Mice with dysfunctional TGF-β signaling develop altered intestinal microbiome and colorectal cancer resistant to 5FU

Abstract

Emerging data show a rise in colorectal cancer (CRC) incidence in young men and women that is often chemoresistant. One potential risk factor is an alteration in the microbiome. Here, we investigated the role of TGF-β signaling on the intestinal microbiome and the efficacy of chemotherapy for CRC induced by azoxymethane and dextran sodium sulfate in mice. We used two genotypes of TGF-β-signaling-deficient mice (Smad4^+/- and Smad4^+/-Sptbn1^+/-), which developed CRC with similar phenotypes and had similar alterations in the intestinal microbiome. Using these mice, we evaluated the intestinal microbiome and determined the effect of dysfunctional TGF-β signaling on the response to the chemotherapeutic agent 5-Fluoro-uracil (5FU) after induction of CRC. Using shotgun metagenomic sequencing, we determined gut microbiota composition in mice with CRC and found reduced amounts of beneficial species of Bacteroides and Parabacteroides in the mutants compared to the wild-type (WT) mice. Furthermore, the mutant mice with CRC were resistant to 5FU. Whereas the abundances of E. boltae, B.dorei, Lachnoclostridium sp., and Mordavella sp. were significantly reduced in mice with CRC, these species only recovered to basal amounts after 5FU treatment in WT mice, suggesting that the alterations in the intestinal microbiome resulting from compromised TGF-β signaling impaired the response to 5FU. These findings could have implications for inhibiting the TGF-β pathway in the treatment of CRC or other cancers.

Keywords: 5FU; Bacteroides; Chemoresistance; Colorectal cancer; Microbiome; TGF-β signaling.

Figure 1.

Significantly altered basal proportions of gut microbiota species in TGF-β signaling-deficient mice (SKO: Smad4+/–/Sptbn1+/– and Smad4+/−). (A) Schematic of AOM/DSS-induced CRC mouse model to explore 5FU response and gut microbiome in WT (n = 16), Smad4+/–Sptbn1+/– (n = 12), and Smad4+/− (n = 4) mice. Fecal samples were collected before CRC induction (basal) and after CRC induction and drug or control treatment (5FU or PBS) at Week 15. (B – D) Relative abundances of bacteria in mice of the indicated genotypes of mice before any treatment (basal). SKO (n = 16), combined results from Smad4+/–Sptbn1+/– (n = 12) and Smad4+/− (n = 4). Data are shown as the average ± SEM. Species shown in B-C are those that exhibited p < 0.01 between WT and Smad4+/–Sptbn1+/– or WT and SKO as determined by unpaired t-test. Fusobacterium species (D) exhibited p < 0.05 between WT and SKO, or Smad4+/–Sptbn1+/– or Smad4+/– as determined by unpaired t-test.

Figure 2.

Smad4+/– and Smad4+/–/Sptbn1+/– mice are more susceptible to AOM/DSS-induced CRC. (A) Representative colon tumors (top, arrow) and pathological analysis by hematoxylin-eosin staining (middle and bottom) in a WT mouse (left) and Smad4+/–/Sptbn1+/– mouse (right). WT mice colon tumor was diagnosed as a high-grade adenoma (i, iii, and v); Smad4+/–/Sptbn1+/– colon tumor was diagnosed as a moderately differentiated adenocarcinoma (ii, iv, and vi). Scale bars, 5 mm, 200 μm, and 100 μm. (B) Bar graph showing summary of histopathological analyses of tumors from TGF-β signaling-deficient mice [Smad4+/–/Sptbn1+/– (n = 5), Smad4+/− (n = 3)] and WT mice (n = 6). Significance testing was performed with Chi-square. (C) Onco-Prints of genomic alterations of SMAD4 and SPTBN1 from TCGA CRC data (n = 526) (top). Kaplan-Meier survival plots of patients stratified according to genetic alterations of SMAD4 and SPTBN1 including Disease-free survival (bottom left) and overall (bottom-right) survival, the P values were calculated by log-rank test.

Figure 3.

Smad4+/– and Smad4+/–/Sptbn1+/– tumors are resistant to 5FU. (A) Mice that are deficient in TGF-β signaling (SKO: Smad4+/–/Sptbn1+/– and Smad4+/−) display significantly increased tumor numbers compared to WT mice. 5FU treatment significantly reduced tumor numbers in WT mice, but not in TGF-β signaling-deficient mice. Significance was performed using an unpaired t-test. (B) SKO mice show the trend of enlarged tumor size compared to WT mice. 5FU treatment significantly reduced tumor size in WT mice, but not in SKO mice. Significance was performed using an unpaired t-test. (C) Tumor size distribution analysis indicated significantly decreased tumor size in WT tumor after 5FU treatment, but no statistical tumor size differences in SKO tumor after 5FU treatment. Significance was performed using Chi-square, *p < 0.05. WT-PBS, n = 6; WT-5FU, n = 6; SKO-PBS (n = 5, Smad4+/–Sptbn1+/–; n = 3, Smad4+/–); SKO-5FU, (n = 7, Smad4+/–Sptbn1+/–; n = 3, Smad4+/–).

Figure 4.

Smad4+/–/Sptbn1+/– tumor after 5FU treatment progresses to liver metastasis. Gross view of primary colon tumors (A) and liver metastasis (B). The histopathological assessment confirmed the tumors in A are primary CRC (C, E) and tumor nodules shown in the B are colon cancer liver metastasis (D, F). Scale bars, 5 mm, 200 μm and 100um.

Figure 5.

