable T3SS1 activity-InvC-FLAG, InvC-FLAG(AAV) and InvC-FLAG(LVA)-to dissect the contribution of T3SS1 to specific stages in the intracellular lifecycle of Salmonella. We precluded the InvC-FLAG(ASV) strain from further analysis because it was indistinguishable from the InvC-FLAG strain in all of the preceding assays ( Fig. 2 and 3). We first used single-cell analysis to monitor bacterial proliferation in epithelial cells by fluorescence microscopy. S. Typhimurium strains carried pFPV-mCherry, and MOIs were adjusted so that equivalent numbers of bacteria were internalized into HeLa epithelial cells. At 1 h p.i., the number of bacteria/cell was comparable for all S. Typhimurium strains, ranging from an average of 2.3 to 3.1 bacteria/cell (Fig. 3A). Wild-type and ΔinvC glmS::tetRA-invC-FLAG bacteria showed similar vacuolar replication at 4 h and 8 h p.i. (Fig. 3A), respectively, and also for the proportion of infected cells containing cytosolic bacteria (Ն100 bacteria) (20,43,44) at 8 h p.i. (6.6% Ϯ 2.4% and 6.3% Ϯ 2.0%) (data are means Ϯ standard deviations [SD]) (Fig. 3A), in agreement with the results from the gentamicin protection assays (Fig. 2B). While vacuolar replication for ΔinvC glmS::tetRA-invC-FLAG(AAV) bacteria was unaffected, the proportion of cells with cytosolic bacteria was significantly reduced at 8 h p.i. (3.3% Ϯ 1.2%) (Fig. 3A). For ΔinvC glmS::tetRA-invC-FLAG(LVA) bacteria, only 0.26% Ϯ 0.58% of infected HeLa cells contained cytosolic Salmonella at 8 h p.i., and the mean number of vacuolar bacteria/cell was also significantly reduced (Fig. 3A), indicating compromised cytosolic and vacuolar proliferation for this strain, consistent with the gentamicin protection assay results (Fig. 2B).
We next used a population-based assay to quantify the proportion of the total bacterial population residing in the cytosol. The chloroquine (CHQ) resistance assay relies on the lysosomotropic properties of CHQ, which accumulates only within the SCV and leads to the killing of vacuolar S. Typhimurium by an as-yet-unidentified mechanism but leaves the cytosolic population unharmed. In conjunction with a gentamicin protection assay, this determines the percentage of internalized bacteria that populate the cytosol at a given time (20). HeLa cells were infected with S. Typhimurium wild type, ΔinvC glmS::tetRA-invC-FLAG, ΔinvC glmS::tetRA-invC-FLAG(AAV), and ΔinvC glmS::tetRA-invC-FLAG(LVA) bacteria such that equivalent numbers were initially internalized, and the proportion of cytosolic bacteria was determined at 1.5 h and 7 h p.i. At 1.5 h p.i., this proportion was equivalent for S. Typhimurium wild type, ΔinvC glmS::tetRA-invC-FLAG, and ΔinvC glmS::tetRA-invC-FLAG(AAV) bacteria, whereas it was significantly reduced for ΔinvC glmS::tetRA-invC-FLAG(LVA) bacteria (Fig. 3B). By 7 h p.i., the phenotype of ΔinvC glmS::tetRA-invC-FLAG(LVA) bacteria was much more dramatic, with only 3.9% Ϯ 1.1% of the total population present in the cytosol, compared to 46% Ϯ 5.7% and 47% Ϯ 11% for wild-type and ΔinvC glmS::tetRA-invC-FLAG bacteria, respectively (Fig. 3B). The ΔinvC glmS::tetRA-invC-FLAG(AAV) bacteria were also defective for cytosolic colonization at 7 h p.i. (29% Ϯ 11%) (Fig. 3B), which is consistent with the single-cell analysis results (Fig. 3A). Importantly, the cytosolic colonization defects for InvC-FLAG(AAV) and InvC-FLAG(LVA) strains were recapitulated in other epithelial cell lines (Fig. S1). Altogether, these results uniformly demonstrated that bacteria producing InvC-FLAG(LVA), the most destabilized variant of InvC, are prominently defective for lysis of the nascent vacuole, which subsequently affects the frequency of cytosolic

S. Typhimurium ΔinvC tetRA-invC-FLAG (InvC), ΔinvC tetRA-invC-FLAG(AAV), and ΔinvC tetRA-invC-FLAG(LVA)
