Overconsumption of low-digestible sugars can increase susceptibility to Salmonella Typhimurium infection

Abstract

Low-digestible carbohydrates (LDCs), including dietary fibers and sugars that resist absorption, are generally considered beneficial for the gut microbiota. However, we show that acute overconsumption of certain LDCs, particularly low-digestible sugars (LDSs), such as arabinose, lactulose, and polyols commonly used as sugar substitutes, can compromise colonization resistance against the common foodborne pathogen Salmonella enterica serovar Typhimurium. LDS overconsumption results in elevated fecal pathogen loads, systemic dissemination, and colitis in infected mice. Mechanistic analysis of arabinose overconsumption revealed that rapid fermentation acidifies the intestinal environment, perturbs the microbiota, alters gut-luminal short-chain fatty acid levels and elevated amino acid availability, which can support pathogen growth. Additionally, this perturbation impacts virulence by broadening tissue tropism (e.g. expansion to small intestine) and selecting for invasive Salmonella. Unlike antibiotic perturbations, Salmonella colonization requires SPI-1 and SPI-2 Type-III-secretion systems, linking virulence to pathogen fitness. These findings reveal how dietary perturbations can impact the gut ecosystem in ways that not only compromise colonization resistance but also enhance pathogen virulence.

Here, the authors show that acute overconsumption of certain poorly absorbed sugars, including some sugar substitutes, can disrupt the gut microbiota, weaken resistance to Salmonella colonization, and enhance pathogen virulence.

Introduction

The gut microbiota plays a vital role in human health, supporting metabolism, immunity, and pathogen defense^1–3. Dysbiosis, an imbalance in the microbial community, has been linked to metabolic disorders, chronic inflammatory diseases, and infections^4–7. Diet is a key factor in microbiota homeostasis, as it directly influences microbial composition and functionality⁸. A critical component of the diet is low-digestible carbohydrates (LDCs), which include dietary fibers and resistant sugars that bypass absorption in the small intestine, becoming accessible for fermentation by the microbiota⁹. High-fiber diets increase microbial diversity and support fiber-degrading bacteria, such as Bacteroides, whereas low-fiber Western diets are associated with a higher abundance of Firmicutes and Proteobacteria^10–13. Western diets are further linked to a higher risk of metabolic and cardiovascular diseases, as well as an increased prevalence of inflammatory bowel disease (IBD) and other gut-related disorders, in part due to a dysbiotic microbiota^14,15. LDCs exert prebiotic effects by serving as substrates for microbial fermentation, resulting in the production of short-chain fatty acids (SCFAs), which play a crucial role in host metabolism, physiology, and immune regulation^16,17. Diets low in LDCs reduce SCFA production, contributing to barrier dysfunction, bacterial overgrowth, and pro-inflammatory states¹⁸.

As LDCs promote the growth of beneficial microbiota, they also influence colonization resistance (CR), the microbiota’s ability to block enteric pathogens^6,19. CR is conferred through competition for nutritional and physical niches, as well as microbiota-mediated immune defenses²⁰. LDCs can accelerate the resolution of Clostridioides difficile infection²¹, whereas high-fat, low-fiber diets exacerbate disease severity through the depletion of Lachnospiraceae and Ruminococcaceae, key competitors for amino acids required by C. difficile²². However, LDCs can also have divergent effects. While inulin promotes beneficial microbes, such as Bifidobacterium and Faecalibacterium prausnitzii²³, it can also promote Citrobacter growth in mice fed a Western-style diet²⁴. Likewise, lactulose and fructooligosaccharides have been reported to enhance Salmonella enterica colonization and invasiveness²⁵ though the basis for these effects is not well understood.

Given these findings, we tested additional dietary compounds for their ability to increase susceptibility to S. enterica serovar Typhimurium (S.Tm) infection in mice. This screen revealed that acute overconsumption of simple LDCs (termed LDSs), including sugar alcohols (polyols), arabinose, and lactulose, compromises CR and promotes colitis with marked systemic dissemination. We focused on arabinose to investigate mechanisms by which certain LDSs can increase susceptibility to enteric infection. Arabinose overconsumption was accompanied by reduced intestinal pH through microbiota fermentation, microflora perturbation and pathogen colonization. Arabinose intake was also associated with enhanced SPI-1 expression, broader tissue tropism (e.g., increased translocation into Peyer’s patches and ileal tissues), and positive selection for virulence. Collectively, these findings demonstrate that dietary perturbations can enhance pathogen virulence.

Results

Acute overconsumption of several LDSs supports S.Tm colonization in the gut

The genomes of Salmonella spp. encode a broad repertoire of carbohydrate utilization pathways, enabling growth on at least 73 different compounds as sole carbon sources^26–28. Despite this versatility, Salmonella still competes with the resident microbiota for these nutrients and is forced to rely on less preferred carbon sources^29,30. To determine whether increased carbohydrate availability promotes Salmonella growth, we screened a panel of digestible and low-digestible carbohydrates for their impact on S.Tm growth in vivo (Fig. 1A). C57BL/6 mice harboring a specified pathogen-free complex (SPF) microbiota with pronounced colonization resistance were gavaged twice with carbohydrate solutions (132 mg per dose) and infected between doses (Fig. 1A). Carbohydrate amounts exceeded the typical daily intake of fermentable short-chain carbohydrates in mice fed on standard chow (16–184 mg/day, assuming 3 g food intake³¹). Mice receiving LDSs (arabinose, lactulose, and sugar alcohols) excreted significantly higher S.Tm fecal loads, more than 100-fold above the PBS controls on day 1 post-infection (p.i.) (Fig. 1B). While fecal S.Tm loads declined by day 2 p.i., a second S.Tm bloom occurred, reaching levels comparable to streptomycin-pretreated control mice (10⁷–10¹¹ CFU/g of feces) by day 3–4 p.i. (Fig. 1B³²). LDS-treated mice also developed systemic infections, with elevated S.Tm counts detected in the liver, spleen, and mesenteric lymph nodes (Fig. 1C). These mice also had colitis as indicated by the inflammatory marker lipocalin in feces (10²–10⁵ ng/g) by days 2–4 and histopathological analysis at day 4 p.i. (Fig. 1D, E). These values exceeded those in PBS controls, but inflammation remained less pronounced than in streptomycin-pretreated mice. These findings suggest that LDSs enhanced S.Tm colonization and systemic dissemination while inflammatory responses remained milder than in antibiotic-pretreated mice.

Fig. 1. Acute overconsumption of low-digestible sugars promotes S.Tm blooms and pathogen-induced colitis.

A Experimental setup. SPF C57BL/6 J mice received two 132 mg doses of the indicated carbohydrate or controls by oral gavage, one hour before and four hours after oral infection with 5 × 10⁷ S.Tm. Streptomycin-pretreated mice were given 25 mg of the antibiotic by gavage one day prior to infection. LDSs are indicated in red text. B–E Effects of LDS overfeeding on S.Tm colonization. n = 8 mice for PBS, 7 for L-arabinose, and 6 each for D-arabinose, lactulose, erythritol, sorbitol, maltitol, and streptomycin. Pathogen loads in B feces collected on days 1–4 (d1–d4) post infection (p.i.), as well as cecum contents (CC) and C internal organs (liver, spleen, and mesenteric lymph nodes (ML)) at the termination of the experiment on day 4 p.i. Gut inflammation was assessed by D histopathology of cecum tissue (taken at d4 p.i.) and E fecal Lipocalin-2 levels by ELISA. F Pathogen fecal loads of mice receiving HDCs. n = 6 mice for PBS, 7 and 5 each for glucose, fructose, galactose, and sucrose. G Fecal loads on day 1 of mice given two doses of PBS, 132 mg PEG4000 or 132 mg PEG400 (n = 8, 5, and 5 mice, respectively). H Competitive index of ΔompR mutant against a wild type strain in mice given streptomycin or arabinose. n = 8 mice per group. I–K Ussing chamber measurements of I short circuit current, J conductance and K paracellular permeability, as assessed by serosal ⁵¹Cr-EDTA recovery, of ileum and cecum from mice receiving either PBS or two doses of 132 mg of arabinose (n = 6 mice per group). L, M Mice were given two doses of 132 mg arabinose and L orally infected with either 5 × 10⁷ CFU of the ΔaraB mutant strain alone (n = 6 mice) or M a 1:1 mixture of wild-type and ΔaraB strains at a total dose of 5 × 10⁷ CFU (n = 5 mice); (M, left) colony-forming units per gram in feces and cecum contents; (M, right) the competitive index (CI), expressed as the ratio of the mutant to wild-type strains. B–K The two-tailed Mann-Whitney U test was used for statistical analysis; *P < 0.05; **P < 0.01; ***P < 0.001. At least two independent experiments were performed. Horizontal lines indicate medians. Abbreviation: hpi hours post infection; dpi days post infection; ns not significant. The source data and P values for panels B–M are provided in the Source Data file.

