Ketogenic diet exacerbates DSS-induced colitis through a β-hydroxybutyrate-Thomasclavelia spiroformis-γδ17 T cell axis in mice

Abstract

Ketogenic diet (KD) is widely recognized for its immunomodulatory and metabolic benefits, but the impact on inflammatory bowel disease remains controversial. Here, we demonstrate that KD maintains homeostasis under physiological conditions but exacerbates colitis by triggering a ketogenesis-microbe-immune cascade upon mucosal injury. Mechanistically, KD elevates luminal β-hydroxybutyrate (β-HB), promoting the expansion of Thomasclavelia spiroformis (T. spiroformis). In turn, T. spiroformis activates colonic γδ17 T cells via cell wall components, ultimately driving IL-17A-mediated iinflammation. Adoptive transfer of γδ17 T cells into Tcrd^-/- mice confirmed their pathogenicity. Ketogenesis or IL-17A blockades abolish KD-exacerbated colitis, whereas β-HB supplementation or ketogenesis activation recapitulated disease exacerbation. Clinically, T. spiroformis abundance correlates with fecal β-HB and serum IL-17A in ulcerative colitis (UC) patients, but not Crohn’s disease, supporting a UC-specific β-HB-T. spiroformis-γδ17 T cell axis. Thus, we identify a diet-induced immunometabolic circuit linking ketogenesis to colitis, highlighting ketone metabolism and IL-17A signaling as potential therapeutic targets.

Ketogenic diet (KD) generates β-hydroxybutyrate metabolites and has been shown to have context-dependent immunomodulatory effects. However, the effects of KD on inflammatory bowel disease (IBD) are controversial. Here, the authors investigate the pathogenic effects of KD in a preclinical model of DSS-induced colitis and identify a β-HB-Thomasclavelia spiroformis-γδ17 T cell axis that exacerbates inflammation and mucosal damage.

Introduction

Inflammatory bowel disease (IBD), including Crohn’s disease (CD) and ulcerative colitis (UC), represents a global health challenge with a rapidly increasing incidence in industrialized nations, particularly in East Asia^1,2. While genetic predisposition and immune dysregulation contribute to IBD pathogenesis, the clinical manifestations differ between subtypes^3,4. CD is characterized by transmural inflammation that may affect any segment of the gastrointestinal tract, while UC manifests as superficial inflammation localized to the colon⁵. Notably, emerging evidence highlights dietary composition as a pivotal environmental driver of colitis progression^6–8. High-fat diets have drawn particular attention for their potential to induce gut dysbiosis, disrupt mucosal homeostasis, and promote chronic inflammation^9,10. Therefore, systematic investigation of dietary intervention strategies targeting inflammatory infiltration has become imperative for optimizing clinical management of IBD.

Ketogenic diet (KD), characterized by a very high-fat and low-carbohydrate composition, induces hepatic production of ketone bodies, primarily β-hydroxybutyrate (β-HB, 78%), acetoacetate (20%), and acetone (2%), by mimicking starvation metabolism. While KD has shown therapeutic efficacy in neurological disorders, metabolic diseases, and some cancers, its immunomodulatory effects appear to be context-dependent^11,12. Specifically, whether KD can be used for IBD intervention is still under debate due to a lack of convincing experimental and clinical data. KD is reported to alleviate acute and chronic colitis by altering the gut microbiota to reduce colonic group 3 innate lymphoid cells (ILC3), but also shows the potential to worsen colitis through p53-mediated cellular senescence and microbiota-dependent barrier disruption^13–15. This contrasting evidence reinforces critical gaps in understanding how KD modulates the inflammatory microenvironments, particularly regarding their interactions with gut microbiota and immune responses.

In this study, we delineate a context-dependent immunometabolic circuit through which KD exacerbates colitis upon mucosal injury. We establish that diet-derived β-HB selectively enriches the gut pathobiont Thomasclavelia spiroformis, which subsequently activates colonic γδ17 T cells to drive IL-17A-mediated pathology. These findings bridge dietary metabolism, gut microbiota, and mucosal immunity, defining a previously unrecognized pathogenic axis. Consequently, this work not only resolves a key controversy regarding KD in IBD but also highlights host-microbe metabolic crosstalk as a fundamental regulatory layer in colitis pathogenesis, revealing the ketone-IL-17A pathway as a potential therapeutic target.

Results

Ketogenic diet feeding exacerbates colitis associated with IL-17A upregulation in the context of mucosal injury

To evaluate the impact of a ketogenic diet (KD) on intestinal homeostasis and colitis susceptibility, we first assessed its effects in healthy mice (Fig. 1a). After 7 days of KD feeding, we observed decreased body weight but a significant increase in colon length compared to the control diet (CD) (Fig. 1b, c). Although KD altered serum levels of glucose, total cholesterol, triglycerides, and high-density lipoprotein (Supplementary Fig. 1a), histopathological analysis revealed no differences in intestinal injury and epithelial barrier integrity (Fig. 1d; Supplementary Fig. 1b), or liver function and histology (Supplementary Fig. 1c, d) compared to controls. Notably, short-term KD reduced colonic Tnf mRNA while increasing Il17a mRNA (Fig. 1e), encoding two markers of intestinal inflammation representing distinct inflammatory pathways, suggesting a diet-induced cytokine imbalance that may prime susceptibility to inflammatory triggers.

Fig. 1. KD aggravates colitis progression in the presence of intestinal mucosal disruption.

a Schematic illustration. Wild-type C57BL/6JNarl mice were fed a CD or KD diet for 7 days. b Body weight loss (n = 8 mice/group). c Colon length (n = 8 mice/group). d Representative H&E, PAS, and MUC2 staining of colon tissues (scale bar, 300 µm), with PAS and MUC2 positive area quantification (n = 8 mice/group). e Relative expression of Tnf and Il17a in colon (n = 8 mice/group). f Schematic illustration. Colitis was induced in mice by 2.5% DSS treatment combined with CD or KD feeding for 7 days. g Body weight loss, stool consistency index, and rectal bleeding index (n = 12 mice/group). h DAI (n = 12 mice/group). i Colon length (n = 12 mice/group). j Representative H&E and PAS staining of colon tissues (scale bar, 300 µm), with histology score and PAS positive area quantification (n = 12 mice/group). k Representative TEM of intestinal epithelial layer (scale bar, 1.0 μm, n = 5 mice/group). l Serum FITC-dextran levels (n = 10 mice/group). m Representative immunofluorescence images of ZO-1 in colon (scale bar, 100 µm), with ZO-1 positive area quantification (n = 5 mice/group). n Representative MUC2 staining of colon tissues (scale bar, 300 µm), with MUC2 positive area quantification (n = 12 mice/group). o Percentages of TNF and IL-17A expressing cells among T cells in the LP of the colon (n = 6 mice/group). p Schematic illustration. Il10 KO mice were given CD and KD for 4 weeks. q Body weight loss (n = 5 mice/group). r Colon length (n = 5 mice/group). s, t Representative H&E staining of colon tissues (scale bar, 300 µm), with histology score (n = 5 mice/group). u Representative MUC2 and IL-17A staining of colon tissues (scale bar, 300 µm), with MUC2 and IL-17A positive area quantification (n = 5 mice/group). v TNF and IL-17A levels in serum (n = 5 mice/group). Data are presented as mean ± SEM. Multiple t-test or two-sided unpaired Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001. Source data are provided as a Source Data file.

We next induced colitis via dextran sulfate sodium (DSS) in mice fed either a KD or CD (Fig. 1f). KD-DSS mice displayed markedly exacerbated disease progression compared to CD-DSS controls, including accelerated weight loss, reduced stool consistency, increased rectal bleeding, elevated disease activity index (DAI), and colon shortening (Fig. 1g–i). Histopathology, Alcian blue staining, ZO-1 immunofluorescence, and MUC2 immunohistochemistry (IHC) demonstrated severe intestinal injury, goblet cell loss, and epithelial barrier disruption in KD-DSS colons. These findings were corroborated by transmission electron microscopy (TEM), which revealed disorganized epithelial junctions, and by elevated serum FITC-dextran levels confirming barrier dysfunction (Fig. 1j–n). Flow cytometry analysis identified an increase in colonic IL-17A in KD-DSS mice, while TNF levels remained comparable between groups (Fig. 1o; Supplementary Fig. 1e). This cytokine profile was similarly observed in splenic cell populations (Supplementary Fig. 1f, g), consistent with the cytokine shift detected in healthy KD-fed mice, and implicating IL-17A as a key mediator of KD-driven colitis exacerbation. The reproducibility of this pro-colitic effect of KD was confirmed across independent facilities (Supplementary Fig. 1h–m).

