key: cord-0319835-9vuek3om authors: Dube, M.-P.; Chazara, O.; Lemacon, A.; Asselin, G.; Provost, S.; Barhdadi, A.; Lemieux Perreault, L.-P.; Mongrain, I.; Wang, Q.; Carss, K.; Paul, D. S.; Cunningham, J. W.; Rouleau, J.; Solomon, S. D.; McMurray, J. J. V.; Yusuf, S.; Granger, C. B.; Haefliger, C.; de Denus, S.; Tardif, J.-C. title: Pharmacogenomic study of heart failure and candesartan response from the CHARM programme date: 2021-09-29 journal: nan DOI: 10.1101/2021.09.28.21263908 sha: c16543e79fa02f73bc0308c198db2edf6101a9cf doc_id: 319835 cord_uid: 9vuek3om Aims. The Candesartan in Heart failure Assessment of Reduction in Mortality and morbidity (CHARM) programme consisted of three parallel, randomised, double-blind clinical trials comparing candesartan with placebo in patients with heart failure (HF) categorised according to left ventricular ejection fraction and tolerability to an ACE inhibitor. We conducted a pharmacogenomic study of the CHARM studies to identify genetic predictors of heart failure progression and the efficacy and safety of treatment with candesartan. Methods. We performed genome-wide association studies (GWAS) with the composite endpoint of cardiovascular death or hospitalisation for heart failure in 2727 patients from CHARM-Overall and stratified by CHARM study according to preserved and reduced ejection fraction. The safety endpoints were hyperkalaemia, renal dysfunction, hypotension, and change in systolic blood pressure. We also conducted a genome-wide gene-level collapsing analysis from whole-exome sequencing data with the composite cardiovascular endpoint. Results. We found the genetic variant rs66886237 at 8p21.3 near the gene GFRA2 to be associated with the composite cardiovascular endpoint in 1029 HF patients with preserved ejection fraction from the CHARM-Preserved study [hazard ratio (HR): 1.91, 95% confidence interval (CI): 1.55-2.35; P=1.7x10-9], but not in patients with reduced ejection fraction. None of the GWAS for candesartan safety or efficacy passed the significance threshold. Conclusions. We have identified a candidate genetic variant potentially predictive of the progression of heart failure in patients with preserved ejection fraction. The findings require further replication and we cannot exclude the possibility that the results may be chance findings. Candesartan is an angiotensin II receptor blocker (ARB) that is widely used alone or in combination with other agents as therapy for hypertension and heart failure (HF). Multiple mechanistic studies alluded to the potential benefits of inhibition of the renin-angiotensinaldosterone system (RAAS) with ARBs in HF, [1] [2] [3] which were subsequently confirmed in large clinical trials, although this benefit varied depending on the population studied and the concomitant medication used. [4] [5] [6] [7] Candesartan is a selective AT1 subtype angiotensin II receptor antagonist, which is orally administered as candesartan cilexetil, a prodrug, which undergoes hydrolysis to candesartan during absorption from the gastrointestinal tract. Candesartan is not significantly metabolised by the cytochrome P450 system and at therapeutic concentrations has no effects on P450 enzymes. Candesartan was shown to be beneficial in patients with heart failure in the Candesartan in Heart failure Assessment of Reduction in Mortality and morbidity (CHARM) programme designed as three parallel, independent, integrated, randomised, double-blind clinical trials comparing candesartan with placebo in three distinct populations of patients with NYHA class II-IV HF based on participants' assessment of left ventricular ejection fraction (LVEF) and history of tolerability to an ACE inhibitor. 8 The primary endpoint of each trial was testing whether the use of candesartan would reduce the risk of cardiovascular death or hospital admission for HF. In CHARM-Alternative and CHARM-Added, the primary endpoint was significantly reduced with candesartan as compared to placebo (HR 0.77, 95% CI 0.67-0.89, p= 0.0004; and HR 0.85, 95% CI 0.75-0.96, p= 0.011, respectively). In CHARM-Preserved, however, the primary composite endpoint did not reach significance (HR= 0.89, 95% CI 0.77-1.03. P= 0.118). CHARM-Overall . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10.1101/2021.09.28.21263908 doi: medRxiv preprint 5 showed a highly significant reduction in the combined incidence of cardiovascular death and hospitalisation for HF (HR= 0.84, 95% CI 0.77-0.91, p<0.0001). Here, we present a post-hoc pharmacogenomic study of the CHARM programme in a subgroup of the original study participants, with the aim of identifying genetic predictors of the progression of HF and for the efficacy and safety of treatment with candesartan. The CHARM programme was designed as three parallel, independent, integrated, randomised, double-blind clinical trials comparing candesartan with placebo in three distinct populations of patients with NYHA class II-IV HF based on participants' assessment of left ventricular ejection fraction (LVEF) and history of tolerability to an ACE inhibitor. 