Tumor-infiltrating bacteria disrupt cancer epithelial cell interactions and induce cell-cycle arrest

Abstract

Tumor-infiltrating bacteria are increasingly recognized as modulators of cancer progression and therapy resistance. We describe a mechanism by which extracellular intratumoral bacteria, including Fusobacterium, modulate cancer epithelial cell behavior. Spatial imaging and single-cell spatial transcriptomics show that these bacteria predominantly localize extracellularly within tumor microniches of colorectal and oral cancers, characterized by reduced cell density, transcriptional activity, and proliferation. In vitro, Fusobacterium nucleatum disrupts epithelial contacts, inducing G0-G1 arrest and transcriptional quiescence. This state confers 5-fluorouracil resistance and remodels the tumor microenvironment. Findings were validated by live-cell imaging, spatial profiling, mouse models, and a 52-patient colorectal cancer cohort. Transcriptomics reveals downregulation of cell cycle, transcription, and antigen presentation genes in bacteria-enriched regions, consistent with a quiescent, immune-evasive phenotype. In an independent rectal cancer cohort, high Fusobacterium burden correlates with reduced therapy response. These results link extracellular bacteria to cancer cell quiescence and chemoresistance, highlighting microbial-tumor interactions as therapeutic targets.

Keywords: Fusobacterium; cancer progression; cell-cycle arrest; chemoresistance; colorectal cancer; epithelial cell-to-cell contacts; host-pathogen interactions; intratumoral bacteria; live-cell confocal imaging; spatial single-cell transcriptomics; tumor microenvironment; tumor-infiltrating bacteria.

Figure 1.. Spatial distribution of cancer cells in bacteria-infected microniches

(A) Top: overlaid image of RNAscope and IHC images of a patient oral squamous cell carcinoma (H_OSCC_01) as indicated in Figures S1A and S1B, respectively. Representative scans from three OSCC patients. Middle: overlaid image of RNAscope and IHC images of a patient colorectal (H_CRC_01) tumor as indicated in Figures S1D and S1E, respectively. Representative scans from three CRC patients. Bottom: IHC scan staining of a mouse colon adenocarcinoma (M_CRC) tumor against Vimentin, Ki67, and F. nucleatum from mouse tumor tissue. Representative scans from 4 mouse colon adenocarcinomas. (B) Confocal images showing the distribution of the cancer nuclei across each region of interest (ROI) for each cancer type as indicated in (A). Color bars indicate average nearest distances among 5 neighboring nuclei. (C) Violin plot shows the distribution of the average distance of each nucleus (data points) to their 5 nearest neighbors in ROIs containing (+) or not (−) bacteria for each cancer type as it is indicated. Pooled data from at least three different tumor samples. (D and E) Boxplots show the distribution of total number of cancer cells PanCK⁺ (D) or percentage of PanCK⁺/Ki67⁺ cells per each field of view (data points; n = 9) comparing ROIs that contain (+) or not (−) bacteria per each cancer type as it is indicated. Pooled data from at least three different tumor samples. In boxplots, whiskers indicate the range of the data, the box encloses the interquartile range, and the line indicates the median. In violin plots, the outline represents the data distribution, dashed lines indicate the interquartile range, and the solid bar indicates the median. No additional error bars (SEM or SD) are shown. Note: In RNAscope and IHC images, variations in shading or faint box outlines reflect whole-slide scanner acquisition and tile stitching in the original image. Also see Figures S1-S3.

Figure 2.. Extracellular bacteria in highly infected tumor regions

(A) RNAscope image staining against eubacteria or F. nucleatum rRNA for each cancer type as it is indicated. Magnified inserts show raw or image-analyzed fluorescence detecting the bacteria foci for ROIs with high (MOI > 80) or low (MOI < 20) bacteria load. Bar plots indicate the percentage of extracellular vs. intracellular bacteria for each ROI. Representative scans from at least three tumor samples. (B) Boxplot indicates the percentage of extracellular bacterial 16S rRNA or F. nucleatum for each ROI (data points) per each cancer type as it is indicated. Pooled data from at least three different tumor samples. p values calculated by mixed-effects model followed by Tukey multiple comparison test. (C and D) x-y plots showing the correlation of extracellular bacterial 16S rRNA (D) or F. nucleatum 23S rRNA (E) with the multiplicity of infection (MOI) per each ROI (data points) for each cancer type as it is indicated. (E and F) Boxplot showing the distribution of cell densities in ROIs (data points) with high or low bacteria load (MOI) for each cancer type as it is indicated (E). Violin plot showing the average distance for each DAPI+ host cell (data points) to their 5 nearest neighbors in ROIs with high or low bacteria load (MOI) for each cancer type as it is indicated (F). Pooled data from at least three different tumor samples. p values were calculated by multiple Mann-Whitney test followed by two-stage step-up method of Benjamini, Krieger, and Yekutieli. In boxplots, whiskers indicate the range of the data, the box encloses the interquartile range, and the line indicates the median. In violin plots, the outline represents the data distribution, dashed lines indicate the interquartile range, and the solid bar indicates the median. No additional error bars (SEM or SD) are shown. Note: In RNAscope images, variations in shading or faint box outlines reflect whole-slide scanner acquisition and tile stitching in the original image. Also see Figures S2 and S3.

