Polygenic risk score for ACE-inhibitor-associated cough based on the discovery of new genetic loci

Abstract

Aims: To search for sequence variants associated with ACEi discontinuation and to test their association with ACEi-associated adverse drug reactions (ADRs).

Methods and results: A genome-wide association study (GWAS) on ACEi discontinuation was conducted, including 33 959 ACEi-discontinuers and 44 041 controls. Cases were defined as persons who switched from an ACEi treatment to an angiotensin receptor blocker. Controls were defined as persons who continued ACEi treatment for at least 1 year. Odds ratios (ORs) and 95% confidence intervals (95% CIs) were computed for ACEi discontinuation risk by mixed model regression analysis. Summary statistics from the individual cohorts were meta-analyzed with a fixed-effects model. To test for association with specific ACEi-associated ADRs, any genome-wide significant (P < 5 × 10-8) ACEi discontinuation variants was tested for association with ACEi-associated cough and angioedema. A polygenetic risk score (PRS) based on ACEi discontinuation GWAS data was constructed and tested for association with ACEi-associated cough and angioedema in two population-based samples. In total, seven genetic genome-wide loci were identified, of which six were previously unreported. The strongest association with ACEi discontinuation was at 20q13.3 (NTSR1; OR: 1.21; 95% CI: 1.17-1.24; P = 2.1 × 10-34). Five of seven lead variants were associated with ACEi-associated cough, whereas none were associated with ACEi-associated angioedema. The ACEi discontinuation PRS was associated with ACEi-associated cough in a dose-response manner but not with ACEi-associated angioedema. ACEi discontinuation was genetically correlated with important causes for cough, including gastro-esophageal reflux disease, allergic rhinitis, hay fever, and asthma, which indicates partly shared genetic underpinning between these traits.

Conclusion: This study showed the advantage of using prescription patterns to discover genetic links with ADRs. In total, seven genetic loci that associated with ACEi discontinuation were identified. There was evidence of a strong association between our ADR phenotype and ACEi-associated cough. Taken together, these findings increase insight into the pathophysiological processes that underlie ACEi-associated ADRs.

Keywords: ACE inhibitors; ACE-inhibitor associated cough; ADR; Adverse drug reaction; Drug discontinuation; GWAS; Genome-wide association study.

Structured Graphical Abstract

A genome-wide association study on ACEi discontinuation, using prescription data from three population-based cohorts. We identified seven genetic loci that associated with ACEi discontinuation, of which six were novel. Using a polygenic risk score approach, we found a dose–response relationship between higher score and risk of ACEi-associated cough but not with ACEi-associated angioedema.

Figure 1

Single nucleotide polymorphism (SNP)-based results from the genome-wide association analysis (GWAS) on ACE-inhibitor discontinuation in 78 000 individuals. (A). Manhattan plot of the results from the GWAS meta-analysis of ACE-inhibitor discontinuation. Lead SNPs in the seven genome-wide significant loci are highlighted with a diamond. Independent genome-wide significant variants are annotated to the nearest gene(s). The y axis represents –log10 of the two-sided P-values for association of variants with ACE-inhibitor discontinuation, from meta-analysis using an inverse variance-weighted fixed-effects model and a total sample size of 78 000. The x-axis represents the genome in physical order. The dashed line represents the threshold for genome-wide significance. (B). Distribution of CADD scores, RegulomeDB categories, chromatin state and functional consequences of all annotated SNPs (ANNOVAR) in linkage disequilibrium of r² ≥ 0.8 with the lead SNPs. CADD scores predict how deleterious the effect of a SNP is likely to be for protein structure/function, with higher scores referring to higher deleteriousness. A CADD score above 12.37 is considered potentially pathogenic. RegulomeDB categories annotate SNPs in genomic risk loci, with a low score indicating a higher likelihood of a SNP having a regulatory function. Chromatin state corresponds to the minimum chromatin state across 127 tissue and cell types for SNPs in the genomic risk loci, with lower states indicating higher accessibility and states 1–7 referring to open chromatin states. ANNOVAR categories identify the variant genic position (e.g. intronic, exonic, intergenic) and associated function.

Figure 2

Genetic correlations of ACE-inhibitor discontinuation with selected neuropsychiatric, cardiometabolic and pulmonary traits. Genetic correlations and two-sided P-values were calculated using linkage disequilibrium score regression. Genetic correlation is presented as a dot and error bars indicate 95% confidence limits. The threshold for significance was set at P < 0.003 (0.05/19 tested traits).

Figure 3

This plot displays the associations between the identified ACE-inhibitor discontinuation loci and ACE-inhibitor associated adverse drug reactions (cough and angioedema), using summary estimates from published GWAS summary data. The points represent the odds ratios (ORs) and lines represent 95% confidence intervals (95% CI). Number in bracket represents number of cases used to derive the GWAS summary statistics. We accounted for multiple testing by Bonferroni correction for the two traits tested per ACE-inhibitor discontinuation locus (P < 0.004). The ACE-inhibitor discontinuation lead SNP associations are displayed for comparison. Where a sentinel variant was not available for all traits, a common proxy was selected (r²>0.7). Each lead (or proxy) variant is annotated to the nearest gene(s).

Figure 4

Odds ratio (OR) for ACEi-associated cough and angioedema by quintiles of ACEi discontinuation polygenetic risk score (PRS). Triangles correspond to OR by PRS quintile estimated for 6008 ACEi-treated individuals, of which 1346 constitute individuals with ACEi-associated cough sampled from the eMERGE consortium. Dots represent the OR by PRS quintile for 24 595 ACEi-treated individuals, of which 201 had a history of ACEi-associated angioedema, sampled from the Copenhagen Hospital Biobank. ORs and 95% confidence limits (error bars) were estimated using logistic regression, adjusted for sex, age and principal components. Diamond represents the reference category (1^st quintile).