Genomic analyses identify 15 susceptibility loci and reveal HDAC2, SOX2-OT, and IGF2BP2 in a naturally-occurring canine model of gastric cancer

Abstract

Gastric cancer (GC) is the fifth most common human cancer worldwide, but the genetic etiology is largely unknown. We performed a Bayesian genome-wide association study and selection analyses in a naturally-occurring canine model of GC, the Belgian Tervuren and Sheepdog breeds, to elucidate underlying genetic risk factors. We identified 15 loci with over 90% predictive accuracy for the GC phenotype. Variant filtering revealed germline putative regulatory variants for the EPAS1 (HIF2A) and PTEN genes and a coding variant in CD101. Although closely related to Tervuren and Sheepdogs, Belgian Malinois rarely develop GC. Across-breed analyses uncovered protective haplotypes under selection in Malinois at SOX2-OT and IGF2BP2. Among Tervuren and Sheepdogs, HDAC2 putative regulatory variants were present at comparatively high frequency and were associated with GC. Here, we describe a complex genetic architecture governing GC in a dog model, including genes such as PDZRN3, that have not been associated with human GC.

Keywords: Bayesian GWAS; Belgian Sheepdog; Groenendael; Malinois; PDZRN3; Tervuren; adenocarcinoma; canine; dog; selection.

Figure 1.. Population structure of the case-control Tervuren and Sheepdog cohort.

A) Box-and-whisker plots of age at diagnosis for cases and age at collection or most recent health update for controls. B) Principal component analysis (PCA) plot for the 470 dogs included in the case vs. control analyses with each point representing an individual.

Figure 2.. Bayesian GWAS and selection scan analysis identify multiple loci associated with gastric cancer.

A) Manhattan plot showing the results of Bayesian GWAS with 200 cases vs. 270 controls. The absolute value of individual variant effect sizes (y-axis) reported by BayesR for 576,754 variants after minor allele frequency and LD pruning is plotted by position (x-axis). The red line is at the moderate effect size (0.001) threshold. B) Top: A regional Manhattan plot of CFA26 shows three independent regions surpassing moderate effect. Pairwise linkage disequilibrium (r²) between the top BayesR variant at CFA26:38Mb (purple diamond) and all other variants is color-coded. Bottom: UCSC image of the human (hg38) syntenic region for the CFA26:38Mb XP-nSL interval that overlaps the GWAS signal. Candidate variants meeting filtering criteria are indicated as blue vertical bars in the top track. GeneHancer double elite v5.21 enhancer (gray bars) and promoter (red bars) regulatory elements are below, with darkest colors indicating highest confidence. Clustered interactions of regulatory elements with protein coding genes and a lncRNA are shown; thick bars correspond to the regulatory element that interacts with the gene. The NCBI RefSeq Select and MANE gene track is at the bottom. C) Normalized XP-nSL values are plotted for 2,039,330 variants with dashed black lines demarcating the 99.99th percentiles. Positive values represent regions with extended haplotype homozygosity among cases, whereas negative values indicate extended haplotype homozygosity among controls.

Figure 3.. Fifteen loci predict gastric cancer phenotype with greater than 90% discriminatory accuracy.

A) Bar graph of the distribution of risk alleles among cases (n=200, red), controls (n=270, gray), and Malinois (n=149, blue). Cases have an average of 20.3 total risk alleles while controls have 15.9 (p=1.31×10⁻⁶¹), and Malinois cluster with controls. B) Receiver Operating Characteristics (ROC) curves for the 5-fold cross-validated 15-locus additive model of GC risk in Tervuren and Sheepdog. Area Under the Curve (AUC) is reported for each fold, with the 95% confidence interval in brackets.

Figure 4.. Belgian shepherd across-breed comparisons identify GC risk and protective loci under selection.

A) Normalized XP-nSL values are plotted for 2,039,330 variants with dashed black lines demarcating the 99.99th percentiles. Positive values indicate regions under selection in an independent Tervuren (n=69) and Sheepdog (n=82) cohort with unknown GC phenotypes, whereas negative values represent regions under selection in Malinois (n=149). B) Left: UCSC image of the human (hg38) syntenic CFA12 region under selection in Tervuren and Sheepdogs. Two variants significantly associated with GC are denoted, as well as all GeneHancer double elite regulatory elements. Right: zooming in to the human syntenic location of the two identified variants shows overlap with a GeneHancer double elite element and ENCODE candidate cis-regulatory elements. The second variant falls within an untranslated region of HDAC2-AS2. Below, the ReMap ChIP-seq density track displays evidence of transcriptional regulators, and the ENCODE H3K27ac track shows enrichment of the H3K27ac histone modification, indicating areas of active enhancers.