Phocaeicola vulgatus induces immunotherapy resistance in hepatocellular carcinoma via reducing indoleacetic acid production

Summary

Immunotherapy has made remarkable achievements in various cancers, but response rates in hepatocellular carcinoma (HCC) remain highly variable. Understanding mechanisms behind this heterogeneity and identifying responsive patients are urgent clinical challenges. In this study, the metagenomic analysis of 65 HCC patients reveals distinct gut microbiota profiles distinguishing responders (Rs) from non-responders (NRs). These findings are further validated through fecal microbiota transplantation (FMT) in mouse models. Notably, Phocaeicola vulgatus (P. vulgatus) is enriched in NRs and diminishes anti-PD-1 efficacy in both syngeneic and orthotopic tumor models. Mechanistically, P. vulgatus suppresses the production of indoleacetic acid (IAA), thereby weakening interferon (IFN)-γ⁺ and granzyme B (GzmB)⁺CD8⁺ T cells and impairing the antitumor immune response. Furthermore, supplementation with IAA restores CD8⁺ T cell cytotoxicity and counteracts the immune-suppressive effects of P. vulgatus. Our findings establish a causal relationship between P. vulgatus and anti-PD-1 resistance in HCC, highlighting IAA as a potential therapeutic target to enhance immunotherapy outcomes.

Graphical abstract

Highlights

• •P. vulgatus enrichment in HCC non-responders is identified through fecal metagenomics
• •P. vulgatus suppresses anti-PD-1 efficacy and CD8⁺ T cell cytotoxicity
• •Targeted metabolomics reveals P. vulgatus-mediated reduction of gut-derived IAA
• •IAA restores CD8⁺ T cell function and overcomes P. vulgatus-induced immunosuppression

Zhao et al. demonstrate that P. vulgatus is enriched in HCC non-responders and reduces anti-PD-1 efficacy by depleting IAA and suppressing CD8⁺ T cell function. IAA supplementation restores CD8⁺ T cell activity and counters microbial immunosuppression, with P. vulgatus and IAA serving as potential biomarkers for immunotherapy resistance and prognosis.

Introduction

Liver cancer is the sixth most common cancer and the third leading cause of cancer-related mortality worldwide.^1,2 Despite advances in treatment strategies, progress in targeted therapies for hepatocellular carcinoma (HCC) has been limited.³ Immunotherapy, particularly immune checkpoint inhibitors (ICIs), has emerged as a key treatment for unresectable HCC, achieving significant success beyond traditional targeted therapies. However, response rates to ICIs remain highly variable among patients, and mechanisms underlying resistance are not fully understood.

The gut microbiota, increasingly recognized as a cancer hallmark, plays a critical role in carcinogenesis and tumor progression.⁴ Recent studies have indicated that the composition of gut microbiota is related to ICI responses, particularly in melanoma, colorectal, and lung cancers.^5,6,7,8,9 Furthermore, preclinical studies using fecal microbiota transplantation (FMT) have established a causal relationship between gut flora and immunotherapy response. Notably, clinical trials have shown that FMT from responders (Rs) can overcome resistance to anti-PD-1 therapy in melanoma, underscoring the therapeutic potential of gut microbiota.^10,11 However, its role in modulating immunotherapy responses in HCC remains poorly understood.

As the largest solid organ, the liver is uniquely positioned at the intersection of the digestive and circulatory systems, facilitating the bidirectional exchange of metabolites and nutrients.^12,13 The portal vein delivers gut-derived nutrients and microbial metabolites such as short-chain fatty acids and indoles to the liver,^14,15,16 while the biliary system returns liver-produced substances to the gut.^17,18 This enterohepatic circulation establishes a dynamic interplay between the gut microbiota and the liver, which can influence HCC development and treatment response. However, while metabolites produced by the gut microbiota are known to impact liver immunity, the precise mechanisms by which gut microbiota modulates immunotherapy responses remain largely unexplored in HCC.

In this study, we aimed to address this knowledge gap by investigating the relationship between gut microbiota composition and immunotherapy outcomes in HCC patients. To this end, we established a cohort of HCC patients undergoing immunotherapy and characterized the gut microbiota profiles of Rs and non-responders (NRs) using metagenomic sequencing and plasma metabolomics. Additionally, we confirmed the causal links between gut microbiota and immunotherapy outcomes through FMT in animal models. Our findings identified that Phocaeicola vulgatus (P. vulgatus) was a key driver of immunotherapy resistance. P. vulgatus-enriched microbiota diminished the production of indoleacetic acid (IAA), which in turn impaired CD8⁺ T cell effector function, ultimately contributing to treatment resistance.

Results

Characteristics and gut microbiota metagenomics profiles of patients with hepatocellular carcinoma

A total of 81 qualified HCC patients receiving immunotherapy for the first time were enrolled in this study (Figure 1A). The detailed patient enrollment scheme is shown in Figure 1B. Five patients were excluded from the analysis for the following reasons: two died after single infusion due to liver failure, two discontinued the therapy because of intolerable immune-related adverse events, and one was lost to follow-up. In 76 qualified patients, 65 patients (50 Rs and 15 NRs) with evaluable fecal samples were ultimately included in the analysis. Among 50 Rs, 4 (8%) achieved a complete response (CR), 19 (38%) had a partial response (PR), and 27 (54%) exhibited stable disease (SD). As shown in Figure 1C, Rs had a significantly longer progression free survival (PFS) of 30.8 months and overall survival (OS) of 94.8 months, compared to NRs (PFS: 3.8 months, p < 0.001; OS: 15.0 months, p < 0.001).

Figure 1.

Gut microbiota influence immunotherapy outcomes in patients with hepatocellular carcinoma

(A) Overview of the patient study design.

(B) Patient enrollment scheme.

(C) PFS (left) and OS (right) in responders (n = 50) vs. non-responders (n = 15).

(D) PCoA of gut microbiota β-diversity using Bray-Curtis dissimilarity metrics for responders (n = 50) and non-responders (n = 15).

(E–H) Experimental design and results from fecal microbiota transplantation (FMT) into antibiotics (ATB)-treated pseudo-germ-free mice bearing subcutaneous RIL-175 tumors. (E) Experimental design: mice received FMT from Rs or NRs, followed by anti-PD-1 or isotype control treatment. (F) Tumor growth curves over time. (G) Representative tumor images (shown from the second of two independent experiments). (H) Final tumor weights. Each group included fecal samples from 6 Rs to 6 NRs, with 4–6 mice per patient. Data represent pooled results from two independent experiments (n = 14–17 mice/group).

(I–K) FMT validation in an orthotopic HCC model via hydrodynamic tail vein injection of oncogenic plasmids. (I) Experimental design. (J) Representative liver tumor images across groups. (K) Representative bioluminescence images and quantification of total tumor flux. Each group included fecal samples from 4 Rs or 4 NRs, with 4–6 mice per patient (n = 10 mice/group).

Data are shown as mean ± SEM. Statistical significance was determined by log rank test (C), PERMANOVA (D), or ANOVA (F, H, and K). ∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, and p < 0.05. irAEs, intolerable immune-related adverse events; R_Isotype, responder FMT + isotype control; R_Anti-PD1, responder FMT + anti-PD-1; NR_Isotype, non-responder FMT + isotype control; NR_Anti-PD1, non-responder FMT + anti-PD-1.

As summarized in Table S1, comparison of the clinical characteristics between the 65 patients revealed that the mean age of Rs was 63.5 ± 9.6 years, while NRs had a mean age of 58.3 ± 9.2 years. The 84% of Rs were male and the remaining 16% were female; the NRs group were 80% male and 20% female, with no significant difference between the two groups. Other clinical factors, including etiology, Eastern Cooperative Oncology Group (ECOG) performance score (PS), alpha-fetoprotein (AFP) level, Child-Pugh (CP) class, Barcelona Clinic Liver Cancer (BCLC) stage, vascular invasion, extrahepatic metastasis, prior history of resection, transarterial chemoembolization (TACE), radiofrequency ablation (RFA), radiotherapy and targeted therapy, immunotherapy regime, and combination therapies, did not show significant differences between two groups.

