Formate Promotes Shigella Intercellular Spread and Virulence Gene Expression

Shigella is an intracellular pathogen that invades the human host cell cytosol and exploits intracellular nutrients for growth, enabling the bacterium to create its own metabolic niche. For Shigella to effectively invade and replicate within the host cytoplasm, it must sense and adapt to changing environmental conditions; however, the mechanisms and signals sensed by S. flexneri are largely unknown. We have found that the secreted Shigella metabolism by-product formate regulates Shigella intracellular virulence gene expression and its ability to spread among epithelial cells. We propose that Shigella senses formate accumulation in the host cytosol as a way to determine intracellular Shigella density and regulate secreted virulence factors accordingly, enabling spatiotemporal regulation of effectors important for dampening the host immune response.

PFL mutant (ΔpflB) with 20 mM formate after S. flexneri invasion and measured plaque formation. In the absence of exogenous formate, the average ΔpflB plaque size was approximately half the size of the WT strain (Fig. 1A), consistent with previous observations (14). However, with the addition of exogenous formate, the plaque size of the ΔpflB mutant was restored to WT size (Fig. 1A), suggesting that S. flexneri-derived formate promotes S. flexneri plaque formation. Additionally, the plaque size of the WT strain increased 2.3-fold in the presence of exogenous formate (Fig. 1A). The formate concentration used in these experiments is in the midrange of the dose-response curve (Fig. 1B) and is within the range found in the mammalian gastrointestinal tract (23)(24)(25)(26).
To determine if other fermentation by-products promote S. flexneri plaque formation, the medium of S. flexneri-infected Henle-407 cells was supplemented with 20 mM formate, acetate, ethanol, or lactate. With the exception of ethanol, these compounds were provided as supplements as sodium salts; thus, we included a control supplemented with 20 mM NaCl. Exclusively in the presence of formate, we observed a 2.7-fold increase in plaque area (Fig. 1C).
Although we cannot completely rule out an effect of formate on Henle-407 cells, experiments to find such an effect have been negative. Examination of the monolayers by light microscopy revealed no changes in morphology or cell density when Henle-407 cells were grown with 20 mM formate, and there was no effect on medium pH. Additionally, formate had no cytotoxic effects on Henle-407 cells, as measured by lactate dehydrogenase release (Fig. S3A), and formate did not alter the growth of Henle-407 cells (Fig. S3B).
Listeria monocytogenes is another pathogen that, similarly to Shigella, accesses the host cell cytosol and exploits host actin polymerization for cell-to-cell spread. We postulated that if formate was affecting host cell physiology to promote bacterial cell-to-cell spread, it would increase the plaque size of L. monocytogenes, as well as Shigella. However, supplementation with exogenous formate had no significant effect on the plaque size of L. monocytogenes in Henle-407 cell monolayers (Fig. S5), further supporting the hypothesis that enhanced spread of intracellular S. flexneri in the presence of formate is a response by the bacteria, rather than the host cell.
The role of formate in increasing Shigella plaque size could be due to faster growth of the bacteria or increased spread. There was no significant difference in S. flexneri intracellular doubling time between the presence and absence of exogenous formate (Fig. 1D). However, we observed an increase in S. flexneri cell-to-cell spread in the presence of exogenous formate (Fig. 1E), suggesting that formate promotes S. flexneri cell-to-cell spread, resulting in increased plaque size in Henle-407 monolayers.
S. flexneri has only one known formate transporter, FocA. FocA is a bidirectional formate transporter that directly interacts with PFL to export formate (27)(28)(29). FocA activity is dependent on pH, and when the pH is below 5.8, the function of FocA switches to a formate importer (28). focA is cotranscribed with pflB and expressed under aerobic or microaerobic conditions under the control of ArcA and the fumarate and nitrate reduction regulator FNR (30)(31)(32). We found that mutating S. flexneri focA reduced plaque size similarly to the ⌬pflB mutant, likely due to reduced formate secretion. However, plaque size of the ⌬focA strain increased with exogenous formate (Fig. S4), indicating that S. flexneri formate import via FocA is dispensable for formate promotion of plaque size and suggesting that formate is sensed by S. flexneri outside the bacterial cytoplasm or, less probably, that formate is imported through an unidentified transporter.