Figure 3.. F. nucleatum induces interepithelial cancer cell detachment

(A) Time-lapse imaging showing HCT-116 cells (Lifeact-GFP⁺/PI⁻) expanding through the field of view following exposure with F. nucleatum (SB010, Fna) at MOI = 100 (F. nucleatum +) or without F. nucleatum (Bacteria−), over a 17 h imaging period. Representative time series from four independent experiments. (B) Quantification of the fold change of the extracellular space (surface area) occupied by Lifeact-GFP⁺/PI⁻ HCT-116 cells over time relative to 0 h as it is indicated in (A). (C) Boxplot shows the fold change of the extracellular surface area at the end of the imaging period (16 h) relative to 0 h with or without F. nucleatum exposure as it is indicated in (B). (D) Magnified confocal images from a time-lapse confocal imaging as indicated in (A) showing epithelial cell-to-cell detachment under F. nucleatum exposure. Dash line measures the Lifeact-GFP (Green) and F. nucleatum (Magenta) fluorescent intensity profile at different time points. (E) Scatterplot shows the distribution of the GFP-Lifeact and F. nucleatum fluorescent intensity through the dash line distance at a given time point at it is indicated in (D). (F) Confocal images show the distribution of cancer nuclei across the field of view of HCT-116 Lifeact-GFP cells infected with F. nucleatum at MOI = 100 (F. nucleatum +) or without F. nucleatum (Bacteria −), for 20 h. Representative images from 12 ROIs from three independent experiments. (G and H) Violin plots show the distribution of the average distance of each nucleus to their nearest neighbors (G) or total number of cancer cells per each ROI (data points; n = 12) (H) as indicated in (F). (I) Confocal images showing HCT-116 Lifeact-GFP spheroids infected with F. nucleatum at MOI = 100 (F. nucleatum +) or without F. nucleatum (Bacteria −), for 20 h. Inserts indicate magnified images. Color bar indicates sphericity. Representative images from at least 12 spheroids per condition from three independent experiments. (J–L) Violin plots show the distribution of sphericities for each spheroid (J), the number of detached objects (K), or the nearest distance of each detached object from the main body mass (L) as it is indicated. Pooled data from at least 12 spheroids (data points) from three independent experiments. (M) TEM images showing interepithelial cell contacts in spheroids exposed to F. nucleatum at MOI = 100 (F. nucleatum +) or without F. nucleatum (Bacteria −). Representative images from three independent experiments. In boxplots, whiskers indicate the range of the data, the box encloses the interquartile range, and the line indicates the median. In violin plots, the outline represents the data distribution, dashed lines indicate the interquartile range, and the solid bar indicates the median. No additional error bars (SEM or SD) are shown. p values were calculated by Mann-Whitney test. Also see Videos S1 and S2.

