Fusobacterium nucleatum interacts with cancer-associated fibroblasts to promote colorectal cancer

Abstract

Gut microbial species contribute to colorectal cancer (CRC) by interacting with tumor or immune cells, however if CRC-associated bacteria engage with stromal components of the tumor microenvironment remains unclear. Here, we report interaction between the CRC-associated bacterium Fusobacterium nucleatum and cancer-associated fibroblasts (CAFs), and show that F. nucleatum is present in the stromal compartment in murine CRC models in vivo and can attach to and invade CAFs. F. nucleatum-exposed CAFs exhibit a pronounced inflammatory-CAF (iCAF) phenotype, marked by elevated expression of established iCAF markers, secretion of pro-inflammatory cytokines such as CXCL1, IL-6 and IL-8, generation of reactive oxygen species (ROS), and an increased metabolic activity. In co-culture experiments, the interaction of cancer cells with F. nucleatum-stimulated CAFs enhances invasion, a finding further validated in vivo. Altogether, our results point to a role for the tumor microbiome in CRC progression by remodeling the tumor microenvironment through its influence on cancer-associated fibroblasts, suggesting novel therapeutic strategies for targeting CRC.

Keywords: Fusobacterium nucleatum; Cancer-associated Fibroblasts (CAFs); Colorectal Cancer; Inflammatory CAF (iCAF); Invasion.

Figure 1. Fn is present in the stromal compartment of CRC where it binds to and invades CAFs.

(A) Re-analysis of the CSI-Microbe Pelka2021 dataset (Robinson et al, ; Pelka et al, 2021) showing Fusobacterium-infected cells per mille (‰) for each cell type in the tumor. (B, C) In situ hybridization staining of human CRC tissue with αSMA in brown (B, C), PDPN in purple (C), and Fn in pink (B, C; indicated by white arrows) in two independent stage III, microsatellite stable CRC patients. Scale bar = 50 µm. The image in (C) was visualized using a gamma correction of 0.5 to enhance visibility. (D) Schematic representation of the germ-free CDX2-CreER^T2Apc^fl/fl experiment. Briefly, tumorigenesis was induced by i.p. injections of tamoxifen (75 mg per kg of body weight) for five daily injections, followed by three times a week oral gavage with 10⁸ CFU/mouse of Fn 71 or PBS control. (E) Representative immunofluorescent images showing Fn 71 (red, white arrows), stained with an OMP Fn-specific antibody, colocalized with the stromal marker PDPN (green), and DAPI (gray) in a dysplastic region of the colon from the mouse germ-free CDX2-CreER^T2Apc^fl/fl experiment. Images are representative of two mice from one experiment. Scale bar = 50 µm. (F) Attachment and binding assay of Fn 71 and Ec (red, indicated by white arrows, DAPI in gray) on human CRC tissue. Epithelium (E) and stromal (S) stromal compartments separated by white dashed line. Images are representative of one patient and two independent experiments. Scale bar = 50 µm. (G) Fluorescence images of CAFs (CT5.3 and CAF05) following co-culture with Fn 25586 for 2 h (MOI 50). Nuclei were stained with DAPI (blue) and bacteria with both CFSE (green, staining intra- and extracellular bacteria) and FMAS (red, staining only extracellular bacteria). Bacteria that potentially invaded CAFs appear as CFSE⁺ and FMAS^- and are highlighted by white arrows on the zoom area of the overlay. Pictures are representative of two independent experiments, with two independent cell lines. Scale bar = 20 μm. (H) Quantification of the binding and invasion of CFSE-labeled Fn 25586 (MOI 50 and 500) or Ec (MOI 500) on CAFs (CAF05 and CT5.3) or HCT116 CRC cells after a 2-h co-culture by flow cytometry (n = 5, n = 2 and n = 3 for CT5.3, CAF05 and HCT116, respectively). (I) Quantification of the binding and invasion of Fn 25586, Fn 23726, Fn 71 and Ec (MOI 50) on CT5.3 CAFs after a 4-h co-culture (n = 3 biological replicates). (J) Representative TEM images of Fn 71 invasion (indicated by white arrows) after a 4-h co-culture with CT5.3 CAFs (MOI 50, n = 1 experiment). Scale bar = 5 μm. PDPN podoplanin, αSMA alpha smooth muscle actin, i.p. intraperitoneal. Bar and error bars in (H, I) show the mean ± SD and data points the values from each biologically independent experiment. Statistically significant differences were determined using a repeated measure ANOVA followed by Tukey’s HSD post hoc test. Source data are available online for this figure.

Figure 2. Fn binds to and invades CAFs, potentially through the Gal-GalNAc-Fap2 axis.