We next examined the effects of long-term KD on gut physiology (Supplementary Fig. 2a). Chronic KD feeding induced sustained weight loss, reduced liver index, and colon shortening without compromising intestinal barrier integrity (Supplementary Fig. 2b–f). Serum IL-1β, IL-6 and IL-17A levels were even reduced following prolonged KD (Supplementary Fig. 2g), suggesting an improved gut immune microenvironment. However, when colitis was induced by DSS after 15 weeks of KD feeding, mice exhibited worsened disease progression characterized by exacerbated epithelial barrier disruption and increased colonic IL-17A levels (Supplementary Fig. 2h–n). Since the DSS-induced colitis model does not fully recapitulate the complex immunopathology of human UC, we employed Il10 knockout mice to validate our finding. When subjected to a 4-week KD, these mice exhibited exacerbated enterocolitis, although this was less severe than the exacerbation seen with short-term KD (Fig. 1p–v). It should be noted that KD exposure exacerbates colitis in both DSS and Il10 knockout models, while prolonged KD feeding also worsens colitis, though to a lesser extent than short-term KD exposure. This temporal pattern suggested that prolonged KD feeding may induce intestinal immune adaptation.

Collectively, these findings demonstrate that while KD induces metabolic changes without compromising intestinal barrier integrity in healthy mice, it exacerbates colitis associated with IL-17A upregulation when mucosal injury is present.

γδ17 T cells are specifically expanded and mediate mucosal crosstalk in colitis exacerbated by KD feeding

To elucidate the mechanisms of KD-driven immune dysregulation in colitis, we performed single-cell RNA sequencing (scRNA-seq) on colon tissues from DSS-treated mice fed either CD or KD. The analysis encompassed epithelial cells, endothelial cells, B cells, macrophages, dendritic cells, granulocytes, monocytes, fibroblasts, and T cell subsets (Fig. 2a, b). Among the 10 distinct T cell clusters identified (Fig. 2c, d), KD-DSS mice showed increased frequencies of γδ17 T cells (Cluster 7), Gzma⁺ CD8⁺ T cells (Cluster 1), FoxP3⁺ CD4⁺ T cells (Cluster 4), and Il22⁺ CD4⁺ T cells (Cluster 9), but reduced frequencies of Klrd1⁺ NK cells (Cluster 0), Tnfsf8⁺ CD8⁺ T cells (Cluster 2), CD28⁺ CD4⁺ T cells (Cluster 3), and Klrc1⁺ NK cells (Cluster 5) compared to CD-DSS group (Fig. 2e).

Fig. 2. IL-17-producing γδ T cells accumulation in KD-DSS group.

a UMAP of all cell types. b Gene expression points map of cell cluster. c UMAP visualization in the CD-DSS and KD-DSS groups. d Distribution of T-cell cluster in the CD-DSS and KD-DSS groups. e Percentage of T-cell cluster. f Enrichment of Il1b, Tnf, and Il17a gene expression in each cluster. g Cluster is sorted by gene expression for enrichment, showing the top 9 genes. h The five most common markers of γδ17 T cells are expressed in cluster 7. i Percentage and number of γδ T cells in the LP of the colon (n = 6 mice/group). j Percentage and number of γδ17 T cells in the LP of the colon (n = 6 mice/group). k Percentage and number of Th17 T cells in the LP of the colon (n = 6 mice/group). l Differential ligand receptor interactions between γδ T cells and other cells. Data are presented as mean ± SEM. Two-sided unpaired student’s t-test. *p < 0.05, ***p < 0.001. Source data are provided as a Source Data file.

scRNA-seq analysis revealed that IL-17A, a pivotal driver of intestinal inflammation, showed predominant expression in γδ17 T cells, while the pro-inflammatory cytokines IL-1β and TNF localized to Cluster 8/9 (Fig. 2f, g). This expression pattern aligns with previous studies establishing γδ17 T cells as key IL-17A producers during colitis initiation^16–18. Signature genes characterizing γδ17 T cells included Cd163l1, Trdv4, Trdc, and Pdcd1 (Fig. 2h), consistent with established γδ T cell markers and further validating their identity¹⁹. Moreover, flow cytometry validated that γδ T cells and γδ17 T cells, but not Th17 cells, another IL-17A-producing population²⁰, were significantly expanded in the lamina propria (LP) of the KD-DSS colons (Fig. 2i–k; Supplementary Fig. 3a). In contrast, splenic cell populations from KD-DSS mice showed expansion of both γδ/γδ17 T cells and Th17 cells (Supplementary Fig. 3b–g). This systemic Th17 cell response may be facilitated through IL-17-dependent crosstalk with γδ T cells²¹.

To explore intercellular crosstalk, we employed CellphoneDB and observed enhanced ligand-receptor interactions between γδ17 T cells and stromal/epithelial cells in KD-DSS colons. The top interacting partners included Lcn2⁺ and Lyz2⁺ epithelial cells, Il1b⁺/Cxcl2⁺ fibroblasts, and Plvap⁺ endothelial cells (Supplementary Fig. 3h). Notably, γδ17 T cell-derived IL-17A engaged IL-17RA/IL-17RC receptors on Lcn2⁺, Cxcl2⁺, Lyz2⁺, and Tff3⁺ epithelial cells (Fig. 2l), suggesting direct modulation of epithelial responses. Correspondingly, Lcn2⁺ epithelial cells in KD-DSS mice showed increased expression of genes involved in mucin biosynthesis (Agr2), complement regulation (Cfh), and chemokine signaling (Ccl11), accompanied by enhanced extracellular matrix remodeling marked by elevated Mmp3 and Timp3 expression (Supplementary Fig. 3i), all processes that are implicated in IBD pathogenesis^22–24.

Taken together, these results identify colonic γδ17 T cells as the dominant source of IL-17A in KD-exacerbated colitis, where they mediated crucial mucosal interactions that amplify epithelial dysfunction.

KD-primed γδ17 T cells drive IL-17A-dependent exacerbation of colitis

To establish the functional role of γδ17 T cells in KD-exacerbated colitis, we employed T cell receptor δ chain-deficient (Tcrd^−/−) mice in combination with adoptive transfer approaches. DSS-induced colitis was modeled in Tcrd^−/− mice, which were then divided into CD-DSS, KD-DSS, or KD-DSS groups receiving γδ T cell transfers (Fig. 3a). Flow cytometry confirmed successful engraftment of γδ T cells in recipient mice (Fig. 3b). No significant differences, except for body weight loss, were observed between fed CD-DSS and KD-DSS-treated Tcrd^−/−mice, while γδ T cell transfer profoundly exacerbated colitis severity in KD-DSS-treated Tcrd^−/− mice. This was revealed by accelerated weight loss, elevated DAI, colon shortening, and worsened histopathological damage and mucosal disruption compared to KD-DSS-treated Tcrd^−/− mice (Fig. 3c–f). These findings underscore that KD acts by amplifying the initial γδ17 T cell response rather than by enhancing the endpoint inflammatory cascade.

Fig. 3. γδ17 T cells are the major player in KD-DSS induced mucosal injury.

a Schematic illustration of the experimental protocol for γδ T-cell adoptive transfer and KD-DSS treatment. b Flow cytometry plots of γδ (CD3⁺γδTCR⁺) T cells in the LP of the colon (left). Percentage of γδ T cells in CD3⁺ cells in the LP of the colon (right, n = 3 mice/group). c, Body weight loss (n = 5 mice/group). d DAI (n = 5 mice/group). e Colon length (n = 5 mice/group). f Representative H&E and MUC2 staining of colon tissues (scale bar, 300 µm), with histology score and MUC2 positive area quantification (n = 5 mice/group). g Schematic illustration. Colitis was induced in mice by 2.5% DSS treatment combined with KD feeding for 7 days, while the mice were administered 200 μg IgG (control) or anti-IL-17A intraperitoneally daily. h Body weight loss and DAI (n = 5 mice/group). i Colon length (n = 5 mice/group). j TNF levels in serum (n = 5 mice/group). k IL-17A levels in serum (n = 5 mice/group). l Representative H&E staining of colon tissues (scale bar, 200 µm) and histology score (n = 5 mice/group). m Representative TNF and IL-17A staining of colon tissues (scale bar, 300 µm). n TNF and IL-17A positive area quantification (n = 5 mice/group). o Schematic illustration. Il17a KO mice were used to establish a colitis model by treating 2.5% DSS combined with KD feeding for 7 days. p Body weight loss and DAI (n = 9 mice/group). q Colon length (n = 9 mice/group). r TNF levels in serum (n = 6 mice/group). s Representative H&E staining of colon tissues (scale bars, 600/300 µm) and histology score (n = 5 mice/group). t Fecal β-HB levels (n = 9 mice/group). Data are presented as mean ± SEM. One-way ANOVA, two-way ANOVA or two-sided unpaired Student’s t-test. *, Tcrd^−/−+CD-DSS vs Tcrd^−/−+KD-DSS; #, Tcrd^−/−+KD-DSS vs Tcrd^−/−+KD-DSS + γδT cells; *, CD-DSS+IgG vs KD-DSS+IgG; #, KD-DSS+IgG vs KD-DSS+anti-IL-17A; */#p < 0.05, **/##p < 0.01; ***/###p < 0.001. Source data are provided as a Source Data file.