In contrast to LDSs, high-digestible carbohydrates (HDCs) had little effect on infection. Fructose, glucose, galactose, and sucrose did not significantly alter pathogen fecal excretion compared to PBS controls (Fig. 1F and Supplementary Fig. 1A), likely because these are efficiently absorbed in the small intestine³³, and therefore do not reach the lower gastrointestinal tract, the primary site of S.Tm growth and disease in mice^32,34.

High carbohydrate doses could induce osmotic stress, potentially disrupting the microbiota and promoting Salmonella growth. However, the administration of osmotically active compounds PEG400 or PEG4000 did not increase fecal Salmonella loads relative to PBS control mice (Fig. 1G). Additionally, the osmotic-sensitive ompR mutant showed no growth attenuation compared to wild-type in arabinose-fed or streptomycin-pretreated mice (Fig. 1H). Intestinal barrier function remained intact in mice that ingested arabinose, as confirmed by Ussing chamber tests assessing ion transport and permeability across the intestinal mucosa for 5 out of 6 parameters. (Fig. 1I–K). In the absence of infection, LDS overconsumption did not alter tissue histopathology, immune cell populations and inflammatory gene expression (Supplementary Fig. 1B–D). These control experiments rule out osmotic stress, barrier dysfunction, and histopathological damage as major contributors to S.Tm overgrowth in LDS-overfed mice.

Among the compounds tested, only LDSs enhanced S.Tm infection. Of these, we chose L-arabinose for further investigation because it had only a modest effect on initial gut colonization yet markedly promoted systemic infection and salmonellosis.

L-arabinose utilization by S.Tm is not required for gut colonization

Since arabinose cannot be absorbed by the host, it remains in the gut lumen after ingestion³⁵. Therefore, S.Tm may consume the supplemented arabinose to fuel its growth. S.Tm possesses the required genes for arabinose uptake and catabolism (araABCD araE) and can use arabinose as a sole carbon source^28,36,37. However, several observations indicate that S.Tm arabinose utilization is dispensable for infection. The D-enantiomer of arabinose, not metabolizable by S.Tm³⁸, produced similar effects to those of L-arabinose (Fig. 1B). Furthermore, the ΔaraB mutant, which cannot catabolize L-arabinose, still established gut colonization and caused systemic infection in mice that ingested L-arabinose (Fig. 1L). In co-infection experiments, the ΔaraB mutant showed a slight growth defect only during the first two days of infection (Fig. 1M). This indicates that while arabinose utilization provides some minor fitness advantage, arabinose overconsumption promotes pathogen expansion indirectly, likely through effects on the microbiota that alter the availability of other nutrients.

Without infection, acute overconsumption of arabinose transiently alters the microbiota composition

We hypothesized that arabinose overconsumption perturbs the gut microbiota and thereby promotes infection. 16S rRNA gene sequencing revealed transient microbiota shifts that return to the pre-ingestion baseline within 96 h in uninfected mice (Fig. 2A). Eight hours post-arabinose administration, Bacteroidota abundance decreased while Bacillota (f. Firmicutes) increased. By 24 h, Bacteroidota levels recovered, Bacillota declined, Verrucomicrobiota increased before all changes resolved by 96 h (Fig. 2A). Delaying infection until one day after arabinose ingestion abolished S.Tm colonization, indicating that restoration of colonization resistance parallels microbiota recovery (Fig. 2B, C). The shifts were specific to arabinose overconsumption, as PBS-treated controls exhibited relatively stable microbiota composition.

Fig. 2. Acute overconsumption of arabinose transiently alters microbiota composition.

A 16S rDNA analysis of fecal microbiota composition in uninfected or infected mice administered two doses of 132 mg arabinose or mock-treated with PBS. Shown is the mean composition (n = 6 mice per group). B, C Mice were given two doses of 132 mg arabinose or PBS one day prior to oral infection with 5 × 10⁷ CFU total S.Tm; n = 10 mice per group. B Pathogen loads in feces (d1–d4) and cecal contents (CC) and C in systemic organs (liver, spleen, and mesenteric lymph nodes (ML)). At least two independent experiments were performed. The two-tailed Mann–Whitney U test was used for statistical analysis; ns, not significant. Horizontal lines indicate medians. The source data for panels A–C are provided in the Source Data file.

In infected arabinose-overfed mice, Verrucomicrobiota gradually declined while S.Tm (Proteobacteria) increased, consistent with overall rising pathogen fecal CFUs (Fig. 2A and Supplementary Fig. 2A). This suggests that arabinose overconsumption opens a niche exploitable by S.Tm, possibly at the expense of Verrucomicrobiota. By days 3 and 4 p.i. major compositional shifts emerged, including a marked decline in both Bacteroidota and Verrucomicrobiota, coinciding with inflammation onset and luminal S Tm densities of 10⁷–10¹⁰ CFUs per gram stool (Fig. 2A and Supplementary Fig. 2A, B). In contrast, mice infected with avirulent ΔinvG or ΔssaV mutants, which cannot trigger inflammation, maintained microbiota compositions similar to uninfected, arabinose-overconsuming mice, with communities trending back toward baseline and an absence of S.Tm growth (Supplementary Fig. 2C). These data suggest that luminal pathogen blooms at days 3–4 p.i. in arabinose-overconsuming mice require virulence-triggered inflammation to overcome colonization resistance from the renormalizing microbiota.

Taken together, these observations explain the infection kinetics (Fig. 1B). Arabinose overconsumption shifts microbiota composition, permitting an initial pathogen bloom on day 1 p.i. that collapses by day 2 p.i. Salmonella virulence subsequently drives more drastic microbiota changes, resulting in a robust S.Tm bloom on days 3–4 p.i. (Fig. 2A and Supplementary Fig. 2A, B).

Fermentation of arabinose by the microbiota leads to rapid intestinal acidification that subsequently reduces colonization resistance

The mechanisms underlying the microbial shifts are not immediately clear, as arabinose is neither inherently toxic nor limited to a narrow subset of microbial taxa. As a component of plant cell walls, arabinose serves as a carbon source for various gut bacteria involved in the digestion and fermentation of dietary fibers^39,40. Arabinose utilization operons are widespread across taxa, including Enterobacteriaceae, Bacteroidota, Bifidobacteriaceae, Enterococcceae, and Oscillospiraceae^41–43. To test whether representative gut bacteria can metabolize arabinose, we grew selected human gut isolates on L-arabinose. Arabinose supplementation enhanced the growth of Bacteroides vulgatus, B. uniformis, B. thetaiotaomicron, B. ovatus, Roseburia intestinalis, and Dorea spp., but not Clostridium ramosum and C. perfringens, Lactobacillus paracasei, and Streptococcus parasanguinis (Supplementary Fig. 3A). Medium acidification was observed in strains that fermented arabinose or glucose.

While arabinose utilization promotes microbial growth, its rapid breakdown can also lead to growth-inhibiting acidification through the accumulation of fermentation byproducts. Acid stress suppresses the growth of several gut commensals in culture⁴⁴. Indeed, a single arabinose dose rapidly lowered the cecum-luminal pH from ~6.8 to ~5.7 within 1 h, reaching 5.0–5.5 by 3 h post-administration before returning to baseline after ~10 h (Fig. 3A). A pH of 5.5 inhibits growth of Bacteroidota in culture while being better tolerated by some members of Bacillota^45,46, consistent with the observed decrease in Bacteroidota and increase in Bacillota 8 h after arabinose administration (Fig. 2A). Because Bacteroides spp. contribute to colonization resistance against Salmonella⁴⁷, their reduction under acidic conditions may explain the initial S.Tm bloom. In vitro, acidity impaired the growth of commensals such as B. thetaiotaomicron and E. rectale more than S.Tm (Supplementary Fig. 3B), suggesting a potential mechanism for the selective advantage of S.Tm under acidic conditions. To assess the link between infection susceptibility and gut acidification, we administered graded doses of arabinose to identify the threshold required to promote S.Tm infection. Higher arabinose doses (≥66 mg) lowered the pH (≈5.5), increased fecal S.Tm, and elevated systemic bacterial loads, whereas lower doses (≤33 mg) did not (pH ≈ 6.8) with little effect on S.Tm colonization resistance (Fig. 3B–D). Mice receiving arabinan (2 × 132 mg), an arabinose polymer that is not rapidly fermented, showed less pronounced acidification and no associated S.Tm blooms (Fig. 3E, F). Arabinan is a plant-derived branched polysaccharide containing arabinofuranose monomers that requires multiple enzymes for its breakdown, and generally has a lower turnover than free monosaccharides⁴⁸. These data suggest an association between gut acidification and reduced colonization resistance.