(A) Schematic representation of the generated RNA-seq dataset. CAFs and tumor spheroids (T) were isolated from fresh human CRC tumor biopsies, cultured and sent for sequencing. (B, C) Heatmap showing the expression of genes (log₂ of median ratio-normalized expression values) responsible for N- and O-glycan biosynthesis (KEGG_‌N_GLYCAN_‌BIOSYNTHESIS and KEGG_‌O_GLYCAN_‌BIOSYNTHESIS gene sets, respectively) (B) and cadherins (C) in the generated RNA-seq dataset. Columns show the expression assessed in independent experiments of patient-derived (T4) and HT-29 tumor spheroids as well as patient-derived CAFs (n = 9 patients: CAF4, CAF12, CAF16, CAF19, CAF20, CAF22, CAF32, CAF41 and CAF42). Expression values are median ratio-normalized counts on a log₂ scale. (D) Binding and invasion of wild-type Fn 23726, fap2 and fadA mutant on CT5.3 cells after a 4-h co-culture (MOI 50) assessed using the CFSE dye by flow cytometry (n = 3 independent experiments). (E) Binding and invasion of Fn 71 on CT5.3 (4-h co-culture, MOI 50) cells in the presence of GalNAc assessed using the CFSE by flow cytometry (n = 3 independent experiments). T tumor. Bar and error bars in (D, E) show the mean ± SD and datapoints the values from each biologically independent experiment. Statistically significant differences were determined using a repeated measure ANOVA followed by Tukey’s HSD post hoc test. Source data are available online for this figure.

Figure 3. Fn treatment polarizes CAFs to an inflammatory phenotype.

(A) ComBat corrected Fn PathSeq scores in TCGA COAD and READ patients (n = 622 patients), ranked by score; the 75^th percentile was used to define Fn^lo and Fn^hi group. (B) Gene expression of iCAF and myCAF markers in Fn^lo and Fn^hi patients, shown as log₂ median ratio-normalized values. (C) Log₂ fold change in IL-6 and ACTA2 expression in Fn-treated versus untreated CAF05 cells (2-h co-culture, MOI 500) (n = 3 biologically independent experiments) as well as in patient-derived CAFs (CAF4 and CAF32, n = 1 for each). (D) Expression of the myCAF markers αSMA (n = 8) and PDGFRβ (n = 5), and the iCAF markers PDGFRα (n = 4) and Lamin A/C (n = 5) 24 h after a 2-h co-culture with either Fn 25586 or Ec (MOI 500) as measured by flow cytometry. (E) Schematic representation of the generated RNA-seq dataset. CT5.3, or the patient-derived CAF180 and CAF181 CAFs were exposed to Fn 71 (MOI 50) for 4 h, washed with penicillin/streptomycin (P/S)-containing medium and incubated for a further 24 h before RNA sequencing (n = 3 independent cell lines). (F) Heatmap of the RNA-Seq expression log₂ fold change in CT5.3 and patient-derived CAFs exposed to Fn 71 compared to untreated CAFs. (G) Kaplan–Meier curve of progression-free survival by Fn load in either the complete CRC cohort or by CMS4/CRIS-B classification. (H) iCAF and myCAF scores (calculated using the R package singscore and the gene sets from (F)) by Fn load and CMS4/CRIS-B classification. Bar and error bars in (D) show the mean ± SD, and datapoints the values from each biologically independent experiment. Statistically significant differences were determined in (A, H) using a two-tailed t test with Holm’s method adjusted P values, in (D) using a repeated measure ANOVA followed by Tukey’s HSD post hoc test and in (G) using an ANOVA on the Cox proportional hazard models to assess the interaction significance. Source data are available online for this figure.

Figure 4. Fn exposure induces CAFs to secret pro-inflammatory cytokines and increases their metabolic activity.