Of 50 Rs, 18 fecal samples were collected at baseline and 32 during immunotherapy. Although most samples were collected after immunotherapy initiation, there was no significant impact of ICI treatment on gut microbiome stability (p = 0.739) (Figure S1A). To better understand the differences in gut microbiota profiles between Rs and NRs, metagenomic sequencing was performed on feces of these 65 HCC patients. As shown in Figure 1D, principal co-ordinates analysis (PCoA) of β-diversity based on Bray-Curtis dissimilarity metrics revealed a significant difference in the composition of gut microbiota between Rs and NRs (p = 0.013). However, the Shannon and Simpson diversity analysis indicated that the α-diversity between groups was not significantly different (Figure S1B). These results indicate that gut microbiota composition differs between Rs and NRs, though intragroup diversity remains unchanged.

Fecal microbiota transplantation from HCC patients into mice dictated outcome after immunotherapy

To study the potential causal relationship between gut microbiota composition and immunotherapy response in HCC, fecal microbiota from Rs and NRs were transferred into antibiotics (ATB)-pretreated mice bearing subcutaneous RIL-175 tumors, followed by anti-PD-1 treatment (Figure 1E). As shown in Figure 1F, irrespective of anti-PD-1 treatment, mice receiving FMT from Rs exhibited smaller tumor volumes compared to those receiving FMT from NRs. After anti-PD-1 administration, tumor volumes decreased in both groups (R_Anti-PD1 vs. R_Isotype; NR_Anti-PD1 vs. NR_Isotype). However, mice transplanted with microbiota from NRs showed a reduced response to anti-PD-1 therapy compared to those transplanted with microbiota from Rs (R_Anti-PD1 vs. NR_Anti-PD1). As shown in Figures 1G and 1H, tumor burden was the lowest in the R_Anti-PD1 group. To better mimic physiological gut-liver interactions, these findings were further validated in a spontaneous orthotopic HCC model by hydrodynamic tail vein injection of oncogenic plasmids (Figure 1I). Results from this model are consistent with the subcutaneous RIL-175 model, reinforcing that gut microbiota composition influences both tumor progression and the efficacy of anti-PD-1 therapy (Figures 1J and 1K).

Increased abundance of P. vulgatus was associated with impaired immunotherapy efficacy in HCC

The significantly enriched taxa in Rs and NRs were identified through linear discriminant analysis effect size (LEfSe) analysis. The Clostridiaceae family, Clostridium and Citrobacter genera were significantly enriched in Rs, while P. vulgatus was the most enriched species in the feces of NRs (Figures 2A and 2B). P. vulgatus was more abundant in NRs compared with Rs (Figures S1C and S1D). To validate the metagenomic findings, quantitative PCR (qPCR) was performed on DNA extracted from stool samples, confirming a significantly increased abundance of P. vulgatus in NRs compared to Rs (Figure 2C). Collectively, both metagenomic and qPCR analyses demonstrated an elevated abundance of P. vulgatus in NRs (Figure 2C). Moreover, Kaplan-Meier plots of PFS and OS further showed that patients with higher P. vulgatus abundance had poorer survival in comparison with lower ones. As shown in Figure 2D, patients with higher P. vulgatus had a mean PFS of 11.1 months and OS of 22.9 months, which were significantly shorter than those with lower P. vulgatus (PFS: 30.8 months, p = 0.047; OS: 63.2 months, p = 0.037). Representative liver magnetic resonance imaging (MRI) images of a responder (low P. vulgatus, top) and a NR (high P. vulgatus, bottom) before and after immunotherapy were shown in Figure 2E.

Figure 2.

Higher abundance of P. vulgatus impairs immunotherapy efficacy in hepatocellular carcinoma

(A) LEfSe analysis shows differentially abundant taxa in responders (n = 50) vs. non-responders (n = 15); taxa with LDA >1.5 are shown.

(B) Cladogram comparing gut microbiota in responders (n = 50) vs. non-responders (n = 15) based on metagenomic sequencing.

(C) Relative abundance of P. vulgatus in feces of responders (n = 50) vs. non-responders (n = 15) detected by metagenome sequencing and qPCR.

(D) PFS (left) and OS (right) in patients with low (n = 33) or high (n = 32) P. vulgatus abundance.

(E) Representative MRI images of a responder (low P. vulgatus, top) and a non-responder (high P. vulgatus, bottom) before and after immunotherapy.

(F–H) In vivo evaluation of P. vulgatus effects in an RIL-175 subcutaneous tumor model (n = 8 mice/group). (F) Experimental design: mice were orally gavaged with P. vulgatus, followed by tumor implantation and anti-PD-1 or isotype treatment. (G) Tumor growth curves. (H) Representative tumor images (left, two tumors in the PBS_Anti-PD1 group had completely regressed and were not visible at necropsy) and corresponding tumor weights (right).

(I–K) Validation in an orthotopic in situ tumor model induced by hydrodynamic tail vein injection of oncogenic plasmids (n = 4–5 mice/group). (I) Experimental design. (J) Representative tumor images. (K) Representative bioluminescence images (left) and quantification of tumor burden by total bioluminescent flux (right).

(L) Relative abundance of P. vulgatus in feces (left) and tumors (right) of PBS_Anti-PD1 and Pv_Anti-PD1 mice one week after gavage, detected by qPCR (n = 5–6).

Data are shown as mean ± SEM. Statistical significance was determined by t test (C and L), log rank test (D), and ANOVA (G, H, and K). ∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, and ∗p < 0.05; ns, not significant. Pv, P. vulgatus.

To determine whether P. vulgatus impairs the efficacy of anti-PD-1, P. vulgatus was gavaged to both syngeneic and orthotopic mouse tumor models (Figures 2F–2L). To avoid the potential confounding effects of antibiotics on immunotherapy, specific-pathogen-free (SPF) mice without antibiotics pretreatment were used. P. vulgatus in the feces was confirmed following one week of oral administration; however, could not be detected in tumors of any groups by qPCR (Figure 2L). Anti-PD-1 monotherapy efficiently inhibited tumor growth in the syngenetic RIL-175 model (Figures 2G and 2H). Although P. vulgatus alone did not affect tumor growth, the combination of P. vulgatus and anti-PD-1 treatment significantly impaired the antitumor efficacy of anti-PD-1. In conclusion, the presence of P. vulgatus impairs the therapeutic effect of anti-PD-1 in preclinical models of HCC.

P. vulgatus impaired immunotherapy efficacy by inhibiting intratumoral CD8⁺ T cell effector function

To investigate the underlying mechanisms by which gut microbiota influence tumor and systemic immune response, we examined the main lymphoid and myeloid cell subtypes in tumors, spleens, and blood of RIL-175 tumor-bearing mice using flow cytometry (Figure 3A). The gating strategy for lymphoid and myeloid cells is shown in Figure S2.

Figure 3.

Gut microbiota and P. vulgatus modulate immunotherapy response by modulating tumoral CD8⁺ T cell function

(A) Schematic of immune profiling study design.

(B) Frequencies of tumoral IFN-γ⁺ and GzmB⁺CD8⁺ T cells in the FMT model.

(C) Frequencies of tumoral IFN-γ⁺ and GzmB⁺CD8⁺ T cells in the P. vulgatus transplantation model.

Data are shown as mean ± SEM. Statistical significance was determined by ANOVA (B and C). ∗∗∗p < 0.001, ∗∗p < 0.01, and ∗p < 0.05.

In the FMT model, intratumoral immune cell profiling revealed significant changes, particularly in CD8⁺ T cell subtypes (Figure 3B; Figure S3). The ratios of IFN-γ⁺ and GzmB⁺CD8⁺ T cells within tumors steadily increased across the NR_Isotype, NR_Anti-PD1, R_ Isotype, and R_Anti-PD1 groups, with the highest ratios observed in mice receiving FMT from Rs in combination with anti-PD-1 (Figure 3B). There were no notable differences in percentages of CD4⁺ and CD8⁺ T cells, or in the proliferation of CD8⁺ T cells marked by Ki-67⁺, nor in myeloid profiles such as MDSCs, macrophages, and DCs (Figure S3). Meanwhile, T lymphocytes in the spleen and blood remained largely unaffected (Figures S4A and S4). Although the frequency and subtypes of certain innate immune cells, such as macrophages and DCs in the spleen (Figure S4A), were altered by FMT, these changes were insufficient to modulate anti-PD-1 response in the absence of adaptive immunity. Thus, we focused on intratumoral T lymphoid profiles in subsequent analyses.