Figure 4.. Cell-cycle dynamics in Fusobacterium-exposed cancer cells

(A) Live-cell confocal imaging of HCT-116 FUCCI⁺ cells growing in 2D surfaces with (+) or without (−) F. nucleatum exposure at MOI = 100 for ~16 h imaging period. Representative time series from four independent experiments. (B) Time-lapse plot shows the fold change of the number of cancer cells over time relative to 0 h incubated with F. nucleatum at MOI = 0, 1, 10, and 100 as it is indicated in (A). (C) Quantification of the fold change of the number of cells relative to 0 h at the end of the imaging at 16 h as indicated in (B). (D) Time lapse plots showing cell-cycle dynamics in cancer cells as describes in (A). (E) Boxplots showing the percentages of G0-G1 or G2-M cells at 0 h or 16 h for each experimental condition as it is indicated. Pooled data from 4 independent experiments (data points). (F) Confocal images showing HCT-116 FUCCI spheroids exposed to (+) or without (−) F. nucleatum at MOI = 100 for 20 h. Representative images from at least 12 spheroids per experimental condition from three independent experiments. (G and H) Violin plots show the distribution of the percentages of G0-G1, S, and G2-M cells (G) and total number of cancer cells (H) per spheroid (data points; n = 12) as indicated in (F). (I) Confocal images showing recovered spheroids after 24 or 72 h of post-F. nucleatum infection. Representative images from three independent experiments. (J) Boxplots show the size (volume μm³) of each spheroid (data points) at different time points per experimental condition as indicated in (I). (K) Time-lapse plot shows the average tumor diameter (data points) growing on mice’s flanks derived from CT-26WT spheroids previously infected with F. nucleatum as it is indicated. Pooled data from 10 mice per experimental condition. Data points represent the mean, and error bars show the standard error of the mean (SEM). (L) Scan images showing HCT-116 GFP-Lifeact spheroids growing in 6-well plates for 1 month after 5-FU post-treatment at 0 or 200 μM and F. nucleatum postinfection. Pooled data from 5 experimental condition. (M) Boxplots show the total GFP-Lifeact area of spheroids growing in each well (data points; n = 5) as indicated in (L). (N) Schematic overview and quantification of total bacterial load in WGSN data from tumor and tumor-adjacent normal tissues in 92 rectal cancer patients with known pathological responses to neoadjuvant therapy. (O) Violin plots showing all bacteria (total bacteria) or Fusobacterium load in tumor and tumor-adjacent normal tissues stratified by MPR status. In boxplots, whiskers indicate the range of the data, the box encloses the interquartile range, and the line indicates the median. In violin plots, the outline represents the data distribution, dashed lines indicate the interquartile range, and the solid bar indicates the median. No additional error bars (SEM or SD) are shown. p values from (C) calculated by One-way ANOVA followed by Tukey multiple comparison test. p values from (E), (G), and (M) were calculated by multiple Mann-Whitney test followed by two-stage step-up method of Benjamini, Krieger, and Yekutieli. p values from (J) calculated by Mixed-effects model followed by Tukey multiple comparison test. p values from (H) calculated by Mann-Whitney test. p values from (O) calculated by Wilcoxon rank sum tests. Also see Figures S4-S6 and Video S3.

Figure 5.. Interepithelial interactions modulate cell-cycle dynamics in cancer cells

(A) Confocal images showing HCT-116 FUCCI spheroids treated for 20 h with trypsin 1%, DMSO 1%, or vehicle as it is indicated. Representative images from at least 12 spheroids per condition from three independent experiments. (B and C) Violin plots show distribution of the percentage of G0-G1, S, and G2-M cells (B) or total number of cancer cells (C) per each spheroid (data points; n = 12) for each experimental condition indicated in (A). (D) Confocal images showing cancer cells growing under ultra-low attachment conditions after 24 or 72 h post-treatment indicated in (A). Representative images from three independent experiments. (E and F) Boxplots show the distribution of the volume of each object (E) or total number of objects per field of view (F) at 24, 48, and 72 h as it is indicated. Pooled data from at least 15 cellular structures per condition from three independent experiments. (G) Confocal images showing HCT-116 FUCCI⁺ cells growing in 3D or 2D cultures with or without F. nucleatum exposure at MOI = 100 for 20 h. Representative images of at least 12 field of view (2D) and 12 spheroids (3D) from three independent experiments. (H and I) Violin plots show the distribution of G0-G1, S, and G2-M cells (H) or total number of cancer cells (I) per spheroid (data points; n = 12) as indicated in (G). (J) Bar plots showing the differentially regulated signaling pathways comparing HCT-116 cells growing in 2D or 3D culture methods treated with or without F. nucleatum at MOI = 100 for 20 h. Upregulation (green) or downregulation (red) of enriched signaling pathways in the first condition listed relative (vs.) to the second condition (condition 1 vs. condition 2) as it is indicated. (K and L) Volcano plots show the upregulation (green) or downregulation (red) of genes in the first condition listed relative (vs.) to the second condition (condition 1 vs. condition 2) as indicated in Figures S7D and S7E. Pooled data from 3 independent experiments. In boxplots, whiskers indicate the range of the data, the box encloses the interquartile range, and the line indicates the median. In violin plots, the outline represents the data distribution, dashed lines indicate the interquartile range, and the solid bar indicates the median. No additional error bars (SEM or SD) are shown. p values from (B) and (C) calculated by Kruskal-Wallis test followed by Dunn’s multiple comparisons test. p values from (E), (F), and (H) calculated by mixed-effects model followed by Tukey multiple comparison test. p values from (I) calculated by Mann-Whitney test. Also see Figure S7 and Tables S1A-S1H.