pflB is required for S. flexneri formate secretion. While S. flexneri metabolism is understudied, metabolism and mixed acid fermentation of the closely related bacterium E. coli have been characterized in detail, and the predominant E. coli fermentation products in vitro are formate, acetate, ethanol, and lactate (33). In limited oxygen, E. coli metabolizes carbon using glycolysis, producing pyruvate for mixed acid fermentation. Pyruvate is converted to lactate via lactate dehydrogenase or to acetyl-CoA via pyruvate dehydrogenase (PDH) or PFL; of these two enzyme complexes which generate acetyl-CoA, only PFL produces formate as a by-product. Acetyl-CoA is then converted to acetate or ethanol, while formate is predominantly secreted through one of two bidirectional transporters, FocA or FocB (28,34). Cytoplasmic formate can be oxidized by the formate hydrogenlyase complex (FHL), while periplasmic formate can be oxidized by one of two formate dehydrogenase complexes (FDH-N or FDH-O). The genomes of E. coli and S. flexneri are highly similar, allowing us to project S. flexneri mixed acid fermentation (summarized in Fig. 2A). Of note, the S. flexneri locus containing focA and the PFL genes (including pflB) is highly conserved among sequenced Shigella species (Fig. S6). However, there are other notable differences in metabolism genes between E. coli and S. flexneri; S. flexneri has 58 metabolism-related pseudogenes compared to E. coli, including the putative formate transporter focB, and fdhF, a gene essential for the activity of FHL (35). We simulated S. flexneri metabolism in silico using a published S. flexneri genome-scale metabolic model (36,37), which integrates S. flexneri genome data and enzyme stoichiometry of metabolic reactions into a metabolic network; constraint-based analysis is then applied to emulate metabolism maximizing biomass production (38,39). To mimic intracellular conditions, we simulated a lowoxygen environment where carbon availability limits total growth. Under these conditions, S. flexneri is predicted to produce formate, acetate, and ethanol in silico at a ratio of approximately 3:2:1 and no lactate ( Fig. 2A). The model predicted that 29.7% of total carbon available is converted to formate under the simulated conditions. This estima-tion is consistent with previous studies examining mixed acid fermentation in S. flexneri and E. coli (16,40).
The amount of formate secreted by the S. flexneri WT and ΔpflB mutant was quantified from cells grown in minimal medium, under the conditions simulated in the in silico model. After 18 h of growth, the concentration of formate secreted by the WT S. flexneri strain was 34.6 Ϯ 0.5 mM, translating to a 38.4% conversion of available carbon to formate (Fig. 2B). In contrast, there was no detectable formate in supernatants from the ΔpflB mutant, confirming that PFL is essential for S. flexneri formate production. At host cytosolic pH, formate exists as a monovalent anion that cannot passively diffuse across bacterial or eukaryotic cellular membranes, with the exception of acidic compartments such as lysosomes (21); we therefore postulated that host cytosolic formate concentration would increase during S. flexneri infection due to S. flexneri metabolism and the spatial restriction imposed by the host cell membrane. We observed a 2.1-fold increase in intracellular formate of infected cells compared to uninfected cells (Fig. 2C). In contrast, Henle-407 cells infected with the S. flexneri ⌬pflB mutant showed no increase in intracellular formate concentration.
FDH-N inhibits S. flexneri plaque formation. S. flexneri encodes two periplasmfacing molybdoselenoformate dehydrogenase complexes, FDH-N and FDH-O, which couple formate oxidation to nitrate reduction. E. coli FDH-N, which catalyzes the conversion of periplasmic formate to H ϩ and CO 2 , is expressed during anaerobic growth and induced by nitrate in a NarL-dependent manner (41); FDH-O is active in the presence of oxygen and accounts for a smaller portion of the total formate dehydrogenase activity (42,43). While S. flexneri PFL and FDH-N levels are elevated within an epithelial cell, expression of FDH-O-related genes is repressed (5,15), suggesting that FDH-N is more important than FDH-O for formate metabolism by intracellular S. flexneri. If S. flexneri FDH-N is responsible for formate catabolism during intracellular growth, deletion of FDH-N would increase intracellular formate in infected Henle-407 cells and thus increase plaque size. We examined the plaque formation of S. flexneri ΔfdnG (catalytic subunit of FDH-N) and ΔfdoG (catalytic subunit of FDH-O) mutants. There was a 3.0-fold increase in plaque size of the S. flexneri ΔfdnG strain over WT (Fig. 3A). Furthermore, in the presence of exogenous formate the plaque size of the ΔfdnG strain increased 1.5-fold over the WT with formate. We measured formate levels of Henle-407 cells infected with S. flexneri WT and the ΔfdnG strain and found that formate was significantly higher in the ΔfdnG strain than the WT strain, consistent with reduced S. flexneri formate oxidation (Fig. 3B). These data indicate that FDH-N reduces S. flexneri spread, presumably by oxidizing formate derived from either S. flexneri or host metabolism and decreasing its concentration within the host cell cytosol. In contrast, the plaque size of the ΔfdoG strain was not significantly different from that of the WT strain, regardless of exogenous formate. This is consistent with the low expression of fdoG in intracellular bacteria (14). However, the ΔfdoG strain formed significantly fewer plaques than the WT strain, indicating that FDH-O, which is more highly expressed extracellularly, may be important for efficient invasion of host cells.
Formate alters S. flexneri virulence gene expression. To identify S. flexneri genes affected by formate that could promote plaque formation, we determined formateinduced differences in the S. flexneri transcriptome using RNA sequencing (RNA-seq). We were able to map an average of 3 ϫ 10 6 unique reads per sample to the S. flexneri genome, and we found that exogenous formate significantly altered the expression of four genes on the virulence plasmid ( Table 1). The S. flexneri gene icsA, which recruits host N-WASP for actin-mediated intercellular spread, was upregulated 2.1-fold by formate. Additionally, two S. flexneri T3SS effectors were upregulated approximately 2-fold by formate: ipaJ and ipgD. Formate also altered the expression of the virulence plasmid maintenance gene ccdB and 13 S. flexneri chromosomal genes (Table 1); however, we focused on icsA, ipaJ, and ipgD, since these genes are known virulence determinants.
