Multi-tissue expression and splicing data prioritise anatomical subsite- and sex-specific colorectal cancer susceptibility genes

Abstract

Genome-wide association studies have suggested numerous colorectal cancer (CRC) susceptibility genes, but their causality and therapeutic potential remain unclear. To prioritise causal associations between gene expression/splicing and CRC risk (52,775 cases; 45,940 controls), we perform a transcriptome-wide association study (TWAS) across six tissues with Mendelian randomisation and colocalisation, integrating sex- and anatomical subsite-specific analyses. Here we reveal 37 genes with robust causal links to CRC risk, ten of which have not previously been reported by TWAS. Most likely causal genes with evidence of cancer cell dependency show elevated expression linked to risk, suggesting therapeutic potential. Notably, SEMA4D, encoding a protein targeted by an investigational CRC therapy, emerges as a key risk gene. We also identify a female-specific association with CRC risk for CCM2 expression and subsite-specific associations, including LAMC1 with rectal cancer risk. These findings offer valuable insights into CRC molecular mechanisms and support promising therapeutic avenues.

Fig. 1. Overview of multi-tissue TWAS, colocalization, and MR-based gene prioritisation for colorectal cancer risk.

a Flowchart showing analysis overview and number of genes/splicing events identified at each stage. “Genes with robust evidence” includes those that had H₄ above 0.8 in colocalisation analyses, and which either passed Bonferroni correction in the relevant MR analysis (p < 4.38 × 10⁻⁵; 0.05/N*G where N is the number of gene-tissue pairs (161) and G is the number of CRC GWAS (7) for genes identified in TWAS analyses or p < 1.32 × 10⁻⁴; 0.05/number of druggable genes with suitable genetic instruments available (380) for genes identified as part of the druggable genome) or which did not have suitable instruments available to be included in the MR analysis. MR Mendelian randomisation. b Manhattan plot showing results of S-MultiXcan and JTI TWAS analyses of colorectal cancer risk, for all anatomical subsites combined. Where genes were identified in multiple TWAS analyses, the one with the lowest p value was retained. Genes labelled are those prioritised following subsequent analyses. All statistical tests were two-sided with the unadjusted p values from S-MultiXcan or JTI plotted. c Venn diagram showing overlap of final prioritised 37 genes identified by each TWAS analysis. JTI joint tissue imputation, eQTLs expression quantitative trait loci, sQTLs splice quantitative trait loci. Source data are provided as a Source Data file.

Fig. 2. Forest plots of JTI effect sizes across colorectal cancer anatomical subsites and sex for anatomical subsite- and sex-specific genes identified by JTI TWAS analysis.

Relevant tissue-specific estimates from JTI for risk of each anatomical subsite are plotted with 95% confident intervals. (sample sizes were 52,775 cases, 45,940 controls for overall; for all anatomical subsites there were 43,099 controls; colon, 28,736 cases; proximal colon, 14,416 cases; distal colon, 12,879 cases; and rectal, 14,150 cases; female, 24,594 cases, 23,936 controls; male, 28,271 cases, 22,351 controls). Solid points indicate the Bonferroni p value threshold of p < 6.01 × 10⁻⁸ was met in the JTI analysis. Errors bars may be hidden by the point estimate where the standard deviation is small relative to effect estimates. A AAMP expression in adipose subcutaneous tissue; B AAMP expression in adipose visceral tissue; C AAMP expression in colon sigmoid tissue; D COLCA1 expression in colon transverse tissue; E EPM2AIP1 expression in adipose visceral tissue; F EPM2AIP1 expression in whole blood; G LAMC1 expression in whole blood; H MLH1 expression in adipose subcutaneous tissue; I MLH1 expression in adipose visceral tissue; J MLH1 expression in whole blood; K MLH1 expression in lymphocytes; L RP11-129K12.1 expression in adipose subcutaneous expression; M RP11-129K12.1 expression in colon transverse tissue; N CCM2 expression in whole blood. Source data are provided as a Source Data file.

Fig. 3. Forest plots of mean Z-score estimates from S-MultiXcan across colorectal cancer anatomical subsites for anatomical subsite-specific genes identified by S-MultiXcan (expression or splicing) TWAS analysis.

Relevant estimates for risk of each anatomical subsite are plotted with 95% confident intervals (sample sizes were 52,775 cases, 45,940 controls for overall; for all anatomical subsites there were 43,099 controls; colon, 28,736 cases; proximal colon, 14,416 cases; distal colon, 12,879 cases; and rectal, 14,150 cases; female, 24,594 cases, 23,936 controls; male, 28,271 cases, 22,351 controls). Solid points indicate the Bonferroni p value threshold of p < 3.91 × 10⁻⁷ was met in S-MultiXcan eQTL analysis or p < 5.49 × 10⁻⁷ was met in S-MultiXcan sQTL analysis. Errors bars may be hidden by the point estimate where the standard deviation is small relative to Z-score scale. A ABCC2 expression; B MLH1 expression; C RP11-129K12.1 expression; D ARPC2 splicing; E GPATCH1 splicing. Source data are provided as a Source Data file.