To directly implicate γδ17 T cell-derived IL-17A in disease pathogenesis, IL-17A was neutralized in KD-DSS mice using an anti-IL-17A antibody (Fig. 3g). IL-17A blockade markedly attenuated colitis progression, reducing DAI, colon shortening, serum and colonic IL-17A concentrations, and histopathological damage (Fig. 3h–n). Confirmatory experiments in IL17a knockout (IL17a^−/−) mice similarly abolished the KD-mediated exacerbation of colitis (Fig. 3o–s), and thus established the central role of IL-17A in KD-exacerbated disease. Collectively, these data delineate a mechanistic axis whereby KD feeding specifically increased γδ17 T cells that produce IL-17A, thereby amplifying colonic injury and inflammatory progression.

β-Hydroxybutyrate drives γδ17 T cell differentiation and IL-17A-dependent colitis exacerbation

The KD stimulates hepatic production of ketone bodies, including AcAc and its downstream product β-HB, which function as alternative energy substrates or are excreted via urine and feces. Consistently, serum β-HB levels were markedly elevated under both short-term and long-term KD regimens compared to CD (Supplementary Fig. 4a–c). More importantly, regardless of DSS treatment, only β-HB showed significant accumulation in the fecal samples from KD-fed mice, while AcAc even decreased (Fig. 3t; Fig. 4a and Supplementary Fig. 4d). These observations implicate β-HB as a potential mediator of KD-driven colitis exacerbation. To confirm the crucial role of β-HB in KD-aggravated colitis progression, we supplemented healthy mice with exogenous β-HB (1.5–12 mg/kg). While β-HB alone did not affect baseline gut physiology, supplementation with 6 mg/kg β-HB in KD-DSS mice was sufficient to recapitulate colitis severity, as shown by worsened mucosal damage, increased colonic IL-17A levels, and enhanced γδ17 T cell differentiation (Fig. 4b–g; Supplementary Fig. 4e–h). These findings support the role of β-HB in driving IL-17A-dependent pathology.

Fig. 4. β-HB is necessary for colitis progression.

a Fecal AcAc and β-HB levels in CD or KD + DSS-treated mice (n = 12 mice/group). b Schematic illustration. 2.5% DSS + CD for 7 days with twice-daily gavage of saline or 6 mg/kg β-HB. c Body weight loss, stool consistency index, rectal bleeding index and DAI (n = 8 mice/group). d Colon length (n = 8 mice/group). e Representative H&E and MUC2 staining of colon tissues (scale bars, 300/100 µm), with histology score and MUC2 positive area quantification (n = 8 mice/group). f Relative expression of Il17a in colon (n = 8 mice/group). g Percentage and number of γδ17 T cells in the LP of the colon (n = 6 mice/group). h Schematic illustration. 2.5% DSS + CD for 7 days with daily gavage of saline or 30 mg/kg EMPA. i Fecal β-HB levels (n = 8 mice/group). j Body weight loss, stool consistency index, rectal bleeding index and DAI (n = 8 mice/group). k Colon length (n = 8 mice/group). l Representative H&E and MUC2 staining of colon tissues (scale bar, 300 µm), with histology score and MUC2 positive area quantification (n = 8 mice/group). m Relative expression of Il17a in colon (n = 8 mice/group). n Percentage and number of γδ17 T cells in the LP of the colon (n = 5 mice/group). o Schematic illustration. 2.5% DSS + KD for 7 days with daily gavage of saline or 2.5 mg/kg TAC. p Fecal β-HB levels (n = 10 mice/group). q Body weight loss, stool consistency index, rectal bleeding index and DAI (n = 10 mice/group). r Colon length (n = 10 mice/group). s Representative H&E and MUC2 staining of colon tissues (scale bar, 300 µm), with histology score and MUC2 positive area quantification (n = 10 mice/group). t Relative expression of Il17a in colon (n = 10 mice/group). u Percentage and number of γδ17 T cells in the LP of the colon (n = 5 mice/group). Data are presented as mean ± SEM. Two-sided unpaired Student’s t-test or multiple t-test. *p < 0.05, **p < 0.01, ***p < 0.001. Source data are provided as a Source Data file.

Next, to further validate the central role of β-HB in KD-aggravated colitis progression, we evaluated three drugs reported to regulate the ketogenic process: the ketogenic enhancer empagliflozin (EMPA)²⁵ and the ketogenic inhibitors tacrolimus (TAC)²⁶ and simvastatin (SIM)²⁷, in DSS-induced colitis models. EMPA enhances ketogenesis by upregulating β-hydroxybutyrate dehydrogenase 1 (BDH1), the rate-limiting enzyme in ketogenesis that converts AcAc to β-HB. In contrast, TAC and SIM suppress ketogenesis by either downregulating HMGCS2 or directly inhibiting HMGCR, which are involved in ketogenesis. In the CD-DSS-induced colitis model, EMPA increased fecal β-HB levels and exacerbated colitis, as evidenced by higher DAI scores, shortened colon length and worsened histological damage (Fig. 4h–l). EMPA treatment also led to elevated colonic Il17a expression and increased γδ17 T cell frequency and numbers in the lamina propria (Fig. 4m, n; Supplementary Fig. 4i). Conversely, TAC significantly inhibited ketogenesis, reduced fecal β-HB levels, and thus attenuated colitis severity, lowering Il17a expression and γδ17 T cell differentiation in KD-DSS mice (Fig. 4o–u; Supplementary Fig. 4j). The therapeutic effect observed may be attributed to β-HB reduction, although a contribution from its established immunosuppressive properties cannot be ruled out. Importantly, SIM, a non-immunosuppressive inhibitor, also alleviated KD-DSS colitis by reducing β-HB and colonic Il17a expression (Supplementary Fig. 5a–f), while having no significant effect in CD-DSS mice (Supplementary Fig. 5g–k). These interventions support the pivotal role that ketogenesis, and specifically β-HB, plays in mediating KD-induced γδ17 T cell differentiation and IL-17A-driven colitis exacerbation.

Gut microbiota remodeling mediates β-HB-driven colitis exacerbation via Thomasclavelia enrichment

To determine whether β-HB directly promotes γδ T cell differentiation or induces intestinal epithelial cell apoptosis, we treated T cells and Caco-2 cells with β-HB. No significant effects on γδ T cell differentiation or epithelial apoptosis were observed (Supplementary Fig. 6a–c), ruling out a cell-autonomous mechanism. Based on previous studies linking KD to gut microbiota alterations^28,29, we next investigated whether microbial remodeling contributes to the pro-colitis effects of KD. Remarkably, depletion of gut microbiota using an antibiotic (ABX) cocktail completely abolished both the KD-induced increase in colonic IL-17A (Supplementary Fig. 7a–f) and the aggravated colitis phenotype in KD-DSS mice (Fig. 5a–c; Supplementary Fig. 7g, h). Histopathological analysis confirmed that microbiota depletion alleviated mucosal damage in both dietary groups, eliminating KD-specific effects (Fig. 5d). We further performed fecal microbiota transplantation (FMT) from KD-DSS donors to recipient mice that were maintained on a control diet. Recipients of KD-DSS microbiota developed exacerbated colitis, showing higher DAI scores, shorter colon lengths, worsened histopathology, and epithelial damage, along with elevated IL-17A levels compared to FMT-CD-DSS recipients (Fig. 5e–i; Supplementary Fig. 7i–k), demonstrating microbiota-dependent pathology.