To validate the findings of our RNA-seq, we quantified S. flexneri icsA, ipaJ, and ipgD transcript levels during growth in vitro or in Henle-407 cells using quantitative reverse transcriptase PCR (RT-qPCR). We observed significant sample-to-sample variation both in vitro and in vivo; however, formate supplementation significantly increased S. flexneri icsA and ipaJ in vitro (Fig. 4). We also observed that exogenous formate increased intracellular S. flexneri ipaJ expression 1.5-fold, while the ΔpflB mutant showed a 1.3-fold decrease in ipaJ expression compared to WT (Fig. 4). Both exogenous formate and pflB had no significant impact on S. flexneri ipgD levels either in vitro or in Henle-407 cells (Fig. 4).
Formate increases surface S. flexneri IcsA. S. flexneri uses polarly localized IcsA to recruit host N-WASP, catalyzing host actin synthesis and propelling the bacterium through the host cell cytosol (44,45). ΔicsA mutants cannot spread and do not form plaques in Henle-407 monolayers (2,46); therefore, to confirm that S. flexneri icsA is differentially regulated by formate, S. flexneri IcsA was visualized in infected Henle-407 cells by fluorescence staining and microscopy. IcsA labeled with fluorescein isothiocyanate (FITC) appeared as green, U-shaped foci on the micrographs that corresponded with the poles of intracellular bacteria, which were stained blue with 4=,6-diamidino-2-phenylindole (DAPI) (Fig. 5). As expected, no green foci were observed in cells infected with the ΔicsA mutant. We observed more S. flexneri with IcsA staining (45.8% Ϯ 0.5%) in infected Henle-407 cells supplemented with formate than in those without formate supplementation (22.0% Ϯ 6%), and the foci appeared brighter. In contrast, a smaller percentage of the intracellular S. flexneri ΔpflB mutant showed IcsA staining (7.0% Ϯ 1.7%), and the foci were dimmer than with the WT strain. These findings together suggest that formate increases the levels of S. flexneri IcsA in infected Henle-407 cells.
ipgD is required for formate-mediated plaque formation. S. flexneri IpgD is a secreted PtdIns(4,5)P 2 4-phosphatase that produces PtdIns5P in the host cell; increased PtdIns5P has widespread effects on the host, including alteration of host membrane tension, activation of host Akt kinase, inhibition of T-cell migration, inhibition of host extracellular ATP (eATP) secretion, activation of host ARF6, and altered host Ca 2ϩ signaling (8,(47)(48)(49)(50). To determine if ipgD is involved in the S. flexneri response to formate, we measured the plaque size of an S. flexneri ⌬ipgD mutant in Henle-407 monolayers with exogenous formate. Compared to the S. flexneri WT strain, the ⌬ipgD mutant formed smaller plaques (Fig. 6). Importantly, the ⌬ipgD mutant showed no significant increase in plaque size when the tissue culture medium was supplemented with 20 mM formate. When S. flexneri ipgD was complemented on a plasmid, the formate-mediated increase in plaque size was restored (Fig. 6). This suggests that ipgD is required for the formate-mediated increase in S. flexneri plaque size. ipaJ contributes to plaque size increase by formate. IpaJ is an S. flexneri-secreted cysteine protease that disrupts the host Golgi apparatus by altering host protein N-myristoylation (11,12,51). We examined how the plaque size of an S. flexneri ΔipaJ mutant changes in response to exogenous formate. We found that in the absence of exogenous formate, the mean plaque size of the ΔipaJ mutant was not significantly different from that of the WT strain ( Fig. 7), consistent with previous reports (52). The ΔipaJ mutant response to formate was diminished compared to WT, with the mutant forming smaller plaques than the WT strain when both were supplemented with exogenous formate; this phenotype was complemented by providing ipaJ on a plasmid (Fig. 7). These results indicate that ipaJ contributes to the formate-mediated increase in S. flexneri plaque size.
ipaJ expression is dependent on S. flexneri cell density and formate. Because S. flexneri secretes formate in the host cytosol as a by-product of carbon metabolism, and this formate accumulates in the host cytosol of S. flexneri-infected Henle-407 cells ( Fig. 2C), we expected that formate-induced virulence gene expression would increase as S. flexneri intracellular density increased, as formate is produced from bacterial metabolism. To test this, we constructed an ipaJ transcriptional reporter, in which the promoter region of ipaJ was fused to gfp on a plasmid, and ipaJ expression was visualized using fluorescence microscopy. We observed increased fluorescence in the intracellular S. flexneri WT cells compared to the ΔpflB mutant (Fig. 8A); likewise, formate supplementation of Henle-407 cells infected with WT S. flexneri increased fluorescence. Green fluorescent protein (GFP) fluorescence was then measured in single bacterial cells from fluorescence micrographs. We observed a 1.8-fold increase in gfp of S. flexneri when exogenous formate was added as a supplement (Fig. 8B), similar to the changes in ipaJ levels observed in previous experiments (Table 1 and Fig. 4). Likewise, we observed a 1.3-fold decrease of ipaJ expression in the ΔpflB mutant compared to the WT strain. We then measured the number of intracellular S. flexneri bacteria and the two-dimensional cell area of Henle-407 cells stained with wheat germ agglutinin (WGA) and regressed mean ipaJ expression of S. flexneri within individual Henle-407 cells on intracellular S. flexneri density. ipaJ expression positively correlated with S. flexneri intracellular density (Fig. 8C), and when cells were supplemented with exogenous formate, ipaJ expression levels were constitutively elevated regardless of intracellular S.