Proliferation and apoptosis in TGF-β signaling–deficient tumors by Ki67 and cleaved caspase 3 immunohistochemistry. (A) Representative colon sections from WT and Smad4+/–/Sptbn1+/– mice stained for Ki67. Scale bars, 100 μm, and 20 μm. The percentage of Ki67+ nuclei per field (5 randomly selected fields) was determined. (B) 5FU treatment significantly decreased the Ki67 index in both WT and TGF-β signaling–deficient colon tumors. Compared to the WT colon tumor, TGF-β signaling–deficient colon tumor showed a significantly higher Ki67 index even after 5FU treatment. (C-D) Representative colon sections from each group with positive staining of cleaved caspase-3 (arrows) and quantification of the number of cells with positive staining of cleaved caspase-3. Significance was performed using a one-way ANOVA test for 4 groups comparison in B (p < 0.0001) and D (p = 0.0013). In B, by Bonferroni’s multiple comparisons test for Ki67: **p<0.005 for WT-PBS vs. WT-5FU; *p<0.05 for SKO-PBS vs. WT-PBS; **p<0.005 for SKO-PBS vs. SKO-5FU; *p<0.05 for WT-5FU vs SKO-5FU. In D, by Bonferroni’s multiple comparisons test for caspase 3: **p<0.005 for WT-PBS vs. WT-5FU; no significance between SKO-PBS vs. WT-PBS or between SKO-PBS vs SKO-5FU; **p<0.005 for WT-5FU vs SKO-5FU. WT-PBS, n = 6; WT-5FU, n = 6; SKO -PBS (n = 5, Smad4+/−Sptbn1+/−; n = 3, Smad4+/−); SKO −5FU, (n = 7, Smad4+/−Sptbn1+/−; n = 3, Smad4+/−).

Figure 5.

Proliferation and apoptosis in TGF-β signaling–deficient tumors by Ki67 and cleaved caspase 3 immunohistochemistry. (A) Representative colon sections from WT and Smad4+/–/Sptbn1+/– mice stained for Ki67. Scale bars, 100 μm, and 20 μm. The percentage of Ki67+ nuclei per field (5 randomly selected fields) was determined. (B) 5FU treatment significantly decreased the Ki67 index in both WT and TGF-β signaling–deficient colon tumors. Compared to the WT colon tumor, TGF-β signaling–deficient colon tumor showed a significantly higher Ki67 index even after 5FU treatment. (C-D) Representative colon sections from each group with positive staining of cleaved caspase-3 (arrows) and quantification of the number of cells with positive staining of cleaved caspase-3. Significance was performed using a one-way ANOVA test for 4 groups comparison in B (p < 0.0001) and D (p = 0.0013). In B, by Bonferroni’s multiple comparisons test for Ki67: **p<0.005 for WT-PBS vs. WT-5FU; *p<0.05 for SKO-PBS vs. WT-PBS; **p<0.005 for SKO-PBS vs. SKO-5FU; *p<0.05 for WT-5FU vs SKO-5FU. In D, by Bonferroni’s multiple comparisons test for caspase 3: **p<0.005 for WT-PBS vs. WT-5FU; no significance between SKO-PBS vs. WT-PBS or between SKO-PBS vs SKO-5FU; **p<0.005 for WT-5FU vs SKO-5FU. WT-PBS, n = 6; WT-5FU, n = 6; SKO -PBS (n = 5, Smad4+/−Sptbn1+/−; n = 3, Smad4+/−); SKO −5FU, (n = 7, Smad4+/−Sptbn1+/−; n = 3, Smad4+/−).

Figure 6.

Shotgun metagenomic sequencing data analyses of the gut microbiome. A) The bioinformatics analysis workflow utilizes two HIVE platform tools, CensuScope and HIVE-Hexagon. CensuScope is used to determine the taxonomic composition of each sample. HIVE-Hexagon is an alignment tool used to remove host DNA and determine the number of alignments or ‘hits’ made to the reference base. After the two-step metagenomics pipeline, relative abundance is calculated using the number of hits. The final abundance profile lists all the bacteria present in the samples and their relative abundances. B-C) Differences in gut microbiota composition on a family-level (B) and a genus-level (C) are observed in mice fecal samples. The bar plots were generated using the Microbiome R package (https://microbiome.github.io/tutorials/). WT-basal, n = 16; WT-tumor-PBS, n = 7; WT-tumor-5FU, n = 7; SKO-PBS n = 8 (n = 5, Smad4+/−Sptbn1+/−; n = 3, Smad4+/−); SKO-5FU, n = 10 (n = 7, Smad4+/−Sptbn1+/−; n = 3, Smad4+/−).

Figure 7.

Unique gut microbiome signature in 5FU resistant SKO, Smad4+/– and Smad4+/–/Sptbn1+/– CRCs. (A) Bacteria species and strains that are significantly reduced in AOM/DSS-induced tumors and recover to basal levels after 5FU treatment in WT mice but not in SKO mice. (B) Bacteria species and strains that are significantly reduced in AOM/DSS-induced tumors and SKO mice at basal levels before any treatment. Data in A-B are plotted as the average of relative abundance ± SEM, the significance is performed using an unpaired t-test, WT-basal, n = 16; WT-tumor + PBS, n = 7; WT-tumor + 5FU, n = 7; SKO-PBS, n = 16; SKO-tumor + PBS, n = 8 (n = 5, Smad4+/−Sptbn1+/−; n = 3, Smad4+/−); SKO-tumor + 5FU, n = 10 (n = 7, Smad4+/−Sptbn1+/−; n = 3, Smad4+/−); *p < 0.05. (C) A Heatmap of Bacteroides and Lactobacillus species and strains that are altered in AOM/DSS induced tumors and recovered to basal levels after 5FU treatment in WT mice but not recovered in SKO mice deficient in TGF-β signaling.