Fig. 7. UC-specific correlations among β-HB, T. spiroformis, and serum IL-17A.

a Flow chart of study design. b Correlation of fecal β-HB concentration with relative abundance of T. spirofomis. c Correlation of fecal AcAc concentration with relative abundance of T. spirofomis. d Correlation of fecal β-HB concentration with serum IL-17A levels. e Correlation of relative abundance of T. spirofomis with serum IL-17A levels. f T. spirofomis abundance in male and female groups across UC and CD cohorts. g Correlation of the relative abundance of T. spirofomis with age. h T. spirofomis abundance in low- and high-level fecal β-HB groups across UC and CD cohorts. i Serum IL-17A levels in low- and high-level fecal β-HB groups across UC and CD cohorts. j Bacterial abundance in low- and high-level fecal β-HB groups in UC cohort. k Bacterial abundance in low- and high-level fecal β-HB groups in CD cohort. l Serum biochemical indices in low- and high-level fecal β-HB groups in UC cohort. m Serthe um biochemical indices in low- and high-level fecal β-HB groups in the CD cohort. n Illustration of the β-HB-driven expansion of T. spiroforme and subsequent IL-17A-mediated inflammation in UC. Data are presented as mean ± SEM. Two-sided unpaired Student’s t-test or multiple t-test. *p < 0.05, **p < 0.01, ***p < 0.001. Source data are provided as a Source Data file.

Fig. 5. Gut microbiota from KD-DSS mice accelerate the colitis progression.

a Schematic illustration. Mice were fed a control diet (CD) or a ketogenic diet (KD) along with an antibiotic cocktail and 2.5% DSS. b Body weight loss, stool consistency index, rectal bleeding index and DAI (n = 5 mice/group). c Colon length (n = 5 mice/group). d Representative H&E and PAS staining of colon tissues (scale bar, 300 µm), with histology score and PAS positive area quantification (n = 5 mice/group). e Schematic illustration. The recipient mice were given a combined antibiotic treatment for 7 days, followed by 2 days of recovery. Fecal microbiota transplantation was then performed using fresh fecal slurry from donor mice that had undergone CD-DSS or KD-DSS treatment during Days 2 to 7. f Body weight loss and stool consistency index (n = 5 mice/group). g Rectal bleeding index and DAI (n = 5 mice/group). h Colon length (n = 5 mice/group). i Representative H&E, MUC2 and IL-17A staining of colon tissues (scale bar, 300 µm), with histology score, MUC2 and IL-17A positive area quantification (n = 5 mice/group). j Bacterial taxonomic profiling at the genus level of gut microbiota on Day 1, 2, and 4 from CD or KD + DSS treatment (n = 3 mice/group). k Alpha diversity analysis based on observed OTUs (n = 3 mice/group). l Relative abundance of genus Akkermansia and genus Thomasclavelia based on 16S rRNA data analysis (n = 3 mice/group). m Relative abundance of genus Thomasclavelia based on qPCR (n = 12 mice/group). n Relative abundance of species T. spiroformis, T. ramose, T. saccharogumia, and T. cocleata based on qPCR (n = 12 mice/group). Data are presented as mean ± SEM. Multiple t-test or two-sided unpaired Student’s t-test. *p < 0.05, **p < 0.01, ***p < 0.001. Source data are provided as a Source Data file.

Furthermore, 16S rRNA sequencing of DSS-treated mice revealed rapid gut microbiota remodeling under KD feeding. KD-DSS mice exhibited significantly increased richness (observed OTU), decreased α-diversity (Chao index), and marked compositional shifts (Fig. 5j, k; Supplementary Fig. 8a–d). At the genus level, KD-DSS mice displayed early dysbiosis characterized by decreased Akkermansia (a mucus barrier-enhancing probiotic)³⁰ and expansion of Thomasclavelia as early as day 2 (Fig. 5l, m). This dysbiotic pattern persisted throughout the dietary intervention period (Supplementary Fig. 8e–i). Thomasclavelia remained elevated in KD-fed mice after 1-, 2-, and 16-week feeding timepoints, while Akkermansia abundance showed context-dependent dynamics, maintaining high levels during KD feeding alone, but suppressed in KD-DSS mice (Fig. 5l; Supplementary Fig. 8g, i). Importantly, mono-colonization with A. muciniphila failed to ameliorate KD-DSS-induced colitis (Supplementary Fig. 9a-f), excluding its causal role in disease exacerbation.

Focusing on Thomasclavelia, qPCR quantification of four species within this genus showed that KD feeding, independent of DSS treatment, selectively amplified T. spiroformis by more than 30-fold, while T. ramose increased by approximately 3-fold, with no alteration in T. cocleata or T. saccharogumia (Fig. 5m, n; Supplementary Fig. 8e). Moreover, the proliferation of T. spiroformis and T. ramose did not directly inhibit the growth of probiotic species that were decreased in DSS (Supplementary Fig. 10a). This species-specific enrichment identifies T. spiroformis, and to a less extent T. ramose, as potential contributors to KD-driven colitis exacerbation.

KD promotes γδ17 T cell differentiation and colitis exacerbation potentially through T. spiroformis-derived cell wall components

To validate T. spiroformis as the primary microbial responder to β-HB in KD-exacerbated colitis, we first conducted in vitro anaerobic co-culture assays. Although low concentrations of β-HB promoted the proliferation of some probiotics, this effect was primarily due to acidification of the culture medium by β-HB, as the growth-promoting effect largely disappeared when the medium was adjusted to the same pH across conditions (Supplementary Fig. 10b–e). In contrast, β-HB enhanced T. spiroformis growth, but not T. ramose, in a dose-dependent manner even in pH-controlled medium (Fig. 6a; Supplementary Fig. 10b). Moreover, both in vivo β-HB supplementation and pharmacological modulation of ketogenesis selectively alter T. spiroformis abundance in the fecal microbiota (Supplementary Fig. 4k–n), confirming its specific responsiveness to β-HB.

Fig. 6. T. spiroformis exacerbates colitis through bacterial cell wall components-mediated innate γδ17 T cell activation.

a Growth curve of T. spiroformis containing β-HB (0-10 mM). Vertical axis shows bacterial DNA levels (n = 3 biological replicates/group). b Schematic illustration. 2.5% DSS + CD for 7 days with daily gavage of saline, T. spiroformis or T. ramose. c Relative abundance of T. spiroformis and T. ramose (n = 5 mice/group). d Body weight loss and stool consistency index (n = 5 mice/group). e Rectal bleeding index and DAI (n = 5 mice/group). f Colon length (n = 5 mice/group). g Representative H&E and MUC2 staining of colon tissues (scale bar, 300 µm), with histology score and MUC2 positive area quantification (n = 5 mice/group). h Flow cytometry plots of IL-17A⁺ T cells, and the percentage in colonic LP (n = 5 mice/group). i Flow cytometry plots of γδ17 T cells, and the percentage in colonic LP (n = 5 mice/group). j Schematic illustration. 2.5% DSS + CD for 7 days with daily gavage of saline, T. spiroformis or pasteurized T. spiroformis. k Relative abundance of T. spiroformis (n = 5 mice/group). l Body weight loss, stool consistency index, rectal bleeding index and DAI (n = 5 mice/group). m Colon length (n = 5 mice/group). n Representative H&E and MUC2 staining of colon tissues (scale bar, 300 µm), with histology score and MUC2 positive area quantification (n = 5 mice/group). o Schematic illustration of bacterial component of T. spiroformis and its effects on T cell differentiation. p Flow cytometry plots of γδ T and γδ17 T cells, and the percentage after treatment with blank medium or T. spiroformis supernatant (n = 4 biological replicates/group). q Flow cytometry plots of γδ T and γδ17 T cells, and the percentage after treatment with blank, live T. spiroformis, pasteurized T. spiroformis or T. spiroformis lysate (n = 4 biological replicates/group). Data are presented as mean ± SEM. Two-way ANOVA, one-way ANOVA or two-sided unpaired Student’s t-test. *, CD-DSS+Veh vs CD-DSS + T. sp; #, CD-DSS+Veh vs CD-DSS + T.r/CD-DSS+pT. sp; */#p < 0.05, **/##p < 0.01, ***p < 0.001. Source data are provided as a Source Data file.

