Multi-omic profiling reveals associations between the gut microbiome, host genome and transcriptome in patients with … – Journal of Translational…

Identification of a set of gut microbes associated with CRC

Most colorectal cancers arise from adenoma to carcinoma as verified by diet, inflammatory processes, gut microbiota, and genetic alterations. Nonetheless, the mechanism by which the microbiota interacts with these etiologic factors to promote CRC is not clear. Therefore, we collected stool samples, tumor and matched normal tissues from 41 CRC individuals, and carried out multi-omics sequencing analyses to evaluate the interplay between cancer cells and gut microbiome (Fig.1 and Additional file 1: Table S1). As shown in Additional file 1: Fig. S1a, the stool samples were subjected to metagenomic sequencing, achieving an average of 7Gb clean data. Additionally, we conducted whole exome sequencing, ensuring a minimum of more than 100X coverage and 20Gb data, respectively (Additional file 1: Table S2).

Metagenomics sequencing of the stool sample and exome and transcriptome sequencing of mucosa tissue in colorectal cancer. We collected stool specimens and matched tumor and normal mucosa tissue from 41 colorectal cancer patients. The former samples were metagenomically shotgun sequenced to yield taxonomic and functional profiles; the latter were processed using exome and transcriptome sequencing technology respectively. Features of the microbiome were correlated with clinic elements somatic mutations, and differentially expressed genes, respectively

We first examined the microbiome dysbiosis by integrating our metagenomic sequencing data with a public Chinese colorectal cancer cohort3 (CRC cohort2 and CON) (Fig.2A). Compared with healthy controls, the CRC patients in our cohort exhibited a significantly decreased alpha diversity (Additional file 1: Fig. S1b), but no obvious difference in the beta diversities (Additional file 1: Fig. S1c). To investigate the alterations in microbiota structure, we conducted the linear discriminant analysis effect size (LEfSe) analysis to compare healthy controls and combined tumor samples. Totally, there were 2 taxa (Viruses_noname and Fusobacteria) at the phylum level and 10 at the genus level significantly altered respectively (Fig.2B and Additional file 1: Table S3). Notably, we figured out 22 species associated with disease status, of which 14 were elevated in CRC group (Fig.2C). Of them, Bacteroides fragilis (LDA=3.897), Parabacteroides spp. (LDA=3.499) and Prevotella intermedia (LDA=3.452) exhibited the highest abundances in CRC patients. In contrast, eight species were enriched in healthy controls, including Faecalibacterium prausnitzii (LDA=4.299), Eubacterium rectale (LDA score=4.255), Eubacterium eligens (LDA=4.002), and so on.

A Microbiome alteration between healthy and CRC subjects. PCoA plot showed the two cohorts used in our project. B Taxonomic profile difference detected with LEfSe. C Differentially abundant species between healthy controls and CRC patients. D Differentially abundant KEGG pathways between healthy controls and CRC patients. E Unsupervised clustering uncovered associations between differentially abundant species and clinic covariates

To further investigate the functions of 22 tumor-associated bacteria, we used HUManN2 to estimate the relative abundance of KEGG ontology (KO) categories. Disease associated KEGG pathway changes were further identified using the method described in Feng Q. et al.4 We observed that bacteria related metabolic pathways were enriched in CRC groups. Especially, one carbon pool by folate metabolic pathway of microbiota was significantly (Reporter score=3.471) higher in CRC patients (Fig.2D). The one carbon pool by folate is a universal cell metabolic process supporting tumorigenesis, obtaining folate (vitamin B9) and cobalamin (vitamin B12) from diet. Furthermore, the cancer enriched species showed positive correlations with the metabolic pathways such as carbon metabolism and oxidative phosphorylation, whereas some well-known beneficial bacteria (including Faecalibacterium prausnitzii), displayed negative correlations (Additional file 1: Fig. S2).

Next, we investigated associations between overall microbiome configuration with CRC clinical covariates. Clinically, of the cohorts 41 individuals (63% male; ages 4679), 26 subjects (63%) belong to COAD and 15 subjects (37%) had carcinomas at rectum. Additionally, 10 subjects were diagnosed at early stage and 31 subjects (76%) at later stage. Among all 41 individuals, we observed that several paraprevotella.ssp were elevated in patients with age<65 (for example, paraprevotella clara, LDA score=3.051; paraprevotella xylaniphila, LDA score=2.478) (Additional file 1: Fig. S3a). Furthermore, Clostridium clostridioforme was predominated found in females (Additional file 1: Fig. S3b, LDA score=3.182). As to Bacteroides genus, the abundance of Bacteroides eggerthii was significantly increased in COAD (LDA score=3.625) whereas Bacteroides massiliensis was enriched in READ (LDA score=4.985) (Additional file 1: Fig. S3c). Bifidobacterium, one of the major probiotics, exhibited a significant increase in the early stage and individuals with age<65 (Bifidobacterium longum, LDA score=3.698; Bifidobacterium dentium, LDA score=2.102) (Additional file 1: Fig. S3a and d).

