Genome-wide association meta-analysis identifies risk loci for abdominal aortic aneurysm and highlights PCSK9 as a therapeutic target

Abdominal aortic aneurysm (AAA) is a common disease with substantial heritability. In this study, we performed a genome-wide association meta-analysis from 14 discovery cohorts and uncovered 141 independent associations, including 97 previously unreported loci. A polygenic risk score derived from meta-analysis explained AAA risk beyond clinical risk factors. Genes at AAA risk loci indicate involvement of lipid metabolism, vascular development and remodeling, extracellular matrix dysregulation and inflammation as key mechanisms in AAA pathogenesis. These genes also indicate overlap between the development of AAA and other monogenic aortopathies, particularly via transforming growth factor β signaling. Motivated by the strong evidence for the role of lipid metabolism in AAA, we used Mendelian randomization to establish the central role of nonhigh-density lipoprotein cholesterol in AAA and identified the opportunity for repurposing of proprotein convertase, subtilisin/kexin-type 9 (PCSK9) inhibitors. This was supported by a study demonstrating that PCSK9 loss of function prevented the development of AAA in a preclinical mouse model.

Supplementary Figure 1: QQ plots of 17 AAA GWAS summary statistics from 14 discovery cohorts.The expected logistic regression association p-values versus the observed p-values for AAA association are displayed.Sample sizes and analytical methods in discovery cohorts are listed in Supplementary Table 1.All p-values are two-sided.The confidence intervals were calculated with assumption that standard uniform order statistics follow a beta distribution (Default settings in https://genome.sph.umich.edu/wiki/Code_Sample:_Generating_QQ_Plots_in_R).
Supplementary Figure 2: Plot of GWAS effect size estimation (X-axis) against -log10p-value (Yaxis) for 9 known index variants of AAA in discovery cohorts.Sample sizes and analytical methods in discovery cohorts are listed in Supplementary Table 1.All p-values are two-sided.
Supplementary Figure 3: Manhattan plot of AAAgen GWAS meta-analysis (Ncase=39,221;  Ncontrol=1,086,107). Meta-analysis of 17 GWAS summary statistics from 14 discovery cohorts was performed by METAL 1 in standard error mode with genomic control correction.Novel loci are in red font and known loci are in black font.Variants with P > 10 -3 and P < 10 -60 (9p21) were excluded for better visualization.All p-values are two-sided.
Supplementary Figure 6: Comparison of effect estimates (95% CI) for 121 index variants between AAAgen GWAS meta-analysis (X-axis; Ncase=39,221; Ncontrol=1,086,107) and leave-onecohort-out meta-analysis for six discovery cohorts (Y-axis).The sample sizes in leave one cohort out meta-analyses can be obtained by subtracting numbers in discovery cohorts in Supplementary Table 1 from AAAgen GWAS meta-analysis.Supplementary Figure 7: Comparison of effect estimates (95% CI) in 121 loci between AAAgen GWAS meta-analysis (X-axis; Ncase=39,221; Ncontrol=1,086,107) and external replication cohorts (Y-axis; Max Ncase=5,451; Ncontrol=299,885; not all variants were present in 3 external replication cohorts; see Supplementary Table 2 for further details).The red dots represent 80 variants with two-sided p-value<0.05 in external cohorts and the grey dots represent other 41 variants.Iceland.The Icelandic controls used were selected from individuals who have participated in various GWA studies and who were recruited as part of genetic programs at deCODE.Individuals with known cardiovascular disease were excluded as controls but controls were unscreened for AAA.Genotyping was performed using various Illumina SNP chips and phased using long-range phasing 22 , followed by imputation of 37.6m high-quality variants from 49,962 whole genomesequenced Icelanders 23,24 .The association analysis of 1,656 AAA cases and 265,410 controls was performed using deCODE genetics software  16 for the usage of UK biobank controls in various analyses.
All participants in UKAGS and VIVA were genotyped using the UKBB Axiom Array.
