Formate oxidation in the intestinal mucus layer enhances fitness of Salmonella enterica serovar Typhimurium

ABSTRACT Salmonella enterica serovar Typhimurium induces intestinal inflammation to create a niche that fosters the outgrowth of the pathogen over the gut microbiota. Under inflammatory conditions, Salmonella utilizes terminal electron acceptors generated as byproducts of intestinal inflammation to generate cellular energy through respiration. However, the electron donating reactions in these electron transport chains are poorly understood. Here, we investigated how formate utilization through the respiratory formate dehydrogenase-N (FdnGHI) and formate dehydrogenase-O (FdoGHI) contribute to gut colonization of Salmonella. Both enzymes fulfilled redundant roles in enhancing fitness in a mouse model of Salmonella-induced colitis, and coupled to tetrathionate, nitrate, and oxygen respiration. The formic acid utilized by Salmonella during infection was generated by its own pyruvate-formate lyase as well as the gut microbiota. Transcription of formate dehydrogenases and pyruvate-formate lyase was significantly higher in bacteria residing in the mucus layer compared to the lumen. Furthermore, formate utilization conferred a more pronounced fitness advantage in the mucus, indicating that formate production and degradation occurred predominantly in the mucus layer. Our results provide new insights into how Salmonella adapts its energy metabolism to the local microenvironment in the gut. IMPORTANCE Bacterial pathogens must not only evade immune responses but also adapt their metabolism to successfully colonize their host. The microenvironments encountered by enteric pathogens differ based on anatomical location, such as small versus large intestine, spatial stratification by host factors, such as mucus layer and antimicrobial peptides, and distinct commensal microbial communities that inhabit these microenvironments. Our understanding of how Salmonella populations adapt its metabolism to different environments in the gut is incomplete. In the current study, we discovered that Salmonella utilizes formate as an electron donor to support respiration, and that formate oxidation predominantly occurs in the mucus layer. Our experiments suggest that spatially distinct Salmonella populations in the mucus layer and the lumen differ in their energy metabolism. Our findings enhance our understanding of the spatial nature of microbial metabolism and may have implications for other enteric pathogens as well as commensal host-associated microbial communities.

T he gut microbiota confers protection from infection with pathogenic organisms, a phenomenon termed colonization resistance (1,2). Colonization resistance comprises both direct microbe-microbe interactions and indirect mechanisms that act via the host. Under homeostatic conditions, nutritional competition for complex polysaccharides is a key driver of the population structure (3,4). As most small nutrient metabolites are absorbed in the small intestine, the primary carbon source for bacterial

Utilization of formate via respiratory FDHs enhances fitness in the presence of specific electron acceptors
Mutants of S. Tm that lack FDH-N activity have been characterized previously (31)(32)(33). For most of these mutants, genetic loci and genes were physically mapped using bacter iophage P22. Whole genome sequencing suggests that these mutations were likely pleiotropic, and mutations may have been in genes encoding L-seryl-tRNA Sec selenium transferase, FDH-O, and/or FDH-H. Furthermore, Salmonella utilizes tetrathionate as a respiratory electron acceptor, while E. coli is unable to do so. We, therefore, sought to assess the physiological functions of the S. Tm respiratory FDHs, FDH-N, and FDH-O (Fig.  1). To address potential redundancy, we generated a mutant lacking the major subunits of FDH-N and FDH-O (fdnG fdoG mutant) and determined fitness in the presence of exogenous formate. To mimic the environment in the intestinal tract, we used broth containing porcine mucin (mucin broth). We inoculated mucin broth with an equal mixture of the S. Tm wild-type strain and an fdnG fdoG mutant. After 16 hours, we enumerated the abundance of each strain and calculated the