FIG 7
The ipaJ mutant has reduced response to formate. Plaque size of S. flexneri WT and ⌬ipaJ mutant was measured in cultured cells in the presence or absence of exogenous formate. (An asterisk indicates statistical significance.) The difference in plaque size caused by the addition of exogenous formate is significantly less in the ⌬ipaJ mutant than in the WT strain, and this phenotype is complemented by providing ipaJ on a plasmid.
Formate and Virulence in Shigella ® flexneri density. Furthermore, ipaJ expression of the ⌬pflB mutant was lower than that of the WT strain, regardless of intracellular S. flexneri density. These data indicate that intracellular ipaJ expression is dependent on the intracellular density of S. flexneri cells, and this regulation is positively associated with formate accumulation.
Henle-407 cells, a HeLa derivative cell line, have altered immune pathways relating to S. flexneri infection response. For example, Henle-407 cells do not express the connexin hemichannels involved in secretion of eATP, a potent inflammatory signal (54). Therefore, we analyzed S. flexneri-infected CoN-841 cells grown with or without exogenous formate and quantified CXCL10, IL-8, TNF-␣, or TNFAIP3 expression levels. We observed more gene expression variation in infected CoN-841 cells than in Henle-407 cells, and TNF-␣ was the most strongly repressed gene in infected CoN-841 cells supplemented with exogenous formate (Fig. 9A, checkered bars), suggesting that the specific genes and the dynamics of the host immune response in regard to formateregulated S. flexneri effectors vary in a different human cell line.
To confirm the effect of formate on S. flexneri ipaJ, we examined host CXCL10 expression in Henle-407 cells infected with the ΔpflB and ΔipaJ mutant strains. In contrast to formate lowering CXCL10 expression in cells infected with WT S. flexneri, CXCL10 expression was increased in Henle-407 cells infected with the S. flexneri ΔpflB mutant, and this difference was abated when exogenous formate was provided as a supplement (Fig. 9B). Similarly to the ΔpflB mutant, CXCL10 expression was elevated in Henle-407 cells infected with the S. flexneri ΔipaJ mutant, and formate reduced this response.

DISCUSSION
Formate is an important biomolecule for both human and bacterial metabolism. In bacteria, formate is a by-product of pyruvate formate lyase (PFL)-mediated conversion of pyruvate to acetyl-CoA. In addition to PFL, Shigella encodes a pyruvate dehydrogenase (PDH) complex that converts pyruvate to acetyl-CoA under aerobic conditions but does not produce formate. An S. flexneri ΔaceE mutant defective in PDH has an intracellular growth defect, indicating that PDH is utilized for acetyl-CoA generation within the host cell (16). Eukaryotic cells contain low levels of cytosolic oxygen, a condition that permits the activity of both PFL and PDH, which explains why the S. flexneri ΔpflB mutant intracellular growth rate is unaffected despite missing this source of acetyl-CoA (16).
Eukaryotic cells metabolize formate via one-carbon metabolism for purine biosynthesis, among other things (55). At biological pH, formate cannot passively cross bacterial or eukaryotic cell membranes (21); however, extracellular formate exchange occurs in cultured human cells (56)(57)(58), presumably facilitated through the formate transporter SLC26A6 (56,59,60). Likewise, formate is shuttled between the mitochondria and eukaryotic cytosol through an unknown mechanism (55). It is unclear how Henle-407 cells react to the influx of cytosolic formate from S. flexneri infection. We predict that a portion of S. flexneri-derived host cytosolic formate is oxidized by S. flexneri FDH-N (Fig. 3), a second portion is converted to purines by host one-carbon metabolism, and a third portion accumulates in the host cytosol.
Formate turnover has been previously linked to the intracellular growth phase of S. flexneri (5,15). Here, we demonstrate that it is not the metabolism of formate but formate itself that promotes S. flexneri plaque formation by increasing intercellular spread via IcsA and modulating the host response to S. flexneri infection. While PFL can be reversed to convert formate to pyruvate for metabolism (61), formate increases ΔpflB mutant plaque size, indicating that formate is not used solely as a substrate for PFL to produce pyruvate for growth (Fig. 2B). And, while formate oxidation is a potential Formate and Virulence in Shigella source of NADH, we also show that formate oxidation is dispensable for intracellular S. flexneri metabolism, since exogenous formate does not increase S. flexneri growth rate within the host cell (Fig. 1C) and knocking out FDH-N promotes S. flexneri plaque formation (Fig. 4). Unlike the closely related E. coli, Shigella spp. (including S. flexneri) have a mutated FHL, suggesting an evolutionary selection against Shigella formate oxidation (35); however, formate dehydrogenases contribute to the fitness of closely related E. coli in the lumen during inflammation-associated colon dysbiosis (62), possibly explaining why S. flexneri maintains its formate dehydrogenases.