8 CHARM-Alternative included patients with depressed LV function (EF ≤ 40%) and who were not treated with ACE inhibitors due to intolerance, 4 CHARM-Added included 2548 patients with depressed LV function (EF ≤ 40%) and treated with ACE inhibitors, 5 and CHARM-Preserved included 3025 patients with preserved LV function (EF > 40%) with or without ACE inhibitors. 6 CHARM-Overall was the joint analysis of all 3 studies combined. 9 The CHARM studies involved 26 countries and 618 sites. The first patient was randomised on March 22 1999 and the last patient completed the study on March 31 2003 , with a median follow-up time of 37.8 months. The active treatment group received candesartan once daily, at a starting dose of 4 mg or 8 mg once daily, which was up-titrated by doubling the dose at 2-week intervals to a maximum of 32 mg once daily or the highest tolerated level. Main exclusion criteria were serum creatinine of 265 μ mol/L or more, serum potassium of 5.5 mmol/L or more, bilateral renal artery stenosis, symptomatic . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10.1101/2021.09.28.21263908 doi: medRxiv preprint 6 hypotension, critical aortic or mitral stenosis, myocardial infarction, stroke, or open-heart surgery in the previous 4 weeks, use of an angiotensin-receptor blocker in the previous 2 weeks. Other exclusion criteria have been previously described. 8 In this pharmacogenetic sub-study, there were 2727 participants in CHARM-Overall, which included 755 participants from CHARM-Alternative, 943 from CHARM-Added, and 1029 from CHARM-Preserved (Figure 1 ). The CV efficacy endpoint for the present pharmacogenomic study was defined identically to the primary endpoint of the individual CHARM studies and consists of a composite of CV death or hospitalisation for the management of chronic HF, whichever occurred first. Potential study endpoints were adjudicated by an independent clinical endpoint committee in the CHARM studies. Event-free patients who completed the study were censored at the date of study completion, and those who did not complete the study were censored at the date of last contact. Patients who died from a non-CV cause were censored at the time of death. The pharmacogenomic safety endpoints considered were hyperkalaemia, renal dysfunction, hypotension, and change in systolic blood pressure. Adverse events were recorded and encoded centrally. Systolic blood pressure was obtained during the physical examination at the scheduled 6-week visit and reported by investigators. Patients were followed up until study completion or date of last contact whichever occurred first. Genome-wide genotyping was performed using 200 ng of genomic DNA extracted from whole blood at the Beaulieu-Saucier Pharmacogenomics Centre (Montreal, Canada). The Illumina . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint Figure 1) . Principal components were generated on the study samples only, and the scree plot and the cumulative explained variance were used to select the principal components to control for confounding due to population structure. 12 Genome-wide imputation was performed using IMPUTE2 (v2.3.2) 13 and phasing was performed using the SHAPEIT2 algorithm (v.2r790). 14 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10.1101/2021.09.28.21263908 doi: medRxiv preprint atrial fibrillation, and prior MI, diabetes to assess possible mediation pathways. All analyses are considered exploratory and hypothesis-generating. Overall, we conducted 18 primary analysis GWAS, which were supported by exploratory and sensitivity analysis GWAS (Supplementary Table 2 ). Each GWAS is assessed at a significance level of 2.8×10 -9 to adjust for the multiple testing of genetic variants and for the 18 primary analyses. Results are reported with point estimates and 95% confidence intervals (CI) which are not adjusted for multiple comparisons. Statistical analyses (except for GWAS and functional annotation analyses) were conducted using SAS versions 9.3 and 9.4, and the top findings from the GWAS were reproduced using SAS. The proportionality of hazards assumption was confirmed for the leading variants identified by GWAS using the Supremum test based on martingale residuals. Genomic DNA was also exome-sequenced at the Columbia Institute for Genomic Medicine. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint Rare variants were analysed with collapsing analyses as previously described, 15 for variants with a minimum read depth of 10, located within the CCDS r22 or the 2 bp canonical splice sites within the introns, and for variants that passed quality score thresholds as described in 16 . Eleven models defining sets of qualifying variants (QV) were used on the basis of the variants' predicted functional consequences and allele frequency (Supplementary Table 6 ). 