Consistent with FMT model data, the administration of P. vulgatus impaired immunotherapy efficacy by reducing CD8⁺ T cell effector function. In the presence of anti-PD-1, co-treatment with P. vulgatus significantly reduced the frequency of IFN-γ⁺ and GzmB⁺CD8⁺ T cells in tumors (Figure 3C).

To elucidate the molecular mechanisms behind this effect, bulk RNA sequencing was employed on tumor tissues from mice treated with anti-PD-1 alone or in combination with P. vulgatus (Figures 4A–4C; Figures S5C and S5D). RNA sequencing data showed significant decreases in the abundance of gzma, gzmg, and gzmf in the P. vulgatus plus anti-PD-1 group compared to the anti-PD-1 alone group (Figure 4A). Gene ontology (GO) analysis revealed a marked reduction in genes associated with “cytolytic granule” pathway in the combination treatment group (Figure 4B). Gene set enrichment analysis (GSEA) showed that “leukocyte mediated cytotoxicity” was similarly inhibited (Figure 4C; Figures S5C and S5D). Immunofluorescence staining further confirmed that the frequency of GzmB⁺ and IFN-γ⁺CD8⁺ T cells was significantly reduced in tumors treated with P. vulgatus (Figures 4D–4G). Moreover, intratumoral IFN-γ expression was significantly lowered in P. vulgatus-treated mice (Figure 3E), while there were no notable differences in concentrations of interleukin (IL)-2, IL-4, IL-17A, IL-10, IFN-γ, and tumor necrosis factor (TNF)-α in both tumors and plasma (Figures S5A and S5B).

Figure 4.

P. vulgatus inhibits CD8⁺ T cell effector function in the tumor microenvironment

(A) Volcano plot of differentially expressed genes in tumors of PBS_Anti-PD1 and Pv_Anti-PD1 mice.

(B) Bubble plot of top five GO terms enriched in tumors from PBS_Anti-PD1 and Pv_Anti-PD1 mice.

(C) GSEA of leukocyte-mediated cytotoxicity in tumors from PBS_Anti-PD1 and Pv_Anti-PD1 mice.

(D and E) Representative immunofluorescence images (D) and quantification (E) of GzmB⁺CD8⁺ T cells in tumors (n = 3 mice/group), Scale bars, 20 μm.

(F and G) Representative immunofluorescence images (F) and quantification (G) of IFN-γ⁺CD8⁺ T cells in tumors (n = 3 mice/group), Scale bars, 20 μm.

(H) IFN-γ levels in tumors of mice treated with P. vulgatus in the RIL-175 subcutaneous model (n = 4–6 mice/group).

Data are shown as mean ± SEM. Statistical significance was determined by ANOVA (E, G, and H). ∗∗∗p < 0.001, ∗∗p < 0.01, and ∗p < 0.05.

In summary, immune cell and cytokines profiling, transcriptomic data and immunofluorescence staining collectively suggest that P. vulgatus plays a critical role in diminishing CD8⁺ T cell cytotoxicity and impairing the efficacy of anti-PD-1 therapy in mice bearing HCC.

P. vulgatus impaired immunotherapy response through reducing indoleacetic acid production

To explore if gut microbes could translocate to tumor and directly influence immunotherapy efficacy, qPCR was utilized to detect P. vulgatus in tumor tissues of orthotopic tumor models. No evidence of bacterial translocation was found (Figure 2L). Therefore, we hypothesized that P. vulgatus inhibited immune responses though microbiota-derived metabolites circulating in the bloodstream, rather than through direct bacterial interaction with tumor-infiltrating CD8⁺ T cells.

To explore this possibility, CD8⁺ T cells were first co-cultured with the culture medium of P. vulgatus. Surprisingly, there was no significant effect on the expression of cytotoxic markers GzmB and IFN-γ on CD8⁺ T cells (Figure 5A). Next, targeted metabolic profiling of feces from P. vulgatus-gavaged mice revealed a significant decrease in tryptophan-derived indole derivates, namely IAA and indolelactic acid (ILA) (Figure 5B). Consistently, plasma metabolomics also revealed the decreased concentrations of IAA and ILA (Figure 5C). These plasma findings likely reflected a reduction in gut-derived metabolites entering circulation via the enterohepatic route. Notably, IAA and ILA were the only metabolites consistently decreased in both plasma and feces (Figure 5D), supporting their relevance in mediating systemic immunomodulatory effects.

Figure 5.

P. vulgatus impairs immunotherapy response through reducing indoleacetic acid production

(A) Frequencies of (left) GzmB⁺ and (right) IFN-γ⁺ CD8⁺ T cells co-cultured with 10% culture medium of P. vulgatus (n = 3).

(B) Heatmap of fecal metabolomic profiles (left) and concentrations of IAA and ILA (right) in feces from PBS and Pv-treated mice (n = 6 mice/group).

(C) Heatmap of plasma metabolomic profiles (left) and concentrations of IAA and ILA (right) in plasma of PBS or Pv-treated mice (n = 11–12 mice/group).

(D) Venn diagram of overlapping metabolites decreased in both plasma and feces of Pv-treated mice.

(E) Concentrations of IAA and ILA in plasma of responders (n = 38) and non-responders (n = 10).

(F) Correlation between P. vulgatus relative abundance and plasma IAA and ILA concentrations in patients (n = 48).

(G) PFS (left) and OS (right) of patients with high (n = 24) or low (n = 24) concentrations of IAA in plasma.

Data are shown as mean ± SEM. Statistical significance was determined by t test (A–C and E), Spearman correlation (F), and log rank test (G). ∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, and ∗p < 0.05; ns, not significant.

These findings were validated in HCC patient plasma samples, where only IAA levels were significantly lowered in NRs, while ILA levels remained unchanged (Figure 5E). Furthermore, plasma IAA concentrations negatively correlated with the relative abundance of P. vulgatus in patient fecal samples (Figure 5F), highlighting IAA as a key metabolite linking P. vulgatus with immunotherapy outcomes. Survival analyses further revealed that higher plasma IAA levels were associated with improved PFS and OS in HCC patients (Figure 5G).

These findings collectively identify IAA as a clinically relevant, microbiota-regulated metabolite that links P. vulgatus abundance with immunotherapy outcomes in HCC.

P. vulgatus impaired anti-PD-1 efficacy by diverting tryptophan metabolism away from IAA synthesis

To explore the underlying mechanisms of IAA depletion, we investigated whether P. vulgatus interfered with Trp-to-IAA conversion. As illustrated in Figure 6A, Trp is first metabolized into indolepyruvic acid (IPyA) by aromatic amino acid aminotransferase. IPyA serves as a precursor for both IAA and ILA, synthesized by indolepyruvate decarboxylase (IPDC) and indolelactate dehydrogenase, respectively.¹⁹ To explore how P. vulgatus influenced IAA production, we analyzed the genome of P. vulgatus strain American Type Culture Collection (ATCC) 8482 (GenBank: GCA_000012825.1). Notably, key enzymes required for IAA synthesis, including IPDC, were not annotated. Moreover, the enzyme responsible for catabolizing Trp (aromatic amino acid aminotransferase) was also not annotated. However, it harbors iorA and iorB, which encode subunits of indolepyruvate oxidoreductase (IOR), an enzyme that catalyzes the conversion of IPyA to indole-3-acetyl-CoA, thereby bypassing IAA synthesis.²⁰

Figure 6.

Mechanistic dissection of IAA depletion by P. vulgatus and its functional consequences

(A) Schematic of the tryptophan (Trp) metabolic pathway and key enzymes for IAA and ILA biosynthesis. P. vulgatus lacks indolepyruvate decarboxylase (IPDC) and instead encodes indolepyruvate oxidoreductase (IOR), redirecting IPyA toward indole-3-acetyl-CoA.

(B) Relative abundance of Trp-metabolizing genes (iorA, iorB, and ipdC) in fecal metagenomes of responders (n = 50) vs. non-responders (n = 15).

(C) Concentrations of IAA in P. vulgatus culture medium compared with BHI medium (n = 3).