To clarify the pathogenic role of T. spiroformis, we mono-colonized CD-DSS mice with either T. spiroformis or the less responsive species T. ramose (Fig. 6b, c). Only T. spiroformis exacerbated colitis, inducing severe histopathology characterized by epithelial barrier disruption (Fig. 6d–g). Colonization with T. spiroformis significantly increased colonic IL-17A levels and promoted γδ17 T cell differentiation (Fig. 6h, i), while T. ramose showed no such effects. To further investigate the underlying mechanism, we administered both live and pasteurized T. spiroforme to CD-DSS mice. Notably, heat-inactivated T. spiroforme produced a similar colitis-exacerbating effect as the live bacteria (Fig. 6j–n), suggesting that bacterial components mediate this effect rather than bacterial viability.

To identify the specific component of T. spiroformis responsible for promoting γδ17 T cell differentiation, we evaluated live bacteria, pasteurized bacteria, culture supernatant, and bacterial lysates that disrupt cell wall integrity (Fig. 6o). Both live and heat-inactivated bacteria consistently enhanced γδ17 T cell differentiation, whereas neither the culture supernatant nor the lysates exhibited this activity (Fig. 6p, q). This pattern, which requires intact bacterial structure but not viability, strongly implicates bacterial cell wall components as the active mediators. Given that T. spiroformis is a Gram-positive bacterium with a cell wall predominantly composed of peptidoglycans, we propose that peptidoglycans of T. spiroformis are the most likely effector molecules driving γδ17 T cell differentiation. Altogether, these findings suggest a mechanistic cascade in which β-HB enriches T. spiroformis, whose peptidoglycan components might drive γδ17 T cell-mediated IL-17A production, thereby exacerbating colitis.

Clinical validation of the β-HB-T. spiroforme-IL-17A axis in UC progression

To investigate the clinical role of the β-HB-T. spiroforme-IL-17A axis in human IBD, we analyzed paired fecal and serum samples from 66 IBD patients (31 UC and 35 CD, Fig. 7a). In UC patients, fecal β-HB, but not AcAc, was positively correlated with T. spiroforme abundance (Fig. 7b, c). Notably, although fecal β-HB levels did not directly correlate with serum IL-17A levels, T. spiroforme abundance was significantly associated with higher serum IL-17A, suggesting a potential microbial-mediated link to intestinal inflammation (Fig. 7d, e). These associations were absent in CD patients, highlighting a UC-specific interaction (Fig. 7b–e). Furthermore, T. spiroforme abundance was independent of sex and age in both diseases, reinforcing its specific association with β-HB and IL-17A (Fig. 7f, g).

We stratified patients by median fecal β-HB levels and found that UC patients in the high-level β-HB group showed significantly elevated T. spiroforme abundance and serum IL-17A levels compared to the low-level group, while no such differences were observed in CD patients (Fig. 7h, i). Microbiota analysis further demonstrated consistent patterns of crucial genus (Thomasclavelia, Akkermansia, Bifidobacterium, Lactobacillus) and their representative species across both diseases, suggesting T. spiroforme may act as a UC-specific core microbe (Fig. 7j, k). Moreover, our analysis revealed that β-HB levels did not influence blood biochemical indices in IBD patients (Fig. 7l, m). These clinical findings corroborate the β-HB-T. spiroforme-IL-17A axis as a disease-specific pathway in UC and align with our mechanistic data from mice models, indicating that β-HB-mediated enrichment of T. spiroforme and subsequent IL-17A production may contribute to UC pathogenesis (Fig. 7n).

Discussion

The KD has gained popularity for its therapeutic benefits in obesity, epilepsy, and polycystic ovary syndrome^13,31. However, its broader health impacts remain under investigation, with potential adverse effects reported in hypertension, heart failure, and autoimmune disorders^32,33, and risks of tumor metastasis and cellular senescence^14,34. These findings suggest the need for a comprehensive understanding of the health implications of KD. Our study aimed to clarify the impact of KD on IBD progression and to explore the underlying mechanisms. Herein, we demonstrate that KD exacerbates colitis under mucosal injury conditions. β-HB, the primary ketone body induced by KD, enriches T. spiroformis within the gut microbiota. This enrichment increases colonic γδ17 T cells, potentially through bacterial peptidoglycans, leading to IL-17A-dependent inflammation and epithelial barrier disruption. Clinical validation in IBD patients further confirmed the relevance of the β-HB-T. spiroformis-IL-17A axis in UC. Notably, while IL-17A is recognized as a driver of inflammatory progression in IBD^35–37, its correlation with disease severity is evident in UC but absent in CD^38,39. This underscores the disease-specific roles of IL-17A in IBD subtypes and positions our mechanistic axis as a UC-selective pathway.

Diet composition strongly shapes gut microbiota and inflammation in a physiopathology-dependent manner. High-protein or high-fat diets often reduce beneficial bacteria and increase pro-inflammatory metabolites, worsening colitis, yet may also alleviate aging-related pathologies by limiting pathogenic translocation^40–43. Similarly, KD modulates gut microbiota and inflammation depending on physiological state. In a healthy gut, KD alone does not damage the epithelium but increases ketone bodies that inhibit bifidobacteria and reduce Th17 cells³¹, while enriching protective taxa such as A. muciniphila and Faecalibacterium prausnitzii, enhancing the intestinal barrier function and reducing seizures^13,44. However, under fiber deficiency, KD can exacerbate colitis by depleting anti-inflammatory microbes or promoting mucus erosion^45–47. Specifically, under epithelial injury, the effects of KD become conflicting. One study reported that KD alleviates acute and chronic colitis by altering the gut microbiota to reduce colonic ILC3¹³, whereas others found that KD exacerbates colitis by disrupting the mucosal barrier and altering gut microbiota¹⁵. These findings demonstrate that diet-induced microbiota shifts are not inherently beneficial or harmful but depend on disease context and microbial niches.

Our study refines this understanding by showing that the effects of KD on colitis depend on dietary duration and the underlying pathophysiological state. Neither short-term nor long-term KD alone damaged a healthy gut as previously reported. Interestingly, different from long-term treatment, short-term KD reduced colonic Tnf mRNA expression but significantly increased colonic Il17a mRNA expression, indicating a cytokine imbalance that may prime susceptibility to inflammatory triggers. However, under epithelial injury, KD exacerbates colitis by impairing epithelial barrier function and promoting inflammation via a β-HB–T. Spiroformis–γδ17 T cell–IL-17A axis. Colonic γδ17 T cells, activated by microbiota-derived signals via RORγt signaling, are early producers of IL-17A and drive acute mucosal injury^16,48, while Th17 cells, dependent on commensals for differentiation, primarily contribute to chronic inflammation in the small intestine^49,50. Together, these findings clarify previous contradictions and suggest the need for timing- and context-specific dietary interventions in IBD management.

KD induces a metabolic shift toward ketogenesis, wherein the ketone bodies β-HB and AcAc serve as alternative energy sources^51,52. Hepatic fatty acid β-oxidation generates AcAc, converted to β-HB by BDH1, with β-HB reaching approximately 70% of total ketones during KD⁵³. β-HB engages in a bidirectional manner with the gut microbiota. It remodels microbial communities by reducing Bifidobacterium³¹ and enriching Lactobacillus⁵⁴, while certain gut bacteria such as Parabacteroides distasonis can, in turn, enhance hepatic β-HB production by upregulating fatty acid oxidation enzymes⁵⁵. Moreover, β-HB regulates cellular functions through histone β-hydroxybutyrylation, which regulates energy metabolism, DNA repair, and tumorigenesis^56–58. These dual roles position β-HB as a central integrator of systemic metabolic adaptations and immunometabolism reprogramming in the intestinal microenvironment. Emerging evidence highlights metabolic homeostasis as a therapeutic target in colitis. Recent studies demonstrated that lipid peroxidation drives colitis progression, while pharmacological inhibition of its crucial regulator, ACSL4 (via pioglitazone or PRGL493), attenuates disease severity⁵⁹. Similarly, pharmacological modulation of ketogenesis using EMPA (ketogenic enhancer) or TAC/SIM (ketogenic inhibitors) allows precise control of cecal β-HB levels and colitis severity in our models. Although TAC’s immunosuppressive properties may have contributed to the observed reduction in IL-17A and T cell infiltration, the fact that SIM, a ketogenesis inhibitor without comparable immunosuppressive activity, produced a congruent protective effect underscores that the reduction of β-HB is the central mechanism. However, the more modest effect of SIM compared to TAC could reflect the complexity of the system. Potential explanations include compensatory mechanisms within the hepatic ketogenesis pathway that bypass HMGCR inhibition, and the limited efficacy of targeting upstream ketogenesis once the downstream inflammatory cascade is fully established. This underscores that optimal therapeutic intervention in complex diseases like colitis may require multi-target approaches within the metabolic-immune network. These results emphasize a critical clinical consideration for colitis management that both KD and pharmacological agents that inhibit hepatic ketogenesis may significantly influence treatment outcomes, and the efficacy of such agents may be modulated by the complexity of the host’s metabolic and immune state.

