Multi-omics Analysis Reveals Adipose-tumor Crosstalk in Patients with Colorectal Cancer

Abstract

Obesity and obesity-driven cancer rates are continuing to rise worldwide. We hypothesize that adipocyte-colonocyte interactions are a key driver of obesity-associated cancers. To understand the clinical relevance of visceral adipose tissue in advancing tumor growth, we analyzed paired tumor-adjacent visceral adipose, normal mucosa, and colorectal tumor tissues as well as presurgery blood samples from patients with sporadic colorectal cancer. We report that high peroxisome proliferator-activated receptor gamma (PPARG) visceral adipose tissue expression is associated with glycoprotein VI (GPVI) signaling-the major signaling receptor for collagen-as well as fibrosis and adipogenesis pathway signaling in colorectal tumors. These associations were supported by correlations between PPARG visceral adipose tissue expression and circulating levels of plasma 4-hydroxyproline and serum intercellular adhesion molecule 1 (ICAM1), as well as gene set enrichment analysis and joint gene-metabolite pathway results integration that yielded significant enrichment of genes defining epithelial-to-mesenchymal transition-as in fibrosis and metastasis-and genes involved in glycolytic metabolism, confirmed this association. We also reveal that elevated prostaglandin-endoperoxide synthase 2 (PTGS2) colorectal tumor expression is associated with a fibrotic signature in adipose-tumor crosstalk via GPVI signaling and dendritic cell maturation in visceral adipose tissue. Systemic metabolite and biomarker profiling confirmed that high PTGS2 expression in colorectal tumors is significantly associated with higher concentrations of serum amyloid A and glycine, and lower concentrations of sphingomyelin, in patients with colorectal cancer. This multi-omics study suggests that adipose-tumor crosstalk in patients with colorectal cancer is a critical microenvironment interaction that could be therapeutically targeted.See related spotlight by Colacino et al., p. 803.

Figure 1.. Schema of tissue transcriptomic, plasma metabolomic, and serum-based biomarker analyses by (A) PPARG VAT expression group and (B) PTGS2 colorectal tumor tissue expression group, for sporadic colorectal cancer patients.

*Regression model coefficients and P-values were estimated for tumor transcripts after adjustment for: patient age, sex, tumor site, and stage.

Figure 2.. Elevated PPARG VAT expression is associated with tissue fibrosis among sporadic colorectal cancer patients.

(A) Serum-based inflammatory and angiogenesis biomarker levels and (B) plasma metabolic concentrations associated with differential PPARG VAT expression. Differential expression of systemic biomarker levels between high versus low PPARG VAT expression patient groups was computed using Wald tests and ranked by fold change. Regression model coefficients and P-values were estimated for serum-based biomarkers after adjustment for: patient age, sex, tumor site, and stage. (C) Volcano plot of gene expression changes in colorectal tumors associated with differential PPARG VAT expression. Differential expression of tumor transcripts between high versus low PPARG VAT expression groups of patients was computed using Wald tests. Transcripts with absolute value fold change (2^β)>1.2 and α<0.05 (dashed lines) were considered to be statistically significant. (D) Ingenuity Pathway Analysis (IPA) results of molecular pathways statistically significantly enriched in the n=249 genes differentially expressed in colorectal tumor tissue by PPARG VAT expression (high versus low). (E) Overrepresentation analysis graphical results of joint gene-metabolite pathways for the significant gene (n=249 colorectal tumor genes from transcriptomic analysis) and metabolite (n=4 metabolites from plasma metabolomics analysis) set. Hypergeometric testing conducted using MetaboAnalyst 4.0 software. Node color shading based on P-value and node radius defined by impact values from pathway topology analysis.

Figure 3.. High PTGS2 expression in colorectal tumor tissue is linked to a fibrotic signature in adipose-tumor crosstalk among sporadic colorectal cancer patients.

(A) Plasma metabolic concentrations and (B) serum-based biomarker levels associated with differential PTGS2 colorectal tumor expression. Differential expression of plasma metabolites and biomarker levels between high versus low PTGS2 tumor expression patient groups was computed using Wald tests and ranked by fold change. Regression model coefficients and P-values were estimated for plasma metabolites and serum-based inflammatory and angiogenesis biomarkers after adjustment for: patient age, sex, tumor site, and stage. Dashed line indicates significance threshold with P-value of 0.05. (C) Volcano plot of VAT gene expression changes associated with differential PTGS2 tumor expression. Differential expression of VAT transcripts between high versus low PTGS2 tumor expression patient groups was computed using Wald tests. Transcripts with absolute value fold change>1.2 and α<0.05 (dashed lines) were considered statistically significant. (D) Ingenuity Pathway Analysis (IPA) results of molecular pathways significantly enriched in the n=129 genes differentially expressed in VAT by PTGS2 tumor expression (high versus low). Dashed line indicates P-value threshold of 0.05.