(D–G) Tryptophan-deprived tumor model (n = 6 mice/group). (D) Experimental design: mice received Trp-deprived or control diets, followed by Pv/PBS gavage and anti-PD-1/isotype treatment. (E) Tumor growth curves. (F) Representative tumor images. (G) Tumor weights.

(H and I) IAA rescue model. (H) Experimental design. (I) Tumor growth curves showing that IAA supplementation restored anti-PD-1 efficacy in Pv-treated mice (n = 6 mice/group).

Data are presented as mean ± SEM. Statistical tests: t test (B and C) and ANOVA (E, G, and I). ∗∗∗p < 0.001, ∗∗p < 0.01, and ∗p < 0.05; ns, not significant. Pv, P. vulgatus; IAA, indoleacetic acid; IPyA, indolepyruvic acid; IPDC, indolepyruvate decarboxylase; IOR, indolepyruvate oxidoreductase.

Consistent with this metabolic profile, metagenomic data showed increased abundance of iorA (K00179) and iorB (K00180), and reduced abundance of ipdC (K04103) in NRs (Figure 6B). In vitro, IAA was undetectable in the culture supernatant of P. vulgatus (Figure 6C), further confirming its inability of P. vulgatus to produce IAA.

To functionally validate the role of Trp metabolism, we performed a P. vulgatus transplantation experiment in a tryptophan-deprived tumor model (Figure 6D). In mice fed with control diet, P. vulgatus significantly suppressed anti-PD-1 efficacy, while this effect was abolished in mice fed with a tryptophan-deprived diet (Figures 6E–6G), indicating that its immunosuppressive effect is Trp-dependent.

Finally, we conducted an IAA rescue experiment to determine whether supplementing IAA could overcome P. vulgatus-mediated resistance (Figure 6H). IAA supplementation fully restored anti-PD-1 efficacy in mice colonized with P. vulgatus, supporting a causal role for IAA depletion in impairing CD8⁺ T cell function and anti-PD-1 efficacy (Figure 6I).

Together, these data demonstrate that P. vulgatus suppresses anti-tumor immunity by diverting tryptophan metabolism away from IAA production, leading to impaired CD8⁺ T cell function and resistance to anti-PD-1 therapy.

Indoleacetic acid enhanced CD8⁺ T cell effector function

To study the role of IAA on CD8⁺ T cells directly, we explored its antitumor effects and synergy with anti-PD-1 in an IAA intervention model (Figure 7A). IAA treatment alone significantly suppressed tumor growth, and combining IAA with anti-PD-1 led to a more pronounced reduction in tumor growth compared to anti-PD-1 monotherapy (Figure 7B). Immune profiling demonstrated that IAA, in combination with anti-PD-1, significantly increased the expressions of key effector molecules (IFN-γ and GzmB) in intratumoral CD8⁺ T cells (Figure 7C). Importantly, depletion of CD8⁺ T cells abrogated the therapeutic benefit of IAA, confirming that its effect is dependent on CD8⁺ T cell-mediated immunity (Figures 7D–7G).

Figure 7.

Indoleacetic acid enhances CD8⁺ T cell effector function and potentiates anti-PD-1 therapy

(A) Experimental design for IAA intervention model.

(B) Tumor growth curves in the IAA intervention model (n = 6 mice/group).

(C) Frequencies of IFN-γ⁺ (left) and GzmB⁺ (right) CD8⁺ T cells in tumors from mice treated with PBS_isotype, PBS-PD1, IAA_isotype, or IAA-PD1.

(D–G) CD8⁺ T cell depletion model (n = 6 mice/group). (D) Experimental design: CD8⁺ T cells were depleted using anti-CD8 antibody to assess dependency of IAA effect. (E) Tumor growth curves. (F) Final tumor weights. (G) Representative tumor images.

(H and I) Frequencies of mouse (H) and human (I) CD25⁺, CD44⁺, CD69^+, GzmB⁺, and IFN-γ⁺ CD8⁺ T cells cultured with increasing concentrations of IAA (n = 3).

(J) Cytotoxicity of human CD8⁺ T cells against liver cancer cells upon IAA treatment (0, 10, 100, and 1,000 μM) (n = 3).

Data are shown as mean ± SEM. Statistical significance was determined by ANOVA. ∗∗∗∗p < 0.0001, ∗∗∗p < 0.001, ∗∗p < 0.01, and ∗p < 0.05; ns, not significant. IAA, indoleacetic acid.

To further validate these findings, we conducted in vitro functional assays using both murine and human-derived CD8⁺ T cells. IAA stimulation increased expressions of activation markers (CD25, CD44, and CD69) and cytotoxicity molecules (IFN-γ and GzmB) in both systems (Figures 7H and 7I; Figure S6). Moreover, IAA significantly augmented the tumor-killing capacity of human CD8⁺ T cells in co-culture with liver cancer cells (Figure 7J). Taken together, IAA not only inhibits tumor growth but also enhances the efficacy of anti-PD-1 therapy, with these effects mediated though CD8⁺ T cells.

Discussion

This study provided comprehensive evidence linking gut microbiota to immunotherapy resistance in HCC. We first observed significant differences in the composition of gut microbiota between Rs and NRs to immunotherapy, with P. vulgatus being more abundant in NRs. We then confirmed the causal relationship between gut microbiota and immunotherapy response using FMT in mouse models. Finally, we identified P. vulgatus as a key driver of immunotherapy resistance and revealed that its effects are mediated through disruption of intestinal IAA synthesis, which ultimately leads to reduced CD8⁺ T cell effector function in the tumor microenvironment (TME).

Evidence has shown that gut microbiota modulates immunotherapy response across various cancers, including melanoma, colorectal, and lung cancers.^5,6,7,8,9 However, these effects vary significantly across cancer types,²¹ emphasizing the necessity of characterizing microbiota in HCC patients specifically. While beneficial microbes such as Bifidobacterium, Roseburia, and Akkermansia have been associated with improved immunotherapy efficacy,²¹ the role of specific bacteria in driving resistance remains unclear. In this study, we identified that P. vulgatus was significantly more abundant in NRs and associated with immunotherapy resistance in HCC.

Phocaeicola vulgatus (formerly Bacteroides vulgatus) is prevalent and abundant member of the human gut microbiome, which plays a critical role in intestinal homeostasis by facilitating the breakdown of complex carbohydrates.^22,23 However, dysregulation of P. vulgatus abundance—commonly referred to as dysbiosis—has been implicated in several gastrointestinal disorders, including inflammatory bowel disease and irritable bowel syndrome.²³ Although P. vulgatus is a common keystone species in the human gut microbiome, its role in modulating treatment responses has been overlooked. Prior studies report its involvement in chemoradiotherapy resistance in rectal cancer patients, and lower risk of immune-related adverse events in melanoma patients receiving immunotherapy, which are typically correlated with poor immunotherapy efficacy.^22,23 These findings are consistent with our observation that P. vulgatus impairs treatment responses in HCC.

In addition to confirming this causality through FMT, our study also elucidated the underlying molecular mechanisms. Our metagenomic analysis indicated that P. vulgatus played a central role in driving resistance. The transplantation of P. vulgatus as a single strain into mouse models verified its role in inducing resistance to anti-PD-1 therapy, both in subcutaneous and in situ tumor models. Importantly, P. vulgatus did not significantly promote tumor growth but inhibited the anti-PD-1 efficacy, suggesting that P. vulgatus primarily exerted its effects by suppressing the immune response rather than directly promoting tumor proliferation. Immune profiling revealed that P. vulgatus significantly reduced the expression of cytotoxicity markers, such as GzmB and IFN-γ, in intratumoral CD8⁺ T cells, highlighting its role in impairing T cell-mediated antitumor immunity.

Our study further explored the molecule mechanisms by which P. vulgatus inhibited CD8⁺ T function. Gut microbiota can influence immune responses through direct interaction with immune cells in the TME via bacterial translocation or by altering circulating metabolites.^24,25 Previous studies have suggested that gut microbiota may translocate to the liver, thereby influence tumor progression and treatment response.^24,26 However, in our study, P. vulgatus was absent from the TME, suggesting that its effects on liver cancer are likely mediated through circulating metabolites. The liver, as the first organ to receive blood from the gut via the portal vein, is especially susceptible to gut-derived metabolites, which may directly influence the TME. It has been reported that P. vulgatus-mediated nucleotide biosynthesis inhibited the response of rectal cancer patients to chemoradiotherapy.²² Interestingly, no direct inhibitory substances secreted by P. vulgatus were identified that could impair CD8⁺ T cell function in vitro. However, targeted metabolomics revealed that over-colonization of P. vulgatus by gavage resulted in a significant decrease in IAA levels in both serum and feces in mice.