UKBB participants were genotyped using the UKBB Axiom array and UKBB BiLEVE Axiom Additional validation cohorts for PRS: Additional AAA case cohorts from Oxford (UK), Uppsala (Sweden) and Utrecht (Netherlands) became available during the conduct of the study.
Those from Oxford and Uppsala were genotyped alongside remaining cases from the UKAGS and VIVA study that had not been included in the discovery study and were analyzed together in comparison to controls taken from UK Biobank.The Aneurysm Consortium AAA GWAS 33 had not met QC criteria for inclusion in the discovery analysis but was available for validation of polygenic risk scores.The UCC-SMART cohort of people with prevalent cardiovascular disease 34 was used for PRS validation to determine discriminatory performance to identify AAA cases amongst populations with cardiovascular disease.or not, a VST expression threshold of 6 was used.This cutoff was chosen based on a density plot of all transcript expression that yielded a bimodal distribution (Supplementary Figure 17).A cutoff of 6 effectively eliminated genes with no expression.

Conventional Mendelian Randomization
We performed conventional two-sample Mendelian randomization to estimate the total effect of each major lipoprotein-related trait on AAA.We constructed genetic instruments from independent (r 2 < 0.001, distance > 10,000kb) genetic variants associated with each major lipoprotein-related trait in the UK Biobank.After identifying the corresponding genetic variants in our GWAS of AAA and harmonizing the effect alleles, we performed MR using the TwoSampleMR package in R 40 .Our primary analysis used the inverse variance weighted method.
In sensitivity analyses, we performed MR-Egger, weighted median, and weighted mode MR, which make different assumptions about the presence of pleiotropy 41 .

MR-BMA
We performed a variable selection method in a multivariable Mendelian Randomization (MR) framework to prioritize the causal lipoprotein determinants of the outcomes.Multivariable MR extends the basic MR framework to include multiple exposures in one joint model, accounting for horizontal pleiotropy among exposures, which is particularly relevant when considering highly correlated traits like blood lipoprotein-related traits as exposures 42 .In order to rank and select the likely causal lipoprotein risk factors for AAA, we employed an extension of multivariable MR called Mendelian randomization Bayesian model averaging (MR-BMA), a Bayesian approach for prioritizing causal exposures in a two-sample multivariable MR setting 43 .MR-BMA performs variable selection by evaluating models with all possible combinations of lipoprotein-related traits as exposures and computing the posterior probability that the model contains the true causal risk factors.Unlike other univariate or multivariable MR methods, MR-BMA aims to identify true causal risk factors among correlated traits, rather than estimate the magnitude of effect.The marginal inclusion probability (level of evidential support for each exposure) is derived from the sum of all posterior probabilities of the models where the specific exposure was included.We removed influential variants based on the Cook's distance and outliers based on the q-statistic as previously recommended 44 .An empirical permutation procedure was performed to calculate p-values.Briefly, the expected marginal inclusion probability distribution for each risk factor under the null hypothesis was generated by performing 1,000 permutations of the MR-BMA analysis, holding the SNP-risk factor associations constant and randomly permuting the SNP-outcome associations.The observed marginal inclusion probabilities for each risk factor were then compared to the expected distribution under the null, with p-values computed by pj = (rj +1)/(nperm + 1), where rj represents the rank of the observed marginal inclusion probability of a given risk factor (j) across all permutations (nperm = 1000).Adjustment for multiple testing was done using the Nyholt correction for correlated traits.