bacterial subcultures were grown in the presence of 100 ng/ml ATc for 1.5 h prior to infection to induce the synthesis of InvC-FLAG. HeLa epithelial cells were infected with the above strains and WT bacteria (harboring pFPV-mCherry). MOIs were adjusted so that equivalent numbers of bacteria were internalized. Monolayers were fixed at 1 h, 4 h, and 8 h p.i., and the number of bacteria in each infected cell was blindly scored by fluorescence microscopy. Cells with Ն100 bacteria contained cytosolic S. Typhimurium. Each dot represents one infected cell, and data from 3 independent experiments are shown (Ͼ180 cells in total). Values at the bottom of the panel indicate the mean number of bacteria per cell for cells containing Ͻ99 bacteria/cell (i.e., cells with Ն100 bacteria were excluded from the statistical analysis). Asterisks at the top and bottom of the panel indicate cytosolic and vacuolar replication that were significantly different from WT bacteria, respectively. (B) CHQ resistance assay. Subcultures were grown as described for panel A and used to infect HeLa epithelial cells. The proportion of cytosolic bacteria was determined at 1.5 h and 7 h p.i. in a CHQ resistance assay. Asterisks denote results that were significantly different from those of WT bacteria. (C) Bacterial egress. HeLa epithelial cells were infected as described for panel B. At 7 h p.i., gentamicin-containing growth medium was replaced with gentamicin-free medium for 1 h, and the number of extracellular bacteria was determined by agar plating. Each point is from an individual well in a 24-well tissue culture plate, and data are from 5 independent experiments. Means are indicated by horizontal lines, and asterisks denote significantly different results from those with WT bacteria.
Controlled Intracellular Activity of Salmonella T3SS1 ® colonization. In contrast, bacteria producing InvC-FLAG(AAV), which is of intermediate stability, are not affected for nascent SCV lysis or vacuolar replication but are compromised for cytosolic growth at later times. This identifies a novel contribution of intracellular T3SS1 activity to the cytosolic proliferation of Salmonella which is distinct from its previously characterized role in nascent vacuole lysis and trafficking.
Finally, we tested the effects of destabilizing InvC on bacterial egress as a potential consequence of cytosolic replication and epithelial cell lysis (19). If cytosolic bacteria are the intracellular population that seed reinfection, we predicted a direct correlation between the number of cytosolic bacteria and the efficiency of their egress. To test this, HeLa epithelial cells were infected with the same strains as for the CHQ resistance assay (adjusting MOIs so that equivalent numbers of bacteria were internalized), and at 7 h p.i., gentamicin-containing medium was removed and replaced with gentamicin-free medium for 1 h, after which the number of extracellular bacteria was enumerated by agar plating. There was a progressive decrease in bacterial egress that correlated with InvC stability (Fig. 1F) and the proportion of cytosolic bacteria (Fig. 3A and B): InvC-FLAG Ͼ InvC-FLAG(ASV) Ͼ InvC-FLAG(LVA) (Fig. 3C). This specifically implicated cytosolic bacteria as the intracellular subset that is released from epithelial cells. Furthermore, all InvC-FLAG variants were affected for bacterial release compared to wild-type bacteria (Fig. 3C), even InvC-FLAG, which is indistinguishable from wild-type bacteria with respect to vacuolar and cytosolic replication at late times postinfection ( Fig. 3A and B). Because the InvC-FLAG strain is not capable of reinducing T3SS1 activity in the absence of the ATc inducer, this also identifies resurgent T3SS1 activity as being requisite for bacterial egress.