The gut microbiota influences intestinal microecology and host immunity through metabolites, niche interactions, and structural components. Microbial metabolites, including short-chain fatty acids and bile acids, act as critical mediators linking microbial activity to host homeostasis^60–62. Clinical studies, such as a double-blind trial in UC patients, have shown that FMT promotes remission by increasing butyrate-producing bacteria and enhancing short-chain fatty acid and bile acid pathways⁶³. Engineered E. coli can generate inositol triphosphate to activate epithelial repair pathways, further illustrating metabolite-driven benefits⁶⁴. Beyond metabolites, live bacteria maintain ecological balance by competing with pathogens. An 18-strain bacterial consortium alleviates colonic inflammation by competitively suppressing Enterobacteriaceae growth while preserving commensal microbiota balance⁶⁵. Interestingly, non-viable bacterial components also modulate immunity through conserved surface molecules (e.g., Amuc-1100 of A. muciniphila) or cell wall components (e.g., peptidoglycan from a gram-positive bacterium)^66,67. Our study extends these insights by showing that T. spiroformis exacerbates colitis and promotes γδ17 T cell differentiation not only when live but also in its heat-inactivated form, indicating that viability is not required for its pro-inflammatory effects. In contrast, bacterial lysates lose this activity. Given the Gram-positive nature of T. spiroformis, we propose that its peptidoglycans potentially act as the effector driving γδ17 T cell-mediated IL-17A production, thereby worsening colitis under KD conditions. In our mono-colonization experiments, the use of broad-spectrum antibiotics to create a microbially depleted gut environment non-specifically removes the entire commensal community. This likely amplifies the observed pro-inflammatory effect of T. spiroformis by eliminating ecological constraints (e.g., nutrient competition) and anti-inflammatory microbes that would otherwise modulate its activity in a complex microbiota. Thus, while this reductionist model powerfully isolates causality, the true pathogenic potential of T. spiroformis in vivo is likely shaped by its interactions with the surrounding microbial network.

There are some limitations to this study. Firstly, prolonged KD worsened colitis less severely than short-term KD, indicating gaps in understanding how diet duration influences immunity, possibly due to adaptations in ketone metabolism, microbiota composition, or immune plasticity. Secondly, the cross-sectional design of our clinical analysis, which lacks detailed dietary intake data, limits causal interpretation. Therefore, the correlations observed in UC patients, while mechanistically supported by animal data, should be considered hypothesis-generating. Future prospective studies incorporating dietary monitoring are necessary to establish causal links in human populations. Lastly, although we identified T. spiroformis peptidoglycan as a potential activator of γδ17 T cells, the structural features distinguishing pathobiont-derived peptidoglycan from commensal variants are undefined, and the specific host recognition mechanism remains elusive. Comparative glycomics and peptidoglycan fragmentomics of T. spiroformis versus non-pathogenic T. ramosa could reveal strain-specific modifications that underpin pro-inflammatory activity. Furthermore, CRISPR-based screening in immune reporter cell lines could unbiasedly identify the host receptor responsible for sensing these unique bacterial features. These limitations warrant future experiments.

Overall, this work advances understanding of diet-microbiota-immune crosstalk by identifying the β-HB-T. spiroformis-γδ17 T cell axis as a conserved driver of KD-exacerbated colitis across mice and humans. Our findings demonstrate that dietary modulation of gut microbiota is a causal, not merely associative, factor in shaping host immunity, with β-HB functioning as both a metabolic intermediary and a pathogenic catalyst. By bridging microbial ecology, immunology, and translational relevance, this study positions the gut microbiome as a therapeutic frontier for diet-modulated diseases.

Methods

Animals

Male, 10 to 12-week-old, wild-type C57BL/6 JNarl mice were purchased from Nanjing Institute of Biomedicine (Catalog#N000013). Il10 knockout mice (C57BL/6J background, 16 to 18-week-old, male, SPF grade) were purchased from Aniphe Biolaboratory Inc (Catalog#NM-KO-19015). Tcrd^−/− mice (C57BL6/J background, 10 to 13-month-old, male, SPF grade) were purchased from Shanghai Model Organisms (Catalog#NM-KO-190435). Il17a knockout mice (C57BL6/J background, 31-week-old, male, SPF grade) were purchased from Shanghai Model Organisms (Catalog#NM-KO-00131). All mice were maintained in a conventional barrier facility under standard conditions, with a 12 h light/dark cycle and sufficient water and food supplies. To isolate microbiota-mediated effects, mice from different experimental groups were housed separately to prevent coprophagy. At the end of all experiments, mice were euthanized by deep anesthesia with isoflurane inhalation followed by cervical dislocation. Ethics approval: All animal experiments were approved by the Animal Ethics Committee of Shanghai Tenth People’s Hospital (Approval No. SHDSYY-2021-3031), and the Eastern Hepatobiliary Surgery Hospital of the Naval Medical University (EDWLL-007).

Human sample collection and ethical compliance

Fecal and serum samples were collected from ulcerative colitis and Crohn’s disease patients under protocols approved by the Ethics Committee of Longhua Hospital (Approval No. 2024LCSY119). Informed consent was obtained from all participants. For the clinical data presented, specific measures have been taken to protect participant anonymity, including the presentation of age as a range in Supplementary Table 1. No combination of data presented allows for the identification of any individual participant.

Diet

The ketogenic diet (KD; 6.4% protein, 0.49% carbohydrate, 93.1% fat) and control diet (CD; 18.26% protein, 65.67% carbohydrate, 16.07% fat) were formulated according to published compositions from Envigo (KD: Catalog#TD.07797, CD: Catalog#TD.150300) and custom-manufactured by Beijing Fubo Biotechnology Co., Ltd. Detailed compositions of CD and KD are shown in Supplementary Table 2.

DSS-induced colitis model

Dextran sodium sulfate (2.5% DSS, MP Biomedical, Catalog# S7102) was given to 10- to 12-week-old SPF-grade C57BL/6JNarl mice for 7 days to induce colitis, and KD or CD was given concurrently for 7 days, with tissue and serum samples from mice collected on the 8^th day.

Antibiotics cocktail intervention

To deplete gut microbiota, mice were administered an antibiotic cocktail containing metronidazole (100 mg/kg), vancomycin (50 mg/kg), and neomycin sulfate (100 mg/kg) via oral gavage every 12 h. Concurrently, ampicillin (1 mg/mL) was supplemented in drinking water⁶⁸. This regimen was maintained throughout the experimental period to ensure sustained microbial suppression.

Fecal microbiota transplantation

Donor C57BL/6JNarl mice were administered DSS to induce colitis, followed by either a KD or CD. Daily fecal samples were collected, homogenized in sterile PBS, centrifuged (500 × g, 5 min), and supernatants administered to recipient mice via oral gavage (200 μL/mouse/day) for 5 consecutive days. Recipient C57BL/6JNarl mice underwent 7-day gut microbiota depletion using the antibiotic cocktail described above, followed by a 2-day washout period. Recipients were maintained on CD plus DSS treatment and received daily FMT from KD- or CD-colitis donors.

Microbe supplementation

Akkermansia muciniphila (A. muciniphila) and Thomasclavelia spiroformis (T. spiroformis) were cultured anaerobically to final concentrations of 4 × 10⁹ CFU/mL and 4 × 10⁸ CFU/mL, respectively. Prior to microbe supplementation, C57BL/6JNarl mice underwent a 7-day antibiotic regimen to deplete the gut microbiota, followed by a 2-day washout period. A. muciniphila (1 × 10⁹ CFU in 250 μL saline), T. spiroformis (1 × 10⁸ CFU in 250 μL saline) or vehicle control (PBS) were administratered via daily oral gavage (at 12 h intervals).

