Intratumoral Parvimonas micra promotes esophageal squamous cell carcinoma via p-cresol-induced Treg differentiation

Abstract

Intratumoral microbiota has emerged as a notable factor influencing cancer initiation and progression. However, its composition and functional impact in esophageal squamous cell carcinoma (ESCC) remain largely unexplored. Here, we performed metagenomic sequencing on 119 paired tumor-normal tissues from patients with ESCC and single-cell RNA sequencing on 45 samples to investigate microbe-host interactions. We identified Parvimonas micra (P. micra), an anaerobic oral-derived bacterium, as significantly enriched in tumor tissues and associated with poor prognosis. Moreover, the abundance of P. micra correlated with increased regulatory T cell (T_reg cell) infiltration in the ESCC tumor microenvironment. Through cellular and animal experiments, we demonstrate that P. micra promotes tumor growth by secreting p-cresol, a metabolite of amino acid fermentation, which elevates reactive oxygen species levels and induces FOXP3⁺ T_reg differentiation, thereby fostering immunosuppression and tumor growth. Our study establishes a mechanistic link between intratumoral microbiota and the immune microenvironment, highlighting the microbial contribution to ESCC progression and prognosis.

Fig. 1.. Characterization of the ESCC intratumoral microbiome.

(A) Schematic representation of the overall study design for multiomics data. (B) Intersample variation in the ESCC intratumoral microbiome explained by the indicated factors, estimated using the PERMANOVA method. Colors represent the value of −log₁₀(P value). The Adonis R² value represents the proportion of variance explained. (C) Representative FISH images of bacterial 16S rRNA from the tumor tissues (n = 18) and matched normal tissues (n = 18). (D) Quantification of (C). The P value is determined using a two-tailed paired Wilcoxon rank-sum test. (E) Comparison of normalized bacterial abundance between tumor tissues (n = 59) and matched normal tissues (n = 59) as measured by qPCR. Bacterial abundance was normalized against total DNA concentration using the formula log₁₀(2^−Ct/DNA concentration × 10¹⁴ + 1). The P value is determined using a two-tailed paired Wilcoxon rank-sum test. (F) Proportion of species annotated with aerophilicity among tumor- and normal-enriched species. P values were assessed using the false discovery rate (FDR)–adjusted Fisher’s exact test. For all panels, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. See also figs. S1 and S2.

Fig. 2.. Identification of tumor-enriched bacterial species and their prognostic relevance in ESCC.

(A) Upper Venn diagram: Species enriched in tumors, associated with poor prognosis, and with >60% prevalence in tumors. Lower Venn diagram: Species enriched in normal tissue, associated with favorable prognosis, and with >60% prevalence in normal tissues. The bar plot shows the median relative abundance of the 25 intersection species—tumor abundance for TPPMs and normal-tissue abundance for NFPMs. (B) Correlation of centered log ratio (clr)–transformed median abundance differences (tumor versus matched normal) with log₂ HR from Cox proportional hazards models. Colors indicate species groups, shapes denote aerophilicity, and point sizes reflect median relative abundance. (C) Representative FISH images of P. micra (P. m.) from the tumor (n = 18) and matched normal tissues (n = 18). (D) Quantification of P. micra abundance based on FISH results in (C). (E) Quantification of P. micra abundance in normal and tumor tissues using qPCR. P. micra abundance was normalized against total DNA concentration using the formula log₁₀(2^−Ct/DNA concentration × 10¹⁴ + 1). (F) Box plot showing the relative abundance of P. micra in tumor tissues compared to normal tissues in the GC cohort (n = 85). (G) Box plot showing the relative abundance of P. micra in tumor tissues compared to normal tissues in the CRC cohorts (PRJNA280026: n = 52; PRJNA383606: n = 55; PRJNA861885: n = 292). (H) Abundance of P. micra in responders (R; n = 81) versus nonresponders (NR; n = 254) in a pancancer cohort treated with ICIs. Whiskers represent the median with the interquartile range. For all panels, the P value is determined using a two-tailed Wilcoxon rank-sum test. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. See also fig. S3.

Fig. 3.. Correlation between the ESCC microbiome and host cells in the tumor microenvironment.

(A) Stick plot showing the proportional difference of the specified cell type between normal and tumor samples. T_reg, regulatory T cells; TAM, tumor-associated macrophage; myCAFs, myofibroblastic cancer–associated fibroblasts; Tex, exhausted T cells; Prolif, proliferating cells; T_H1, T helper 1 cells; iCAFs, inflammatory cancer-associated fibroblasts; TEC, tumor endothelial cells; Tn, naive T cells; RTMs, resident tissue macrophages; GC B, germinal center B cells; Trm, tissue-resident memory T cells; NEC, normal endothelial cell; NFs, normal fibroblasts; NAFs, normal-associated fibroblasts; Tem, effector memory T cells; MD, mucosal defense; Teff, effector T cells; cDC2, type 2 conventional dendritic cells. (B) Kaplan-Meier plot comparing the OS of patients with ESCC stratified by low or high proportions of the specified cell type. HR and 95% CI are calculated by the Cox proportional hazards model. The P value is determined using a log-rank test. AP, antigen presenting; Bmem, memory B cells; HY, hypoxia-related stress; T_FH, T follicular helper cells. (C) Dot plot showing Spearman correlation between the relative abundance of specified bacterial species and the proportion of cell subtypes. Colors indicate the Spearman correlation coefficient (r), while sizes represent the −log₁₀(P) value. (D) Representative multiple immunofluorescence images of KRT6A, CD4, and FOXP3 in tumor tissues (n = 18) and matched normal tissues (n = 18). (E) Quantification of FOXP3⁺CD4⁺ cells in (D). The P value is determined using a two-tailed paired Wilcoxon rank-sum test. (F) Spearman correlation between P. micra abundance and the proportion of FOXP3⁺ CD4⁺ T cells in tumor tissues (n = 18) and matched normal tissues (n = 18). The gray area represents 95% CI. ****P < 0.0001; n.s., not significant. See also figs. S4 and S5.