Fig. 3. Gut acidification correlates with S.Tm blooms following arabinose ingestion.

A Cecal pH was measured at the indicated time points following a single 132 mg dose of arabinose or untreated controls. n = 5 mice per time point. B Cecal pH measured 4 h after administration of two doses of arabinose at the indicated amounts. n = 7, 6, 5, and 5 mice for the untreated, 16.5-, 33-, 66-, and 132-mg groups, respectively. C, D Fecal and organ S.Tm loads in mice given two doses of varying amounts of arabinose (n = 5, 8, and 6 mice for the 33-, 66-, and 132-mg groups, respectively) or mock treatment (PBS; n = 7 mice). E Cecal pH measured 4 h after two doses of 132 mg arabinan, arabinose, or PBS. n = 5 mice per group. F Fecal S.Tm loads in infected mice administered two doses of 132 mg arabinan (n = 6 mice), or PBS (n = 8 mice). At least two independent experiments were performed. The two-tailed Mann-Whitney U test was used for statistical analysis; *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant. Horizontal lines indicate medians. The source data and P values for panels A–F are provided in the Source Data file.

Arabinose overconsumption alters metabolite concentrations in the gut lumen

Carbons from arabinose enter glycolysis via the pentose phosphate pathway and are incorporated into fermentation byproducts in the form of organic acids, such as SCFAs, succinate, and lactate^49–51. Targeted liquid chromatography-mass spectrometry (LC-MS) detected significant increases in propionate, succinate, and lactate in the mouse gut four hours after the second dose of arabinose, whereas butyrate and acetate decreased, and formate remained unchanged (Fig. 4A). These shifts in organic acid levels are consistent with the observed reduction in luminal pH.

Fig. 4. The microbiota metabolize arabinose, altering gut metabolite profiles.

A, B Metabolite measurements of short-chain fatty acids (SCFAs), carboxylates, and amino acids in cecal contents from mice administered two doses of 132 mg arabinose or mock treatment (PBS). n = 5 mice per group. Samples were collected 4 h after the final dose. A LC-MS quantification of formate, acetate, butyrate, propionate, succinate, and lactate. B LC-MS quantification of relative amino acid levels. C Arabinose-overfed mice (two oral doses of 132 mg arabinose; n = 6 mice) were infected with a consortium of tagged lysine-, arginine-, and phenylalanine-auxotrophic mutants plus three isogenic wild-type strains at equal titers (5 × 10⁷ total CFU). Strain abundances were quantified by qPCR using tag-specific primers. Competitive indices (CI) were calculated relative to the mean abundance of the wild-type strains. The two-tailed Mann–Whitney U test was used for statistical analysis; *P < 0.05; **P < 0.01; ***P < 0.001 ns, not significant. n = 6 mice per group. Horizontal lines indicate medians. The source data and P-values for panels A–C are provided in the Source Data file.

Arabinose ingestion also caused significant increases in several amino acids, namely alanine, lysine, arginine, proline, leucine, isoleucine, tyrosine, glutamine, and phenylalanine in the cecal contents of mice (Fig. 4B). These increases may reflect reduced microbial uptake or release of intracellular metabolites through microbial lysis. Elevated amino acid availability may contribute to S.Tm blooms as excess amino acids promote S.Tm colonization in complex or dysbiotic microbiota^52,53. Indeed, the high competitive fitness of auxotrophic strains suggests that amino acid import was sufficient to support S.Tm growth in arabinose-overfed mice (C.I. ≈ 1.0; Fig. 4C). Overall, these findings support a model in which gut acidification disrupts the microbiota, subsequently reshaping the metabolic landscape to favor S.Tm expansion.

Impact of virulence factors on gut luminal growth

Colonization kinetics and marked systemic pathogen loads point to enhanced pathogen invasiveness in LDS overconsuming mice (Fig. 1B, C). Increased invasion into intestinal tissue and subsequent reseeding may contribute to the luminal S.Tm blooms at days 3-4 p.i., while enhanced dissemination likely explains the elevated pathogen burdens in the mesenteric lymph nodes, spleen and liver, which exceeded those in streptomycin-pretreated controls. Virulence genes carried on two pathogenicity islands (SPI-1 and SPI-2) enable S.Tm to invade and replicate within intestinal and immune cells, such as dendritic cells and macrophages, facilitating systemic spread^54–58. We hypothesized that LDSs induce SPI-1 expression and thereby promote tissue invasion. Using a transcriptional GFP reporter for the SPI-1 chaperone SicA⁵⁹, we observed a fivefold higher percentage of sicA + (GFP+) S.Tm in arabinose-overfed mice compared to streptomycin-pretreated mice (Fig. 5A), indicating increased SPI-1 expression and invasive potential.

Fig. 5. In arabinose-overfed mice, virulence is induced and required for luminal growth.

A Percentage and total number of S.Tm expressing sicA at 8 h p.i. in feces from mice pre-treated with streptomycin or administered two doses of 132 mg arabinose. n = 7 mice per group. B–E Mice given arabinose were orally infected with 5 × 10⁷ CFU of one of the indicated strains: wild-type (WT; n = 8 mice), SPI-1 mutant (ΔinvG; n = 5 mice), SPI-2 mutant (ΔssaV; n = 6 mice), or SPI-1 SPI-2 double mutant (ΔinvG ΔssaV; n = 6 mice). B Pathogen loads in feces (d1-d4) and cecal contents (CC). C Bacterial loads at systemic sites (liver, spleen, and mesenteric lymph nodes (ML)). D Fecal lipocalin-2 (LCN2) levels. E Histopathological assessment of cecal tissue collected on day 4 p.i. F–H Arabinose-fed mice were orally infected with 5 × 10⁷ CFU of one of the indicated strains: wild type (WT; n = 5 mice), ΔinvG (n = 5 mice), Δ4 (ΔsipA, ΔsopB, ΔsopE, ΔsopE2; n = 5 mice), ΔsipA (n = 6 mice), ΔsopB (n = 6 mice), ΔsopE (n = 8), or ΔsopE2 (n = 5 mice). F S.Tm loads in feces and cecal contents. G Bacterial loads in systemic organs. H Fecal LCN2 levels. The two-tailed Mann-Whitney U test was used for statistical analysis; *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant. Horizontal lines indicate medians. The source data and p-values for panels A–H are provided in the Source Data file.

While SPI-1 and SPI-2 and their effectors are essential for inflammation and disease progression, they are not inherently required for gut luminal growth in antibiotic-pretreated mice, where the microbiota is strongly perturbed. Accordingly, SPI-1 and SPI-2-deficient mutants colonize the guts of streptomycin-pretreated mice as efficiently as wild type^56,60. By contrast, both SPI-1 and SPI-2 virulence were required for colonization and pathogen growth in mice receiving arabinose. Mutants lacking essential components of either the SPI-1 or SPI-2 type III secretion systems (ΔinvG and ΔssaV, respectively) or both (ΔinvG ΔssaV) colonized the gut similarly to wild-type at day 1 p.i. but became severely attenuated thereafter (days 2–4 p.i.), with fecal loads 5–6 logs lower than wild-type (Fig. 5B). Infection with these mutants also yielded reduced organ counts, lower lipocalin levels, and milder mucosal pathology, consistent with a marked defect in virulence (Fig. 5C–E). These observations suggest that the requirement for SPI-1 and SPI-2 in S.Tm colonization is context dependent, and that arabinose overconsumption provides a robust model for studying the role of SPI-1 and -2 virulence in infection, transmission or virulence evolution.