19ference panel backbone used for imputation18.The association analysis of 3,079 AAA cases and 180,236 controls was performed with SAIGE19using year of birth, sex and 10 PCs as covariates.The Cardiovascular Health Improvement Project (CHIP) is a cohort of individuals treated at Michigan Medicine with linked genotype, EHR, and family history data.The Michigan Genomics Initiative (MGI) is a hospital-based cohort with linked genotype and EHR data from participants recruited during pre-surgical encounters at Michigan Medicine.Both studies were approved by the Institutional Review Board of the University of Michigan Medical School 5-1.5)1.8 (1.1-2.9)1.7 (1.1-2.7)2.3(1.4-3.6)2.2(1.4-3.4)2.9 (1.9-4.6)3.2(2.1-5)3.6(2.3-5.6)5.4 (3.5-8.3)1.2 (0.8-1.7) 1.6 (1.1-2.2) 1.7 (1.2-2.3)1.9 (1.4-2.6)1.7 (1.3-2.9)2.1 (1.3-0.8)2.1 (0.9-0.6) 2.2 (1.6-1.1)2.7 (1.7-1.1)2.5 (1.8-1.2) 2.8 (1.8-1.2) 3.1 (2.6-1.8)3.9 (2.9-1.9)Supplementary Figure 8: Performance of the PRS constructed by this AAAgen GWAS metaanalysis was compared with the Klarin et al.PRS (MVP) 2 , the largest GWAS of AAA previously published.Three validation datasets were used.The UKAGS-UKB dataset is a comparison of screen-detected AAA from the UK AAA screening programmes with controls from UK Biobank.The AC-WTCCC data is a case-control study of AAA.The SMART study is an observational cohort of patients presenting with cardiovascular disease and represents a case-control study where controls had prevalent cardiovascular disease.Three different shapes were used to denote odds ratio (95% CI) observed three validation datasets.We observed higher odds ratios by AAAgen compared to Klarin et al. in all validation datasets.*Decile 1 is the reference decile against which all other deciles were compared.Screening Array from Illumina was used for genotyping samples from CHB-CVDC and DBDS.Whole-genome sequence data from 8429 Danes along with 7146 samples from North-Western Europe forms a CHIP+MGI: (IRBMED) (HUM00052866, HUM00071298) and informed consent was obtained from study participants.534 cases from CHIP were identified as aneurysm in abdominal aorta following diagnosis by Cardiologists and after excluding cases with known dissection.From MGI, 749 cases were identified using ICD codes (ICD9: 441.3/4441.4;ICD10: I71.3/I71.4)after excluding cases with known dissection (ICD9: 441.00-03;ICD10: I71.00-03).After removing samples with related phenotypes (phecodes 440-449.99), a case-control matching strategy was used to identify controls from MGI. Case-control matching was performed using MatchIt package 20 in R using birth year, gender, array version and 4 genotype PCs.Samples in CHIP and MGI were genotyped using two imputation.Imputation was performed with the Haplotype Reference Consortium (HRC) panel using Minimac4 21 .Association analysis of 1,283 AAA cases and 12,202 controls was performed using SAIGE 19 with birth year, gender, array version and 4 genotype PCs as covariates.deCODE: Icelandic individuals with AAA were identified from a registry of individuals, using ICD codes (ICD9: 441.3, 441.4,ICD10: I71.3, I71.4) who were admitted at Landspitali University Hospital, in Reykjavik, Iceland, 1980-2016, or diagnosed at private clinic in Reykjavik, 19, and we adjusted for gender, county of origin, current age or age at death (first and second order term included), blood sample availability for the individual, and an indicator function for the overlap of the lifetime of the individual with the time span of phenotype collection.We used LD score regression to account for distribution inflation due to cryptic relatedness and population stratification25.We identified 3092 cases and 15,025 controls in the EUR dataset and 119 cases and 603 controls in the AFR dataset, and logistic regression was performed on unrelated samples within ancestry groups using PLINK 2.0. Alls were adjusted by clinical site and first 5 principal