15 We defined credible candidate variants as those located within 500 kb of the leading variants and with P values within two orders of magnitude of the lead variant. We used the software GCTA-CoJo 18 to conduct a conditional analysis to identify independent signals. We used PAINTOR 19 to identify credible sets of causal variants based on the magnitude and direction of the association . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10.1101/2021.09.28.21263908 doi: medRxiv preprint and the pairwise linkage disequilibrium structure at the loci, and we used RegulomeDB 20 and DSNetwork 21 to assign a relative ranking to variants. We used in silico functional annotations from the public databases Open Target Genetics 22 , ExPheWAS 23 and PhenoScanner v.2 24, 25 to identify potential functional mechanisms and target genes. We tested the colocalization between the CHARM GWAS signals and clinically relevant phenotypes using the COLOC R package v3.2-1 26 (refer to Supplementary Material). The data underlying this article cannot be shared publicly to preserve the privacy of study participants. The analytic methods and study materials may be made available to other researchers for purposes of reproducing the results or replicating the procedure. Summary The clinical trials upon which this post-hoc study is based had been registered on clinicaltrials.gov as NCT00634309, NCT00634712, and NCT00634400. There were 2727 participants available for the pharmacogenomic analysis of CHARM ( Figure 1 ). The baseline characteristics of patients according to the study treatment groups are shown in We conducted 3 GWAS to identify genetic variants associated with the composite of CV death or hospitalisation for HF as an indicator of disease progression using 1) patients from the CHARM-Overall programme, 2) the subset of patients with reduced ejection fraction from the CHARM-Alternative and CHARM-Added studies, and 3) the subset of patients with preserved ejection fraction from the CHARM-preserved study. We found a significant association signal in the GWAS of HF patients with preserved ejection fraction (Figure 2 Table 3 ). The effect of rs66886237 was not modulated by sex or by treatment with candesartan. Sensitivity analyses were conducted to assess the association of rs66886237 with the composite of CV death or hospitalisation for HF in CHARM-Preserved with further adjustment (individually) for BMI, left ventricular ejection fraction, history of atrial fibrillation, and prior MI, diabetes, and in all of the tested models, the genetic variant remained significant at P<2.8×10 -9 . Baseline characteristics and study outcomes are . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10.1101/2021.09.28.21263908 doi: medRxiv preprint 14 presented according to rs66886237 genotypes in Supplementary Table 3 . The cumulative incidence of the composite of CV death or hospitalisation for heart failure is presented in Figure 3 stratified according to rs66886237 genotypes. The minor allele (A) had a frequency of 14% in the study population. The variant is located in an intron of the GFRA2 gene encoding the glial cell line-derived neurotrophic factor (GDNF) family receptor alpha 2. We conducted 3 GWAS to discover genetic variants predictive of candesartan efficacity with the composite CV endpoint using the candesartan arm of the CHARM-Overall programme, the combined CHARM-Alternative and CHARM-Added studies, and the CHARM-Preserved study. No results passed the multiple-testing corrected GWAS significance threshold (P<2.8×10 -9 ). One region on chromosome 3 at 3q13.13 passed the unadjusted threshold of P<5×10 -8 , with variant rs664669 leading the signal in the CHARM-Overall programme in 1371 participants randomised to candesartan (Supplementary Figure 2, Supplementary Table 4 ). The minor allele (C) had a frequency of 44% and was associated with the composite endpoint of CV death or hospitalisation for HF in the candesartan group (HR= 1.48, 95% CI 1.29-1.69, P= 2.63×10 -8 ) with no effect in the placebo group (HR= 1.11, 95% CI 0.97-1.26; P= 0.125) and the interaction term between rs664669 and candesartan treatment was significant (P= 0.003) (Supplementary Table 4) . Variant rs664669 is intronic to the long interspersed non-coding RNA RP11-457K10.1, a processed pseudo gene of 763 base pairs with a transcript of 262 base pairs. Exploratory GWAS testing for the interaction effect between genome-wide variants and treatment also did not find significant signals (Supplementary Online Material) . . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10.1101/2021.09.28.21263908 doi: medRxiv preprint We conducted 15 GWAS to identify genetic variants predictive of hyperkalaemia, renal dysfunction, hypotension, and change in systolic blood pressure in candesartan-treated patients. There were 1371 participants randomised to candesartan in all three CHARM studies who were included in the genetic analyses of the GWAS with safety endpoints, of those, 89 (6.5 %) patients had a report of hyperkalaemia adverse events, 240 (17.5 %) with renal dysfunction, and 275 (20.1 %) with hypotension. None of the tested genetic variants passed the GWAS significance threshold (P<2.8×10 -9 ). Exploratory GWAS testing for the interaction effect between genome-wide variants and treatment also did not reveal significant signals We conducted a genome-wide gene-level collapsing analysis for the composite of CV death or hospitalisation for HF in the CHARM-Preserved study and in the CHARM-Overall study to support the GWAS findings. None of the 18,500 tested coding genes was significantly associated with the composite cardiovascular endpoint in candesartan treated patients (Supplementary Table 7 ). In the candidate regions at 8p21.3 and 3q13.13, none of the genes within the flanking regions were significantly associated (Supplementary Table 8 ). Multiple genetic variants related to the renin-angiotensin system have been proposed to modulate the effects of ARBs in HF and other cardiovascular diseases. The AGTR1 rs5186 genetic variant, which results in an A to C substitution at position 1166 in the 3'-untranslated region of the . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. In this pharmacogenomic study of the CHARM programme, we have identified a candidate genetic predictor of HF progression in patients with preserved ejection fraction from the CHARM-Preserved study with the composite of CV death or hospitalisation for HF. The genetic variant rs66886237-A at 8p21.3 was associated with time to hospitalisation for HF or CV death with an allelic HR of 1.91. The lead variant at this locus is located in an intron of the GFRA2 gene. Functional analysis of the variant did not find it to be a regulator of gene expression and it was not previously strongly associated with phenotypes in queried databases. However, the genetic region identified was concordant by colocalization analysis with a locus associated with cardiomyopathy in the FinnGen project, based on hospital discharge and cause of death registries (ICD-10 code I42). 27 Gfra2 was found to be a specific marker for cardiac progenitors among mesodermal cells and mice, and GFRA2 expression marks human developmental cardiac progenitor cell populations in embryonic stem cells/induced pluripotent stem cells differentiation. 28 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10.1101/2021.09.28.21263908 doi: medRxiv preprint We also report a non-significant association signal for the efficacy of candesartan at 3q13.13. Individuals with the rs664669-TT genotype represented approximately 31% of the population and they had a 41% reduction (95% CI: 0.45-0.76) in the combined incidence of CV death and hospitalisation for HF when treated with candesartan as compared to placebo. This may represent further benefit to that observed in the overall population with 16% reduction in CV death and hospitalisation for HF reported in the CHARM-Overall programme. However, the finding is hypothesis-generating, and replication of the association in an independent population sample is necessary. The previously reported candidate gene for response to candesartan, the angiotensin II receptor type 1 gene (AGTR1), 29, 30 was not associated with the tested endpoints, except for a possible association with hypotension in the HFrEF patient population. Based on our data, it is unlikely that single common genetic variants will have a major impact on response to candesartan. Nonetheless, given the limited size of the study population and the complexity of the HF phenotype, we cannot exclude the existence of a modest effect of several other gene variants including that of AGTR1. Our study had limitations. There may be volunteer bias in the pharmacogenomic subgroup compared to the main trial population, but the subgroup of participants was representative of the main trial in terms of the composite cardiovascular endpoint of time to CV death or hospitalisation for HF. Although we have corrected for the testing of several phenotypes, the results are to be interpreted with care and should be considered strictly as hypothesis generating. Results will have to be replicated and demonstrate biological plausibility before considering any changes in clinical practice. We have also limited our analyses to study subjects genetically predicted to be of European ancestry by comparison to HapMap CEU . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted September 29, 2021. ; https://doi.org/10.1101/2021.09.28.21263908 doi: medRxiv preprint samples. This was done to protect from population structure which may bias results. We did not have enough study participants from other population groups to assess the generalisability of our finding across other ancestries. The leading variant rs66886237 had a minor allele frequency of 0.14 in our study population, which compares to the frequency of 0.15 in samples from European populations in the 1000 Genomes reference, 31 and minor allele frequencies of 0.41 and 0.09 in samples from African and East Asian populations respectively. Importantly, the results have not been replicated in an independent population sample, and we cannot exclude the possibility that the results may be chance findings. In conclusion, we have identified a candidate genetic region at 8p21.3 near gene GFRA2 associated with the progression of HF in patients with preserved ejection fraction from the CHARM-Preserved study. The findings will need to be replicated in an independent population. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint Position on build37. Chr: chromosome; CI: confidence interval; EA: effect allele; EAF: effect allele frequency in the analysis dataset; HF: heart failure; RA: reference allele; HR: hazard ratio. * Allelic dosage effect assessed using Cox proportional hazards regression adjusted for age, sex, 10 principal components and candesartan treatment. . It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) Comparison of candesartan, enalapril, and their combination in congestive heart failure: randomized evaluation of strategies for left ventricular dysfunction (RESOLVD) pilot study. The RESOLVD Pilot Study Investigators Angiotensin II type 1 receptor antagonist decreases plasma levels of tumor necrosis factor alpha, interleukin-6 and soluble adhesion molecules in patients with chronic heart failure Effects of combined candesartan and ACE inhibitors on BNP, markers of inflammation and oxidative stress, and glucose regulation in patients with symptomatic heart failure Effects of candesartan in patients with chronic heart failure and reduced left-ventricular systolic function intolerant to angiotensin-converting-enzyme inhibitors: the CHARM-Alternative trial Effects of candesartan in patients with chronic heart failure and reduced left-ventricular systolic function taking angiotensin-converting-enzyme inhibitors: the CHARM-Added trial Effects of candesartan in patients with chronic heart failure and preserved left-ventricular ejection fraction: the CHARM-Preserved Trial A randomized trial of the angiotensin-receptor blocker valsartan in chronic heart failure Candesartan in heart failure--assessment of reduction in mortality and morbidity (CHARM): rationale and design. Charm-Programme Investigators Effects of candesartan on mortality and morbidity in patients with chronic heart failure: the CHARM-Overall programme pyGenClean: efficient tool for genetic data clean up before association testing PLINK: a tool set for whole-genome association and population-based linkage analyses Principal components analysis corrects for stratification in genome-wide association studies Genotype imputation with thousands of genomes A linear complexity phasing method for thousands of genomes Spontaneous Coronary Artery Dissection: Insights on Rare Genetic Variation From Genome Sequencing International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted Spontaneous Coronary Artery Dissection: Insights on Rare Genetic Variation from QQperm: permutation based QQ plot and inflation factor estimation GCTA: a tool for genome-wide complex trait analysis Improved methods for multi-trait fine mapping of pleiotropic risk loci Annotation of functional variation in personal genomes using RegulomeDB DSNetwork: An Integrative Approach to Visualize Predictions of Variants Open Targets Genetics: systematic identification of traitassociated genes using large-scale genetics and functional genomics ExPheWas: a browser for gene-based pheWAS associations PhenoScanner V2: an expanded tool for searching human genotypephenotype associations PhenoScanner: a database of human 26 Eliciting priors and relaxing the single causal variant assumption in colocalisation analyses Documentation of R4 release GFRA2 Identifies Cardiac Progenitors and Mediates Cardiomyocyte Differentiation in a RET-Independent Signaling Pathway Effects of AGTR1 A1166C gene polymorphism in patients with heart failure treated with candesartan A1166C on the effects of candesartan in patients with heart failure ) 366 (76.4%) 374 (80.6%) 293 (57.3%) 270 (52.1%) Heart disease risk factors NYHA class II 3%) 238 (17.6%) 3%) 139 (10.3%) 33 (8.7%) 3%) 239 (49.9%) 214 (46.1%) 343 (67.1%) 351 (67.8%) Atrial fibrillation PCI, n (%) 212 (15.5%) 227 (16.7%) 1%) Diuretic 1185 (86.4%) 1155 (85.2%) 330 (86.6%) 327 (87.4%) 438