β-HB supplementation

To determine the optimal β-HB dose, C57BL/6JNarl mice fed a CD received daily oral gavage of β-HB at gradient concentrations (1.5, 3, 6, 12 mg/kg) for 6 days. Fecal samples were collected 4 h post-administration to monitor microbial composition. Based on dose-response outcomes, 6 mg/kg β-HB was selected for subsequent colitis studies. In the DSS induced-colitis model, mice were administered 6 mg/kg β-HB via oral gavage every 12 h for 7 consecutive days.

Il10 knockout mouse model

Il10 knockout (Il10 KO) mice 16- to 18-week-old with a C57BL/6 J background, were given CD and KD for 4 consecutive weeks. Il10 KO mice developed spontaneous enteritis with age.

γδ T cells adoptive transfer

Spleens from wild-type C57BL/6 J mice were aseptically dissected, mechanically dissociated, and filtered through a 70 μm strainer. γδ T cells were isolated using a BD FACS Aria II flow cytometer. Cells were centrifuged and resuspended in PBS. Tcrd^−/− mice received 1 × 10⁶ γδ T cells in 200 μL PBS via tail vein injection. Recipient mice were fed a CD or KD for 7 days, with 2.5% DSS administered in drinking water for the same duration. Samples were collected on day 8.

IL-17A neutralization

Mice received intraperitoneal injections of either control IgG or IL-17A neutralizing antibody (200 µg per mouse) on experimental days 2, 4, and 6. Antibodies were diluted in sterile PBS to a final volume of 200 µL per injection.

Il17a KO mouse model

Il17a^−/− mice were fed either CD or KD for 7 days. Concurrently, mice received 2.5% DSS in drinking water for 7 days. All samples were collected on day 8 post-treatment initiation.

Disease activity index assessment

To evaluate DSS-mediated intestinal damage and inflammatory responses, daily assessments were performed, measuring three parameters: body weight loss, stool consistency, and rectal bleeding. The scoring system was implemented as follows: body weight loss was calculated relative to the weight at baseline (100%). Weight loss of 1–5%, 5–10%, 10–20%, and >20% was scored as 1, 2, 3, and 4, respectively. Stool consistency was scored 0 for normal-formed pellets, 1 for soft but still formed stools, 2 for soft stools and 3 for watery diarrhea. To assess the occult blood in the stool, urine fecal occult blood test kit (Nanjing Jiancheng Bioengineering Institute, Catalog#C027-1-1) was used. 0 was scored for negative hemoccult, 1 for weakly positive hemoccult, 2 for positive hemoccult, 3 for visible blood traces in stool, and 4 for gross bleeding. The disease activity index was calculated as the total of these three scores, ranging from 0 (no inflammation) to 11 (severe colitis). The serum, liver, intestine and cecal content were collected at the endpoints of the experiments.

FITC-dextran permeability assay

Mice were fasted overnight and deprived of water for 4 h prior to oral gavage with 100 μL FITC-dextran (60 mg/mL, Sigma-Aldrich, Catalog#68059). Serum was collected 4 h post-gavage. Fluorescence intensity (490 nm excitation) was measured in 50 μL serum samples using a black 96-well plate.

Immunofluorescence staining

Frozen tissue sections (5–8 μm) were air-dried for 20 min at room temperature, washed with PBS (3 × 5 min), and blocked with 5% fetal bovine serum in PBS containing 0.1% Triton X-100 for 60 min at room temperature. Sections were incubated with primary antibodies (1:200 dilution in blocking buffer) for 60 min at room temperature, followed by PBS washes (3 × 5 min). Secondary antibodies (1:100 dilution) were applied for 60 min in the dark, with subsequent PBS washes (3 × 5 min). Nuclei were counterstained with DAPI (1:1000, 60 s), followed by final PBS washes (3 × 5 min). Slides were mounted with anti-fade medium and stored at 4 °C protected from light until imaging. Antibody details are provided in Supplementary Table 3.

Immunohistochemical staining

Paraffin-embedded tissue sections were deparaffinized, rehydrated, and endogenous peroxidase activity was blocked with 3% H₂O₂ for 20 min. Antigen retrieval was performed by heating in citrate buffer (pH 6.0). After blocking with 5% BSA at 37 °C for 30 min, sections were incubated with primary antibodies overnight at 4 °C. Following PBS washes, sections were incubated with HRP-conjugated secondary antibodies at 37 °C for 30 min. DAB (1:50 dilution) was applied for 30–120 s for chromogenic detection, followed by hematoxylin counterstaining, dehydration, and mounting. Antibody details are provided in Supplementary Table 3.

Fecal DNA extraction

Human or mice fecal DNA was extracted using the HiPure Stool DNA Kit (Magen, Catalog#D3141-03F) according to the manufacturer’s protocol. Briefly, 200 mg of stool samples were homogenized in lysis buffer, followed by bead-beating mechanical disruption. DNA was purified through spin-column binding, washed, and eluted in nuclease-free water. Quality and concentration were assessed by spectrophotometry and fluorometry. The primer details of the strain are shown in Supplementary Table 4.

16S rRNA gene sequencing

The V3-V4 hypervariable regions of bacterial 16S rRNA genes were amplified using barcoded universal primers (515 F: 5’-CCTAYGGGRBGCASCAG-3’, 806 R: 5’-GGACTACNNGGGTATCTAAT-3’). Pooled libraries were subjected to paired-end sequencing on an Illumina MiSeq (Illumina) platform.

RNA isolation, DSS removal and cDNA quantification

Colonic isolation and DSS removal were performed as previously described⁶⁹. Briefly, 30-50 mg tissues were homogenized in 1 mL TRIzol reagent (Invitrogen, Catalog#15596026), mixed with 200 μL chloroform, and centrifuged (12000 × g, 4 °C, 15 min). The aqueous phase was precipitated with isopropanol (−20 °C overnight), washed with 75% ethanol, and dissolved in DEPC water (final concentration >400 ng/μL). Residual DSS was removed by two rounds of LiCl precipitation (0.1 vol 8 mol/L LiCl, 2 h on ice; 14000 × g, 4 °C, 30 min), followed by NaAc/ethanol precipitation (20 μL 3 mol/L NaAc, 400 μL cold ethanol; −20 °C, 30 min). RNA pellets were washed with ethanol and resuspended in DEPC water. cDNA was synthesized with Evo M-MLV RT Mix Kit with gDNA Clean for qPCR Ver.2 (Accurate Biotechnology, China) according to the manufacturer’s instructions. Real-time PCR was performed using SYBR Green Premix Pro Taq HS qPCR Kit (Accurate Biotechnology, China) on a CFX384 Touch Real-Time PCR Detection System (Bio-Rad, USA). The results were analyzed by the ΔCt method and normalized to reference genes. The sequence of primers is shown in Supplementary Table 3.

Bacteria culture and preparation

Thomasclavelia spiroformis (T. spiroformis, ATCC 29900) was cultured anaerobically in Sweet E broth at 37 °C. Thomasclavelia ramose (T. ramose, ATCC 25582) was grown in Gifu Anaerobic Medium (HB8518-1, Hopebio) supplemented with 5 mg/L hemin and 1 mL/L 0.1% vitamin K1. Akkermansia muciniphila (A. muciniphila, ATCC BAA-835) was cultured in Brain Heart Infusion broth (BD 237500). All strains were maintained in an anaerobic workstation (E500, Gene Science) at 37 °C. All procedures were performed under strict anaerobic conditions unless specified.

For bacterial suspensions, cultures were centrifuged (4000 × g, 20 min, 4 °C), washed twice with sterile saline, and adjusted to 3 × 10⁸ CFU/mL. T. spiroformis culture supernatant was collected by centrifugation. For inactivation, T. spiroformis was pasteurized (75 °C, 30 min), with viability confirmed by culture. The bacterial lysates were prepared by subjecting bacterial suspensions to three freeze-thaw cycles (each cycle consisting of freezing in liquid nitrogen for 1 min followed by thawing in a 100 °C water bath for 2 min), with the resulting product subsequently filtered through a 0.22 μm membrane.

Serum β-HB assay

Serum β-hydroxybutyrate (β-HB) was assessed in a 96-well microplate by using a commercial β-HB assay kit (Nanjing Jiancheng Bioengineering Institute, Catalog#E030-1-1) and monitored at 340 nm for 5 min by using a Versamax microplate reader (Molecular Devices, CA, USA).