components.The Trøndelag Health (HUNT) Study 28 is a population-based health study from the Trøndelag region in Norway.Individuals have been enrolled in the study though multiple phases (HUNT1, HUNT2, HUNT3, and HUNT4) beginning in 1984.During each enrollment period, individuals living in the region who are 20 years or older are invited to participate.To date, over 120,000 individuals have participated and have longitudinal health record data available from surveys, medical records, and biological samples.AAA cases were defined by ICD-9 codes 441.3 and 441.4 and ICD-10 codes I71.3 and I71.4.HUNT samples were genotyped using the Illumina HumanCoreExome array and imputed with Minimac3 21 using a combined reference panel consisting of the HRC v1.1 combined with whole genome sequencing of 2,201 HUNT participants.CPT codes, and laboratory measurements) and natural language processing of unstructured data elements, such as vascular laboratory and imaging reports in the EHR.The study was approved by the Mayo Clinic Institutional 74 Review Boards (IRB # 08-008355).After exclusion of overlapping samples between Mayo VDB Dataset and eMERGE III V3 Imputed Array Dataset 26 , AAA cases were defined as having an infrarenal abdominal aortic diameter ≥3 cm or a history of open or endovascular AAA repair based on an electronic phenotyping algorithm using Natural HumanCoreExome Beadchip, and Human 610 Quad V1 platforms.We combined samples genotyped on different platforms by genotype imputation based on a HRC v1.1 panel.Association testing of 771 AAA cases and 4,913 controls was performed on the imputed dataset using SAIGE software with the following covariates: study enrollment age, sex, genotyping platform, first 5 principal components of ancestry.In the Million Veteran Program (MVP), individuals aged 18 to over 100 years have been recruited from 63 VA Medical Centers across the United States.MVP received ethical/study protocol approval by the VA Central Institutional Review Board, and informed consent was obtained for all participants.Each additional study received approval from their local institutional review board.From the participants passing quality control in MVP, individuals were defined as having AAA or being a disease-free control using a previously utilized 2 definition initially proposed by Denny et al 29 .AAA cases were defined as the presence of two instances of any of the Imputation was then conducted using IMPUTE 2.2 run on the BCISNPmax database platform (version 3.5, BCI Platforms, Espoo, Finland).The reference haplotypes were based on the 1000 Genomes June 2011 release.Imputed calls were filtered by quality score (>0.9)TheTripleABarcelonaStudy (TABS) is a hospital-based study recruiting individuals with AAA treated in the Hospital de la Santa Creu i Sant Pau in Barcelona, Spain.All individuals have repetitive measurements of the abdominal aortic diameter, either by CT-scan or by ultrasound along with anthropometric and clinical information.DNA, RNA, and plasma samples were collected from all individuals.All participants gave written informed consent.All procedures were approved by the Institutional Review Board at Hospital de la Santa Creu i SantPau.Cases were defined as those with dilation of the abdominal aorta with a diameter higher than 30 millimeters.Type B dissections that progress to AAA, saccular aneurysms or thoracic aneurysms were excluded.In addition to 42 cases from TABS, 10 cases and 82 controls were leveraged from the Triple A Genetic Study (TAGA)31.Also, 401 controls were leveraged from the RETROVE study 32 .Genome-wide genotyping was performed using the Infinium Global Screening Array-24 v2.0 from Illumina (coverage 665,608 variants) and imputed to the HRC reference panel using the Michigan server.Association analysis of 52 AAA cases and 483 controls was performed using SAIGE19with birth year, sex, batch and PCs as covariates.