competitive index as the ratio of the two strains in the media corrected by the ratio of the two strains in the inoculum. The wild-type strain was marked with a mutation in phoN, encoding an acidic phosphatase with no known role in virulence, to facilitate identification. Both strains were equally fit in the absence of exogenous electron acceptors during anaero bic growth (Fig. 1A). The wild-type strain outcompeted the fdnG fdoG mutant in the presence of the electron acceptors tetrathionate (anaerobic), nitrate (anaerobic), and oxygen (microaerobic). To evaluate the individual contribution of FDH-N and FDH-O, we repeated this experiment and determined the fitness of an fdoG mutant relative to the fdnG fdoG mutant (Fig. 1B) and an fdnG mutant relative to the fdnG fdoG mutant (Fig.  1C), respectively. The fdoG mutant outcompeted the double mutant in the presence of tetrathionate, nitrate, and oxygen, while the fdnG mutant only displayed a significant fitness advantage under microaerobic conditions. No fitness advantage was apparent for either single mutant over the double mutant in the absence of an exogenous electron acceptor ( Fig. 1B and C). For genetic complementation, we introduced the fdnG and fdoG genes, under control of their native promoters, in a neutral locus in the chromo some (phoN); this complementation strategy restored fitness to wild-type levels (Fig. S2). Collectively, these results suggest that S. Tm FDH-O primarily couples formate oxidation to oxygen respiration, while the electrons liberated by the S. Tm FDH-N enzyme are donated to various electron acceptors under these experimental conditions.

Respiratory formate dehydrogenases enhance fitness of S. Tm in the murine large intestine
We next investigated S. Tm formate metabolism in a murine model of Salmonellainduced colitis. Oral treatment of C57BL/6 mice with streptomycin increases susceptibil ity to S. Tm infection (34,35). In this model, the inflammatory infiltrate in the intestinal tissue is dominated by neutrophils (35), akin to human infection with non-typhoidal Salmonella (36,37). We infected groups of streptomycin treated mice with an equal mixture of the S. Tm wild-type strain and the fdnG fdoG mutant, the wild-type strain and a fdnG mutant, or the wild-type strain and a fdoG mutant and determined the abundance of each strain in the cecal and colon contents 4 days after infection ( Fig. 2A and B). The wild-type strain was recovered in higher numbers than the fdnG fdoG double mutant, indicating that respiratory FDH activity enhances fitness of S. Tm during infection. The single mutants (fdnG mutant and fdoG mutant) were as fit as the wild-type strain ( Fig. 2A and B), suggesting that FDH-N and FDH-O may fulfill redundant roles during S. Tm infection in this animal model.
We had previously shown that formate concentrations in the murine lumen increase in a mouse model of non-infectious colitis (38). To assess whether formate levels increase during S. Tm-induced colitis, we monitored formate concentrations in the cecal lumen by gas chromatography-mass spectrometry (Fig. 2C). Mice that were treated with strepto mycin and infected with the S. Tm wild-type strain exhibited significantly elevated levels of formate (1.5 µmol/g) compared to mock-treated animals (0.48 µmol/g). The concentra tion of formate was further elevated (3.6 µmol/g) when we inactivated the ability of S. Tm to utilize formate through respiration (fdnG fdoG mutant), a finding that is consistent with consumption of formate by S. Tm (Fig. 2C).
Oral administration of streptomycin disturbs the composition of the gut microbiota (16,39). We thus determined whether S. Tm to utilizes formate in the presence of an unperturbed microbiota. Unlike C57BL/6 mice, CBA mice survive infection with S. Tm and develop a neutrophilic inflammatory response in their large intestine between 7 and 10 days after infection (21). In the CBA mouse model, the formate concentration in the lumen of the colon rises significantly 7 days after S. Tm infection (Fig. 2D). Furthermore, FDH-N and FDH-O provide a fitness advantage in the cecum and colon contents (Fig. 2E). We conclude that utilization of formate through respiratory formate dehydrogenases enhances S. Tm fitness in the murine large intestine.

