Gut microbial metabolite 4-hydroxybenzeneacetic acid drives colorectal cancer progression via accumulation of immunosuppressive PMN-MDSCs

Abstract

Colorectal cancer (CRC) is characterized by an immune-suppressive microenvironment that contributes to tumor progression and immunotherapy resistance. The gut microbiome produces diverse metabolites that feature unique mechanisms of interaction with host targets, yet the role of many metabolites in CRC remains poorly understood. In this study, the microbial metabolite 4-hydroxybenzeneacetic acid (4-HPA) promoted the infiltration of PMN myeloid-derived suppressor cells (PMN-MDSCs) in the tumor microenvironment, consequently inhibiting the antitumor response of CD8+ T cells and promoting CRC progression in vivo. Mechanistically, 4-HPA activates the JAK2/STAT3 pathway, which upregulates CXCL3 transcription, thereby recruiting PMN-MDSCs to the CRC microenvironment. Selective knockdown of CXCL3 resensitized tumors to anti-PD-1 immunotherapy in vivo. Chlorogenic acid reduces the production of 4-HPA by microbiota, likewise abolishing 4-HPA-mediated immunosuppression. The 4-HPA content in CRC tissues was notably increased in patients with advanced CRC. Overall, the gut microbiome uses 4-HPA as a messenger to control chemokine-dependent accumulation of PMN-MDSC cells and regulate antitumor immunity in CRC. Our findings provide a scientific basis for establishing clinical intervention strategies to reverse the tumor immune microenvironment and improve the efficacy of immunotherapy by reducing the interaction among intestinal microbiota, tumor cells, and tumor immune cells.

Keywords: Cancer immunotherapy; Colorectal cancer; Gastroenterology; Immunology; Oncology.

Figure 1. F. nucleatum and E. coli stimulated cytokine secretion in CRC cells.

(A) α Diversity analysis using the Chao1 and the Simpson indexes in the CRC (n = 20) and HC (n = 20) groups. (B) Principal coordinate analysis at the species level between the CRC (n = 20) and HC (n = 20) groups. (C) The discriminant analysis effect size method identified marker species between the CRC (n = 20) and HC (n = 20) groups. Blue and red bars represent markers enriched in the CRC and HC groups, respectively. (D) A human cytokine antibody array was applied to detect the changes of inflammatory factors in CM of HCT116 cells treated with F. nucleatum or E. coli. Differential cytokines associated with immune cell chemotaxis are shown in the black boxes. A cytokine chip Wayne diagram is shown below those boxes. (E and F) qPCR analysis revealed changes in cytokine expression after co-culture with F. nucleatum and E. coli for 6 hours (n = 3). (G) ELISA detection of CCL20 and CXCL3 secretion from CRC cells after co-culture with E. coli and F. nucleatum (n = 5). All numerical data and error bars represent the mean ± SD of 3 independent experiments. Statistical analyses were conducted using 1-way ANOVA with Dunnett’s T3 correct multiple-comparison test. *P < 0.05, **P < 0.005, ***P < 0.0005.

Figure 2. F. nucleatum and E. coli mediate immunosuppressive microenvironment in CRC.

(A) Schematic diagram of the microbiota-treated Apc^min/+ tumorigenesis mouse model administration method (n = 5). (B) Colonization efficiency of E. coli and F. nucleatum was assessed by metagenomic sequencing analysis of Apc^min/+ mice after 8 days (5 days of antibiotics treatment to deplete their gut microbiota, followed by 3 days of F. nucleatum, E. coli, or PBS orally administration for microbiota colonization) . The relative abundance of Operational Taxonomic Units (OTUs) in fecal bacterial is shown. (C) Representative images of tumors in the intestines of Apc^min/+ mice are shown. (D) Statistics of tumor load of the intestines derived from Apc^min/+ mice treated with PBS, E. coli, or F. nucleatum (n = 5). (E) Representative images of tumorigenesis of intestines in Apc^min/+ mice visualized by H&E staining. (F) H&E scoring of tumor-related lesions (including inflammation, adenoma, atypical hyperplasia, and crypt fusion) (n =5). (G) The percentages of PMN-MDSCs (CD11b⁺Ly6G⁺Ly6C^low) and M-MDSCs (CD11b⁺Ly6G^-Ly6C^hi) in TILs (CD45⁺) of Apc^min/+ mice were determined by flow cytometry sorting. (H and I) Statistical chart of PMN-MDSCs and M-MDSCs (n = 5). (J) The percentage of Tregs (CD4⁺Foxp3⁺) in TILs (CD45⁺) of Apc^min/+ mice, detected by flow cytometry sorting. (K) Statistical chart of Tregs (n = 5). (L) Tumor-infiltrating CD8⁺T cells in TILs (CD45⁺) of Apc^min/+ mice, detected by flow cytometry sorting. (M). Statistical chart of CD8⁺T cells (n = 5). (N) The granule productions (GzmB⁺) of CD8⁺T cells. (O) Statistical chart of GzmB⁺ cells (n = 5). Data and error bars represent the mean ± SD of 3 independent experiments. Statistical analyses were conducted using 1-way ANOVA with Dunnett’s T3 correct multiple-comparison test. *P < 0.05, **P < 0.005, ***P < 0.0005.