Fig. 4.. Intratumoral injection of P. micra promotes tumor growth and modulates the immune microenvironment.

(A) Experimental design for intratumoral injection of P. micra using a subcutaneous tumor model. (B) Excised tumor images of allografts derived from subcutaneous transplantation of mEC25 cells treated with PBS or P. micra (n = 6 per group). (C) Tumor growth curves of (B). (D) Representative images of FISH of tumor tissues in (B). (E) Quantification of (D). (F) Representative multiple immunofluorescence images of KRT6A, CD4, and FOXP3 of tumor tissues in (B). (G) Quantification of FOXP3⁺CD4⁺ cells in (F). (H) Representative multiple immunofluorescence images of KRT6A, CD8, GZMB, and PD-1 of tumor tissues in (B). (I) Quantification of PD-1⁺CD8⁺ cells and GZMB⁺CD8⁺ cells in (H). For all panels, data are presented as the means ± SD. P values are determined using a two-tailed Student’s t test. ***P < 0.001 and ****P < 0.0001.

Fig. 5.. P. micra induces T_reg cell differentiation and promotes tumor growth by secreting metabolites.

(A) Experimental design for coculture systems. h, hours. (B) Western blot analysis of the indicated proteins in Jurkat cells treated with P. micra culture medium (CM) of different concentrations. (C) Western blot analysis of the indicated proteins in Jurkat cells treated with heat-killed P. micra of different MOIs. (D) Flow cytometry of FOXP3 and CD25 in mouse spleen CD4⁺ T cells treated with P. micra CM or BHI (n = 3 biological replicates). (E) Quantification of (D). (F) Flow cytometry of FOXP3 and CD25 in mouse spleen CD4⁺ T cells treated with heat-killed P. micra of different MOIs (n = 3 biological replicates). (G) Quantification of (F). (H) Western blot analysis of the indicated proteins in Jurkat cells treated with heat-inactivated P. micra CM of different concentrations. (I) Flow cytometry of FOXP3 and CD25 in mouse spleen CD4⁺ T cells treated with heat-inactivated P. micra CM or BHI (n = 3 biological replicates). (J) Quantification of (I). (K) Excised tumor images of allografts derived from subcutaneous transplantation of mEC25 cells treated with BHI or P. micra CM (n = 4 per group). (L) Tumor growth curves of (K). (M) Representative multiple immunofluorescence images of KRT6A, CD4, and FOXP3 of tumor tissues in (K). (N) Quantification of FOXP3⁺CD4⁺ cells in (M). (O) Representative multiple immunofluorescence images of KRT6A, CD8, GZMB, and PD-1 of tumor tissues in (K). (P) Quantification of PD-1⁺CD8⁺ cells and GZMB⁺CD8⁺ cells in (O). For all panels, data are presented as the means ± SD. P values are determined using a two-tailed Student’s t test. **P < 0.01, ***P < 0.001, and ****P < 0.0001; n.s., not significant. See also fig. S6.

Fig. 6.. Metabolomic analysis reveals altered metabolite profiles in the tumor interstitial fluid of P. micra–treated allografts.

(A) Schematic of metabolomics analysis in the interstitial fluid of tumor following intratumoral injection of P. micra. (B) PCA plot showing overall metabolite patterns in tumor interstitial fluid of the P. micra group and PBS group. (C) Stick plot showing differential levels of bacterium-derived metabolites between the P. micra group and the PBS group. (D) Stick plot showing the top 10 differential metabolite levels in the P. micra group and PBS group. (E) Bar plot showing the quantification of p-cresol in P. micra CM or BHI (n = 3 biological replicates). The P value is determined using a two-tailed Student’s t test. **P < 0.01. See also fig. S7.

Fig. 7.. P. micra–derived metabolite p-cresol induces T_reg cell differentiation through ROS-mediated signaling.

(A) Western blot analysis of T_reg-related proteins in Jurkat cells treated with p-cresol of different concentrations. (B) Relative ROS level of Jurkat cells treated with p-cresol of different concentrations (n = 3 biological replicates). (C) Relative ROS level of Jurkat cells treated with p-cresol, NAC, or a combination of p-cresol and NAC (n = 4 biological replicates). (D) Western blot analysis of T_reg-related proteins in Jurkat cells treated with p-cresol, NAC, or a combination of p-cresol and NAC. (E) Flow cytometry of FOXP3 and CD25 in mouse spleen CD4⁺ T cells treated with p-cresol, NAC, or a combination of p-cresol and NAC (n = 3 biological replicates). (F) Quantification of (E). (G) Box plot showing the ROS signature score of CD4⁺ T cells subtypes. Data are presented as the median with 25th to 75th percentiles with a 1.5× quantile range represented by whiskers and outliers. (H) Spearman correlation between the P. micra abundance and the ROS signature score of CD4⁺ T cells. The gray area represents 95% CI. (I) Excised tumor images of allografts derived from subcutaneous transplantation of mEC25 cells treated with PBS, p-cresol, NAC, or a combination of p-cresol and NAC (n = 4 per group). (J) Tumor growth curves of (I). (K) Representative multiple immunofluorescence images of KRT6A, CD4, and FOXP3 of tumor tissues in (I). (L) Quantification of FOXP3⁺CD4⁺ cells in (K). For all panels, data are presented as the means ± SD. P values are determined using a one-way analysis of variance (ANOVA). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. See also fig. S7.