19,27e Geisinger Institutional ReviewBoard (2006-0258).We included individuals with EHR data, genotype data, and those determined to be of primarily European descent based on SNP-derived principal components analysis.We defined AAA cases based on the presence of at least two instances of any of the following ICD10 codes: 441.3, 441.4,I71.3,I71.4,andrequiredcontrols to lack any occurrence of the aforementioned ICD10 codes, in addition to the ICD10 codes I71-75, I77-79, K55.We further removed individuals who contributed to the study via the eMERGE consortium.DiscovEHR participants were genotyped on either the Illumina Omni Express Exome or Global Screening Array and imputed from the HRC reference panel using the Michigan Imputation Server.We performed association analysis of 2,238 AAA cases and 105,433 controls using whole genome regression in REGENIE (v1.0), including Age, Age 2 , Sex, Age×Sex, Age 2 ×Sex, 10 common variant (MAF>1%) derived principal components, and an array batch indicator as covariates in the model.R program.Samples in eMERGE Phase III were genotyped on 78 different Illumina and Affymetrix SNP array platforms.Data from different sites and genotyping platforms were imputed using the HRC reference panel on Michigan Imputation Server and then merged into one set26,27.HUNT:Association testing of 734 cases and 68,901 controls was performed using SAIGE19with batch, sex, birth year, and PC1-4 as covariates.Mayo VDB: Mayo Vascular Disease Biorepository (VDB) at the Mayo Clinic was established to archive DNA, plasma, and serum from patients with suspected atherosclerotic cardiovascular disease (ASCVD) referred for noninvasive vascular evaluation and exercise stress testing at the Mayo Clinic Gonda Vascular Center from January 14, 2006 to July 24, 2020.ASCVD phenotypes and related comorbidities were ascertained using previously validated electronic phenotyping algorithms based on both structured data elements (ICD diagnosis codes, Language Processing on ultrasound reports, validated by manual chart review.Controls were not known to have AAA and had no ICD-9 diagnosis codes for AAA.Genotyping was performed using three different Illumina platforms, including Illumina Human660W-Quad V1, MVP: following ICD-9/10 codes in a participant's EHR: 441.3, 441.4,I71.3,I71.4.Controls were defined as possessing zero occurrences of the aforementioned ICD codes, as well as zero occurrences of the ICD-9 codes 440-448, or ICD-10 codes I71-75, I77-79, K55.In our MVP analysis, we evaluated 17,672 AAA cases and 303,695 controls of European ancestry, and 1,888 AAA cases and 87,728 controls of African ancestry.Genotyped and imputed DNA sequence variants in individuals of were tested for association with AAA using logistic regression adjusting for age, sex, and 5 principal components of ancestry assuming an additive model using the PLINK2.0statisticalsoftwareprogram.NZ AAA Genetics Study:The Vascular Research Consortium of New Zealand recruited (ultrasound) and other cardiovascular risk factors.All participants were genotyped in two separate case-control cohorts using the Affymetrix SNP6 (cohort 1; 608 AAA cases and 612 controls) or Illumina Omni2.5 (cohort 2; 397 cases and 384 controls) GeneChip arrays and had call rates >95% (mean 99.2%).analysis.19,515recruitsweregenotyped on three different arrays, Illumina Quad Omni, Global Screening Assay v1 and Global Screening Assay v2.Genotype imputations were performed using the Eagle2 30 and Minimac softwares 21 on the Michigan Imputation server and were completed for all autosomes with the HRC reference panel.AAA cases were found using the AAA phecode definition of 442.11, meaning at least 1 encounter with abdominal aortic aneurysm diagnosis codes and excluding other confirmed disease of the arteries.Controls were defined as all others with genotype data and no evidence of AAA.GWAS of 388 cases and 9879 controls was performed using plink software with age at recruitment, gender, genetic determined ancestry and the first 10 principal components.TABS:(UKAGS) is a prospective study of men attending the NHS aneurysm screening programmes in the UK at reaching age 65.Men recruited into the UKAGS completed a postal questionnaire to obtain information on smoking, comorbidities and medications.Screening outcomes (ultrasound measured AAA diameter) were obtained directly from the AAA screening programmes.Ethical approval was granted by an NHS research ethics committee.The study was funded by the British Heart Foundation (CS/14/2/30841 and RG/18/10/33842) and the Circulation Foundation.The VIVA screening trial is a randomized, clinically controlled study designed to evaluate the benefits of vascular screening and modern vascular prophylaxis in a population of 50,000 men aged 65-74 years, randomized to either receive an invitation for vascular screening