Although it is clear that extracellular formate can enter Henle-407 cells when provided as an exogenous supplement, as shown by uptake of radiolabeled formate (see Fig. S2 in the supplemental material) and alterations in S. flexneri gene expression ( Table 1; Fig. 4 and 8), we believe that Shigella-derived formate is sufficient to differentially regulate expression related to intercellular spread and host immune dampening because we observe these phenotypes in S. flexneri formate metabolism mutants. We propose a model in which, following entry into the host cell cytoplasm, S. flexneri begins to replicate and metabolize host sugars and pyruvate, producing and secreting formate as a by-product. Formate accumulates in the infected cell cytosol rapidly due to the spatial constrictions of the host cell membrane and induces icsA to promote S. flexneri intercellular spread. Formate also induces the expression of ipaJ and possibly ipgD to alter the host immune response to S. flexneri infection. Because formate regulates S. flexneri icsA and ipaJ expression in vitro (Fig. 4), S. flexneri likely possesses a formate sensory mechanism; this S. flexneri formate sensor would likely act at the outer membrane or periplasm, as formate increases plaque size of the ⌬focA formate transport mutant (Fig. S2). One possible sensing mechanism is the BarA-UvrY twocomponent regulatory system, which senses formate and short-chain fatty acids to modulate the activity of the Csr regulatory system in E. coli (63) and regulates virulence in S. flexneri and Salmonella enterica (25,64). However, the fact that acetate fails to increase plaque size indicates that BarA-UvrY is not the sensory mechanism; consistent with this, we found that S. flexneri ΔbarA or ΔuvrY mutants still increase plaque size in response to formate (Fig. S5).
Previous studies demonstrate that there is spatiotemporal regulation of S. flexnerisecreted effectors during host cell infection (5,6,(65)(66)(67). Our data indicate that formate is an S. flexneri signal to differentially regulate icsA and ipaJ expression in the host cell, and this regulation is correlated with increased intracellular S. flexneri density. The dynamic nature of formate-mediated S. flexneri virulence gene regulation could explain the variation in both host and S. flexneri gene expression that we observed in response to formate ( Fig. 4 and 9) and why there were inconsistencies such as formate-altered S. flexneri ipgD expression (Table 1 and Fig. 4). The production of formate is not essential to S. flexneri virulence, evidenced by the fact that the ΔpflB mutant is still able to form plaques, albeit smaller than those of the WT. Rather, formate-mediated regulation of icsA/ipaJ appears to be a way to fine-tune expression in response to intracellular bacterial density or to facilitate spread once a threshold density has been achieved. There is currently a high degree of interest in the complex relationship between metabolism and pathogenesis (68,69), and the concept of a metabolic by-product regulating bacterial virulence is not new; one example is indole, a byproduct of tryptophan hydrolysis that regulates virulence phenotypes and has been reported as a social signal in enteric microbial communities (70)(71)(72). It is easy to draw comparisons between indole and formate signaling; however, one notable difference between indole and formate signaling is that indole is diffusible across cellular membranes while formate is not (73), allowing cytosolic formate to accumulate during intracellular infection.
As the bacterial load within a host cell increases, so does the need of the bacteria to dampen the host immune response in order to evade host cytosolic defenses such as autophagy, as well as systemic defenses such as neutrophil recruitment and inflammation. Consistent with other pathogens that suppress host immune response, immunomodulating effectors secreted by a single S. flexneri cell, which are costly to produce, benefit the entire intracellular S. flexneri population (74). Thus, we consider modulating host response to be a cooperative social behavior which conveys a benefit to S. flexneri intracellular populations. If formate sensing is social in nature, it could convey local information about both intracellular S. flexneri cell density and S. flexneri spatial constraint, as formate accumulation differs in the intracellular and extracellular environments. It should also be noted that inflammation in the colon results in higher levels of formate in the lumen (59). During Shigella infection, the Shigella-induced inflammatory response could lead to increased formate in the lumen and subsequently in the epithelial cells, promoting bacterial spread and production of anti-inflammatory effectors. A more complete understanding of the role of formate production and sensing in infected host cells will provide a better understanding of this complex host-pathogen interaction.

MATERIALS AND METHODS
Media and growth conditions. Bacterial strains, plasmids, and primers used in this study are listed in Table S1 in the supplemental material. S. flexneri was cultured on tryptic soy broth agar plates with 0.01% (wt/vol) Congo red (TSBA-CR), and red colonies were selected to ensure the presence of the S. flexneri virulence plasmid. Overnight bacterial cultures were grown in Luria-Bertani (LB) broth at 30°C, subcultured 1:100, and grown at 37°C to an optical density at 650 nm (OD 650 ) of 0.5 to 1.0 (mid-log phase) prior to infection. Deoxycholate (DOC) was added as a supplement at 0.1% (wt/vol) where indicated to increase efficiency of invasion of Henle-407 cultured cells (75). Antibiotics were added as supplements where indicated at the following concentrations: 25 g/ml ampicillin, 50 g/ml kanamycin, 6 g/ml chloramphenicol, and 20 g/ml gentamicin. Sodium formate was added as a supplement where indicated at 20 mM.