Given the strong contribution of SPI-1 to S.Tm luminal blooms at days 3–4 p.i., we next dissected the role of individual SPI-1 effectors in this process. SPI-1 effectors, namely SipA, SopB, SopE, and SopE2, manipulate the host cytoskeleton to enable S.Tm entry into non-phagocytic cells. Among these, SipA plays the most significant role in the invasion of enteric epithelial cells in vivo and polarized cells in tissue culture⁶¹. However, contributions of SPI-1 effectors to gut-luminal colonization were not resolved. In arabinose-overconsuming mice, the ΔsipA mutant behaved similarly to mutants lacking all four effectors or a SPI-1 secretion-deficient mutant, exhibiting an initial bloom on day 1 p.i., followed by a marked attenuation in fecal counts (Fig. 5F). ΔsipA infection also resulted in reduced dissemination to the liver (day 4 p.i.; Fig. 5G) and lower levels of fecal lipocalin throughout the infection (days 1–4 p.i.; Fig. 5H). The ΔsopB mutant was also attenuated, albeit less severely than ΔsipA, and showed moderate fecal shedding and reduced inflammation (Fig. 5F, H). By contrast, SopE played only a minor role, as ΔsopE-infected mice resembled wild-type infection, except for a slight delay in inflammation (i.e., at day 2 p.i.; Fig. 5F, H). SopE2 appeared redundant, as the ΔsopE2 mutant was phenotypically indistinguishable from wild type.

These observations highlight a functional hierarchy among SPI-1 effectors in vivo that mirrors prior invasion studies, with SipA being most critical for mucosal infection, followed by SopB, whereas SopE/SopE2 appear largely dispensable^61,62. Importantly, the effects of arabinose overconsumption tie invasion to gut-luminal blooms, a link not captured in earlier in vivo models⁶⁰.

The inability of virulence-attenuated mutants to maintain gut colonization in arabinose-overconsuming mice might stem from their failure to induce inflammation. Inflammation promotes Enterobacteriaceae blooms by generating niche-specific nutrients, including tetrathionate and nitrate, that S.Tm can exploit for anaerobic respiration and oxidative TCA cycle metabolism^63–67. To test this, we performed a 1:1 co-infection with a SPI-1, SPI-2-secretion-deficient (ΔinvG ΔssaV) mutant and a congenic wild-type strain to trigger inflammation in trans. On day 1 p.i., both strains bloomed to similar levels as indicated by fecal shedding (Fig. 6A). Consistent with previous observations (Fig. 1B), the total fecal pathogen loads plummeted by day 2 and rebounded to 10⁶–10¹⁰ CFU per gram feces by days 3–4 p.i. (Fig. 6A). Strikingly, by day 4 p.i., the wild-type strain had completely outcompeted the ΔinvG ΔssaV mutant, whose fecal counts fell below detection in 6 of 12 mice (Fig. 6A, B). The onset of inflammation, indicated by elevated lipocalin levels and histopathology, coincided with a rise of the wild-type strain and a sharp decline of the ΔinvG ΔssaV mutant population (Fig. 6C, D). These findings suggest that SPI-1 and SPI-2 virulence not only trigger mucosal inflammation but also confer a strong competitive advantage during the days 3–4 p.i. luminal bloom. The reduced fitness of the ΔinvG ΔssaV mutant was not due to an intrinsic growth defect, as it grew comparably to wild type in streptomycin-pretreated mice (Fig. 6A, B). The absence of the mutant at systemic sites confirms that it remained confined to the gut lumen (Fig. 6E). Thus, the dominance of wild-type S.Tm in the lumen at days 3–4 p.i. further suggests that the luminal population is substantially replenished by re-seeding wild-type S.Tm from tissue reservoirs established via SPI-1 and/or SPI-2 virulence. Previous work in antibiotic-pretreated mice suggests such re-seeding events are comparatively rare⁶⁸. We substantiated this hypothesis using pools of barcoded wild-type S.Tm strains in arabinose-overfed mice (Supplementary Fig. 4). Overall, the data suggest that virulence and gut colonization are tied in the diet-driven infection model, but uncoupled in the antibiotic-pretreatment model, underscoring how infection context shapes the selective pressures governing pathogen fitness.

Fig. 6. Arabinose overconsumption selects against avirulent S.Tm mutants.

Mice pre-treated with streptomycin (n = 5 mice) or administered two doses of 132 mg arabinose (n = 14 mice) were orally infected with a 1:1 mixture of wild-type (WT) S.Tm and the avirulent mutant strain (ΔinvG ΔssaV). A Pathogen loads in feces (d1-d4) and cecal contents (CC). B Competitive indices (CI) comparing the relative fitness of the wild-type and the avirulent strain. C Fecal lipocalin-2 (LCN2) levels. D Histopathological assessment of cecum tissue collected on day 4 p.i. The two-tailed Wilcoxon and Mann–Whitney U tests were used for statistical analysis; *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant. E Pathogen loads at day 4 p.i. in the liver, spleen, and mesenteric lymph nodes (ML). Abbreviations: inoc inoculum. Horizontal lines indicate medians. The source data and p-values for panels A–E are provided in the Source Data file.

Arabinose overconsumption influences tissue tropism

To identify intestinal reservoirs, we performed gentamicin protection assays on tissues from infected mice receiving either streptomycin or arabinose. This assay eliminates extracellular bacteria, allowing quantification of tissue-lodged bacteria. Sampled tissues included typical invasion and persistence sites, such as the cecum, ileum, and Peyer’s patches. In the arabinose group, S.Tm was distributed similarly across these tissues whereas in the streptomycin-pretreated mice, S.Tm localized preferentially in cecum tissue (Fig. 7A). Tissue loads in arabinose-overfed mice rise until day 3 of infection and plateaued (or slightly decline) by day 4 p.i. unlike the streptomycin-pretreatment group, where tissue burdens changed little over time (Fig. 7A). Systemic pathogen loads (liver, spleen, and mesenteric lymph nodes) increased in both groups, but by day 3 p.i. were higher in the arabinose-overfed mice than in the streptomycin-pretreatment controls (Fig. 7B).

Fig. 7. S.Tm tissue tropism upon arabinose overconsumption.

A–C Mice were pretreated with streptomycin (n = 7, 7, 6, and 6 mice on days 1–4 p.i., respectively) or given two doses of 132 mg arabinose (n = 7, 9, 5, and 5 mice, respectively) and given 5 × 10⁷ S.Tm by oral gavage. A S.Tm extracted from indicated tissue (CT cecum tissue; IT ileum tissue; PP Peyer’s Patches) by gentamicin protection assay and enumerated by plating on MacConkey agar. B Pathogen loads in systemic organs: liver (L), spleen (S), and mesenteric lymph nodes (ML). C Pathogen loads in intestinal contents: feces (F), cecal contents (CC), and ileum contents (IC). D Heterozygous (n = 16 mice) and homozygous (n = 14 mice) Il2rg/Rag2 mice were given two doses of 132 mg arabinose and orally infected with 5 × 10⁷ CFU of S.Tm. S.Tm counts in fecal (d1–d4) and cecal contents (CC). The two-tailed Mann–Whitney U and Wilcoxon tests were used for statistical analysis; *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant. Horizontal lines indicate medians. The source data and P values for panels A–D are provided in the Source Data file.

Early in infection (days 1–2 p.i.), the luminal S.Tm population was concentrated in the cecum and colon with lower titers in the ileum in both treatments (as inferred from cecal and ileal contents and feces; Fig. 7C). In arabinose overconsuming mice, this pattern shifted by day 3 p.i., when ileal S.Tm titers became comparable to cecal titers. In contrast, streptomycin-pretreated mice exhibited only limited S.Tm expansion in the lumen of the ileum, with titers remaining significantly lower than in cecal contents (Fig. 7C). These data may suggest that reseeding from ileal tissues contributes to S.Tm shedding in arabinose overconsuming mice, as substantial bacterial titers were detected in both ileal tissue and Peyer’s patches following blooms in ileal contents. To investigate the contribution of Peyer’s patches to reseeding, we tested Il2rg^−/− Rag2^−/− mice, which lack Peyer’s patches. These mice exhibited lower fecal S.Tm loads compared to +/− littermate controls (Fig. 7D). In contrast, S.Tm systemic burdens were much higher in the knockout mice, likely due to the absence of NK cells, which release pro-inflammatory cytokines that accelerate gut inflammation while restricting systemic spread^69–72 (Supplementary Fig. 5A). The reduction in fecal shedding may suggest that Peyer’s patches contribute to luminal reseeding. However, we cannot directly attribute the decreased fecal loads to the absence of Peyer’s patches, as knockout mice also exhibited reduced inflammation compared to heterozygous controls, which could independently limit luminal S.Tm growth (Supplementary Fig. 5B).

