Proinflammatory and Cancer-Promoting Pathobiont Fusobacterium nucleatum Directly Targets Colorectal Cancer Stem Cells

Abstract

Intestinal bacterial communities participate in gut homeostasis and are recognized as crucial in bowel inflammation and colorectal cancer (CRC). Fusobacterium nucleatum (Fn), a pathobiont of the oral microflora, has recently emerged as a CRC-associated microbe linked to disease progression, metastasis, and a poor clinical outcome; however, the primary cellular and/or microenvironmental targets of this agent remain elusive. We report here that Fn directly targets putative colorectal cancer stem cells (CR-CSCs), a tumor cell subset endowed with cancer re-initiating capacity after surgery and chemotherapy. A patient-derived CSC line, highly enriched (70%) for the stem marker CD133, was expanded as tumor spheroids, dissociated, and exposed in vitro to varying amounts (range 100-500 MOI) of Fn. We found that Fn stably adheres to CSCs, likely by multiple interactions involving the tumor-associated Gal-GalNac disaccharide and the Fn-docking protein CEA-family cell adhesion molecule 1 (CEACAM-1), robustly expressed on CSCs. Importantly, Fn elicited innate immune responses in CSCs and triggered a growth factor-like, protein tyrosine phosphorylation cascade largely dependent on CEACAM-1 and culminating in the activation of p42/44 MAP kinase. Thus, the direct stimulation of CSCs by Fn may contribute to microbiota-driven colorectal carcinogenesis and represent a target for innovative therapies.

Keywords: PTPase; bacterial adhesins; cancer stem cells; carcino-embryonic antigen cell adhesion molecule-1; colorectal cancer; fusobacterium nucleatum; microbiota; tumor microenvironment; tumor spheroids.

Figure 1

Direct interaction between Fn and spheroidal colon cancer stem cells. (A) Flow cytometry analysis for the stem cell marker CD133 in spheroidal CSC-P cells. The percentage of the CD133+ population is indicated. (B) Relative expression level of NANOG and POU5F1/OCT4 mRNA in CRC lines CaCo₂ and HT29 compared with primary colonsphere cells (CSC) grown in proliferative (S) or differentiative (D) medium. Values are the mean ± SD of qPCR experimental duplicates. (C) Flow cytometry plot of CSC-associated fluorescence following incubation with 100 MOI of red-labeled bacteria. Black trace: baseline; Fn: Fusobacterium nucleatum; Sg: Streptococcus gallolyticus. (D) Representative confocal images of red-fluorescent Fn adhering to freshly dissociated CSCs. Left: phase-contrast image; right: transmittance/fluorescence overlay. (E) Flow cytometry plot of FITC-PNA (peanut agglutinin)-stained CSC-P and SA-22 cells. Numbers indicate the % of cells exposing the Gal-GalNac sugar moiety, recognized by PNA. The red dashed line marks the positivity threshold (autofluorescence). (F) Single-parameter flow cytometry histogram for surface CEACAM-1 expression in spheroidal cultures of CSC-P and SA-22 cells. Numbers denote the % of cells above the isotype control (vertical dashed line) threshold. (G) CEACAM-1 and E-cadherin immunodetection in protein homogenates from CSC-P and SA-22 cells. β-actin serves as a control for equal sample loading. Relevant bands are indicated by arrows. (H) Immunoblot analysis of bacteria-adsorbed proteins after 30 min of incubation with homogenates from GFP-transduced CSCs. The input lysate is shown as a positive control for staining. The different bacterial strains used as baits are indicated in italics. Picture representative of 2–3 independent experiments with comparable results.

Figure 2

Proinflammatory signaling of Fn in CSC. (A) Immunoblot analysis of CSC-P homogenates following short (10′ and 30′) exposure to 100 MOI Fn. Bands corresponding to the bacterial sensor NOD2, the phosphorylated (Ser 32) form of the NF-kB inhibitor IkB, total IkB, and β-actin (loading control) are highlighted by arrows. Picture representative of at least three independent experiments. (B) NF-kB luciferase reporter assay in CSC-P cells transfected with a 3kB-luc reporter and exposed to whole bacteria (Fn or E. coli DH5alpha, 100 MOI each), 10 μg/mL of MDP, or TNFα + IL1β (10 ng/mL each, positive control) for 24 h. Values are the mean ± S.E.M. luminescence of duplicate/triplicate samples, normalized to the mean of untreated controls (PBS). * p < 0.05 compared with PBS (ANOVA/Tukey HSD). ** p < 0.005, single-sample t-test, two-tailed. * between columns: p < 0.05, unpaired t-test. (C) Representative flow cytometry histogram of CSC-P cells loaded with the nitric oxide-sensitive dye DAF-FM. Fluorescence distribution of unstimulated (empty histogram) and Fn-treated (200 MOI, 24 h, gray-filled histogram) cells are overlayed. (D) Mean cell fluorescence as in C, averaged from two independent experiments. Values are the mean ± S.E.M. of fluorescence intensities in the FL-2 channel, normalized to the value of the untreated (PBS) sample. * p < 0.05 by single-sample Student’s t-test (two-tailed). (E) Protein array hybridization comparing 42 inflammatory cytokines in the supernatants of CSC-P cells exposed to 100 MOI Fn, or left untreated. Each cytokine was spotted in duplicate. Spots corresponding to IL-8 (CXCL-8, red) and Gro-α (CXCL-1, blue) are circled. Picture representative of two independent hybridizations.