Figure 3. Knockdown CXCL3 inhibits tumor growth by preventing PMN-MDSC accumulation and activating CD8⁺ T-cell infiltration.

(A) Knockdown of Ccl20 and Cxcl3 inhibits the growth of CT26 subcutaneous tumors in nude mice (n = 5). A photograph of CT26 subcutaneous tumors in nude mice and a graph of tumor growth are shown. (B) Knockdown of Ccl20 and Cxcl3 inhibits the growth of CT26 subcutaneous tumors in BALB/c mice (n = 5). A photograph of CT26 subcutaneous tumors in BALB/c mice and a graph of tumor growth are shown. (C) Neutralizing antibodies of Ccl20 and Cxcl3 inhibited subcutaneous tumors in BALB/c mice(n = 6). A photograph of CT26 subcutaneous tumors in BALB/c mice and a graph of tumor growth are shown. (D) Knockdown of Ccl20 and Cxcl3 inhibits the progression of CT26 orthotopic implanted tumor in BALB/c mice (n = 5). Representative tumor images and tumor load are shown. (E) The percentage of PMN-MDSCs (CD11b⁺Ly6G⁺Ly6C^low) in TILs (CD45⁺) of orthotopic implanted CRC mice detected by flow cytometry sorting. A bar chart indicating statistical values is presented (n = 5). (F) The percentage of Tregs (CD4⁺Foxp3⁺) in TILs (CD45⁺) of orthotopic implanted CRC mice detected by flow cytometry sorting. A bar chart indicating statistical values is presented (n = 5). (G and H) Tumor-infiltrating CD8⁺T cells and their granule production (GzmB⁺) in TILs (CD45⁺) of orthotopic implanted CRC mice detected by flow cytometry sorting. Bar charts indicating statistical values are presented (n = 5). (I) MDSCs (Gr-1⁺), Tregs (CD4⁺, Foxp3⁺), and CD8⁺ T-cell infiltration in tumor tissues of orthotopic implanted CRC mice. Representative IHC images are shown. (J) Histogram showing the number of Gr-1⁺, CD4⁺, Foxp3⁺, and CD8⁺ cells per ×20 objective lens visual field (n = 5). Data represent the mean ± SD of 3 independent experiments. We used 2-way ANOVA to determine statistical significance of subcutaneous tumor volume. The remaining statistical methods were conducted using 1-way ANOVA with Dunnett’s T3 correct multiple-comparison test. *P < 0.05, **P < 0.005, ***P < 0.0005. sh, short hairpin.

Figure 4. The CXCL3/CXCR2 axis mediates MDSC recruitment and inhibits T-cell effector function.

(A) IF assays were performed to detect CXCR2 and Gr-1 in orthotopic cecal tumor of BALB/c mice. Scale bar: 50 μm. (B and C) The CXCL3-CXCR2 axis promoted the migratory abilities of MDSCs, as detected by transwell assays (n = 5). (D and E) Representative flow cytometry data show that MDSCs cells isolated from C57 mice inhibited cytokine and cytolytic granule production in CD8⁺T cells (D); the summarized result is presented in (E) (n = 5). (F) Effect of short hairpin Cxcl3 (sh-Cxcl3) and PD-1 immunotherapy on subcutaneous tumor of BALB/c mouse: CT26 subcutaneous tumors (n = 5). The CD279 anti–PD-1 antibody or isotype control (IgG) was i.p. injected three times daily (G and H) Tumor growth (G) and weight (H) were monitored (n = 5). Data represent the mean ± SD of 3 independent experiments. Statistical analyses were conducted using Student’s t test (2-comparison test) and 1-way ANOVA with Dunnett’s T3 correct multiple-comparison test. We used 2-way ANOVA to determine statistical significance of tumor volume. *P < 0.05; **P < 0.005; ***P < 0.0005.

Figure 5. Nontargeted metabolomics reveal key metabolites that mediate CXCL3 secretion.

(A) Identification of differential metabolites in the E. coli and F. nucleatum imbalance models using nontargeted metabolomics, presented volcano plots. (B) Principal component analysis (PCA) comparing the E. coli or F. nucleatum imbalance groups with the control group. (C) A Wayne chart illustrating differential metabolites between the E. coli or F. nucleatum imbalance models and the control group. (D and E) Heatmaps depicting differential metabolites in the E. coli (D) and F. nucleatum (E) imbalance groups compared with the control groups. P < 0.05, 2-tailed Mann-Whitney U test. (F and G) ELISA was used to assess the impact of differential metabolites (1 mM; 48 hours) on CXCL3 levels (n = 5). (H–K) ELISA assays measuring the effects of the 4-HPA concentration gradient and time gradient on CXCL3 secretion (n = 5). Data represent the mean ± SD of 3 independent experiments. Statistical analyses were conducted using 1-way ANOVA with Dunnett’s T3 correct multiple-comparison test. *P < 0.05, ***P < 0.0005. PC2, principal components 2; QC, quality control; VIP, variable importance in projection.

