Deep coverage whole genome sequences and plasma lipoprotein(a) in individuals of European and African ancestries

Abstract

Lipoprotein(a), Lp(a), is a modified low-density lipoprotein particle that contains apolipoprotein(a), encoded by LPA, and is a highly heritable, causal risk factor for cardiovascular diseases that varies in concentrations across ancestries. Here, we use deep-coverage whole genome sequencing in 8392 individuals of European and African ancestry to discover and interpret both single-nucleotide variants and copy number (CN) variation associated with Lp(a). We observe that genetic determinants between Europeans and Africans have several unique determinants. The common variant rs12740374 associated with Lp(a) cholesterol is an eQTL for SORT1 and independent of LDL cholesterol. Observed associations of aggregates of rare non-coding variants are largely explained by LPA structural variation, namely the LPA kringle IV 2 (KIV2)-CN. Finally, we find that LPA risk genotypes confer greater relative risk for incident atherosclerotic cardiovascular diseases compared to directly measured Lp(a), and are significantly associated with measures of subclinical atherosclerosis in African Americans.

Fig. 1

Schema of overall study design Analyses were stratified by phenotype, Lp(a) (mass) and Lp(a)-C, where available. Lp(a)-C analyses were performed using the following individuals with WGS data: 2284 individuals from the Estonian Biobank (EST) and 3418 individuals from Jackson Heart Study (JHS). Lp(a) mass analyses were performed using the same Jackson Heart Study participants, as well as array-derived genotypes from 27,344 Finnish FINRISK (FIN) individuals with imputation performed using 2690 FIN individuals with WGS and 5093 FIN individuals with WES. After quality control filters, 119,401,837 SNPs and 7,207,350 indels were discovered genome-wide across individuals analyzed. Structural variant discovery at the LPA locus was performed, finding KIV2-CN and eight additional rare CNVs. An imputation model was developed to impute KIV2-CN using 60 LPA locus variants. Three overarching analyses were subsequently performed: (1) Common variant analyses, (2) Rare variant analyses, and (3) Mendelian randomization. Among common and low-frequency variants with MAF > 0.1%, we performed single variant analysis, and separately, analyzed genetic modifiers of KIV2-CN’s effect on Lp(a) and Lp(a)-C concentrations. We also performed rare variant analyses, aggregating rare variants (MAF < 1%) in (1) coding sequence and (2) putative functional non-coding sequence, and associated with Lp(a)-C. Lastly, we performed Mendelian randomization, using different classes of variants associated with Lp(a) as genetic instruments and associating these with incident clinical cardiovascular events in FIN and prevalent subclinical atherosclerosis in JHS, MESA, FHS, and OOA. CNV copy number variant, EST Estonian biobank, FHS Framingham heart study, FIN FINRISK, JHS Jackson Heart Study, KIV2-CN kringle IV-2 copy number, Lp(a) lipoprotein(a), Lp(a)-C lipoprotein(a) cholesterol, MAF minor allele frequency, MESA Multi-ethnic study of atherosclerosis, MR Mendelian randomization, OOA Old-Order Amish

Fig. 2

Structural variant discovery at the LPA locus and KIV2-CN imputation. a Nine separate copy number variants were discovered across the EST, JHS, and FIN whole genome sequences. Here, these are shown by plotting sample-level normalized read depth against the position along the hg19 reference genome at the LPA locus (with the black line denoting median read depth across all individuals). The KIV2-CN is shown in the highlighted region and each unique non-gray line outside of this region depicts a discovered structural variant (described further in Supplementary Table 4). b The random forest importance of each variant in the 61-variant KIV2-CN imputation model developed in FIN is shown against its genomic position, with KIV2-CN region highlighted and the top five rsIDs labeled. c Correlation of directly genotyped KIV2-CN and imputed KIV2-CN from 738 FIN individuals with WGS in the validation dataset (with Pearson correlation, R_p = 0.78). EST Estonian biobank, FIN FINRISK, JHS Jackson Heart Study, KIV2-CN kringle IV-2 copy number