SopB and SipA are required for efficient cytosolic proliferation by Salmonella. Our results designated a newly appreciated role for intracellular T3SS1 activity with regard to cytosolic proliferation of Salmonella. T3SS1 effectors can be detected at later times (Ն8 h p.i.) in epithelial cells, albeit to a lesser extent than shortly after initial bacterial entry (Fig. 2) (23,(40)(41)(42)45). To investigate whether any of the known T3SS1 effectors are involved in bacterial colonization of the epithelial cell cytosol, HeLa cells were infected with S. Typhimurium wild-type, ΔsipA, ΔsopA, ΔsopB, ΔsopD, or ΔsopE ΔsopE2 bacteria, and the CHQ resistance assay was used to quantify the proportion of cytosolic bacteria in the total populations at 1.5 h and 7 h p.i. Compared to wild-type bacteria, no effector deletion mutants were compromised for nascent vacuole lysis at 1.5 h p.i., but two mutants had a significantly reduced proportion of cytosolic bacteria at 7 h p.i.: ΔsipA and ΔsopB (Fig. 4A). Notably, this defect was restored to wild type levels upon complementation with a single gene copy in the respective mutant (Fig. 4B). A ΔsipA ΔsopB double mutant was not more affected than either single deletion mutant at 7 h p.i. (Fig. 4A). Single-cell analysis corroborated that ΔsipA, ΔsopB and ΔsipA ΔsopB mutants were all affected for the proportion of infected cells containing cytosolic bacteria (Ն100 bacteria/cell; 9.5 Ϯ 1.5% for wild type, 6.0 Ϯ 1.6% for ΔsopB, 4.0 Ϯ 0.93% for ΔsipA, and 3.0 Ϯ 1.6% for ΔsipA ΔsopB bacteria) (Fig. 4C). ΔsipA and ΔsipA ΔsopB mutants were also compromised for replication in the SCV (Fig. 4C). Altogether, these data implicate the activities of SipA and SopB, two T3SS1-dependent effectors, as being required for the efficient cytosolic proliferation of Salmonella, likely via a common or converging mechanism.

DISCUSSION
Here, we have used a number of different approaches to affect T3SS1 inactivation dynamics, and in order to analyze the intracellular contributions of this invasionassociated injectisome to S. Typhimurium. One approach we chose was an inducible expression system whereby the conditional mutant acts like wild-type bacteria when in the presence of an inducer (extracellularly) and ideally like a T3SS1 mutant in the absence of inducer (inside mammalian cells). A similar tactic has been used before to investigate the postinvasion roles of the Shigella T3SS. By placing various S. flexneri genes encoding T3SS structural components under the control of an arabinoseinducible promoter, two studies have demonstrated that the S. flexneri Mxi-Spa T3SS is required for different aspects of cell-to-cell spread (33,46). Recently, Finn et al. (23) employed arabinose-inducible expression of S. Typhimurium invA, which encodes a T3SS1 structural protein, to generate a T3SS1-defective mutant in the intracellular environment and showed that this strain was compromised for SopB and SipA translocation into HeLa cells, but not in Caco-2 C2BBe1 cells, specifically by cytosolic S. Typhimurium. Inducible gene expression has also been exploited to define the temporal contribution of the T4SS to intracellular colonization. Using isopropyl-␤-Dthiogalactopyranoside (IPTG)-inducible expression of dotA, which encodes a Legionella pneumophila Dot/Icm T4SS structural component, Roy et al. showed that DotA was required for intracellular replication of L. pneumophila, but only at time points after 24 h p.i. (47). Similar conclusions were drawn using a Cre/loxP in vivo gene deletion system in L. pneumophila (48). Most