Construction of S. flexneri mutants and plasmids. Strains, primers, and plasmids used in this study are listed in Table S1. The ΔfdnG, ΔfdoG, and ΔfocA mutants were created using P1 bacteriophage transduction of the genes from E. coli strains JW1470, JW3865, and JW0887 (respectively) from the Keio collection (76). The ΔipaJ mutant was generated using a modified method of Datsenko and Wanner (77,78). Overlap extension PCR was used to fuse 399 bp upstream and 383 bp downstream of ipaJ to each side of the chloramphenicol resistance cassette of pKD3 (77). This product was amplified by PCR using the ipaJKO-1 and ipaJKO-4 primers, ethanol precipitated, and brought to a concentration of Ͼ500 ng/l in water. S. flexneri containing the plasmid pKD46 (77) was grown to mid-log phase in 25 ml LB with no NaCl at 30°C, and then 2 mM arabinose was added and the culture was brought to 37°C for 30 min. Cells were pelleted and resuspended in 500 l warm water, mixed with 10 l ipaJ knockout (KO) PCR product, and immediately electroporated. Mutants were selected on TSBA-CR with chloramphenicol. All mutants were verified by PCR and DNA sequencing at the University of Texas at Austin DNA sequencing facility.
The ipaJ complement plasmid (pBK24) was constructed by amplifying the ipaJ locus beginning ϳ400 bp upstream of the annotated ipaJ gene and ending at the ipaJ stop codon from S. flexneri DNA by PCR with primers ipaJ-30c-fw and ipaJ-30c-rv, containing 5= KpnI and XbaI sites, respectively. After restriction digest, the product was ligated into the corresponding sites of the plasmid pWKS30 (79). The ipaJ-gfp transcriptional reporter plasmid (pBK25) was constructed by amplifying the ipaJ promoter region, beginning ϳ500 bp upstream of ipaJ and ending at the ipaJ start codon, by PCR using primers ipaJ-gfp-fw and ipaJ-gfp-rv containing 5= SmaI and XbaI sites. After restriction digest, the product was ligated into the corresponding sites of the plasmid pLR29 (14). Both plasmids were confirmed by PCR and DNA sequencing at the University of Texas at Austin DNA sequencing facility.
In silico analysis of S. flexneri metabolism. In silico metabolism simulations were performed using a previously published S. flexneri genome-scale metabolic network reconstruction (36) and OptFlux (37). We removed external boundary metabolites and used the core biomass production as our objective function and the pFBA simulation method. To simulate anaerobic growth in M63 medium, default environmental conditions were used, except that the lower bound of O 2 exchange was set to 0, the lower bound of nicotinate exchange was set to Ϫ0.162, and the lower bounds of glucose and pyruvate exchange were set to Ϫ10. Total biomass production was limited by carbon availability under these simulated conditions. Tissue culture assays. Plaque assays were performed as described previously (18) with modifications (17). Bacteria at mid-log growth were centrifuged and resuspended in sterile saline to a final concentration of 5 ϫ 10 4 CFU/ml, and 100 l of this suspension was added to Henle-407 monolayers grown to confluence. Plates were centrifuged for 10 min at 1,000 ϫ g and incubated for 60 min at 37°C and 5% CO 2 . Monolayers were then washed four times with phosphate-buffered saline (PBS) and incubated in MEM supplemented with 0.2% glucose, gentamicin, and formate where indicated for either 48 or 72 h. Monolayers were stained with Wright-Giemsa stain (Camco) and imaged using an Alpha Innotech AlphaImager (Protein Simple), and plaque area was measured using ImageJ (80). Statistical significance For L. monocytogenes plaque assays, L. monocytogenes was grown in brain heart infusion (BHI) broth for approximately 18 h at 30°C without shaking. One milliliter of culture was centrifuged at 13,000 ϫ g for 2 min, and then the supernatant was removed and the pellet was resuspended in 1 ml PBS. Two microliters of L. monocytogenes suspension was added to Henle-407 monolayers grown in a 6-well plate and gently shaken for 1 min. The plate was incubated for 1 h at 37°C and 5% CO 2 . The monolayers were then washed four times with PBS and incubated in MEM supplemented with 0.2% glucose, gentamicin (10 g/ml), and formate where indicated for 48 to 72 h. Monolayers were stained with Wright-Giemsa stain (Camco) and imaged using an Alpha Innotech AlphaImager (Protein Simple), and plaque area was measured using ImageJ (80). Statistical significance was determined by Student's t test (P Ͻ 0.05; GraphPad Prism); figures are representative of 2 independent experiments.
To determine bacterial doubling time, Henle-407 monolayers in 6-well plates were infected as described for plaque assays with S. flexneri grown in deoxycholate and added at a multiplicity of infection (MOI) of 20. At 1 h postinfection (hpi) and every 30 min thereafter, one well was trypsinized and Henle-407 cells were then lysed in 1% deoxycholate, diluted, and spot plated on TSBA-CR. Doubling time (r) was determined using the formula t ϫ log 2 /(log y 2 Ϫ log y 1 ), where t is time (150 min), y 1 is Shigella at 1 hpi, and y 2 is Shigella at 3.5 hpi (n ϭ 2).
Cell-to-cell spread assays were performed as previously described (81). S. flexneri was added to Henle-407 cells at approximately 65% confluence at an MOI of 10. Cells were stained with Wright-Giemsa stain at 3 hpi, and cell-to-cell spread rates were calculated by counting approximately 100 infected Henle-407 cells adjacent to other Henle-407 cells. The proportion of S. flexneri spreading events was scored by determining if any Henle-407 cells adjacent to an infected cell also contained 3 or more internal bacterial cells. Statistical significance was determined by Student's t test (n ϭ 3, P Ͻ 0.05).