Formate oxidation in the murine gut is coupled to several inflammationderived electron acceptors
We next sought to determine which electron acceptor(s) enable formate oxidation in the murine gut. We performed competitive colonization assay in strains lacking the three nitrate reductases (NR mutant), the tetrathionate reductase (ttrA mutant), and oxygen respiration (cydA mutant) ( Fig. 3A and B). The growth advantage conferred by FDH-N and FDH-O was still observed in the absence of nitrate and tetrathionate respiration and was even more pronounced in the absence of cytochrome bd-II oxidase-mediated oxygen respiration. One explanation for these findings could be potential redundancy in the utilization of these different electron acceptors. Consistent with this idea, FDH-N and FDH-O ceased to provide a fitness advantage when we removed the ability to utilize nitrate respiration, tetrathionate reduction, and cytochrome bd-II oxidase-mediated oxygen respiration ( Fig. 3A and B).
Nitrate and tetrathionate have been shown to be generated as byproducts of inflammatory reactive oxygen and nitrogen metabolism in the gut lumen (18,19). Furthermore, oxygen availability in the gut lumen increases during inflammation (9,20). We, therefore, tested whether inflammation is required for S. Tm formate utilization replicate. *P < 0.05; **P < 0.01; ***P < 0.001; ns, not statistically significant.
Research Article mBio during infection. Inactivation of the two type III secretion system renders S. Tm noninvasive and unable to replicate in tissues (40). The fitness advantage conferred by formate oxidation was abolished in the absence of both type III secretions systems (T3SS1/2) ( Fig. 3C and D), consistent with the idea that the availability of nitrate, tetrathi onate, and oxygen increases when inflammation reshapes the metabolite landscape in the large intestine.

S. Tm utilizes formate produced by its own metabolism as well as by the microbiota
In mouse models of non-infectious colitis, commensal E. coli utilizes primarily microbiotaderived formate (38) and we hypothesized that the formate utilized by S. Tm is also generated by the gut microbiota. We colonized gnotobiotic mice with Bacteroides thetaiotaomicron, an organism that supports cross feeding of formate to E. coli and to Methanobrevibacter smithii (41). We then infected groups of B. thetaiotaomicroncolonized and mock-treated mice with the S. Tm wild-type strain and the fdoG fdnG mutant, and assayed Salmonella colonization. Formate oxidation provided a similar was determined by gas chromatography-mass spectrometry. Bars represent the mean ± standard error. (D) Groups of CBA mice were orally infected with the wild-type strain. After 7 days, the concentration of formate in the colon content was quantitated by gas chromatography-mass spectrometry. Bars represent the mean ± standard error. (E) CBA mice were orally infected with an equal mixture of the wild-type (WT) strain (AJB715) and a mutant lacking respiratory formate dehydrogenase activity (SW1197). Competitive fitness in the cecum content (gray bar) and the colon content (black bar) was determined 7 days after infection.
Bars represent the geometric mean ± geometric standard deviation. Each dot represents data obtained from one animal. *P < 0.05; ***P < 0.001.
Research Article mBio fitness advantage in the presence and absence of B. thetaiotaomicron (Fig. 4A), suggest ing that S. Tm either accesses dietary formate or produces formate as part of its central metabolism.
To better understand the origin of formate, we generated a S. Tm mutant (pfl mutant) lacking the three pyruvate-formate lyases (pflB, pflD, pflF) as well as the bifunctional pyruvate-formate lyase/2-ketobutyrate formate lyase (tcdE). In vitro, FDH-N and FDH-O provide a fitness advantage to wild-type cells under anaerobic conditions in the presence of tetrathionate, even when no exogenous formate is added (Fig. S3). This growth advantage is not observed in the pfl mutant background (pfl mutant vs pfl fdnG fdoG mutant), and it is rescued with the addition of formate to the growth media (Fig.  S3). This implies that formate is released into the growth media as part of mixed acid fermentation, and then utilized by FDH-N and FDH-O.