IAA is a tryptophan-derived indole derivative produced by intestinal microorganisms and cannot be synthesized by the human body.²⁷ Tryptophan-derived indole derivatives are well recognized to enhance CD8⁺ T cell mediated antitumor immunity,^28,29 Previous studies have shown that other indole derivatives, such as indole-3-aldehyde and indole-3-propionic acid, enhance CD8⁺ T cell cytotoxicity and improve efficacy of immunotherapy.^24,30 In agreement with these findings, we demonstrated that IAA supplementation promoted the expressions of cytotoxicity effectors (GzmB and IFN-γ) on CD8⁺ T cells in vitro and improved immunotherapy efficacy in vivo, which was completely opposite to the inhibitory effects of P. vulgatus on immune. Notably, the supplementation of IAA restored the anticancer effects of anti-PD-1 even over-colonization of P. vulgatus. These findings provide direct evidence that P. vulgatus induces immunotherapy resistance by depleting IAA, thereby impairing CD8⁺ T cell function.

The molecular basis for IAA depletion involves the capacity of P. vulgatus to alter tryptophan metabolism. Trp is mainly metabolized to IPyA, which is then converted to IAA by enzyme IPDC.¹⁹ However, our genomic analysis revealed that P. vulgatus lacks IPDC gene but encodes IOR, which diverts IPyA into indole-3-acetyl-CoA.²⁰ This metabolic shift exhausts IPyA, limiting the production of IAA. Consistent with this, metagenomic data from NRs showed an increased abundance of IOR genes and reduced expression of IPDC genes in the gut microbiota. These findings provide a molecular explanation for how P. vulgatus disrupts IAA synthesis and contributes to immunotherapy resistance.

Our findings have significant clinical implications for the treatment of HCC. The overabundance of P. vulgatus and the associated depletion of IAA could serve as predictive biomarkers for immunotherapy resistance. Importantly, IAA supplementation showed potential as a therapeutic strategy to overcome resistance driven by P. vulgatus over-colonization. This highlights the feasibility of targeting gut microbiota-produced metabolites as adjuvant therapies to improve immunotherapy outcomes. Moreover, the use of gut microbiota profiling could help facilitate personalized immunotherapy strategies by identifying patients at risk of resistance based on their microbial composition.

In conclusion, this study establishes a causal link between gut microbiota and immunotherapy resistance in HCC, with P. vulgatus identified as a key driver through its disruption of IAA synthesis. Our findings suggest that P. vulgatus impairs CD8⁺ T cell function, thereby reducing the efficacy of anti-PD-1 therapy. Moreover, we highlight the potential of gut microbiota-targeted interventions, including metabolite IAA supplementation, to enhance immunotherapy outcomes. By integrating gut microbiota profiling and metabolite analysis into clinical practice, personalized strategies can be developed to overcome resistance and improve treatment efficacy for HCC patients.

Limitations of the study

Despite the robust evidence provided, our study is limited by a relatively small cohort size of 65 patients. This limitation may introduce potential biases and restrict the generalizability of our findings. Further studies with larger, multi-center cohorts that encompass diverse populations are necessary to validate our results. Additionally, given the role of P. vulgatus in mediating immunotherapy resistance, future research should focus on identifying supplements or natural products capable of inhibiting P. vulgatus to mitigate its effects on treatment outcomes.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Feng-Ming (Spring) Kong (kong0001@hku.hk).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •The metagenomic sequencing data have been deposited at the NCBI SRA database: PRJNA1211300. The metabolomic data have been deposited at the National Genomics Data Center database: PRJCA044933.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.

Acknowledgments

The authors are grateful for all the subjects who participated in the study. We thank Emily Liu and Erica Yau from the Imaging and Flow Cytometry Core, Centre for PanorOmic Sciences, University of Hong Kong for providing and maintaining the equipment and technical support needed for flow cytometry analysis. This work was supported by grants from Shenzhen Science and Technology Program (KQTD20180411185028798 and ZDSYS20210623091811035), Shenzhen Medical Research Funds (A2303060).