We also assessed the connections between clinical characteristics and 22 cancer associated bacteria in our subjects through unsupervised clustering (Fig.2E and Additional file 1: Table S4). Of note, we observed significant gender differences (p=0.01) among the C3 community type (Additional file 1: Fig. S4a). Tumor locations (colon or rectum; p=0.01) were linked to the C4 community type, which primarily consisting of the beneficial species (Additional file 1: Fig. S4b).

Previous studies indicated that gut microbes may induce DNA damage, thereby accelerating cancer development [29]. Consequently, we detected somatic mutations using exome sequencing technology from 41 CRC tumors and idntified 4 significantly mutated genes with MutSigCV, including TP53 (Q value=0), APC (Q value=1.26E-11), KRAS (Q value=1.11E-10) and SMAD4 (Q value=7.37E-04) (Fig.3A and Additional file 1: Table S6).

An overview of the associations between cancer genome and microbiome genomes. A Bar plots illustrate the frequently mutated genes in 41 tumor tissues. B The interaction between gut microbial taxa and somatic altered genes

To explore their associations with microbiota composition, we conducted the LEfSe analysis to compare tumors with or without mutated genes (Fig.3B). TP53 is the most prevalent somatic altered genes in our cohort. In TP53 mutated subjects, an enrichment of several disease-associated species, including Alistipes putredinis (LDA score=4.402), Porphyromonas asaccharolytica (LDA score=3.816), and Prevotella intermedia (LDA score=3.795) (Fig.4A). Previous observations uncovered that butyrate treatment could activate the TP53 pathway [30]. Consistently, the abundance of butyrate-producing bacteria, Butyricicoccus pullicaecorum, exhibited a significant reduction in TP53 mutation carriers (LDA score=2.395). Interestingly, Roseburia inulinivorans (LDA score=3.96) and Ruminococcus gnavus (LDA score=3.426), two other butyrate producers, were also significantly depleted in APC mutation carriers (Fig.4B). Besides, the relative abundance of Enterococcus genus was enriched in subjects with KRAS and SMAD4 mutations (Enterococcus faecalis, LDA=2.217; Enterococcus avium, LDA score=3.075) (Fig.4C, D). We also performed similar analysis between gut microbiota and other frequently mutated genes (Additional file 1: Fig. S5). In stool samples, probiotics, including Ruminococcus lactaris(LDA score=3.405), Bifidobacterium bifidum (LDA score=2.425), were dramatically elevated in MUC5B or MUC16 mutated individuals (Additional file 1: Fig. S5e, f). Barnesiella intestinihominis, acting as an enhancer for anticancer therapy, was proven enriched in TNN mutation carriers (LDA score=3.156) (Additional file 1: Fig. S5m).

AD Significantly mutated genes related taxonomic difference. Differentially abundant species between tumors with and without TP53 (A), APC (B), KRAS (C), SMAD4 (D) alterations, respectively

We further characterized the differences of microbial pathways between subjects with specific mutations and control group. Interestingly, the most abundant pathways were generally housekeeping processes encoded by microbes, such as one carbon metabolism, aromatic amino acids, branched chain amino acid and so on (Additional file 1: Fig. S6). One-carbon (1C) metabolism, consistently overexpressed in cancer, supports multiple biological processes, including nucleotides synthesis, methionine recycling pathway and redox defense [31]. An increased level of bacterial purine (reporter score=2.909) and pyrimidine (reporter score=3.188) metabolism were found in TP53 mutation carriers (Additional file 1: Fig. S6a). Similarly, bacterial cysteine-methionine metabolism (reporter score=3.246) and folate biosynthesis (reporter score=1.949) exhibited significant alterations in individuals with APC mutations (Additional file 1: Fig. S6b). Bacteria can synthesize different amino acids. Compared to control group, we found APC (Additional file 1: Fig. S6c) and SMAD4 mutation carriers (Additional file 1: Fig. S6d) were significantly associated with high levels of bacterial tryptophan (Trp) metabolism pathway (reporter score=3.045 and 2.732, respectively). Moreover, we observed an elevated abundance of bacterial phenylalanine metabolism correlated with KRAS mutations (reporter score=4.345) (Additional file 1: Fig. S6c).