To assess the contribution of different sources of formate in the streptomycin-treated mouse model, we analyzed the competitive fitness of a pfl mutant and a pfl fdnG fdoG mutant in the presence or absence of a strain that produces formate but that does not orally infected with an equal mixture of the S. Tm wild-type strain (IR715) and the respiratory formate dehydrogenase-deficient mutant (SW1197) or a mixture of an avirulent strain (T3SS1/2; SW1401) and an avirulent strain lacking respiratory formate dehydrogenases (T3SS1/2 ΔfdnG ΔfdoG; SW1201). Competitive fitness in the cecum content (C) and the colon content (D) was determined 4 days after infection. Bars represent the geometric mean ± geometric standard error. Each dot represents one animal. *P < 0.05; **P < 0.01; ***P < 0.001.

Research Article mBio
interfere with consumption (fdnG fdoG mutant). Each strain carries a different antibiotic resistance marker, enabling us to monitor bacterial colonization for each strain (Fig. 4B). The fitness advantage conferred by FDH-N and FDH-O was significantly decreased in the cecal content in the absence of pyruvate-formate lyase/2-ketobutyrate formate lyase (Fig. 4C). A similar trend was observed in the colon content ( Fig. 4D), however, this difference was not statistically significant. Cross-feeding by the fdnG fdoG mutant to the pfl-deficient strains rescued the fitness phenotype conferred by formate oxidation in the pfl mutant background ( Fig. 4C and D). We thus conclude that formate produced through pyruvate-formate lyase/2-ketobutyrate formate lyase activity is excreted by S. Tm, and then utilized by FDH-N and FDH-O on the periplasmic side of the inner membrane. S. Tm also likely accesses microbiota-derived formate since a modest growth advantage for formate oxidation remains in the pfl mutant background ( Fig. 4C and D).

Formate oxidation preferentially occurs in Salmonella cells residing in the mucus layer
Salmonella colonizes both the lumenal space and the mucus layer. Salmonella subpopu lations differ in their virulence gene expression (42), but it is unclear whether they differ in their metabolism. To address this question, we collected the luminal content and the mucus layer of S. Tm-infected gnotobiotic mice 1 and 2 days after infection, and assessed mRNA levels of pflB, fdoG, fdnG, and napA by RT-qPCR ( Fig. 5A and B). No significant differences in mRNA levels of all these genes were found at the early time point, while mRNA levels were markedly increased in the mucus-associated subpopulation compared to the luminal population after 2 days (Fig. 5B).
To determine whether differences in formate utilization would result in altered fitness, we infected gnotobiotic mice with a mixture of the wild-type strain and the fdnG fdoG mutant and assessed fitness in the mucus layer and the luminal content 1 and 2 days after infection ( Fig. 5C and D). Consistent with the transcriptional analysis, the wild-type strain and the formate oxidation deficient mutant were equally fit in the lumen and mucus layer at day 1. At day 2, the wild-type strain outcompeted the formate oxidation deficient mutant in both settings; however, the magnitude of the phenotype was significantly higher in bacteria residing in the mucus layer. These experiments suggest that formate consumption via FDH-N and FDH-O occurs preferentially in bacteria associated with the mucus layer.

DISCUSSION
During gut colonization, S. Tm exhibits a versatile respiratory metabolism that depends on the increased availability of tetrathionate, nitrate, and oxygen (43,44). Tetrathionate and nitrate are generated as byproducts of reactive nitrogen and oxygen metabolism, while oxygen influx during S. Tm infection is a consequence of inflammation-associated changes to host cell metabolism. The ability to reduce a vast number of exogenous electron acceptors is a property that sets S. Tm apart from most commensal bacteria and enables S. Tm to outcompete the gut microbiota (45). Here, we demonstrate that respiratory formate dehydrogenases are an important component of this inflammation adapted metabolism by donating electrons to the quinone pool and supporting respiration.