recently, via ATc-regulatable expression of the Brucella abortus T4SS ATPase VirB11, Smith et al. demonstrated that bacterial egress (at 72 h p.i.) was dependent on the VirB T4SS (49). The premise of the inducible expression approach used in all these studies is that, in the absence of gene expression, HeLa cells were infected with S. Typhimurium WT or ΔsipA or ΔsopB deletion mutants and these deletion mutants complemented with a single gene copy via Tn7 integration at the attTn7 site. The proportion of cytosolic bacteria at 7 h p.i. was determined using the CHQ resistance assay. Asterisks indicate results significantly different from those for WT bacteria. (C) Single-cell analysis of intracellular proliferation. HeLa epithelial cells were infected with S. Typhimurium WT, ΔsopB, ΔsipA, and ΔsipA ΔsopB bacteria (harboring pFPV-mCherry). Monolayers were fixed at 1 h, 4 h, and 8 h p.i., and the number of bacteria in each infected cell was blindly scored by fluorescence microscopy. Cells with Ն100 bacteria contain cytosolic S. Typhimurium. Each dot represents one infected cell, and data from 3 independent experiments are shown (Ͼ180 cells in total). Values at the bottom of the panel indicate the mean number of bacteria per cell for cells containing Ͻ99 bacteria/cell (i.e., cells with Ն100 bacteria were excluded from the statistical analysis). Asterisks at the top and bottom of the panel indicate cytosolic and vacuolar replication, respectively, that was significantly different from that of WT bacteria.
Controlled Intracellular Activity of Salmonella T3SS1 ® there is sufficient protein turnover to intracellularly inactivate the secretion system being studied. However, several lines of evidence suggest this is not necessarily so for intracellular bacteria. For example, L. pneumophila IcmQ, a Dot/Icm T4SS structural protein, appears to be an extremely stable protein in infected macrophages (48). Furthermore, the half-life of the B. abortus VirB11 ATPase is considerably extended inside macrophages (80 min) compared to that in liquid broth (16 min) (49). Similarly, we have shown that the half-life of S. Typhimurium InvC, the ATPase for the invasionassociated injectisome, is increased when bacteria are inside epithelial cells (140 min) compared to bacteria in broth culture (55 min) (Fig. 1). Not all T3SS structural proteins are long-lived inside mammalian cells, however. While the half-life of SipC, one of the T3SS1 translocator proteins, is Ͻ10 min for broth-grown S. Typhimurium (50), it is 25 min inside epithelial cells (Fig. S3). SipB, the other T3SS1 translocator protein, and SipD, the T3SS1 tip complex protein, similarly have short half-lives in mammalian cells (Fig. S3). Altogether, these findings suggest that the T3SS basal body is a more stable structure than the T3SS needle-associated complexes. With such short half-lives, SipB, SipC, or SipD would seem logical targets to manipulate intracellular T3SS1 activity. However, the ΔsipC pBAD30-sipC strain lacked a phenotype in HeLa epithelial cells upon arabinose withdrawal ( Fig. S2B and C). This result could be interpreted as showing that T3SS1 has little impact upon the intracellular colonization of Salmonella. However, it was recently shown that SopB and SipA delivery by cytosolic Salmonella are independent of SipB (23). Taken together, we favor the explanation that T3SS1 delivery of effectors by cytosolic Salmonella is independent of SipB and SipC translocator proteins, perhaps more akin to a contact-independent secretion-like process that is triggered in broth culture.