Figure 6. 4-HPA promotes the transcription of CXCL3 regulated by STAT3 in CRC cells.

(A) Chemical structures of 4-HPA, FITC, and FITC-labeled -HPA. (B) Detection of FITC-labeled 4-HPA by fluorescence confocal microscopy in RKO and SW480 cells. Scale bar: 50 μm. (C) Predicted transcription factors for CXCL3. (D) WB analysis of the effects of 4-HPA on transcription factors in CRC cell lines. (E) The binding sites of STAT3 and CXCL3 were confirmed using a dual-luciferase reporter assay (n = 3). (F) Transcriptional regulation of CXCL3 by p-STAT3 was detected using ChIP assays. (G) WB analysis of the JAK2/STAT3 signaling pathway and CXCL3 in SW480 cells. (H–J) WB analysis of CXCL3 in SW480 cells. (K and L) IF assays visualizing the subcellular localization of STAT3 and p-STAT3 in SW480 cells treated with 4-HPA. Scale bar: 50 μm. Data represent the mean ± SD of 3 independent experiments. Statistical analyses were conducted using Student’s t test. ***P < 0.0005. MUT, mutation; NC, negative control.

Figure 7. 4-HPA correlates with PMN-MDSC accumulation in CRC.

(A and B) Orthotopic implanted CRC mice were fed 4-HPA (1 mM) or control water (DMSO). (A) Bar chart of PMN-MDSCs and M-MDSCs measured by flow cytometry (n = 5). (B) Bar chart of CD8⁺ T cells and GzmB⁺ CD8⁺ T cells measured by flow cytometry (n = 5). (C–J) Results from the Apc^min/+ tumorigenesis mouse model. (C) Tumors in the intestines of Apc^min/+ mice are shown, as is a bar chart of tumor load of intestines derived from Apc^min/+ mice colonized with F. nucleatum and fed with or without 4-HPA (1 mM) and CGA (1 mM) (n = 5). (D) Tumors in the intestines of Apc^min/+ mice are shown, as is a bar chart of tumor load in intestines derived from Apc^min/+ mice colonized with E. coli and fed with or without 4-HPA (1 mM) and CGA (1 mM) (n = 5). (E and H) Bar chart of PMN-MDSCs and M-MDSCs measured by flow cytometry (n = 5). (F and I) Bar chart of CD8⁺ T cells and GzmB⁺ CD8⁺ T cells measured by flow cytometry (n = 5). (G and J) Detection of 4-HPA in tumor tissues of Apc^min/+ mice by HRGC-MS (n = 5). Data represent the mean ± SD of 3 independent experiments. Statistical analyses were conducted using 1-way ANOVA with Dunnett’s T3 correct multiple-comparison test. *P < 0.05, **P < 0.005, ***P < 0.0005. NC, negative control; sh, short hairpin.

Figure 8. 4-HPA correlates with PMN-MDSC accumulation in CRC.

(A) Tumors in the intestines of Apc^min/+ tumorigenesis mice (n = 5). (B). Bar chart of tumor load in intestines derived from Apc^min/+ mice treated with PD-1 immunotherapy or IgG, with or without 4-HPA (1 mM) (n = 5). (C) Representative images of intestines tumorigenesis visualized by H&E staining. (D) H&E scoring of tumor-related lesions (including inflammation, adenoma, atypical hyperplasia, and crypt fusion) (n = 5). (E and F) Detection of 4-HPA in tumor tissues of patients with CRC by HRGC-MS (n = 12). (G–K) Expression of CD8, CD33, CD11b, and CXCL3 in CRC tissues of patients with CRC analyzed by multiple IF. Visualization of 3 representative cases is shown. Scale bar: 50 μm. Multiple IF detection was performed on the tumor tissues of patients with CRC (G). Bar charts of CD8⁺ T cells (H), CD11b⁺ cells (I), CD33⁺ cells (J), and MDSCs (CD11b⁺CD33⁺) (K) in CRC tissues (n = 5). Data represent the mean ± SD of 3 independent experiments. We used 2-way ANOVA to determine the significance of tumor volume of PD-1 treated mice and H&E staining. The remaining statistical analyses were conducted with Student’s t test (2-comparison test) and 1-way ANOVA with Dunnett’s T3 correct multiple-comparison test. *P < 0.05, **P < 0.005, ***P < 0.0005. rel., relative.

Figure 9. A schematic of the microbial metabolite 4-HPA mediating CRC immunosuppression.

The gut microbiome uses 4-HPA as a messenger to regulate chemokine CXCL3 level in CRC cells, thereby controlling the accumulation of CXCR2⁺ PMN-MDSCs. The accumulated PMN-MDSCs inhibit the antitumor effect of CD8⁺ T cells.