Respiratory formate dehydrogenases form electron transport chains with terminal reductases. The activities of these two enzymes are not only linked via the quinone pool but these protein complexes also physically interact and form supercomplexes (46). Since neither the enterobacterial nitrate reductases, cytochrome bd-II oxidase, tetrathionate reductase, nor the respiratory formate dehydrogenases are thought to exhibit true proton pump activity, proton translocation is likely achieved through two half redox loops in which proton consuming and generating reactions occur on separate sides of the cytoplasmic membrane (scalar chemistry) (47).
Respiration is not only more efficient in generating cellular energy, but it also enables S. Tm to access poorly fermentable carbon sources such as ethanolamine, 1,2 propane diol, and lactate (21,23,24,48). These compounds are eventually converted to key molecules in the intermediary metabolism, such as acetyl-CoA and pyruvate, which can be used as building blocks for biosynthesis. In contrast, formate is oxidized to carbon dioxide, which is presumably lost to the environment. As such, the fitness advantage conferred by FDH-N and FDH-O is likely to be related to energy metabolism as an electron donor for respiration.
Respiratory formate dehydrogenases couple with specific terminal reductases according to substrate availability. For example, FDH-N couples with nitrate reductases, in particular NarGHI, when nitrate is present in the culture, while FDH-O is the dominant enzyme under microaerobic conditions in vitro. In E. coli, coupling is primarily due to gene expression (33). In mouse models of non-infectious colitis, commensal E. coli relies on FDH-N for optimal gut colonization. Unlike under in vitro conditions, FDH-N couples to cytochrome bd-II-mediated oxygen respiration in the murine inflamed gut (38). In the current study, we observed that S. Tm FDH-O was the predominant enzyme under microaerobic conditions in vitro, while formate oxidation via FDH-N was coupled to the reduction of nitrate, tetrathionate, and oxygen. In contrast to E. coli, S. Tm uses all three of these electron acceptors for formate oxidation in the murine gut. The cues and genetic factors that regulate expression of FDH-N and FDH-O in Salmonella are incompletely understood and require further investigation.
Formate utilization occurs in different animal models of Salmonella infection. A genetic screen for Salmonella colonization factors identified selenocysteine biosynthesis as a critical factor for chick colonization (49). The three FDHs are the only selenoproteins known to be produced by E. coli and Salmonella (50)(51)(52). Mutants unable to produce selenocysteine (selD) cannot metabolize formate in vitro. The colonization defect of selD mutants in chicks suggests that formate metabolism likely enhances fitness of S. Tm in this animal. Further work is needed to define whether the defect of the selD mutant in the chick gut is due to a lack of respiratory FDH-N and FDH-O, or fermentative FDH-H activity.
Formate is not only important for energy metabolism, but also serves as a cue to regulate expression of virulence factors. Formate induces production of the invasionassociated (SPI-1) type III secretion system in S. Tm by targeting hilA and hilD, which encode for key regulators of SPI-1 transcription (53,54). Furthermore, formate induces virulence gene expression of Shigella during its intracellular stage (55).
Local availability of nutrients shapes microbial metabolism in the intestinal lumen and contribute to microhabitat formation. Both epithelial cells and infiltrating phago cytes contribute to nitrate production when homeostasis is perturbed (56,57). During S. Tm-induced colitis, S. Tm utilizes primarily phagocyte-derived nitrate, while E. coli utilizes Research Article mBio epithelial-derived nitrate during antibiotic-induced dysbiosis (57). These niches appear to be unique since an avirulent S. Tm strain is unable to access epithelial-derived nitrate in a setting of antibiotic-induced dysbiosis. Epithelial cells undergoing cell death release pyruvate into the gut lumen, thus providing S. Tm bacteria with an energetically valuable carbon source (58). In our study, pyruvate formate lyase was a contributor to the formate utilized by S. Tm. Host-derived pyruvate was metabolized in S. Tm by pyruvate formate lyase PflB (58). It is also conceivable that L-lactate, released by epithelial cells during S. Tm infection (24,25), is taken up by Salmonella and converted to pyruvate. Our finding that formate utilization preferentially occurs in S. Tm bacteria associated with the mucus layer suggests the existence of a disease-and habitat-specific metabolism relying on formate oxidation via respiration. One limitation of our studies is that administration of streptomycin changes the composition of the gut microbiota, which may obfuscate the exact origin of formate during S. Tm infection of antibiotic-naïve animals. Genetic evidence suggests that Citrobacter rodentium, a mouse pathogen that is closely associated with the colonic epithelium, uses FDH-N to colonize the murine intestinal tract (59). Our study on Salmonella, supported by this observation in Citro bacter, raises the possibility that formate utilization via respiratory formate dehydrogena ses near the intestinal lining might be a common metabolic feature of enteric pathogens.