Figure 3

Constitutive Wnt activity in CSC-P cells and modulation by Fn. (A) Flow cytometry analysis of red/green fluorescence distribution in CSC-P cells transduced with the Wnt reporter TopGFPmC. Percentages in the upper quadrants refer to the mCherry+ population expressing the reporter. (B) Effect of bacterial infection on green fluorescence intensity within the GFP+/Red+ population (UR quarter). Fn, or E. coli were administered at 100 MOI for 48 h in a growth-factor-free medium. Values are the GFP mean fluorescence intensity (MFI) normalized to the untreated sample. Representative of two–three independent experiments; * p < 0.01 and ** p < 0.005 by a single-sample t-test (two-tailed). * between columns: p < 0.05, unpaired t-test. (C) Immunoblot detection of phospho(Ser 9) GSK3β and GFP in lysates from reporter-transduced CSC-P under the indicated treatments. Arrows highlight relevant bands and equal loading. (D,E) band densitometry from multiple (n = 3–4) experiments as in (C). Values are the mean ± S.E.M of the band intensities normalized to the untreated sample’s mean (Fn 0 MOI). * p < 0.05 vs. 0 MOI by ANOVA for correlated samples. (F) Cell-Glo^® luminescent survival test for the effect of heat-killed Fn (250 MOI) on CSC-P cell sensitivity to Oxaliplatin. Columns are the mean ± SD of triplicate samples, normalized to the mean of the untreated wells (100% survival). Raw values for the untreated samples in the PBS and Fn groups were 5,035,374 ± 676,612 and 4,732,458 ± 89,854 (p = 0.48), respectively. Picture representative of two independent experiments. * p < 0.05; ** p < 0.01; *** p < 0.005, unpaired Student’s t-test on normalized values.

Figure 4

Fn triggers CEACAM1-dependent protein tyrosine phosphorylation signals in spheroidal CSCs. (A) Representative Western blot analysis of total homogenates from control (/) and Fn-stimulated (Fn) CSC-P. Cells were stimulated for 20 min with 200 MOI Fn in serum-free RPMI. Membranes were stained with antisera directed against tyrosine-phosphorylated proteins (upper) and phosphor-ERK1/2 (Thr202/Tyr204) (lower). Immunoblots for β-actin and total ERK1/2 are reported as a control for equal protein loading. Relevant patterns and bands are indicated by braces and arrows, respectively. (B) Lane (pTyr, range 60–140 kD) and band (pERK) quantitations from 7–8 independent experiments as in (A). Represented values are the mean fold-induction ± S.E.M. Statistics were determined by the Wilcoxon signed-rank test (* p < 0.02). (C,D) Representative immunoblots and relative lane/band quantitations of tyrosine-phosphorylated proteins (upper) and phospho-ERK1/2 (Thr202/Tyr204) (lower) in homogenates of CSCs treated with CbpF-expressing E. coli (E_CbpF), the non-recombinant control strain (E.C.), or no bacteria (/). MOI (100 or 500) are indicated. Loading controls are as in (A). Values are the mean ± S.E.M. of band intensities from multiple (range 2 to 6) experiments, normalized to the mean of the untreated samples. Statistics (* p < 0.05; ** p < 0.01 vs. untreated; black line = EcbpF vs. E.C.) were determined by ANOVA (Tukey HSD post hoc test). (E) Immunoblot analysis of protein phosphorylation signals in CSC cells depleted of CEACAM-1 (#77692) and their mock-infected (pLKO) controls. Cells were stimulated as in (A) with the indicated MOI of Fn or E_CbpF. Stainings and loading controls as in A and C. Successful silencing of CEACAM1 in #77692 cells was verified by anti-CEACAM1 immunoblotting (top panel). (F,G) Lane/band densitometry from multiple (2 to 3) experiments. Values are the mean ± S.E.M of the band intensities, normalized across the treatments to the mean of the pLKO (/) sample. Statistics were performed on non-normalized values by either two-way ANOVA (panel (F), * p < 0.01 (#77692 vs. pLKO; pLKO treated vs. untreated), or a paired Student’s t-test (panel (G), ** p < 0.005 vs. pLKO; * p < 0.05 vs. pLKO). n.s. denotes non-significance.

Figure 5

Effect of Fn on the CEACAM1-SHP2 complex in spheroidal cells. (A) Immunoblot detection of phospho-SHP-2 (Tyr542) and total SHP-2 in total cell lysates from Fn-treated (100 MOI for 20 min) CSC-P cells. Relevant bands around 70 kD are indicated by arrows. (B) Band densitometry for p(Tyr542)SHP-2 from n = 7 independent experiments. Values are the mean ± S.E.M. of stimulation indexes (stimulated/unstimulated). Statistics were determined by the Wilcoxon signed-rank test performed on raw data.(* p < 0.02). (C) CEACAM1/SHP-2 co-immunoprecipitation assay in CSC-P cells exposed to Fn as in (A). The abundance of the two proteins in anti-SHP2 immunoprecipitates (upper and lower left quadrants) and anti-CEACAM-1 immunoprecipitates (upper and lower right quadrants) was determined by standard immunoblotting. (D) CEACAM-1 relative abundance in anti-SHP-2 immunoprecipitates from Fn-stimulated compared with untreated cells over 3 independent experiments. The error bar is the SD. (E) Pull-down assay on protein lysates obtained as in (A,C). A GST fusion protein encompassing the 2 (N terminal and C-terminal) SHP2 phosphotyrosine binding domains (depicted in panel (F)) was used as bait. CEACAM-1 in input lysates (top) and GST-SHP2 precipitates (middle) was revealed by immunoblotting. Equal amounts of immobilized bait protein across different samples were confirmed by reversible Ponceau-S staining (bottom panel).