Formate-induced cytotoxicity was determined by measuring secreted lactate dehydrogenase of Henle-407 cells supplemented with 20 mM formate every 24 h over 3 days using a lactate dehydrogenase assay (Sigma MAK066) according to the manufacturer's instructions. Statistical significance was determined using ANOVA with a Bonferroni posttest.
Formate uptake. The formate uptake assay was modified from a previous study (22). Briefly, Henle-407 cells were cultured in 30-by 10-mm plates, and 20 M [ 14 C]formate (Moravek) was added to the tissue culture medium. Over the course of 48 h, cells were washed with PBS, trypsinized, and counted. Cells were then lysed directly in Optiphase HiSafe 3 (PerkinElmer). Counts per minute (cpm) was determined by scintillation, and counts were normalized to cell number and are relative to untreated cell lysates.
Formate quantification. For supernatant formate quantification, overnight cultures of S. flexneri WT and the indicated mutants were diluted 1:1,000 in M63 medium [100 mM KH 2 PO 4 , 15 mM (NH 4 ) 2 SO 4 , 1.8 M FeSO 4 , 1 mM MgSO 4 , 162.5 nM nicotinic acid, 10 mM glucose, 10 mM pyruvate]. Cultures were grown for approximately 18 h at 37°C in sealed anaerobe jars containing anaerobic gas packs (BD 260651). The OD 650 was recorded to determine bacterial growth. Cultures were then centrifuged, supernatants were filter sterilized prior to analysis using a formate assay (Sigma MAK059) according to the manufacturer's instructions, and samples were diluted 1:100 in formate assay buffer. Statistical significance was determined by ANOVA with a Bonferroni posttest (P Ͻ 0.05, n ϭ 3) using GraphPad Prism.
For intracellular formate quantification, Henle-407 cells were infected with S. flexneri grown with deoxycholate as described above at an MOI of 5. At 5 hpi, Henle-407 monolayers were trypsinized and Henle-407 cells were counted using a Countess II cell counter (ThermoFisher). Where indicated, bacteria were quantified by plating. Cell pellets were then resuspended in 1 ml cold methanol-water (50:50) and incubated for 20 min on ice. Five hundred microliters of cold chloroform was added, and the mixture was vortexed and then centrifuged at 13,000 ϫ g for 15 min. The aqueous phase was transferred to a new tube, liquid was evaporated using a vacuum manifold, and then samples were rehydrated in 100 l formate assay buffer. Formate was quantified using the assay described above, and statistical significance was determined by either ANOVA with a Bonferroni posttest (P Ͻ 0.05, n ϭ 3) or Student's t test (P Ͻ 0.05, n ϭ 3). Results are representative of 2 independent experiments.
Phylogenetic analysis. For the phylogenetic analysis of the PFL locus, the PFL locus was identified in the 79 Shigella complete genomes currently represented in the NCBI database. The sequences were aligned using Geneious and had 96.2% identical sites. The PFL loci from other related gastrointestinal pathogens were included for comparison. A tree was built using the Geneious Tree Builder, using the Tamura-Nei model and neighbor joining method, with global alignment with free end gaps and a cost matrix at 51% similarity. The tree was visualized using FigTree (http://tree.bio.ed.ac.uk/software/figtree).
RNA extraction. For in vitro RNA extractions, S. flexneri was grown in intracellular salts medium (ISM) (4) supplemented with 10 mM pyruvate, RDM supplement (82), 20 mM bicarbonate, and 100 M nitrate. Five-milliliter cultures were grown statically at 37°C with 5% CO 2 for 7 h. One milliliter of cold RNA-Stay (95% ethanol, 5% phenol) was added to each culture, and the entire volume was pelleted by centrifugation. One milliliter of cold RNA-Bee (Amsbio) was pipetted over the pellet. For RNA extractions of S. flexneri-infected Henle-407 cells, the Henle-407 cells were infected as described above with S. flexneri WT and mutants grown in deoxycholate and added at an MOI of 20. At 3 hpi, monolayers were washed once with PBS, and 1 ml of cold RNA-Bee (Amsbio) was pipetted over the monolayer. The cell lysates from either bacterial pellets or Henle-407 monolayers were transferred to new tubes, and 200 l chloroform was added. The mixture was vortexed and after 5 min on ice was centrifuged at 13,000 ϫ g, and the upper aqueous phase was transferred to a new tube. RNA was then precipitated with isopropanol, treated with DNase I (Invitrogen), and solubilized in water.
cDNA library generation and RNA-seq. RNA from infected Henle-407 cells grown with and without formate and a mock uninfected sample was used to generate cDNA by the Genomic Sequencing and Analysis Facility (GSAF) at the University of Texas at Austin. Single samples from each condition tested were analyzed. RNA was checked for quality using a Bioanalyzer (Agilent); samples had an RNA integrity number (RIN) of Ͼ9. One microgram of total RNA was used for ribosome depletion using the Ribo-Zero Gold rRNA removal kit (Illumina) according to the manufacturer's instructions to eliminate both eukaryotic host rRNA and Shigella rRNA. The depleted RNA was fragmented to ϳ200 bp, and cDNA was generated using the NEBNext Ultra II directional RNA kit. Samples were barcoded, and single-end sequencing (75 cycles) of cDNA libraries was performed using a NextSeq 500 system (Illumina) at the GSAF. Sequencing generated 1.1 ϫ 10 8 total reads for infected Henle-407 cells, and 1.0 ϫ 10 8 total reads for infected Henle-407 cells grown with formate.