Curiously, formate and other short chain fatty acids are added to animal feed to prevent infection of livestock (60,61). This notion suggests that not only the mere availability of any given metabolite determines microbial utilization, but also the local context matters. It is conceivable that formate, added in the diet, could change the composition of the gut microbiota. For example, Enterobacteriaceae use formate as an electron donor to support growth in the mammalian gut through respiration (38). E. coli respiration occurs in various settings in the murine gut, such as colitis and antibiotic treatment (38,62,63). The presence of Enterobacteriaceae interferes with Salmonella colonization in mice (15,(64)(65)(66). It is possible that administering subtherapeutic doses of antibiotics to livestock, commonly added to the feed to promote animal growth, enables respiration of Enterobacteriaceae in the gut. As such, a combination of dietary formate and subtherapeutic doses of antibiotics could favor colonization by commensal Enterobacteriaceae, which in turn could increase resistance to Salmonella colonization. Also, high levels of dietary formate could raise concentrations of this metabolite in different portions of the intestine, thus interfering with appropriate expression of T3SS-1 (53). Our study suggests that during S. Tm infection, the electron acceptors required for respiration emanate from the tissue, contributing to a local metabolic microenvironment suitable for S. Tm to perform formate oxidation.

Bacterial strains and mutants
The bacterial strains used in this work are shown in Table 1. All S. Tm mutants were generated in IR715, a nalidixic acid resistant derivative of 14028S (67). Unless noted otherwise, we cultured S. Tm and E. coli strains aerobically in lysogeny broth (LB; 10 g/L tryptone, 5 g/L yeast extract, and 10 g/L sodium chloride) or on LB plates (10 g/L tryptone, 5 g/L yeast extract, 10 g/L sodium chloride, and 15 g/L agar) at 37°C. Antibiot ics were added to LB and LB plates at the following final concentrations: carbeniciliin (Carb), 100 mg/L; nalidixic acid (Nal), 50 mg/L; kanamycin (Kan), 100 mg/L; and chlor amphenicol (Cm), 15 mg/L (all Sigma-Aldrich, St. Louis, MO, USA). The phoN gene is a neutral locus in the S. Tm genome, and we used acidic phosphatase-deficient strains in all competitive fitness experiments (21). To detect acidic phosphatase (PhoN) activity, 5-bromo-4-chloro-3-indolyl phosphate (X-phos, Chem-Impex, Wood Dale, IL, USA) was added at a concentration of 100 mg/L to LB plates.
The plasmids used in this study are listed in Table 2. New plasmids were generated using Gibson Assembly (New England Biolabs, Ipswich, MA, USA). DNA fragments were amplified using Q5 Hot Start High Fidelity DNA Polymerase (New England Biolabs) and S.