or being a control.Ethical approval was granted by the research ethics committee of Mid Denmark (M20080028), and funded by the FP7, EU, and the Region of Mid Denmark.ClinicalTrials.govNCT00662480.Individuals included from both studies were of white British/Danish ancestry, with PCA identified outliers excluded; related individuals were identified with PLINK v1.9 IBD computation (--genome), and one of any pairwise kinships removed.forsex and age, and without overlap of subjects between analyses.See Supplementary Figure 35AAA: Multimodal Assessment of Aortic Aneurysm Disease Pathogenesis: Oxford Abdominal Aortic Aneurysm Study (OxAAA).Single centre study at the Oxford University Hospitals NHS Trust, funded by the NIHR Oxford Biomedical Research Center.IRAS ID: 122591.REC name: South Central -Oxford C Research Ethics Committee.REC reference: 13/SC/0250.Uppsala Abdominal Aortic Aneurysm Cohort Study, part of the collection Swedish Cohort Consortium (Cohorts.se).A prospective population-based case-control study which aims to find biomarkers for the cause and progression of AAA and investigate how persons with or without an AAA experience their health.All 65-year-old men, identified through the National Population Registry, were invited to an ultrasound examination.Financial support was provided by the Swedish Research Council (grant 2012-1978), the Swedish Heart and LungFoundation, the King Gustaf V's and Queen Victoria's Freemason Foundation, and Regional Uppsala-Örebro Research Grant AC: Aneurysm Consortium (AC), cases of AAA together with available DNA recruited in eight centers in the UK, Australia, and New Zealand and also samples from the United Kingdom Wellcome Trust Case Control Consortium 2 data used included samples from the 1958 British Birth Cohort and from the UK National Blood Service.Further details in Bown et al.33and https://www.wtccc.org.uk/.This set includes 808 cases from UKAGS and 4 cases from VIVA that were not used in the meta-analysis, 247 cases from UppsalaAAA, 71 cases from OxAAA, and 5810 controls from UKBB.All participants in UKAGS, UppsalaAAA, OxAAA and VIVA were genotyped using the UK Biobank Axiom Array.UKBB participants were genotyped using the UK Biobank Axiom array and UK Biobank BiLEVE Axiom array.The following QC filters were used: This set includes 1887 cases from AC and 5437 controls from WTCCC.The case and control cohorts were separately genotyped with the Illumina 670K BeadChips.Raw intensity data were normalized with BeadStudio, and genotypes were called concurrently from the combined case control data set with the Illuminus algorithm.The following a large number of samples or when samples were >5 standard deviations from median for inbreeding coefficient, with a sex mismatch between genotype and phenotype35.For the PRS validation, further exclusions were made on the following basis: those with missing AAA or CVD phenotype information, controls that were aged 43 or under at the initial intake, controls that were lost to follow up, and cases that were under the age of 40 at intake.AAA cases were determined on 1) self-reported operations in AAA, 2) self-reported history of AAA, 3) inclusion in the SMART study because of an AAA, 4) infrarenal AAA measure >30mm or ratio of infrarenal:suprarenal aortic diameter > 1.5, or 5) finding of an AAA in the echo.This resulted in the inclusion of 663 cases and 3,105 controls in this validation dataset.
UppsalaAAA:WTCCC2: low quality genotyping or those who were related to each other were excluded.In case of the latter, the patient with the latest (most recent) date of inclusion was excluded.Other reasons for exclusion during quality control were samples with likely sample contamination based on high degree of relatedness with and stored at room temperature between 1-3 days.The RNAlater solution was removed, and the tissue was stored at -80 °C.Tissues were cryopulverized using a CP02 instrument (Covaris), and an aliquot of the powdered tissue was used for isolation of total RNA using Trizol (Invitrogen) and RNeasy Mini Kit (Qiagen, GmbH, Hilden, Germany).Briefly, the RNA in the aqueous phase from the Trizol extraction was transferred to a new tube and mixed with an equal volume of 70% ethanol before processing on an RNeasy Mini kit column according to the manufacturer's instructions.RNA was eluted in 50ul of RNAse-free water.Concentrations were measured using the Qubit RNA Broad Range kit (Thermo Scientific, USA), followed by RNA integrity (RIN) evaluation by RNA TapeScreen on the Agilent 2200 TapeStation (Agilent Technologies, USA). centered around 50%. 15 abdominal aortic aneurysm RNA samples passed quality control.FeatureCounts 38 was used to aggregate a gene counts matrix.DESeq2 39 was used for variance stabilization transformation (VST) normalization.To determine if a gene was expressed