Abstract
Genome-wide association studies (GWAS) have discovered 27 loci associated with glioma risk. Whether these loci are causally implicated in glioma risk, and how risk differs across tissues, has yet to be systematically explored. We integrated multi-tissue expression quantitative trait loci (eQTLs) and glioma GWAS data using a combined Mendelian randomisation (MR) and colocalisation approach. We investigated how genetically predicted gene expression affects risk across tissue type (brain, estimated effective n = 1194 and whole blood, n = 31,684) and glioma subtype (all glioma (7400 cases, 8257 controls) glioblastoma (GBM, 3112 cases) and non-GBM gliomas (2411 cases)). We also leveraged tissue-specific eQTLs collected from 13 brain tissues (n = 114 to 209). The MR and colocalisation results suggested that genetically predicted increased gene expression of 12 genes were associated with glioma, GBM and/or non-GBM risk, three of which are novel glioma susceptibility genes (RETREG2/FAM134A, FAM178B and MVB12B/FAM125B). The effect of gene expression appears to be relatively consistent across glioma subtype diagnoses. Examining how risk differed across 13 brain tissues highlighted five candidate tissues (cerebellum, cortex, and the putamen, nucleus accumbens and caudate basal ganglia) and four previously implicated genes (JAK1, STMN3, PICK1 and EGFR). These analyses identified robust causal evidence for 12 genes and glioma risk, three of which are novel. The correlation of MR estimates in brain and blood are consistently low which suggested that tissue specificity needs to be carefully considered for glioma. Our results have implicated genes yet to be associated with glioma susceptibility and provided insight into putatively causal pathways for glioma risk.
Highlights
Genome-wide association studies (GWAS) have discovered 27 loci associated with glioma risk
Use of a combined Mendelian randomisation (MR) and colocalisation pipeline can strengthen causal inferences by discounting MR results which arise due to confounding through linkage disequilibrium (LD)—which can arise when another genetic variant in high LD with the variant of interest is associated with the outcome conditional on the exposure and which MR struggles to differentiate[17,18,19]
Our MR analysis of brain and blood eQTLs identified 34 associations that met at least the suggestive P value threshold (P < 9.49 × 10–5) for 17 genes associated with risk of glioma, GBM or non-GBM (Fig. 2)
Summary
Genome-wide association studies (GWAS) have discovered 27 loci associated with glioma risk. We investigated how genetically predicted gene expression affects risk across tissue type (brain, estimated effective n = 1194 and whole blood, n = 31,684) and glioma subtype (all glioma (7400 cases, 8257 controls) glioblastoma (GBM, 3112 cases) and non-GBM gliomas (2411 cases)). Examining how risk differed across 13 brain tissues highlighted five candidate tissues (cerebellum, cortex, and the putamen, nucleus accumbens and caudate basal ganglia) and four previously implicated genes (JAK1, STMN3, PICK1 and EGFR) These analyses identified robust causal evidence for 12 genes and glioma risk, three of which are novel. Whilst including genetic factors into the classification criteria has seen measurable benefits for patients, functional studies have been limited and it is not known if certain mutations are merely correlated with gliomagenesis and subtype differentiation or play a causal role in risk. These data are linked to eQTLs derived from blood to determine how the risk profile for glioma differs between brain tissue and whole blood
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