We have overcome the technical limitation imposed by protein stability by the addition of C-terminal destabilizing peptide tags. By using tags that confer different protein stability, we can control how long the InvC ATPase persists and thereby "tune" intracellular T3SS1 activity. For S. Typhimurium producing the most destabilized version of InvC, InvC(LVA), we saw the most dramatic intracellular phenotypes. Compared to wild-type bacteria, this strain was defective for the translocation of SipC and SopB (Fig. 2), nascent vacuole lysis (Fig. 3), and cytosolic and vacuolar replication and bacterial egress (Fig. 3). Studies using surrogate-mediated entry of T3SS1 null mutants (cointernalization with wild-type bacteria or invasin-mediated entry) have led to the idea that T3SS1 damages some nascent vacuoles, which leads to the cytosolic access of bacteria within compromised SCVs, followed by their autophagy-dependent elimination or autophagy-promoted repair of the damaged SCV (21,51). Our findings with this T3SS1 "low-activity" mutant further advocate for the level of T3SS1 activity influencing the extent of nascent vacuole lysis. Upon invasin-mediated entry of an S. Typhimurium ΔinvA mutant into HeLa epithelial cells, this T3SS1 null mutant is defective for intracellular replication at 6 h p.i. compared to wild-type bacteria (25); whether this is due to a replication defect in the vacuole and/or cytosol was not investigated. Our T3SS1 low-activity strain clarifies that bacterial replication within both intracellular niches is T3SS1-dependent (Fig. 3A). S. Typhimurium with the intermediate destabilized InvC, InvC(AAV), was only affected for later intracellular events, i.e., cytosolic proliferation and bacterial egress (Fig. 3), which allowed us to assign a role for T3SS1 in the cytosolic proliferation of Salmonella that is independent of defects in nascent vacuole lysis and trafficking.
After internalization of S. Typhimurium into nonphagocytic cells, SPI1 gene expression and T3SS1 activity are downregulated (13,14). While numerous environmental cues have been identified that activate T3SS1, comparatively little is known about what down-modulates its activity, or how its function is terminated, once bacteria are inside mammalian cells. One study showed that the downregulation of hilA and hilD gene expression after bacterial entry into epithelial cells is retarded for an S. Typhimurium Δlon mutant (13). In support of this, HilC and HilD are stabilized in a Δlon mutant (52), although an independent study found only a minor effect of Lon on HilD stability (53). Therefore, whether Lon protease is involved in the postinvasion degradation of tran-scriptional regulators of SPI1 gene expression remains an open question. Interestingly, while HilE has been shown to negatively affect hilA transcription in vivo (26), a ΔhilE mutant is comparable to wild-type S. Typhimurium for the timing and extent of hilA and hilD downregulation in epithelial cells (13). This suggests that HilE is not responsible for turning off SPI1 gene expression intracellularly.
The InvC-FLAG strain, which is indistinguishable from wild-type bacteria with regard to the kinetics of downregulation of intracellular T3SS1 activity (Fig. 2) and load of cytosolic bacteria at later times (Fig. 3), is defective for bacterial release from epithelial cells. Since this strain is unable to reinduce intracellular T3SS1 activity in the absence of ATc, this suggests that resurgent T3SS1 activity is required for efficient bacterial egress (Fig. 3). While little is known about the conditions within the intracellular microenvironment that trigger the cessation of SPI1 gene expression, even less is known about how Salmonella reactivates SPI1 gene expression (19,20,23) and T3SS1 activity (Fig. 2) (23) once it reaches the epithelial cell cytosol. For example, is T3SS1 degraded in vacuolar bacteria or diluted by bacterial growth in the vacuole and then new T3SS1 complexes are synthesized once Salmonella has reached the cytosol? This would seem to be very energetically costly to a bacterium. A more likely scenario is that T3SS1 remains intact but T3SS1 type III effectors are no longer synthesized, or translocation of T3SS1-dependent effectors is somehow prevented and then reengaged when required. Since many bacteria rely on secretion systems to deliver effectors at specific doses and times during their infectious cycle, understanding how bacteria control the stability, turnover, and dynamics of these nanomachines are important questions to address in the future (5).