Discussion

Foodborne pathogens are frequently encountered through contaminated food, yet most exposures do not result in disease due to microbiota-mediated CR. Antibiotics are among the best-established risk factors for enteric infection, but many cases occur without recent antibiotic exposure, suggesting additional susceptibility determinants. In humans, dietary contributions remain poorly defined, as studies largely focus on identifying contaminated food sources. Foodborne infections caused by Salmonella enterica and Escherichia coli exhibit seasonal spikes, including during holiday periods, typically attributed to temperature, food handling, and storage practices^73–79. Such periods can also involve short-term dietary alterations and excess food intake, though whether these contribute to infection susceptibility remains unknown. However, animal studies show that diet-elicited perturbations can affect the gut microbiota and weaken CR^2,80. High-fat Western diets promote excessive bile salt secretion, which most Bacteroidetes and Firmicutes strains cannot tolerate, permitting blooms of tolerant Enterobacteriaceae such as S.Tm and E. coli⁸¹. Here, we broaden current understanding of diet-mediated perturbations by showing how acute overconsumption of specific carbohydrates can not only disrupt CR, but also reshape pathogen–microbiota interactions in ways that enhance pathogen virulence.

Arabinose intake induced reversible changes in gut metabolites, pH, and microbiota composition. Metabolite and pH analyses revealed that arabinose was rapidly fermented, causing organic acid accumulation and subsequent gut acidification (Fig. 3). This pH shift may have reversibly altered the microbiota composition (Fig. 2A) and promoted pathogen growth through increased availability of nutrients, including amino acids that supply carbon and nitrogen (Fig. 4B). S.Tm possesses multiple acid resistance systems^82–84, enabling it to withstand gastric acidity (pH 2–3)⁸⁵ and proliferate within macrophages’ acidic compartments (pH 4.8)^86–90. This resistance may give S.Tm a competitive edge over the microbiota under mild pH stress (Supplementary Fig. 3B).

Recent studies identified arabinose as a key carbon source for pathogen growth. S.Tm acquires arabinose via a genome-encoded alpha-N-arabinofuranosidase that liberates the sugar from dietary polysaccharides to support its expansion in superspreader mice⁹¹. Additionally, continuous supplementation of free L-arabinose in drinking water throughout the infection increased pathogen burdens^91,92. In these settings, arabinose catabolism contributed substantially to pathogen fitness, as mutants defective in arabinose utilization were attenuated. Surprisingly, in our model, defective arabinose catabolism did not prevent S.Tm colonization or disease when arabinose was administered as two high doses around the time of infection (Fig. 1L). With this regimen, the ∆araB mutant featured merely a slight disadvantage (Fig. 1M). This suggests that S.Tm expansion in our model primarily depends on a dysbiotic microbiota rather than direct utilization of a supplemented sugar. Differences in experimental design, particularly acute overconsumption versus continuous intake, likely explain the varying reliance on arabinose catabolism.

The mechanisms by which other LDSs tested, namely lactulose, erythritol, maltitol, and sorbitol, compromise colonization resistance remain unexplored. Sorbitol and erythritol did not alter intestinal pH (Supplementary Fig. 6), suggesting that acidification does not underlie their effects. However, lactulose and maltitol appear to act through a similar mechanism, as both induce a comparable drop in pH, albeit less pronounced with maltitol (Supplementary Fig. 6). This raises the possibility that gut acidification may represent one broader mechanism by which certain rapidly fermentable foods weaken colonization resistance. Further studies should address whether different LDSs elicit distinct microbiota perturbations and assess how different perturbations reduce CR.

Overconsumption alone does not inherently compromise colonization resistance, as arabinan, glucose, fructose, galactose, and sucrose did not support S.Tm colonization (Figs. 1F and 3F). This may reflect efficient host absorption of preferred sugars, limiting their availability to the microbiota³³, or the slower breakdown of dietary fibers, resulting in only modest pH changes insufficient to substantially alter the microbial community⁴⁸. Thus, the risk associated with overconsumption depends on what is overconsumed.

We demonstrated that dietary components could influence pathogen virulence in several ways, including increased virulence expression, broader tissue invasiveness, and selection against avirulent mutants. Despite milder disease sequelae and slow infection kinetics, mice overfed on arabinose or other LDSs unexpectedly exhibited high pathogen loads at systemic sites (Fig. 1B, C). The increased fraction of SPI-1⁺ S.Tm cells likely contributes to increased dissemination (Figs. 1C and 5A, 7B). This enrichment of SPI-1⁺ cells cannot be directly attributed to L-arabinose, as previous studies have shown that L-arabinose represses SPI-1 by inhibiting the SPI-1 master regulator HilD and reducing cell invasion in vitro⁹³. Instead, altered gut metabolites likely contributed to SPI-1 induction. Microbiota-derived metabolites, such as SCFAs, are known to modulate virulence gene expression. Formate and acetate upregulate HilD expression via BarA/SirA signaling⁹⁴, whereas propionate and butyrate exert inhibitory effects^47,95,96. Metabolite analysis revealed increased propionate and decreased butyrate and acetate levels (Fig. 4A). Although these changes would generate a mix of opposing regulatory signals for SPI-1, we observed an overall increase in SPI-1⁺ cells, suggesting that SPI-1 induction may result from the collective shift or from additional unidentified microbiota-derived signals. Future work should address how these metabolites interact to regulate SPI-1 expression, as this interplay may have important implications for pathogen virulence in vivo.

Increased lactate and succinate were also observed (Fig. 4A). These metabolites, along with low pH, are known inducers of SPI-2 expression intracellularly and in culture^97–100 suggesting that SPI-2 may likewise be induced in vivo. While we did not assess SPI-2 expression in the gut lumen, where its role remains less defined, prior studies show that SPI-2 transcription is primed before invasion^101,102, potentially enhancing S.Tm survival within host tissues¹⁰³. Follow-up studies should determine whether luminal SPI-2 expression contributes to early-stage virulence.

Arabinose intake broadens tissue tropism, likely contributing to higher systemic burdens. It increased pathogen loads in ileal tissue, particularly Peyer’s patches, whereas significantly lower loads were observed in antibiotic-pretreated mice (Fig. 7A). Peyer’s patches function in immune surveillance but also serve as entry points for several bacterial pathogens, including other S. enterica spp.^104–106, Yersinia enterocolitica^107,108, and Vibrio cholerae¹⁰⁹, which exploit the microfold cells to access the lamina propria for dissemination. These specialized lymphoid tissues are enriched in macrophages and dendritic cells, capable of trafficking Salmonella to mesenteric lymph nodes, the bloodstream, and distal organs such as the spleen and liver^110,111. In contrast, the cecum becomes the primary site of infection in antibiotic-pretreated mice, explaining the comparatively lower pathogen burdens at systemic sites. In arabinose overfed mice, elevated virulence expression may promote access to Peyer’s patches or other immune inductive sites where SPI-2 virulence factors are especially crucial for evading innate immune responses. The exact mechanisms underlying this altered tissue tropism remain unclear and warrant further investigation.

Experimental evolution studies reveal that the microbiota selects against virulence-attenuating mutations^112–116. During S.Tm infection of antibiotic-pretreated mice, the combined effects of antibiotics and inflammation on the microbiota mediate a trade-off between pathogen replication and virulence loss¹¹⁵. S.Tm achieves optimal growth by depleting the microbiota through SPI-1 and SPI-2-driven inflammation⁶⁷, but this also promotes the luminal overgrowth of virulence-attenuating hilD mutations¹¹⁷. While the mechanisms underlying the selection for virulence-attenuated mutants remain poorly understood, the arabinose infection model provides a setting in which virulence is positively selected in the gut-luminal pathogen population. S.Tm required SPI-1 and SPI-2 for sustained gut colonization in arabinose overfed mice (Fig. 5B), whereas these virulence factors were dispensable under antibiotic pretreatment⁵⁶ (Fig. 6A, B). While inflammation promoted S.Tm growth in both models, only in the arabinose infection model was the growth of avirulent mutants significantly constrained (Figs. 5B and 6A, B). This can be explained by the microbiota’s control over pathogen population dynamics in the gut lumen between days 1–3. Notably, gut acidification triggers transient dysbiosis that allows an initial S.Tm bloom but also permits fast microbiota recovery that suppresses further growth and renders the luminal pathogen population susceptible to clearance by subsequent inflammatory responses (Fig. 2). As a consequence, avirulent mutants, confined to the gut lumen, are selected against, while reseeding tissue-resident Salmonella (carrying functional SPI-1 and SPI-2 virulence factors) establish a luminal population that retains its full virulence (Fig. 6). In contrast, antibiotic pretreatment extensively depletes the microbiota, allowing the luminal S.Tm population to expand to high densities that resist clearance even when SPI-1 and SPI-2 are defective (Fig. 6A, B). Incomplete clearance of luminal S.Tm permits the overgrowth of avirulent mutants, particularly hilD mutants, which can arise through spontaneous mutation¹¹⁷ and replicate faster than wild-type⁵⁹. Our study suggests that diet can mediate pathogen:microbiota interactions that favor virulence.