Read mapping and gene expression analysis. Sequence reads were aligned to the S. flexneri 2457T genome (GenBank accession no. AE014073.1) and the S. flexneri virulence plasmid pCP301 (GenBank accession no. AF386526.1) using CLC Genomics Workbench (Qiagen). Multimapped reads were excluded, as were genes containing reads that mapped to the S. flexneri genome or virulence plasmid in the mock treatment (listed in Table S2). Values of reads per kilobase per million (RPKM) were used for expression analysis, and genes with an RPKM value of Ͻ20 (arbitrary cutoff) were excluded. A Baggerley proportions test with a false-discovery rate correction was used to determine statistical significance (P Ͻ 0.05). Genes with a fold change greater than 2 were considered significantly different. A complete list of read mappings can be found in Table S2.
Quantitative reverse transcriptase PCR. RNA was quantified using a ND-1000 spectrophotometer (NanoDrop), and a total of 1 g RNA was used with a high-capacity cDNA reverse transcription kit (Applied Biosystems). Two microliters of cDNA was then used as the template for SYBR green quantitative PCR (qPCR) (Applied Biosystems) using a ViiA 7 real-time PCR system (Applied Biosystems) at the University of Texas at Austin GSAF. Human gene expression was normalized to actB, and Shigella gene expression was normalized to rssA. All primers used for RT-qPCR are listed in Table S1. Statistical significance was determined using Student's t test of threshold cycle (ΔC T ) values compared to the S. flexneri WT-infected Henle-407 cells or S. flexneri WT grown in ISM.
Fluorescence microscopy. For IcsA labeling, Henle-407 cells were grown on glass coverslips in 6-well plates and infected with the S. flexneri WT, ΔpflB, or ΔicsA strain grown with deoxycholate (DOC) at an MOI of 1, as described above. At 3 hpi, cells were washed with PBS and then fixed with formaldehyde. Cells were then permeabilized with 0.5% Triton X-100, blocked with 5% bovine serum albumin (BSA) in PBS, and then incubated overnight with rabbit polyclonal antibody to IcsA (rabbit no. 5) provided by Edwin Oaks (Walter Reed Army Institute of Research) diluted 1:50 in PBS with 2.5% BSA. Cells were then washed with 0.05% Tween 20 in PBS and then blocked with 5% BSA in PBS for 30 min. The cells were then incubated with goat anti-rabbit IgG-FITC (sc-2012; Santa Cruz Biotechnology) diluted 1:1,000 and DAPI in PBS with 2.5% BSA for 2 h and then washed with 0.05% Tween 20 in PBS. Coverslips were then mounted on slides with Prolong Diamond antifade mountant (ThermoFisher). Images were then acquired using an Olympus BX41 microscope (100ϫ objective) with a DP73 digital camera (Olympus) and processed using cellSens software (Olympus) or using a Zeiss LSM 710 confocal microscope (63ϫ objective) at the University of Texas Microscopy and Imaging Facility. All exposures were identical, and images were processed in Photoshop (Adobe) to enhance contrast; all images were processed identically. Micrographs are representative of at least 3 independent experiments. To determine the proportion of intracellular bacteria with polar IcsA, the total number of bacteria and the number with polar IcsA staining were counted. Results shown are means Ϯ standard deviations (SD) for the percent IcsA-positive bacteria in at least 4 fields.
S. flexneri strains containing the ipaJ-gfp reporter were grown with deoxycholate and ampicillin and diluted to an MOI of 1. Infections were carried out as described above, and at 3 hpi, cells were washed with PBS and formaldehyde fixed. Fixed cells were stained with wheat germ agglutinin (WGA)-Alexa Fluor 555 at 1 g/ml. Slides were mounted with Vectashield with DAPI (Vector Laboratories), and images were acquired using an Olympus BX41 microscope (100ϫ objective) with a DP73 digital camera (Olympus) and processed using cellSens software (Olympus). Acquisition of FITC (gfp) was fixed at 300 ms. Twodimensional Henle-407 cell area was measured from the WGA membrane-labeled red channel, and gfp intensity of individual bacteria was quantified from the green channel using ImageJ (80). Background gfp fluorescence was measured in portions of each cell without bacteria and subtracted from individual bacterial cell fluorescence. A D'Agostino-Pearson normality test indicated that measurements did not conform to a Gaussian distribution. Therefore, statistical significance of gfp differences was determined by a Kruskal-Wallis test with a Dunn posttest (P Ͻ 0.05, n Ͼ 300). Regressions were performed using GraphPad Prism (n ϭ 12 to 20), and slopes were found to be significantly different (P Ͻ 0.05).
Accession number. Sequencing data were deposited in GEO under accession number GSE119622.