Author contributions

C.-N.Z., S.-S.L., X.-Y.G., and F.-M.K. participated in the conception and design of the study. T.Y., W.-Q.C., and R.J. recruited patients. C.-N.Z. and S.-S.L. collected samples and conducted experiments. C.-N.Z. and S.-S.L. participated in data analysis and interpretation. X.-Y.G. and F.-M.K. provided laboratory resource and fundings. The manuscript was drafted by C.-N.Z. and S.-S.L., and reviewed by all authors.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-Mouse PD1 (RMP1-14)	Bio-X-Cell	Cat#BE0146; RRID: AB_10949053
Rat IgG2a Isotype Control (2A3)	Bio-X-Cell	Cat#BE0089; RRID: AB_1107769
Anti-Mouse CD8α (2.43)	Bio-X-Cell	Cat#BE0061; RRID: AB_1125541
Anti-Mouse CD45 (30-F11)	BD Biosciences	Cat#557659; RRID: AB_396774
Fixable Viability Stain 700	BD Biosciences	Cat#564997; RRID: AB_2869637
Anti-Mouse CD3e (145-2C11)	BD Biosciences	Cat#562600; RRID: AB_11153670
Anti-Mouse CD4 (RM4-5)	BD Biosciences	Cat#563151; RRID: AB_2687549
Anti-Mouse CD8α (53–6.7)	BD Biosciences	Cat#553030; RRID: AB_394568
Anti-Mouse Ki-67 (SolA15)	eBioscience	Cat#25-5698-82; RRID: AB_11217689
Anti-Mouse Granzyme B (NGZB)	eBioscience	Cat#12-8898-82; RRID: AB_10870787
Anti-Mouse IFN-γ (XMG1.2)	BD Biosciences	Cat#566097; RRID: AB_2739549
Anti-Mouse IL-17A (eBio17B7)	eBioscience	Cat#61-7177-82; RRID: AB_2574656
Anti-Mouse Foxp3 (MF23)	BD Biosciences	Cat#560401; RRID: AB_1645201
Anti-Mouse CD25 (PC61)	BD Biosciences	Cat#566498; RRID: AB_2744345
Anti-Mouse CD44 (IM7)	BioLegend	Cat#103028; RRID: AB_830785
Anti-Mouse CD69 (H1.2F3)	BD Biosciences	Cat#560689; RRID: AB_1727506
Anti-Mouse CD11b (M1/70)	BD Biosciences	Cat#562950; RRID: AB_2737913
Anti-Mouse CD11c (N418)	BD Biosciences	Cat#565586; RRID: AB_2869690
Anti-Mouse Ly-6G and Ly-6C (RB6-8C5)	BD Biosciences	Cat#563299; RRID: AB_2738126
Anti-Mouse F4/80 (T45-2342)	BD Biosciences	Cat#565410; RRID: AB_2687527
Anti-Mouse CD86 (GL1)	BD Biosciences	Cat#742120; RRID: AB_2871388
Anti-Mouse CD206 (MMR)	BioLegend	Cat#141717; RRID: AB_2562232
Anti-Mouse CD16/32 (93)	BioLegend	Cat#101320; RRID: AB_1574975
Anti-Mouse CD3e (145-2C11)	BD Biosciences	Cat#553057; RRID: AB_394590
Anti-Mouse CD28 (37.51)	BD Biosciences	Cat#553294; RRID: AB_394763
Anti-Human GzmB (GB11)	BD Biosciences	Cat#561142; RRID: AB_10561690
Anti-Human IFN-γ (B27)	BD Biosciences	Cat#557643; RRID: AB_396760
Anti-Human CD8 (SK1)	BD Biosciences	Cat#563919; RRID: AB_2722546
Anti-Human CD25 (BC96)	Biolegend	Cat#302630; RRID: AB_11126749
Anti-Human CD44 (G44-26)	BD Biosciences	Cat#555478; RRID: AB_395870
Anti-Human CD69 (FN50)	BD Biosciences	Cat#555533; RRID: AB_398602
Rabbit Anti-CD8α (EPR21769)	Abcam	Cat#ab217344; RRID: AB_2890649
Rabbit Anti-Granzyme B (D2H2F)	CST	Cat#17215; RRID: AB_2798780
Rat Anti-IFN-γ (37895)	Novus Biologicals	Cat#MAB485; RRID: AB_3714769
Goat anti-Rabbit IgG-Alexa Fluor 488	Invitrogen	Cat#A-11008; RRID: AB_143165
Goat anti-Rat IgG-Alexa Fluor 594	Invitrogen	Cat#A-11007; RRID: AB_10561522
Goat anti-Rabbit IgG-HRP	Servicebio	Cat#GB23303; RRID: AB_2811189
Bacterial and virus strains
Phocaeicola vulgatus	ATCC	Cat#8482
Biological samples
Feces and plasma samples of HCC patients	Queen Mary Hospital and University of Hong Kong-Shenzhen Hospital	N/A
Chemicals, peptides, and recombinant proteins
Mouse IL-2 Recombinant Protein	PeproTech	Cat#212-12
Recombinant Human IL-2	R&D systems	Cat#202-IL-010
Human T cell TransAct	Miltenyi	Cat#130-111-160
Neomycin	Sigma-Aldrich	Cat#N5285-25G
Metronidazole	Sigma-Aldrich	Cat#M1547-25G
Vancomycin	Sigma-Aldrich	Cat#V8138-1G
Ampicillin	Sigma-Aldrich	Cat#A1593-25G
Indole-3-Acetic Acid Sodium Salt	Sigma-Aldrich	Cat#I5148
Dulbecco’s Modified Eagle Medium	Gibco	Cat#C11995500BT
RPMI 1640 Medium	Gibco	Cat#C11875500BT
PBS	Gibco	Cat#C20012500BT
TRYPSIN 0.25% EDTA	Gibco	Cat#25200072
Fetal Bovine Serum	Thermo Fisher	Cat#10270-106
Penicillin-Streptomycin	Gibco	Cat#15140-122
TexMAC GMP Medium	Miltenyi	Cat#170-076-309
2-Mercaptoethanol	Gibico	Cat#21985023
D-Luciferin	PerkinElmer	Cat#122799
DNase I	Roche	Cat#10104159001
Liberase TM Research Grade	Roche	Cat#05401119001
Percoll	GE Healthcare	Cat#17-0891-01
RBC Lysis Buffer	BioLegend	Cat#420301
Cell Staining Buffer	BioLegend	Cat#420201
Leukocyte Activation Cocktail	BD Biosciences	Cat#550583
Protein Transport Inhibitor	BD Biosciences	Cat#555029
Triton X-100	Thermo Fisher	Cat#85111
DAPI	Thermo Fisher	Cat#S36938
BSA	Sigma‒Aldrich	Cat#A9418
L-Amino Acid Rodent Diet without Added Tryptohan	SYSE Bio-Tec. Co., Ltd.	Cat#A22071501
L-Amino Acid Standard Rodent Diet	SYSE Bio-Tec. Co., Ltd.	Cat#A17100503
Critical commercial assays
Mouse CD8+ T cell Isolation Kit	Miltenyi Biotec	Cat#130-104-075
Human CD8+ T cell Isolation Kit	Miltenyi Biotec	Cat#130-096-495
Mouse Th1/Th2/Th17 CBA kit	BD Biosciences	Cat#560485
MD5115MagPure Stool DNA KF Kit B	Guangzhou Magen Biotechnology Co., Ltd.	Cat#MD5115-02B
QIAamp PowerFecal Pro DNA Kit	QIAGEN	Cat#51804
QuantiNova SYBR PCR Mix Kit	QIAGEN	Cat#208054
Transcription Factor Buffer Set	BioLegend	Cat#424401
Fix/Perm Buffer Set	BioLegend	Cat#426803
Deposited data
Metagenomic sequencing data	NCBI SRA database	PRJNA1211300
Metabolomic data	NGDC	PRJCA044933
Experimental models: Cell lines
RIL-175	kindly provided by Prof. Xinyuan Guan	N/A
Huh7	kindly provided by Prof. Xinyuan Guan	N/A
Experimental models: Organisms/strains
C57BL/6J mouse	Charles River	N/A
Oligonucleotides
16S V4-F (515F): 5′-GTGYCAGCMGCCGCGGTAA-3′	Sangon Biotech	N/A
16S V4-R (806R): 5′-GGACTACNVGGGTWTCTAAT-3′	Sangon Biotech	N/A
P. vulgatus-F: 5′-GCCGACGCTTTCTGACAAAA-3′	Sangon Biotech	N/A
P. vulgatus-R: 5′-GAGGCGGCTTTCCATTGTTC-3′	Sangon Biotech	N/A
pCMV(CAT)T7-SB100	Addgene	Cat#34879
pT3-EF1A-MYC-IRES-luc	Addgene	Cat#129775
pX330-p53	Addgene	Cat#59910
Software and algorithms
GraphPad Prism 8.3.0	GraphPad Software	N/A
FlowJo 10.0	BD Biosciences	N/A
ImageJ	NIH	N/A
R	R Development Core Team	N/A
ZEN	ZEISS	N/A
IVIS imaging	PerkinElmer	N/A
MEGAHIT	MEGAHIT development team	N/A
Bracken	Bracken development team	N/A
TMBQ	Metabo-Profile	N/A
Salmon	COMBINE lab	N/A

Experimental model and study participant details

Human participants

The biomarker study (NCT05061342) was approved by the Institutional Review Board (IRB No. UW 19–565). Patients with HCC receiving immunotherapy for the first time were enrolled from Queen Mary Hospital and University of Hong Kong-Shenzhen Hospital between August 2020 and December 2022. Written informed consent was obtained from all patients prior to the collection of specimens and questionnaires. Fecal and blood samples were collected for metagenomics and metabolomics analyses, respectively.

Inclusion criteria: 1) Male or female, age 18–80 years; 2) Diagnosed with HCC; 3) Immunotherapy-naive; 4) At least one measurable lesion present; 5) Ineligible for surgical or local therapies, or with progressive disease (PD) following surgical or locoregional therapies; 6) Local therapy must have been completed at least 4 weeks prior to the baseline imaging; 7) ECOG performance status score of 0, 1 or 2; 8) Receiving either monotherapy or combination ICIs, with or without tyrosine kinase inhibitors; 9) Either on clinical trials or as standard-of-care therapy; 10) No antibiotic use within 30 days prior to stool sample collection to avoid interference with gut microbiota.

Exclusion criteria: 1) Fibrolamellar HCC, sarcomatoid HCC, or mixed cholangiocarcinoma -HCC; 2) History of invasive malignancy within the previous 2 years, except for noninvasive malignancies such as cervical carcinoma or prostate cancer in situ, non-melanomatous carcinoma of the skin, surgically cured lobular or ductal carcinoma in situ of the breast; 3) Prior organ allograft or allogeneic bone marrow transplantation; 4) Active, known, or suspected autoimmune disease; 5) Receipt of adjuvant immunotherapy following complete surgical resection.

Patient characteristics, including demographics, hepatitis history, ECOG performance status and Child-Pugh score, were recorded. Tumor characteristics (tumor stage, number and size of lesions, intrahepatic vascular invasion, and metastasis), treatment history and immunotherapy regimes were documented.

Surveillance imaging was performed every 3 months. Patients were categorized into responders (Rs) and NRs groups based on their radiological response after 6 months of immunotherapy. The primary endpoint was evaluated according to modified Response Evaluation Criteria for Solid Tumors (mRECIST) guideline.³¹ Rs was identified as patients achieving CR, PR or stable disease (SD) for at least 6 months. NRs was identified as patients with PD within 6 months.

Cell lines and culture

Human liver cancer cells (Huh7) and mouse liver cancer cells (RIL-175) were kindly provided by Prof. Xinyuan Guan (The University of Hong Kong). Prior to use, all cell lines in this study were routinely authenticated by Short Tandem Repeat profiling to ensure identity and absence of cross-contamination, in addition to being confirmed mycoplasma-free. All cell lines were maintained in Dulbecco’s Modified Eagle Medium (DMEM) supplemented 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin at 37°C under 5% CO₂. All cell culture reagents were purchased from Gibco-Invitrogen (Carlsbad, USA).