Tm IR715 as a template. The primers used for mutagenesis are listed in Table 3. Suicide plasmids were propagated in E. coli DH5α λpir. To generate pCG15, pCG25, and pCG27, the upstream and downstream regions of narG, narZ, and napA were PCR amplified and inserted into SphI-digested pRDH10 using the Gibson Assembly reaction. Similarly, for plasmids pMW301, pMW302, pMW303, DNA fragments comprising the upstream and downstream regions of pflD, pflF, and tdcE were introduced into SphI-digested pGP706. To generate pMW304, an internal fragment of the S. Tm pflB gene was amplified by PCR and cloned into the SphI site in pGP704 using a Gibson Assembly reaction. The promoter and coding sequence of fdoG and fdnG were PCR amplified and ligated into SphI-digested pSW327 to generate pRC29 and pRC30, respectively. The DNA sequence of inserts and key fragments was verified by Sanger sequencing. For mutagenesis, plasmids were introduced into E. coli S17-1 λpir, which served as the donor strain for conjugation into the appropriate S. Tm strains. A 1:1 mixture of the S17-1 λpir donor and the S. Tm recipient was spread on an LB plate and incubated overnight at 37°C. After conjugation, bacteria were spread on LB plates supplemented with Cm (pRDH10 derivates), Carb (pGP704 and pSW327 derivatives), or Kan (pGP706 derivatives) to select for clones in which a single crossover event had occurred. Pure cultures of these strains were then grown in LB broth overnight at 37°C. Counterselection was performed by plating on sucrose plates (5% sucrose, 15 g/L agar, 8 g/L nutrient broth base; Thermo Fisher, Waltham, MA, USA) to recover clones with second crossover events. Clean, unmarked deletions of narG, narZ, and napA in IR715 were generated through repeated use of this mutagenesis strategy with plasmids pCD15, pCG25, and pCG27, respectively, giving rise to CG40. MW472 and MW559 were created by applying this mutagenesis strategy to IR715 and CG40, using pCG124. Similarly, unmarked deletions of pflF, pflD, and tdcE in IR715 were generated using plasmids pMW301, pMW302, and pMW303, giving rise to MW364. To generate the strains MW519, MW525, RC141, RC142, the plasmids pMW304, pRC29, and pRC30 were conjugated into MW548, MW557, SW2182, SW1195, respectively. Clones in which a single crossover (plasmid insertion) event had occurred were obtained by plating on LB agar containing Carb and Nal. Clean deletions and insertion mutations were confirmed by PCR.

In vitro growth experiments
Competitive growth experiments were performed as described in (18,21). Briefly, raw porcine stomach type II mucin (Sigma-Aldrich) was suspended in 70% (vol/vol) ethanol, incubated at 65°C for 2 hours, and then incubated at room temperature overnight. The ethanol was evaporated in a vacuum centrifuge with mild heat. The sterilized mucin preparation was then resuspended in no-carbon E medium (0.2 g/L MgSO 4 heptahy drate, 3.9 g/L KH 2 PO 4 , 5.0 g/L anhydrous K 2 HPO 4 , and 3.5 g/L NaNH 4 HPO 4 tetrahydrate; Sigma-Aldrich) (78,79). The final concentration of mucin was 0.5% (wt/vol). Sodium formate, sodium nitrate, and potassium tetrathionate (all Sigma-Aldrich) were dissolved in sterile water, filter-sterilized, and added at a final concentration of 2 mM (sodium formate) or 4 mM each (sodium nitrate, potassium tetrathionate), as indicated. For anaerobic growth experiments, the media was pre-incubated overnight in an anaerobic chamber (Bactron EZ; Sheldon Manufacturing, Cornelius, OR, USA). The atmosphere in the anaerobic chamber was composed of 5% carbon dioxide, 5% hydrogen, and 90% nitrogen. The absence of oxygen was confirmed on a routine basis (Oxoid, Basingstoke, Hampshire, United Kingdom). To prepare the inoculum, bacterial strains were pre-cul tured in LB overnight under aerobic conditions at 37°C. Two milliliters of mucin broth was inoculated with 1 × 10 3 CFU/mL of each strain incubated in an anaerobic chamber or a hypoxic chamber (1% oxygen, 99% nitrogen; Coy Lab Products, Grass Lake, MI, USA) at 37°C for 16 hours. Serial dilutions of the inoculum and the final culture were spread on LB plates supplemented with X-phos. The competitive index was calculated by correcting the ratio of the two strains of interest in the final culture by the corresponding ratio in the inoculum.