We recently identified three S. Typhimurium genes, corA, recA, and ydgT, that are required for efficient bacterial colonization of the epithelial cell cytosol (44). Here, we identified two more cytosolic colonization factors, SipA and SopB (Fig. 4). Unlike what we observed with the ΔsopB mutant (Fig. 4), Finn et al. found no difference between cytosolic replication of wild-type and ΔsopB bacteria in HeLa cells (23); the reason for this discrepancy remains unknown. Both SopB and SipA are type III effectors that continue to be synthesized and translocated by T3SS1 for many hours after bacterial internalization (23,(40)(41)(42)45). In the case of SopB, this type III effector localizes to the plasma membrane and nascent SCV shortly after bacterial internalization (40,41,54), on mature SCVs (40), and unidentified structures in cells containing cytosolic Salmonella cells (23). SopB is known to activate the mammalian prosurvival kinase Akt and thereby delay the onset of apoptosis in epithelial cells (55). Recently SopB was shown to specifically delay the onset of cell death in epithelial cells containing cytosolic Salmonella (23), which occurs by pyroptosis (56). It is not known how SopB might delay epithelial cell pyroptosis or whether the reported SopB-dependent 70-min delay in cell death requires Akt activation, as no direct connection between Akt and pyroptosis has been established. SipA is an actin-binding protein that promotes actin polymerization (57)(58)(59)(60)(61) and accumulates as discrete foci near bacteria shortly after their internalization (62,63), on the mature SCV (45), and as undefined puncta in cells harboring cytosolic Salmonella (23). As assessed in the gentamicin protection assay, a sipA deletion mutant is attenuated for total bacterial replication in NIH 3T3 fibroblasts (45). We have shown that SipA contributes to proficient replication in both the SCV and cytosol at later times (Fig. 4). This poses the question of how the activity of SipA contributes to bacterial proliferation in such distinct intracellular niches. SipA might alter membrane stability of the mature SCV due to changes in actin accumulation, as vacuoles surrounding ΔsipA and sipA ϩϩ bacteria (sipA ϩϩ bacteria express enhanced levels of SipA) are associated with poor and enriched F-actin recruitment at 6 h p.i. in fibroblasts, respectively, and sipA ϩϩ SCVs are inherently unstable (45).
In summary, here we have shown that newly appreciated key stages in the infection cycle of Salmonella, lysis of the internalization vacuole and subsequent hyperreplication in the cytosol of epithelial cells, are dependent on T3SS1 activity. While T3SS1 was originally designated the "invasion-associated injectisome," based on its absolute requirement for directing bacterial entry into nonphagocytic cells, cumulative work from others and us emphasizes that this designation is no longer all-encompassing. We envision that disengaging the activity of the secretion machinery by targeting the energy-providing ATPase will prove to be a useful strategy for pinpointing the temporal contributions of secretion systems to the infectious cycle of other pathogenic bacteria.
The ΔsopB mutant was complemented by Tn7 integration of sopB-sigE at the attTn7 site. The sopB-sigE coding regions (sigE encodes the cognate chaperone of SopB) and 414 bp of upstream region were amplified from S. Typhimurium SL1344 genomic DNA with Xho-SopBcomp-F and Sma-SigEcomp-R. The amplicon was digested with XhoI/SmaI and ligated into XhoI-SmaI-digested pGP-Tn7-Cm (Amp r Cm r ) (35). The pGP-Tn7-sopB-sigE construct was transferred to E. coli SM10pir cells and conjugated into the S. Typhimurium ΔsopB strain bearing pSTNSK, which carries the Tn7 transposase-encoding genes tnsABCD (35). Chloramphenicol-resistant S. Typhimurium colonies were screened for the presence of sopB-sigE by PCR with sopB-specific primers and primers flanking glmS and sl3827 (35). A similar strategy was used to complement the ΔsipA::FRT deletion mutant. The sicA promoter (sipA is transcribed as part of the sicA-sipBCDA operon) was amplified from SL1344 genomic DNA with XhoI-SipAcomp-F and SipACyaA-OLR, and the sipA open reading frame was amplified with SipA-CyaA-OLF and SmaI-SipAcomp-R. These two amplicons were mixed and reamplified with XhoI-SipAcomp-F and SmaI-SipAcomp-R, followed by digestion with XhoI/SmaI and ligation into XhoI/SmaI-digested pGP-Tn7-Cm to create pGP-Tn7-Cm-sipA. This construct was used to insert sipA at the attTn7 site located downstream of the glmS gene in the SL1344 ΔsipA::FRT chromosome.