(A) Cytokine profiling of medium collected 24 h after a 2-h co-culture of Fn 25586 with CT5.3 CAFs (MOI 500) using the Proteome Profiler Human XL Cytokine Array Kit (R&D Systems, n = 1 experiment). (B) CXCL1 concentration in CM 24 h after a 2-h co-culture of Fn 25586 and Ec (MOI 500) with CAF05 and CT5.3 CAFs, as determined by ELISA (n = 3 independent experiments with two independent cell lines). (C, D) IL-6 (C) and IL-8 (D) levels in CM 24 h after a 4-h co-culture of CT5.3, or the patient-derived CAF180 and CAF181 CAFs with Fn 71 and Ec (MOI 50), measured by ELISA (n = 3 independent experiments with three independent cell lines). (E) ROS geneset scores in iCAF and myCAF in the Lee CRC scRNA-Seq dataset. Scores were calculated using UCell on previously labeled CAFs (Koncina et al, 2023). (F) ROS geneset scores in the TCGA CRC dataset segregated into Fn^lo and Fn^hi (Fig. 3A). Scores were calculated using singscore. (G) Cytoplasmic (n = 5 for CAF05 and n = 4 for CAF4, H₂DCFDA) and mitochondrial (n = 6 for CAF05 and n = 4 for CAF4, MitoSOX) ROS levels as analyzed by flow cytometry, in CAFs 24 h after a 2-h treatment with Fn 25586 or Ec (MOI 500). Data in (B–D, G) is shown as the mean ± SD, and the datapoints represent the values from each biologically independent experiment. The horizontal lines in (E, F) show the median. Statistically significant differences in (B–D, G) were determined using a nested ANOVA followed by Tukey’s HSD post hoc test and in (E, F) using a two-tailed t test. Source data are available online for this figure.

Figure 5. Fn promotes CAF-induced migration and invasion of tumor cells.

(A) Schematic representation of the generated RNA-seq dataset. Tumor spheroids (HT-29) were treated with CM collected 24 h after a 2-h co-culture of Fn 25886, MOI 500 with CAF05 or the patient-derived CAF32 and CAF20 CAFs (n = 3 independent experiments with three different cell lines) compared to non-treated CAF-CM. (B) Volcano plot showing genes being differentially expressed (DESeq2 ashr shrunken |fold change | >1.5 and Benjamini and Hochberg method adjusted Wald test P value < 0.05) in Fn-CAF-CM- vs ctr-CAF-CM-treated tumor cells as outlined in (A). (C) Ingenuity pathway analysis of diseases and biological functions of the differentially expressed genes. Activation z-scores, P values and the number of genes are shown for migration and infiltration-related pathways. P values were obtained from the IPA implemented right-tailed Fisher’s exact test. (D) Schematic representation of the Transwell assay. (E) Migration of HCT116 cells after 24 h of co-culture in a Transwell assay with CT5.3 CAFs that had been pretreated for 2 h with either no bacteria or with bacteria (Fn 25586 and Ec, MOI 50 or 500) (n = 9 independent experiments). (F) Representative images of crystal violet-stained HCT116 cells that migrated in the Transwell setup at endpoint. Scale bar = 500 µm. (G) Schematic representation of the in vitro complex 3D spheroid co-culture experimental setup. mCherry-labeled HCT116 CRC cells are co-embedded in collagen with GFP-labeled CAFs (CT5.3). These multi-component spheroids are then treated with CM from CT5.3 CAFs cultured alone (blank medium control) or CM from CT5.3 CAFs co-cultured with Fn 71 or Ec (4-h co-culture, MOI 50), in the presence or absence of NAC (5 mM). (H) Quantification of invasion in the 3D complex spheroid model measured by HCT116 outgrowth (mCherry) over 6 days (n = 3 independent experiments, 2–5 technical replicates included per experiment). (I) Quantified invasion of HCT116 (mCherry) outgrowth at endpoint (day 6 from the data shown in panel 5H). CM conditioned medium, NAC N-acetyl-l-cysteine. The bar chart and error bars in (E, I) show the mean ± SD. Statistically significant differences were determined using a nested ANOVA followed by Tukey’s HSD post hoc test. The error bars in (H) show the mean ± SD and statistically significant differences determined using pairwise two-way repeated measure ANOVAs (time × treatment). The interaction term P values were adjusted using Holm’s method. Source data are available online for this figure.

Figure 6. Fn promotes CAF-induced metastatic spreading of tumor cells in vivo.

(A) Schematic representation of the in vivo experimental setup. SPF NSG mice were tail vein injected with 1× 10⁶ HT-29 cells, which were pretreated with CM of CT5.3 CAFs cultured alone or CM of CT5.3 CAFs co-cultured with Fn 71 or Ec (4-h co-culture, followed by a 24-h incubation, MOI 50). Lungs were harvested after 30 days (n = 8 mice for the Ctrl and Ec conditions and n = 7 for the Fn condition). (B) Quantification of the relative tumor surface area in percentage per imaged lobe (four lobes were analyzed per mouse). (C) Representative H&E images of tumors (indicated by black arrows) in the lungs on day 30. CM conditioned medium. Scale bar = 100 µm. Statistically significant differences were determined using an ANOVA on a robust linear model fit (using the lmRob function in the R package robust) followed by Tukey’s multiple comparison method (using the R package multcomp). Source data are available online for this figure.