Mouse experiments
All experiments were conducted in accordance with the policies of the Institutional Animal Care and Use Committees at UT Southwestern and UC Davis. Conventional C57BL/6 and CBA mice were originally obtained from The Jackson Laboratory. Animals were subsequently bred at UT Southwestern and UC Davis in barrier facilities under specific pathogen-free conditions. Germ-free Swiss Webster mice were maintained in plastic gnotobiotic isolators. All rooms were on a 12 hours light/dark cycle. Animals consumed food and water ad libitum throughout the experiment. Both male and female mice, aged 8-10 weeks, were used for experiments. We strived for equal representation of both sexes in each treatment group. No overt sex-specific differences were noted.
Naïve CBA mice were infected with 1 × 10 9 CFU of S. Tm by gavage or mock treated with LB broth. C57BL/6 mice received 20 mg streptomycin sulfate (VWR) in water by gavage (35). Three days later, we administered 1 × 10 9 CFU for single strain infection experiments, 5 × 10 8 CFU of each S. Tm strain for a total of 1 × 10 9 CFU for competitive infection experiments, or 3.3 × 10 8 CFU of each S. Tm strain for a total of 1 × 10 9 CFU for infection experiments involving three strains. Mice were euthanized at the indicated time points. Cecal and colonic tissues were flash frozen in liquid nitrogen and stored at −80°C. Cecal and colonic contents were dispersed in ice-cold, sterile phosphate buffered saline (PBS), the weight determined, and serial dilutions plated on LB plates supplemented with Nal and X-phos. The competitive index was calculated by dividing the ratio of the two strains of interest in the intestinal content by the corresponding ratio in the inoculum.
For the experiment shown in Fig. 4A, gnotobiotic Swiss Webster mice received 1 × 10 9 CFU of B. thetaiotaomicron in 0.1 mL PBS. After 7 days, both groups were infected with 5 × 10 4 CFU of each S. Tm strain for a total of 1 × 10 5 CFU. The competitive fitness was determined as described above. For the experiment shown in Fig. 5, gnotobiotic Swiss Webster mice were either infected with 1 × 10 5 CFU of the S. Tm wild-type strain   Research Article mBio RNA preparation, a mock-RT-PCR, lacking reverse transcriptase, was performed for each sample and target gene. SYBR green-based real-time PCR was performed using the primers listed in Table 3 on a QuantStudio 6 Flex instrument (Thermo Fischer). Data were analyzed using the comparative Ct method. Gene expression was normalized to S. Tm gmk mRNA levels. Each mucus scrap sample was compared to the luminal sample from the same mouse.

Quantitation of formate by GC/MS
GC/MS analysis was performed as described in reference (38). Briefly, luminal content from the cecum and colon was placed into ice-cold, sterile PBS and carefully dispersed by vortexing. The suspension was centrifuged at 20,000 g for 15 minutes at 4°C and the supernatant stored at −80°C. This procedure does not lyse representative gut commen sal bacteria and thus likely reflects mostly extracellular metabolites (38

Statistical analysis
Data were analyzed and processed in Microsoft Excel and GraphPad Prism v.9. All raw data were transformed with the natural logarithm prior to statistical analysis. Animals that were euthanized for health reasons prior to the end of the experiment were excluded from analysis. Similarly, mice that were insufficiently colonized (<10 colonies in 100 µL of undiluted sample) in competitive infection experiments were excluded from analysis. To determine statistical differences between groups of mice or treatment regimens in vitro, a two-tailed, unpaired Student's t-test was applied to the logarithmi cally transformed data. For formate measurements by GC/MS, an unpaired Student's t-test was used. P values less than 0.05 were considered significant.