Figure 6.. Extracellular bacteria modulate cancer epithelial cell cycle and distribution

(A) Confocal images showing HCT-116 infected for 3 h with (+) or without (−) F. nucleatum or F. necrophorum at MOI = 100. Representative images from at least 10 fields of view per condition from three independent experiments. (B and C) Violin plots show the distribution of the percentage of extracellular bacteria (B) or percentage of HCT-116 cells with intracellular bacteria (C) for each field of view (data points; n = 10) as indicated in (A). (D) Confocal images showing HCT-116 FUCCI spheroids infected with (+) or without (−) F. nucleatum or F. necrophorum at MOI = 100 for 20 h. Representative images from at least 12 spheroids per condition from three intendent experiments. (E–G) Violin plots show the distribution of the percentage of G0-G1 and G2-M cells (E), total number of cancer cells (F), and percentage of dead cells (G) for each spheroid (data points; n = 12) as indicated in (D). (H) Confocal images showing HCT-116 LBR⁺ growing in 2D surfaces exposed with F. nucleatum at MOI = 100 for 3 h. (I) Confocal images showing HCT-116 LBR⁺ spheroids exposed to (+) or without (−) F. nucleatum at MOI = 100 for 20 h. Representative images from at least 12 spheroids per condition from three independent experiments. (J and K) Violin plots show the distribution of the average distance of each cancer nuclei to their 5 nearest neighbors (J) or the total number of cancer cells (K) per each spheroid (data points; n = 12) as indicated in (I). (L) Higher-resolution confocal images showing HCT-116 LBR⁺ spheroids exposed to F. nucleatum as it is indicated in (I). Representative images from at least 12 ROIs from three independent experiments. (M) Violin plot shows the distribution of HCT-116 nuclei (data points; n = 12) containing F. nucleatum as it is indicated in (L). (N) Confocal images showing HCT-116 spheroids embedded in FFPE blocks stained against p-γH2AX and F. nucleatum as it is indicated. Representative images from at least 12 spheroids (data points) per condition from three intendent experiments. (O) Violin plots showed the distribution of the percentages of cancer cells p-γH2AX⁺ per spheroid (data points; n = 12) as indicated in (N). (P) Image scans showed RNAscope and IHC staining of a mouse colon adenocarcinoma tumor. Representative scans of at least 12 ROIs per condition from four mouse colon adenocarcinomas. (Q and R) Boxplots show the distribution of the percentage of cells for each host cell population and for each ROI (data points; n = 12) as it is indicated in (P). In boxplots, whiskers indicate the range of the data, the box encloses the interquartile range, and the line indicates the median. In violin plots, the outline represents the data distribution, dashed lines indicate the interquartile range, and the solid bar indicates the median. No additional error bars (SEM or SD) are shown. p values for (B), (C), (E), (F), and (G) were calculated by Kruskal-Wallis test followed by Dunn’s multiple comparisons test. p values for (J) and (K) were calculated by Mann-Whitney test. p values for (Q) and (R) were calculated by multiple Mann-Whitney test followed by two-stage step-up method of Benjamini, Krieger, and Yekutieli. Also see Figure S8.

Figure 7.. Transcriptomic characterization of cancer cells in bacteria-infected microniches

(A) Scan images showing: overlaid images of IHC and RNAscope scans (top left) of a patient’s colorectal tumor, H_CRC_03. Spatial distribution of bacteria and host transcriptomic signals (top right), annotated tissue spatial niches (bottom left), and cell types (bottom right) following host-bacterial single-cell spatial transcriptomics, as it is indicated. (B) UMAP plots indicating niche composition (top left), host cell types (top right), and distribution of eubacteria (bottom left) and Fusobacterium (bottom right) signals in single-cell spatial transcriptomic data from patient tumor H_CRC_03. (C and D) Dot plot indicating the top five most expressed genes for each host cell population (C) and for each niche (D) as it is indicated in (A). (E and F) Dot plot indicating the percentage and cell counts of each host cell population within each niche (E) and the percentages of expressed genes and average of total reads per each host cell population within each niche (F). (G) Left panel: IHC scan showing the distribution of epithelial cancer cells (PanCK⁺) and immune cells (CD45⁺) across different types of niches as it is indicated. Right panel: distribution of the number of total transcripts per each host cell across the niches. Eubacteria and Fusobacterium spatial signals, as shown. Colored bar indicates the average number of reads. (H) Volcano plots showing the upregulation (green) or downregulation (red) of genes in tumor tissue niche II or IV relative (vs.) to niche III, from patient tumor H_CRC_03, as it is indicated in Figure S9H. (I) Heatmap plot showing the fold change of the production of cytokines in supernatants from HCT-116 and HL-60 cells exposed with F. nucleatum at MOI = 100 for 20 h relative to respective uninfected control condition. Note: In RNAscope and IHC images, variations in shading or faint box outlines reflect whole-slide scanner acquisition and tile stitching in the original image. Also see Figure S9 and Tables S2 and S3.