The LDSs under study are commonly used as medications or food additives. Arabinose has antidiabetic properties, including sucrase inhibition¹¹⁸, while lactulose is widely used as both a laxative and a treatment for hyperammonemia^119,120. Maltitol, erythritol, and sorbitol are common sugar substitutes found in low-calorie diet foods¹²¹. Although our findings may raise food safety concerns, infection susceptibility was dose-dependent and occurred only at large, acute doses of ~ 5 g/kg body weight. Moderate intake did not produce the same effect. A human clinical study found that a 50 g dose of lactulose did not alter gut microbiota composition or trigger E. coli blooms in healthy adults¹²², suggesting that tolerable doses are unlikely to compromise CR. In our study, lower arabinose doses failed to weaken CR and failed to induce pronounced gut acidification. Thus, moderate fermentation was insufficient to overwhelm host buffering capacity. Organic acids are neutralized by host-derived bicarbonate, secreted via the SLC26 family of Cl⁻/HCO₃⁻ and SCFA/HCO₃⁻ exchangers, thereby maintaining pH along the intestinal tract^123,124. Disruption in these transporters causes colitis and microbiota dysbiosis in mice¹²⁵ and is associated with increased inflammatory bowel disease incidence in humans^126–128. These studies underscore that pH homeostasis supports a resilient gut microbiota.

While evidence supports the safety of LDSs at moderate doses, the literature indicates that outcomes can vary depending on dietary context and duration of intake. For example, fructo-oligosaccharides and xylo-oligosaccharides, touted as prebiotics for promoting Lactobacillus and Bifidobacterium growth, suppress the growth of E. coli, C. jejuni, and S. Enteritidis in vitro¹²⁹. However, prolonged supplementation in mice with 10% fructo-oligosaccharides in mice chow or water increased systemic Salmonella loads and fecal shedding when consumed two weeks prior to pathogen challenge¹³⁰. Moreover, fructo-oligosaccharides consumption amplified Salmonella blooms in rats fed on a low calcium, high-fat Western diet¹³¹. Histopathology was not assessed in these studies, making it unclear whether there was overt salmonellosis, as hallmark symptoms like colitis and diarrhea were not directly evaluated. Together, these findings emphasize that although moderate intake of LDSs may generally be safe, specific contexts can disrupt microbial homeostasis and promote pathogen expansion.

Several limitations should be considered. Our findings are based on animal models, and the contribution of CR to human enteric infection susceptibility remains inferred rather than directly established. We also examined only a limited set of individual carbohydrates, whereas human diets involve complex mixtures that may exert synergistic or buffering effects on the microbiota. Whether similar mechanisms operate in more physiologically relevant dietary contexts remains to be determined. Nevertheless, our study highlights the importance of evaluating not only the nutritional but also the microbiological consequences of emerging dietary interventions. Given the impact of dietary compounds on pathogen virulence, we recommend that prebiotic tolerability testing include assessments of infection susceptibility.

Methods

Experimental model and subject details

Mouse lines

All mice used in this study originated from C57BL/6 J breeders, initially obtained from Jackson Laboratories and housed at the EPIC mouse facility of ETH Zurich, Switzerland. Mice were maintained on a standard chow diet (Kliba Nafag, 3537; autoclaved), consisting of 4.5% fat, 18.5% protein, approximately 50% carbohydrates, and 4.5% fiber by weight. Mice were bred and maintained under specific pathogen-free, full-barrier conditions in individually ventilated cages at the ETH Zurich EPIC mouse facility, with a 12-h light/12-h dark cycle, a room temperature of 21 ± 1 °C, and a relative humidity of 50 ± 10%. For our experiments, we used 8- to 12-week-old mice, randomly assigning animals of either sex to experimental groups. All studies were conducted in accordance with ethical and legal requirements and were reviewed and approved by the Kantonales Veterinäramt Zürich under the licenses: ZH193/2016, ZH158/19, and ZH109/22.

Bacteria strains

All Salmonella enterica serovar Typhimurium (S.Tm) strains used in this study are isogenic to S.Tm SB300, a re-isolate of SL1344¹³² from a systemic mouse infection in the Galán lab¹³³. This strain was genome re-sequenced and is naturally resistant to streptomycin.

Method details

Bacteria strains and culture conditions

Strain construction. Novel single-gene deletion strains were constructed using the λ Red recombination system¹³⁴. PCR products amplified from pKD3 or pKD4 plasmid template using primers with 40-nt flanking sequences of the target gene were electroporated into wild-type S. Typhimurium harboring the pKD46 plasmid. Recombinant strains were then selected on LB agar supplemented with appropriate antibiotics. P22 phage lysates were prepared from the original mutant strains and used to transduce the mutation into the ancestral strain. A complete list of strains is provided in Supplementary Data.

S.T m strains were routinely grown overnight at 37 °C in Lysogeny broth (LB,¹³⁵) with agitation under ambient conditions. Strains were stored at −80 °C in 1₃% peptone 5% glycerol broth.

Mouse infection experiments

Preparation of the inoculum

Bacteria were grown overnight for 12 h in high salt (0.3 M NaCl) LB medium supplemented with the appropriate antibiotics, then sub-cultured for 4 h in antibiotic-free (0.3 M NaCl) LB, washed twice in PBS, and resuspended in PBS. Mice were inoculated via oral gavage with 5 × 10⁷ bacterial and housed in individually ventilated cages.

Carbohydrate administration

Carbohydrates (listed in Supplementary Data) were dissolved in water, gently heated until fully dissolved, cooled to room temperature, and administered to mice via oral gavage.

Sample collection

Feces and intestinal contents were weighed and suspended in PBS. Liver, spleen segments, and mesenteric lymph nodes were homogenized in PBS/Tergitol (0.5% v/v)/BSA (0.5% w/v) using a TissueLyser (Qiagen) and plated on MacConkey agar supplemented with the appropriate antibiotics.

Gentamicin protection assay

Intestinal tissue segments were cut longitudinally, washed three times in PBS to remove luminal contents, and incubated in 400 μg/mL gentamicin (in PBS) for 45 min to kill extracellular bacteria. Samples were then washed six times in PBS to remove residual gentamicin and subsequently homogenized in PBS/Tergitol (0.5% v/v)/BSA (0.5% w/v) using a TissueLyser to release intracellular bacteria. Tissue suspensions were plated on MacConkey agar supplemented with the appropriate antibiotics to quantify viable S.Tm.

Inflammation assessments

Intestinal tissue samples were cryo-embedded for histopathological scoring as described in ref.³². 5 μm cryosections were prepared and stained with hematoxylin and eosin (H&E)¹³⁶ to access enteropathy severity. Fecal Lipocalin-2 levels were measured using ELISA (DuoSet Lipocalin ELIZA kit, DY1857, R&D Systems, Minneapolis, MN, USA).

Intestinal barrier function

Using chambers were used to evaluate intestinal permeability as previously described¹³⁷. Paracellular and transcellular permeability were assessed using 6 μCi/ml of 51Cr-EDTA and 5 × 10⁵ M HRP (type II; Sigma Aldrich) probes, respectively. Serosal 51Cr-EDTA was quantified using a liquid scintillation counter and reported as percent recovery/cm²/h.