Fecal ketone bodies measurement

Fecal β-HB and AcAc levels were quantified using a modified LC-MS/MS protocol⁷⁰. Briefly, 10 mg of fecal sample was homogenized in 200 μL of ice-cold extraction solution containing 200 μM d3-acetoacetate internal standard in 50% aqueous methanol (1:20, w/v) and vortexed for 5 min. After centrifugation (14000 × g, 10 min, 4 °C), 80 μL of supernatant was transferred to a clean tube and derivatized with 10 μL each of 0.1 M butylhydroxylamine (BHA) and 0.25 M EDC in methanol at 25 °C for 1 h. The reaction mixture was then diluted 1:2 with 50% aqueous methanol and extracted with 600 μL dichloromethane by vigorous shaking for 10 min. Following phase separation by centrifugation, 400 μL of the lower organic layer was collected and evaporated to dryness using a vacuum concentrator (MiniVac, Gene Company Limited). The residue was reconstituted in 100 μL of 50% aqueous methanol, vortexed for 30 s, and centrifuged (14000 × g, 5 min, 4 °C) prior to LC-MS/MS analysis.

Chromatographic separation was performed on a Q Exactive HF hybrid quadrupole-orbitrap mass spectrometer (Thermo Fisher Scientific) equipped with an ACQUITY UPLC BEH C18 (2.1 × 50 mm, 1.7 μm, Waters Co.). The Mobile phases consisted of mobile phase A (80%): 0.1% formic acid in H₂O with 10 mmol/L NH₄COOH, mobile phase B (20%): 0.1% formic acid in MeOH: IPA (9:1, v/v). The flow rate was maintained at 0.3 mL/min, and the column temperature was set to 45 °C. Mass spectrometric detection was performed in positive electrospray ionization (ESI⁺) mode using an Orbitrap Elite mass spectrometer. Quantification was performed using external calibration curves, allowing absolute quantification of β-HB and AcAc concentrations in fecal samples.

Bacterial proliferation

The compound used for the bacterial growth assay was DL-β-hydroxybutyric acid (β-HB, Sigma Catalog#166898). β-HB was dissolved in water and sterilized by filtration to prepare a 1 M stock solution. β-HB was added during the preparation of the liquid medium to yield final concentrations of 1, 3, and 10 mM; the control was a medium without β-HB. For specific experimental groups, the medium was neutralized to pH 7.0. The strains T. spiroformis, T. ramose, A. muciniphila, B. pseudolongum, and L. reuteri were cultured in medium supplemented with β-HB or neutralized β-HB, and incubated in an anaerobic workstation at 37 °C. The growth-promoting effect of β-HB was determined by extracting microbial DNA and quantifying its concentration.

Splenocytes preparation and differentiation assay

C57BL/6 mice were euthanized, and their spleens were isolated aseptically. The spleens were ground, and cell debris and clumps were removed to get the single-cell suspension. Erythrocytes were depleted with ammonium chloride buffer solution, and then the cells were washed and resuspended in RPMI 1640 media containing 10% FBS. For the differentiation assay, splenocytes (2 × 10⁶ per well) were cultured in 24-well plates in triplicate in the presence or the absence of 1 mM or 3 mM β-HB for 24 h at 37 °C with 5% CO₂. The differentiation of splenocytes was detected by flow cytometry.

Cell proliferation assay

Cell proliferation was assessed by CCK8. Briefly, Caco-2 cells (2 × 10⁴ per well) were cultured in 96-well plates in triplicate for 48 h in the presence or the absence of 1 mM or 3 mM β-HB for 24 h at 37 °C with 5% CO₂. The cells cultured with media alone were used as controls. Twenty μL solution reagent was added 2 h before the end of culturing. At the end of culturing, the OD value was measured at 450 nm using the microplate reader (MD SpectraMAX190, Sunnyvale, CA, USA).

Colon tissue dissociation and single-cell isolation

Colon tissues were washed with PBS, minced into 0.5-1 cm pieces, and digested in 0.5 M EDTA (20 mL/tube, 4 °C, 5000 × g, 15 min; repeated once). Tissue fragments were then enzymatically digested with 3 mL Type IV collagenase (37 °C, 30 min). The cell suspension was washed 3× with PBS, treated with DNAzyme (15 mL, 20 min), and filtered through a 70 μm strainer. After centrifugation (150 × g, 5 min, 4 °C), the pellet was resuspended for flow cytometric sorting (BD FACS Aria II) to isolate viable single cells.

Single-cell RNA sequencing

Cell suspensions were adjusted to 700–1200 cells/μL and processed using microfluidic technology. Single cells were co-encapsulated with barcoded gel beads in droplets to generate Gel Bead-in-Emulsions (GEM). Within each GEM, reverse transcription was performed to produce barcoded cDNA. The cDNA was then fragmented (200–300 bp), adapter-ligated, and PCR-amplified to construct sequencing libraries. Final libraries were sequenced on an Illumina NovaSeq platform (paired-end). Experimental procedures were conducted by Personalbio, with subsequent data analysis performed by Genergy Biotech.

Isolation of colonic lamina propria cells

Colonic tissues were harvested and rinsed in ice-cold PBS before being cut into 0.5 cm segments. Tissue fragments were first digested in pre-warmed Digestion Buffer I (HBSS with 5% FBS and 2 mM EDTA) at 37 °C for 30 min with shaking at 250 rpm, followed by PBS washing. The tissue was then subjected to two sequential enzymatic digestions in fresh Digestion Buffer II (RPMI-1640 with 5% FBS, 0.1% collagenase IV, and 25 μg/mL DNase I) under identical conditions. The resulting cell suspensions were pooled, filtered through a 40 μm strainer, and centrifuged at 600 × g for 20 min at 4 °C.

After PBS washing and centrifugation, cells were resuspended in 1 mL PBS and layered onto a discontinuous Percoll gradient (2 mL 80% overlaid with 4 mL 40% Percoll in 1 × PBS). Gradient separation was performed at 1100 × g for 20 min at 4 °C with minimal brake application. The intermediate cell layer was collected, washed with 12 mL PBS, and centrifuged at 600 × g for 15 min at 4 °C. The final pellet containing purified lamina propria cells was resuspended for flow cytometry assay.

Flow cytometry assay

Single suspensions of mouse colonic lymphocytes were prepared. Then the single cells were blocked with purified anti-mouse CD16/CD32 mAb (2.4G2) and stained with BV510-conjugated anti-CD45 mAb, FITC-conjugated anti-CD3 mAb, BV421-conjugated anti-γδ TCR mAb and Percp-cy5.5-conjugated anti-CD4 mAb for 20 min in the dark. For intracellular staining, cells were first stained with surface markers, followed by fixation and permeabilization using the Foxp3 Staining Buffer set. Subsequently, cells were labeled intracellularly with PE-conjugated anti-IL-17A for 40 min in the dark. Flow cytometry was performed on a BD LSR Fortessa Flow Cytometer, and the data were analyzed with FlowJo V10 software. Antibody details are provided in Supplementary Table 3.

Quantification and statistical analysis

The experimental results were statistically analyzed using Prism version 10.2.3 (GraphPad software), and the results were expressed as mean ± standard error of the mean (SEM). Two-sided unpaired Student’s t-test, multiple t-test (followed by two-stage step-up method of Benjamin, Krieger, and Yekutieli⁷¹), one-way ANOVA (followed by Dunnett’s multiple comparisons test), and two-way ANOVA (followed by two-stage step-up method of Benjamin, Krieger, and Yekutieli) are used to determine significance. A p-value < 0.05 was considered to be statistically significant.

Reporting summary

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

Author contributions

W.L., Lei.C. and C.X. conceived the project. Y.Liu, X.W., Li.C. and S.W. were responsible for the methodology. Li.C. performed flow cytometry. S.W. and X.W. conducted animal experiments. W.W. and X.Y. were responsible for clinical sample collection and information compilation. J.X., W.X. and X.Z. performed microbial cultivation. S.L. performed computational analysis. Y.Liu, X.W., S.W., Y.Li and L.H. conducted formal analysis. Y.Liu, X.W., Li.C., S.W., J.X., W.X., X.Z., S.L., Y.Li. and L.H. performed the investigation. Y.Liu, X.W. and Li.C curated the data. Y.Liu, X.W. and Li.C wrote the manuscript. F.G. supervised the research.

Peer review

Peer review information

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

Data availability

The single-cell RNA sequencing in this study has been deposited in the Gene Expression Omnibus (GEO) database under the accession code GSE312416. 16 s rRNA data generated in this study have been uploaded to the NCBI Sequence Read Archive (SRA) database with the accession numbers PRJNA1392765 and PRJNA1394685. 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 10.1038/s41467-026-72044-0.