Bacteria strain and culture

The P. vulgatus (ATCC 8482) was purchased from ATCC. The bacteria were cultured overnight at 37°C under anaerobic conditions in brain heart infusion (BHI) broth.

Mouse models

All animal experiments were conducted using male C57BL/6J mice (6–8 weeks old), housed under SPF conditions. Environmental parameters were maintained at a constant temperature of 20 ± 2°C and relative humidity of 55 ± 10%, with a 12-h light/dark cycle. All animal experiments were approved by the Shenzhen TopBiotech Co., Ltd. Animal Ethics Committee (Approval No. TOP-IACUC-2020-10-0014).

For RIL-175 syngeneic subcutaneous tumor model, RIL-175 cells with 1 × 10⁵ cells in 50 μL PBS per mouse were subcutaneously injected into the right flanks of mice. Once tumors reached 20–35 mm² in size (approximately one week), mice received anti-PD-1 treatment twice weekly. Tumor dimensions were measured twice weekly, and tumor volume was calculated as length × width² × 0.5.

For the in situ HCC model, hydrodynamic tail vein injection was used to deliver oncogenic plasmids enabling gene expression in hepatocytes. Mice were injected with a plasmid mixture (total 22 μg) containing MYC, mutant p53 and SB100 (Addgene), in a mass ratio of 10: 10: 2. The plasmids were diluted in 2 mL sterile sodium chloride solution and passed through a 0.22 μm filter prior to injection. The solution was rapidly injected into the lateral tail vein within 5–8 s to ensure efficient hepatocyte transfection via hydrodynamic pressure. Tumors developed two weeks after plasmid delivery. Mice then received anti-PD-1 antibody treatment (200 μg/mouse, intraperitoneally, twice weekly) or its isotype control beginning at week 2 post-injection. Tumor growth was monitored using luciferase bioluminescence imaging.

For FMT model, mice were treated with an antibiotic solution containing ampicillin, metronidazole and neomycin (1 mg/mL), along with vancomycin (5 mg/mL) added to sterile drinking water for two weeks prior to FMT. Then 400 μL of the bacteria suspension was administered daily via oral gavage to each antibiotic-pretreated mouse for one week, then had one week of intestinal resting before tumor inoculation.

For P. vulgatus transplantation model, P. vulgatus was administered via oral gavage at a dose of 400 μL suspension containing 1 × 10⁹ CFU live bacteria in PBS for one week before tumor inoculation. After tumor developed, P. vulgatus was gavaged on the same day as anti-PD-1 treatment. The bacterial suspension was quantified using a spectrophotometer, measuring optical density at 600 nm.

For tryptophan-deprived mouse tumor model, mice were fed with a tryptophan-deprived diet or an isocaloric L-amino acid control diet (Trp = 0.21%) for one month prior to tumor inoculation and maintained throughout the experiment. P. vulgatus was gavage one week before tumor inoculation. Anti-PD-1 treatment began once tumors reached 20–35 mm² in size and was administered concurrently with P. vulgatus gavage.

For IAA rescue model, the P. vulgatus intervention was conducted as described above. IAA (100 mg/kg B.W. in 100 μL PBS) was injected intraperitoneally on the same day as P. vulgatus gavage and anti-PD-1 treatment after tumors reached 20–35 mm² in size.

For IAA intervention model, RIL-175 cells were subcutaneously injected into the right flanks of mice. After tumors reached 20–35 mm² in size, IAA was injected intraperitoneally at a concentration of 100 mg/kg B.W. in 100 μL PBS. The injections were administered on the same day as anti-PD-1 treatment.

For CD8⁺ T cell depletion model, mice bearing established tumors (20–35 mm²) were intraperitoneally injected with anti-CD8 monoclonal antibody (200 μg/mouse). Antibody administration was initiated on the same day as IAA intervention and anti-PD-1, and was repeated twice weekly thereafter in alignment with the treatment schedule.

Method details

Fecal sample collection, DNA extraction and metagenomics sequencing

Fecal samples (3–5 g) were collected in sterile tubes. For inpatients, samples were stored at 4°C immediately after collection and delivered to the laboratory within 4 h. Outpatient samples were collected in tubes containing DNA stabilizer (INVITEK, 1038111200) and delivered within 3 days. All tubes were labeled with the study case number, collection date and treatment time point, and stored at −80°C. DNA extraction was performed using the MD5115MagPure Stool DNA KF Kit B (Guangzhou Magen Biotechnology Co., Ltd., MD5115-02B) following manufacturer’s instructions. Shotgun metagenomics sequencing was performed using DNBSEQ platform (BGI, Shenzhen, China).

Metagenomics analysis

Raw sequencing data were trimmed using SOAPnuke (v.2.2.1), and host-origin reads were removed by mapping trimmed reads to human genome using SOAP2. High-quality clean reads were de novo assembled using MEGAHIT software. Contigs shorter than 300 bp were excluded from further analysis. Gene prediction was performed on contigs using MetaGeneMark, and redundant genes were removed using CD-HIT with an identity cutoff of 95% and a coverage cutoff of 90%. Gene abundance was estimated using Salmon software. Taxonomic annotation was assigned using the Kraken LCA algorithm, and Bracken software was used to generate the taxonomic abundance profiles.

The α-diversity was calculated using the Shannon index and Simpson index, based on species-level abundance profiles. The β-diversity was calculated using Bray-Curtis distance, and compared by PERMANOVA. Linear discriminant analysis effect size (LEfSe) was applied to identify species that best explained differences between groups.

Bacterial load quantification by qPCR

Bacterial DNA in fecal or tissue samples was extracted using QIAamp PowerFecal (pro) DNA kit (QIAGEN, 51804). Quantitative PCR (qPCR) were performed using CFX Opus Real-Time PCR Systems (Bio-Rad) with the QuantiNova SYBR PCR Mix kit (QIAGEN, 208054) according to the manufacturer’s instructions. The abundance of P. vulgatus in fecal or tissue samples was determined using the qPCR with P. vulgatus-specific primers and normalized to 16S rRNA levels. Primers were designed according to previous studies²² as follows: 16S V4 specific primers, F (515F): 5′-GTGYCAGCMGCCGCGGTAA-3′ and R (806R): 5′-GGACTACNVGGGTWTCTAAT-3’; P. vulgatus, F: 5′-GCCGACGCTTTCTGACAAAA-3′ and R: 5′-GAGGCGGCTTTC CATTGTTC-3’. DNA concentration was measured using Nanodrop One (Thermo), and 50 ng of DNA was loaded per well for qPCR. Each sample was analyzed in triplicate, and the average cycle threshold (Ct) value was calculated. The relative abundance of P. vulgatus was determined using the ΔCt method, defined as ΔCt = Ct (P. vulgatus) - Ct (16S rRNA genes).

Targeted metabolomics profiling

Targeted metabolomic profiling was performed on patient and mouse plasma, mouse feces and culture medium of P. vulgatus as previously reported.³² Plasma and culture medium (25 μL) were extracted using ice-cold methanol for protein precipitation. For fecal samples, 5 mg of lyophilized feces was homogenated using zirconium oxide beads and methanol. After centrifugation, 30 μL of supernatant was mixed with derivative reagents and diluted with methanol. The supernatant was analyzed using ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) on an ACQUITY UPLC-Xevo TQ-S system (Waters Corp., Milford, MA, USA) by Metabo-Profile Biotechnology Co., Ltd. (Shanghai, China). Standards of targeted metabolites were sourced from Sigma-Aldrich (St. Louis, USA), Steraloids Inc. (Newport, USA) and TRC Chemicals (Toronto, Canada). UPLC-MS/MS raw data were processed using TMBQ software (v1.0, Metabo-Profile, Shanghai, China) for peak integration, calibration, and quantitation of each metabolite. Multivariate statistical analysis was performed using orthogonal partial least square discriminant analysis.

Bacteria suspension preparation

For the FMT model, 10% sterile glycerol was added to patient fecal samples and stored at −80°C. Frozen stool samples were thawed in a 37°C water bath for 10 min. Once thawed, fecal samples were resuspended in sterile PBS at a ratio of 200 mg feces to 2 mL PBS. The feces suspensions were stirred until no obvious large particles, then sequentially filtered through 200-mesh sterile filters to remove undigested food particles. The filtered suspensions were centrifuged at 600 g for 5 min to remove insoluble debris. Finally, the bacteria suspension was supplemented with 10% sterile glycerol and stored at −80°C. The bacteria suspensions were thawed in 37°C for 10 min before gavage.