Metabolite measurements

Sample preparation

Cecum contents were collected, suspended in PBS and homogenized (without beads) using a Tissuelyzer (Qiagen) at 25 Hz for 2.0 min, followed by centrifugation for 5 min at 20,000 x g to pellet cells and particulate matter. Cell-free supernatants were collected and then centrifuged for 40 min at 20,000 x g, 4 °C to remove any remaining suspended particulate matter. Clear supernatants were transferred to clean microfuge tubes and stored at −80 °C until thawed for LC-MS processing. For acetate, butyrate, formate, lactate, and succinate, samples were derivatized using methods from¹³⁸. In brief, samples were mixed with isotopic internal standards and reacted with the derivatizing reagents, 3NPH (3-nitrophenylhydrazone) and N3D (3-dimethylaminopropyl-N’-ethylcarbodiimide (EDC)·HCl) for 45 min at 40 °C. The reaction was quenched with trifluoroacetic acid (TFA) at a final concentration of 0.08% and then diluted in 70% isopropanol to a final concentration of 250 µg cecal content per ml. Derivatized samples were stored at −80 °C until measurement. For amino acids, the samples were mixed with an isotopic standard (¹³C5-labeled glutamate) and diluted in HILIC buffer to achieve a final concentration of 250 µg cecal content per ml. Samples were then stored at −80 °C until measurement.

Liquid chromatography–mass spectrometry

For each compound, calibration curves were made to determine its quantifiable concentration range. LC-MS analyses were carried out with an Ultimate 3000 ULPC instrument (Thermo Scientific, Waltham, MA, USA) hyphenated to a QExative plus mass spectrometer (Thermo Scientific, Waltham, MA, USA). In case of short-chain fatty acid and carboxylate analysis, we applied a previously described method with slight modifications¹³⁸. For LC separation, we used a Kinetex XB C18 column (particle size 1.7 µm; 50 × 2.1 mm, Phenomenex, Torrance, CA, USA) with 0.1% formic acid in water (solvent A) and 0.1% formic acid in acetonitrile as mobile phases. The sample injection volume was 2 µl. At a flow rate of 500 µl min⁻¹, we run a linear gradient with solvent B changing as follows: 0 min: 2%; 3 min: 95%; 5 min: 95%; 5.3 min: 2%. Subsequently, the column was equilibrated for 2 min at the initial condition. The applied amino acid analysis method was based on a previously described Method¹³⁹ adapted to the MS instrument. For LC separation, we used an AQUITY BEH NH2 column (particle size 1.7 µm; 100 × 2.1 mm, Waters, Milford, MA, USA) with a buffer consisting of 10 mM ammonium formate (short buffer) in water: acetonitrile (50:50) (solvent A) and the same buffer in acetonitrile: water: methanol (95:5:5) as mobile phases. To shift pH, 1.76 ml of formic acid was added to both mobile phases. The sample injection volume was 2 µl. At a flow rate of 500 µl min⁻¹ we run a linear gradient with solvent B changing as follows: 0 min: 84.3%; 1.5 min: 84.3%; 5.5 min: 5.3%; 7.5 min: 5.3%; 8 min: 84.3%, 10 min: 84.3%. Heated electro spray ionization was performed in both cases by FTMS at a mass resolution of 70’000 (at m/z 200). SCFAs and carboxylates were measured in the negative mode applying a spray voltage of −2.7 kV and amino acids in the positive mode with a spray voltage of +3.4 kV. All other source parameters were the same and set as follows: spray voltage: 2.7 kV; S-lens RF level: 50; sheath gas: 50; aux gas: 20; sweep gas: 0; aux gas heater: 350 °C.

Bacterial flow cytometry

SicA promoter activity was assessed in mice using a plasmid-based PsicA-gfp reporter⁵⁹. Fresh feces collected 8 h.p.i. were diluted in PBS and incubated for 1 h at room temperature with 2 μg ml⁻¹ chloramphenicol (AppliChem) to inhibit protein synthesis and allow GFP maturation. S.Tm was stained using a human anti-O12 antibody (hSTA5, gift from A. Lanzavecchia, IRB Bellinzona) and a goat anti-human IgG AF647 secondary antibody (Jackson ImmunoResearch). Fluorescence was measured on a Cytoflex flow cytometer (Beckman Colter) using CytExpert v2.5, with gating as described in ref.¹⁴⁰.

16S rRNA gene sequencing

Fecal pellets and cecum contents were snap-frozen and stored at −20 °C. For gDNA extraction, samples were homogenized in RLT buffer with 3 mm metal beads using a TissueLyser II (Qiagen) for 3 min at 30 Hz, followed by two rounds of bead-beating with 0.1 mm zirconia beads. Lysates were processed using the AllPrep DNA/RNA Mini Kit (Qiagen) per the manufacturer’s protocol. The V4 region of the 16S rRNA gene was amplified using primers 806 R (5’-GGACTACNVGGGTWTCTAAT-3’) and 515 F (5’-GTGYCAGCMGCCGCGGTAA-3’)^141,142, and libraries were multiplexed with Illumina Unique Dual Indexing Primers. Sequencing was performed on an Illumina NextSeq2000 (2 × 300 bp, 450 bp inserts) at BMK, Germany. Reads were trimmed with cutadapt¹⁴³ and processed using the DADA2 pipeline¹⁴⁴; ASVs were annotated using IDTAXA with the SILVA v138 database^145,146. Raw data were deposited at the European Nucleotide Archive (ENA), Accession numbers: PRJEB115268; ERP195487

Carbohydrate utilization of human microbiota

Unless otherwise stated, Bacteroides ovatus ATCC-8483, Bacteroides thetaiotaomicron DSM No.: 2079, Bacteroides vulgatus DSM No.: 1447, Bacteroides uniformis DSM No.: 6597, Clostridum perfringens DSM No.: 11782, Clostridum ramosum DSM No.: 1402, Coprococcus comes ATCC No.: 27758, Dorea formicigenerans DSM No.: 3992, Lactobacillus paracasei ATCC No: SD5275, Odoribacter splanchnicus DSM No.: 20712, Roseburia intestinalis DSM No.: 14610, Ruminococcus gnavus ATCC No.: 29149, S.Tm SB300, and Streptococcus parasanguinis DSM No.: 6778 were routinely grown in YCFA medium (YCFA—Yeast extract-Casein Hydrolysate-Fatty Acids; DSMZ medium 1611) supplemented with 0.5% (w/v) glucose at 37 °C in the anaerobic chamber (Coy Laboratory Products Inc., 2% H2, 12% CO2, and the balance N2). YCFA supplemented with glucose promotes the growth of all the previously mentioned bacterial strains and allows us to replace the carbohydrate source. Therefore, YCFA medium was used as a base medium to examine carbohydrate utilization of human microbial strains. YCFA medium was prepared without a carbohydrate and supplemented with 2x concentrations (1%) of arabinose or glucose (AppliChem, LOT: 64015153. Following supplementation and filtering of the YCFA medium, supplemented and unsupplemented YCFA medium were pipetted into a deep well plate (Thermo Fischer Scientific). Then, 50 µl of medium was distributed to each well of a 96-well u-bottom screening plate (Thermo Fischer Scientific) by using Epmotion 96 (Eppendorf) and stored at −30 °C until use. The human microbial strains were inoculated in YCFA (supplemented with 0.5% glucose) and cultured anaerobically at 37 °C before being subcultured. On the day of subculturing, the stored 96-well plates were thawed, brought into the anaerobic chamber, and pre-reduced. Following O/N incubation, 1.5 mL of each subculture was aliquoted in Eppendorf tubes and centrifuged at RT for 1 min. Supernatants were removed from the Eppendorf tubes, and the bacterial pellets were resuspended in YCFA medium without a carbohydrate source, allow to avoid glucose residues to influence the growth in other carbohydrate sources. OD578 measurements of the resuspended bacteria culture were taken, and the cultures were diluted to an OD578 of 0.02 in YCFA medium (without carbohydrate). Furthermore, 50 µl of the diluted culture per well was used to inoculate the wells of the 96-well plate containing 50 µl of the YCFA supplemented with/without carbohydrate sources, resulting in a starting of a OD578 of 0.01 bacterial culture supplemented with/without 0.5% of carbohydrates.