We also investigated the relationship between the microbiome composition and the gene expression patterns in CRC. We observed that certain bacterial species were significantly correlated with the gene expression pattern (Additional file 1: Fig. S7 and Table S6). The differentially expressed functional genes were clustered according to their correlation with differentially abundant species, following by annotation with DAVID (Fig.5). We observed that Fusobacterium nucleatum, along with some Clostridium spp. exhibited positive associations nitrogen metabolism and bile secretion pathways, but negatively with cytokine-cytokine receptor interaction pathway.

Correlation of differentially abundant species and deregulated genes. Tumor associated deregulated genes were clustered and annotated with DAVID. The X axis illustrated the DAVID functional annotation and Y axis showed differentially abundant species. Red color represents positive association while green color means negative association

Subsequently, the interaction between 22 bacterial species and up-regulated oncogene expression was explored. As shown in Fig.6A, Fusobacterium nucleatum was positively correlated with PKM (p=0.03), SCD (p=0.0186), FASN (p=0.014), which are key enzymes in glycolysis and fatty acid metabolism. Consistent with the findings, we categorized patients into high and low Fusobacterium nucleatum groups, and found that various metabolism related pathways were significantly enriched in the high groups (pentose and glucuronate interconversions, p=0.026; starch and sucrose metabolism, p=0.007; porphyrin and chlorophyll metabolism, p=0.023; oxidative phosphorylation, p<0.00001) (Fig.6B). Taken together, the intestinal microbiota promotes CRC progression by shaping the expression of host gene expression, especially metabolic pathways.

Gene expression signature and metabolic pathways reprogramming associated with microbial shifts. A The association between up regulated oncogene expression and cancer related species. The X axis represents up regulated cancer genes. Significant associations were highlighted below the heatmap. B Pathway difference between high and low Fusobacterium nucleatum groups

The composition of immune and stromal cell types was identified by XCELL, a gene signature-based method that integrates the advantages of gene set enrichment with deconvolution approaches. Compared with adjacent normal tissues, the overall immune score was significantly lower in tumor tissue (Fig.7A). Especially, the abundance of most B cells and CD8+T cells elevated in tumors while regulatory T cells and T helper cells exhibited a decreasing trend (Additional file 1: Fig. S8), indicating the important role of the immune microenvironment in the progression of CRC. The associations between different microbial species and immune cell types in the CRC were shown in Fig.7B. Fusobacterium nucleatum was negatively associated with dendritic cells and CD8 T cells (Fig.7C). While Faecalibatcerium prausnitzii were significantly positively correlated with dendritic cells and Macrophages M1 (Additional file 1: Fig. S9a).

Fusobacterium nucleatum promoted CRC by modifying the tumor immune environment and TNFSF9 expression. A Comparison of immune cell scores between tumor and adjacent normal tissues. B The heatmap illustrates the correlations between differential abundant species and immune cells. The stars indicate the level of statistical significance. C Significant association of F. nucleatum and aDC and CD8 T cells. D Pathway alteration between normal and tumor tissue. E Significant association between Fusobacterium nucleatum and TNFSF9 gene expression

Interestingly, Gene set enrichment analysis of revealed that the cytokine-cytokine receptor interaction (p<0.001) was significantly altered in CRC (Fig.7D). Correlation analysis of genes related to cytokine-cytokine receptor interaction pathway related genes and 22 species uncovered several significant associations (Additional file 1: Fig. S9b). Among them, Fusobacterium nucleatum exhibited a positive association with TNFSF9, a member of TNF (tumor necrosis factor) family members (r=0.443, p=0.0037) (Fig.7E). Previous study showed that Fusobacterium nucleatum autoinducer-2 (AI-2) enhanced the mobility and M1 polarization of macrophages, possibly through TNFSF9/TRAF1/p-AKT/IL-1 signaling. Our results further suggested that pathogenic bacteria, like Fusobacterium nucleatum, may interact with CRC cells and modify the tumor immune environment by TNFSF9, finally facilitating the tumor development.

See the rest here:
Multi-omic profiling reveals associations between the gut microbiome, host genome and transcriptome in patients with ... - Journal of Translational...