Single cell isolation

Mice were anesthetized with pentobarbital sodium and sacrificed for the collection of blood, spleen and tumor tissues. Blood samples were collected from the retro-orbital sinus. Excised tumors were cut into small pieces and digested in RPMI 1640 medium containing DNase1 and Liberase (Roche, 05401119001) for 30 min. Tumor suspensions were passed through 70 μm cell strainers (Becton & Dickinson), then mixed with 40% Percoll (GE Healthcare, 17-0891-01) and layered onto 80% Percoll. Following centrifugation at 400g for 30 min, the immune cell layer between 40% and 80% Percoll layers was carefully collected. Spleens were crushed in RPMI 1640 medium and filtered through 70 μm cell strainer to obtain single-cell suspensions. Red blood cells in spleen and blood samples were lysed using RBC lysis buffer (BioLegend, 420301).

Whole-transcriptome RNA sequencing (bulk RNA-seq) analysis

Total RNA was extracted using TRIzol following the manual instruction. The cDNA library was prepared and sequenced on the BGISEQ-500 (BGI, Shenzhen, China). Reads were aligned to the mouse reference genome (GRCm39) using STAR (v.2.5.3a) with default settings. Normalization and differential analysis were performed using DESeq2 R package.

Immunofluorescence staining

FFPE tumor tissue sections were blocked with 3% BSA for 30 min at room temperature and incubated at 4°C overnight with the primary antibodies: CD8a mAb (Abcam, 217344), GzmB mAb (CST, 17215) and IFN-γ mAb (Novus Biologicals, MAB485). Following primary antibody incubation, the slides were washed and incubated with secondary antibodies for 2 h. The slides were mounted with antifade reagent with DAPI. Images were captured using an ultra-high-resolution confocal microscope (Zeiss LSM 900) and analyzed using ZEN.

Flow cytometry and antibodies

Single-cell suspensions were blocked by TruStain FcX (anti-mouse CD16/32) antibody (BioLegend, 101320) with 2 μL per sample for 10 min. Cell viability was assessed using Fixable Viability Stain 700 (BD Biosciences, 564997). For intracellular cytokine staining, cells were stimulated with Leukocyte Activation Cocktail (BD Biosciences, 550583). For intranuclear and intracellular staining, the True-Nuclear Transcription Factor Buffer Set (BioLegend, 424401) and Cyto-Fast Fix/Perm Buffer Set (BioLegend, 426803) were used. Samples were stained with the following antibodies for 30 min in the dark at room temperature, including CD45-APC-Cy7 (clone 30-F11, BD), CD3e-BV421 (clone 145-2C11, BD), anti-CD4-BV605 (clone RM4-5, BD), CD8a-FITC (clone 53–6.7, BD), anti-PD-1-BB700 (clone J43, BD Biosciences), IL-17A-PE-CF594 (clone eBio17B7, BD), Ki-67-PE-cy7 (clone SolA15, eBioscience), FOXP3-APC (clone MF23, BD), GzmB-PE (clone NGZB, eBioscience), IFN-γ-BV510 (clone XMG1.2, BD), CD11b-BV510 (clone M1/70, BD), CD11c-BB515 (clone N418, BD), Gr-1-BV605 (clone RB6-8C5, BD), F4/80-PE (clone T45-2342, BD), CD86-BB700 (clone GL1, BD), and CD206-BV421 (clone MMR, BioLegend). Flow cytometry was performed using BD Aria III, and data were analyzed with FlowJo software v.10.8. The hierarchical gating strategy of lymphocyte and myeloid lineages is shown in Figure S2.

Cytokine detection

Mouse blood and tumor cytokines were measured using cytometric bead array with mouse Th1/Th2/Th17 Cytokine kit (BD Biosciences, 560485). The levels of IL-2, IL-4, IL-6, IFN-γ, TNF, IL-17A and IL-10 were quantified according to the manufacturer’s instruction.

In vitro CD8⁺ T cell stimulation with indoleacetic acid

Mouse CD8⁺ T cells were isolated from the spleen using mouse CD8⁺ T cell isolation kit (Miltenyi, 130-104-075) according to the manufacturer’s instruction. Human CD8⁺ T cells were isolated from peripheral blood mononuclear cells of a health donor using the human CD8⁺ T cell isolation kit (Miltenyi, 130-096-495). CD8⁺ T cells were cultured in TexMACS GMP Medium (Miltenyi, 170-076-309), supplemented with 10% heat-inactivated FBS, 1% P/S and 55 mM 2-mercaptoethanol (Gibico, 21985023). In 96-well plate, 3 x 10⁵/well mouse CD8⁺ T cells were stimulated with pre-coated anti-CD3 (BD, 553057) and soluble anti-CD28 (BD, 553294) antibodies (each 1 μg/mL) in the presence of 10 ng/mL IL-2 (PeproTech, 212-12) and varying concentrations of 0, 1, 10, 100, 1000 μM IAA (Sigma-Aldrich, I5148). For human CD8⁺ T cell activation, 3 x 10⁵ cells per well were cultured in the same medium supplemented with T cell TransAct (1:100 dilution; Miltenyi, 130-111-160) and 10 ng/mL human IL-2 (R&D systems, 202-IL-010) for at least 3 days. Following activation, cells were treated with 0, 10, 100, or 1000 μM IAA for downstream analysis.

Tumor-CD8⁺ T cell co-culture assay

GFP-expressing human liver cancer cells (Huh7) were seeded at a density of 1 × 10⁴ cells per well in a 96-well plate and allowed to adhere for 12 h. Activated human CD8⁺ T cells were then added at an effector-to-target (E:T) ratio of 5:1 (5 × 10⁴ T cells/well). Co-cultures were maintained in complete T cell culture medium supplemented with varying concentrations of IAA (0, 10, 100, or 1000 μM). After 36 h of co-culture, tumor cell viability was quantified by measuring GFP fluorescence using a microplate reader. A reduction in fluorescence was interpreted as increased tumor cell killing by CD8⁺ T cells.

CD8⁺ T cell activation analysis

The stimulated CD8⁺ T cells were subjected to flow cytometry to assess activation and cytotoxicity markers. After 20 h of stimulation, cells were incubated with protein transport inhibitor containing Brefeldin A (BD Biosciences, 555029) for 4 h to block intracellular cytokine release. Mouse CD8⁺ T cells were stained by the following activation markers, including CD25 (BB700, clone PC61, BD), CD44 (APC/Cy7, clone IM7, Biolegend) and CD69 (APC, clone H1.2F3, BD). After 48 h of stimulation, CD8⁺ T cells were collected to detect cytotoxicity markers, including IFN-γ (BV510, clone XMG1.2, BD) and GzmB (PE, clone NGZB, eBioscience). Human CD8⁺ T cells were stained by the following antibodies: CD8 (BV510, clone SK1, BD), CD25 (BV421, clone BC96, Biolegend), CD44 (FITC, clone G44-26, BD) and CD69 (APC, clone FN50, BD), IFN-γ (PE-Cy7, clone B27, BD) and GzmB (PE, clone GB11, BD). The FACS staining procedures were the same as mentioned previously.

Quantification and statistical analysis

Patient characteristics between groups were calculated using CBCgrps2.8 R package.³³ Categorical variables were compared using the χ² test, while continuous variables were assessed for normality and analyzed using T test for normally distributed data, or Wilcoxon rank-sum test for non-normally distributed data. Patient characteristics were reported as mean ± standard deviation (SD), median with interquartile range, or as percentage, as appropriate. A two-side α level of 0.05 was considered statistically different. All statistical analyses were conducted using R software (v.4.2.1). Animal model data were analyzed using GraphPad Prism (v.9) and presented as mean ± standard error of mean (SEM). Differences between groups were evaluated using T test or ANOVA, with statistically significant set at p < 0.05. Spearman correlation analysis was used to assess relationships between gut microbiota and other experimental parameters. All statistical details of experiments can be found in the figure legends.

Published: September 25, 2025