After inoculation, plates were sealed with breathable membranes (Breathe-Easy) to prevent evaporation and cross-contamination between wells. OD578 was measured hourly for 24 h after 30 s of linear shaking using a microplate spectrophotometer (Epoch2, BioTek, Gen5 software, version 3.05) and an automated microplate stacker (BioStack 4, BioTek) installed in a custom-built incubator (EMBL Mechanical Workshop). At least three biological replicates of each compound-strain combination were tested. Data analysis was performed as previously described (Maier et al.¹⁴⁷). Briefly, growth curves were blanked, truncated at stationary phase entry, AUCs calculated, and normalized to control growth curves (YCFA without carbohydrate source).

pH measurements of human microbiota

A photometric pH assay for microplate bacterial cultures was modified to measure the pH of human microbial strains growing in different carbohydrate sources, resulting in a simple and affordable pH measurement protocol (Takao et al., 2020). The following stock solutions were prepared in dimethyl sulfoxide (DMSO; Sigma-Aldrich): 2.5% methyl orange (MO; TCI chemicals), 1% methyl red (MR; Sigma-Aldrich), and 20% bromothymol blue (BTB; TCI chemicals). The following volumes of the stock solutions were then freshly mixed in acetonitrile (AcCN; TCI chemicals) to prepare pH indicator mixtures: 360 µl (0.01%) MO, 480 µl (0.0053%) MR, 180 µl (0.04%) BTB, 5490 µl (0.63 mM) 10 mM NaOH, 5490 µl water, 75000 µl (86.2%) AcCN. The pH standards were simultaneously prepared using a series of McIlvaine buffers (pH 4.0–7.8) (McIlvaine, 1921). A 40 μL aliquot of either buffer/medium/culture and a 90 μL aliquot of AcCN-pH indicator solution were mixed in a 96-well v-bottom plate (Greiner Bio-One). After centrifugation at 500 x g for 1 min in a plate centrifuge (Thermo Fisher Scientific), each 60 μL of supernatant was transferred to a well of a black, clear-bottom plate (Greiner Bio-One) and the plates were sealed with breathable membranes (Breathe-Easy) to prevent evaporation. The absorbance values at 540 nm and 620 nm were then measured by a plate reader (Tecan), and Δ(540–620) was calculated. The estimation of the pH of each sample was done by obtaining the calibration curve of the measured pH values against Δ(540–620). At least three biological replicates of each compound-strain combination were tested.

Growth rates of B. thetaiotaomicron, E. rectale, and S.Tm in varying pH

Data for the dependence of growth rates on pH for B. thetaiotaomicron (ATCC 29148) and Eubacterium rectale (ATCC 33656) are published in ref. ¹⁴⁸, and are shown here for comparison. For this study, we have measured the growth of S.Tm at different pH values under the same conditions described in ref. ¹⁴⁸. In short, strains were cultured under anaerobic conditions (10% CO2, 5% H2, 85% N2) in growth medium containing 1% tryptone, 20 mM glucose, 100 mM KH₂PO₄/K₂HPO₄ adjusted to the desired pH, 50 mM NaCl, 0.5 mM CaCl₂, 0.4 mM MgCl₂, 50 μM MnCl₂, 50 μM CoCl₂, 4 μM FeSO₄, 5 mM cysteine, 20 mM NaHCO₃, 5 mM Na₂SO₄, 20 mM NH₄Cl, 1.2 mg/L hemin, 1 mg/L menadione, 2 mg/L folinic acid, and 2 mg/L vitamin B12. To reach lower pH values, pH was adjusted using HCl.

WITS-distribution analysis

Genomic DNA was prepared from overnight cultures, using commercial kits (DNA Mini Kit, Qiagen). qPCR analysis was performed using the FastStart Universal SYBR Green Master Mix (Roche). The thermocycling programs and primers were described previously^149,150. The population size of each tagged strain was calculated by taking the product of the number of kanamycin-resistant colonies/g feces and the ratio of WITS determined by qPCR.

Flow cytometry analysis of cecum mucosa immune cell composition

Flow cytometry analysis of cecum mucosa immune cell composition was performed as previously described⁸¹. Specifically, mice were euthanized, and the cecum was excised and opened longitudinally. To detach epithelial cells, the cecal tissue was cut into small fragments and incubated twice in 14 ml PBS containing 5 mM EDTA, 15 mM HEPES, and 10% FCS for 20 min at 37 °C with gentle agitation. The samples were then washed in RPMI supplemented with 30% FCS and transferred to RPMI containing 1 mg/ml Collagenase VIII (Sigma) and 0.2 mg/ml DNase I (Roche). The tissue fragments were digested for 1 h at 37 °C with mild shaking. Following digestion, the suspension was passed through a 70 μm cell strainer and washed with RPMI. Cells were resuspended in RPMI and gently layered onto a NycoPrep 1.077 gradient (Progen). After centrifugation for 30 min at 400 g, cells at the interphase were collected, rinsed with RPMI containing 15% FCS, and resuspended in PBS with 1% FCS. For intracellular staining, surface marker staining was performed first. Cells were then fixed and permeabilized for 40 min at room temperature in the dark using Foxp3 Fix/Perm buffer (eBioscience/Thermo Fisher). After washing with Permeabilization/Wash buffer (eBioscience/Thermo Fisher), cells were stained for an additional 40 min at room temperature in the dark. Antibodies and dyes used in this study included: CD45 (30-F11, BioLegend), CD4 (RM4-5, Thermo Fisher), Foxp3 (FJK-16s, Thermo Fisher), F4/80 (BM8, BioLegend), CD103 (2E7, BioLegend), MHCII (M5/114.15.2, BioLegend), CD11b (M1/70, BioLegend), CD11c (N418, BioLegend), Sytox Blue (Invitrogen), and Zombie NIR (BioLegend).

Real-Time qPCR for analysis of transcript levels

Real-Time qPCR was performed as previously described⁸¹. Mid-segments of the cecum were rinsed in cold sterile PBS to remove luminal contents, shock-frozen in RNAlater (Qiagen) using liquid nitrogen, and stored at −80 °C until RNA isolation. Total RNA was purified with the RNeasy Mini Kit (Qiagen) following the manufacturer’s protocol. RNA concentrations were adjusted to 0.125 µg µl⁻¹, and cDNA was synthesized using the RT2 HT First Strand cDNA Kit (Qiagen). Gene expression analysis was performed with custom RT2 Profiler PCR arrays and primer assays (Qiagen) using Universal FastStart SYBR Green Master (ROX; Roche) on a StepOne Plus Real-Time PCR System (Applied Biosystems). The thermal cycling program consisted of an initial denaturation at 95 °C for 14 min, followed by 40 amplification cycles of 94 °C for 15 s, 61 °C for 30 s, and 72 °C for 20 s. Transcript levels were normalized to actin-b expression, and a Cq cut-off of 38 cycles was applied.

Quantification and statistical analysis

Statistical analysis

The Mann–Whitney or Wilcoxon test was performed using GraphPad Prism Version 10 for Windows (GraphPad Software, La Jolla, California, USA, www.graphpad.com). Statistical details of each experiment are provided in the corresponding figure legends and in the Source Data File, which contains the underlying raw data. Where sample sizes (n) or P-values are not explicitly indicated in the figure legends, they are provided in the Source Data File. Horizontal bars in the graphs indicate medians where applicable. P values of <0.05 (two-tailed) were considered statistically significant. ^∗ P < 0.05, ^∗∗ P < 0.01, ^∗∗∗ P < 0.001.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceived and designed the experiments, B.D.N., S.Y.W., L.M., and W.-D.H.; performed the experiments, B.D.N., S.Y.W., L.M., C.F., Y.S., S.K., E.F.V, N.A., M.A., D.L.B., A.H., C.S., P.K., P.C., M.B., and L.F.; provided protocols, A.S.; provided conceptual input, E.G; analyzed the data, B.D.N., S.Y.W., L.M., C.F., Y.S., S.K. E.F.V, N.A., M.A., D.L.B., A.H., C.S., P.K., L.F., and W.-D.H.; wrote the paper, B.D.N., and W.-D.H.; supervision, B.D.N., E.S., J.A.V., and W.-D.H.

Peer review

Peer review information

Nature Communications thanks anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was funded by grants from the Swiss National Science Foundation (407240_167121, 310030B_173338, 310030_192567, 310030B_201265, 10.001.588, and NCCR Microbiomes grant 51NF40 180575 and 51NF40_225148). Open access funding provided by Swiss Federal Institute of Technology Zurich.

Data availability

The raw sequencing data generated in this study have been deposited in the European Nucleotide Archive (ENA) under accession numbers PRJEB115268 and ERP195487. Source data are provided with this paper (Source Data File). Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-78218-0.