Fig. 3

KIV2-CN association with Lp(a) phenotypes. Directly genotyped KIV2-CN (in EST and JHS) and imputed KIV2-CN (in FIN) are inversely associated with Lp(a) and Lp(a)-C. EST Estonian biobank, FIN FINRISK, JHS Jackson Heart Study, KIV2-CN kringle IV-2 copy number, Lp(a) lipoprotein(a), Lp(a)-C lipoprotein(a) cholesterol; R_p = Pearson correlation; R_S = Spearman correlation

Fig. 4

Trans-ethnic LPA and non-LPA loci associations with lipoprotein(a) phenotypes. In trans-ethnic meta-analysis of single variant results adjusted for KIV2-CN, we observed two associations (P < 5 × 10⁻⁸) at loci distinct from LPA and independent of other conventional lipid measures: SORT1 for Lp(a)-C and APOE for Lp(a). a–b Associations (Betas in SD and 95% CI) for top variants at the SORT1 and APOE loci are shown by ethnicity. The SORT1 and APOE loci have been previously associated with LDL cholesterol. Thus, associations conditional on LDL cholesterol are also presented. The effect size for SORT1 is preserved after conditioning on LDL cholesterol while the effect size for APOE is slightly reduced but still genome-wide significant. c Standardized effect estimates for variants at the LPA locus (LPA TSS ± 1 Mb) attaining P < 5 × 10⁻⁸ in JHS are shown comparing effects in JHS (African Americans) with FIN (European Americans). Color demonstrates inter-ethnic effect difference as measured by heterogeneity P. Similar effects are observed for a known null (splice donor) mutation in LPA but strongly diverging effects are observed for a distinct nearby LPAL2 intronic variant. d Genetic heritability estimates using variants with MAF > 0.001 for normalized Lp(a) were acquired for African Americans in the whole-genome sequenced JHS cohort and for Europeans in the genotyped and imputed FIN cohort. Here, heritability and 95% CI are shown without adjusting for KIV2-CN. KIV2-CN kringle IV-2 copy number, HetP heterogeneity P, Lp(a) lipoprotein(a), Lp(a)-C lipoprotein(a) cholesterol, MAF minor allele frequency, TSS transcription start site

Fig. 5

Genetic modifiers of KIV2-CN’s effect on lipoprotein(a) cholesterol. Three independent genetic modifiers of KIV2-CN’s effect on Lp(a)-C were discovered at the LPA locus. Regional association plots showing the variant-by-KIV2-CN interaction P values of all variants within a 1 Mb window of the LPA TSS are shown for African Americans (top) and Europeans (bottom), highlighting variants in linkage disequilibrium with rs1810126 (green), rs13192132 (red), and rs1840445 (blue), the top independent genome-wide significant variants (interaction P < 5 × 10⁻⁸) upon meta-analysis. KIV2-CN kringle IV-2 copy number, Lp(a)-C lipoprotein(a) cholesterol, TSS transcription start site

Fig. 6

Rare variant non-coding burden analyses. A schematic of rare variant association results from (1) aggregating rare variants in adult liver enhancers or promoters and strong DHS (P < 10⁻¹⁰) within 3 kb sliding windows, and (2) aggregating rare variants in liver enhancers grouped to LPA via the “By expression” in silico prediction method. In the top two panels, each red diamond represents the meta-analyzed mixed-model SKAT P-value with Lp(a)-C of rare (MAF < 1%), non-coding variants overlapping liver enhancer or promoter annotations in strong DHS (P(DHS) < 1e-10) grouped in a 3 kb window, before adjusting for KIV2-CN (top, “Original”) and after adjusting for KIV2-CN (bottom, “Conditioned on KIV2-CN”). The horizontal red lines denote the genome-wide Bonferroni significance threshold given the number of unique windows analyzed. The horizontal gray lines denote the Bonferroni significance threshold within this 1MB region around LPA. The regions incorporated into the “By Expression” grouping to LPA are shown in aqua, along with the respective associations of rare non-coding variants in these regions before and after conditioning on KIV2-CN. Annotated adult liver enhancers (green bars) and promoters (red bars) overlapping strong DHS are included above protein-coding genes from Ensembl. DHS DNAse hypersensitivity sites, Lp(a)-C lipoprotein(a) cholesterol, MAF minor allele frequency, SKAT Sequence Kernal Association Test

Fig. 7

Association of LPA variant classes with atherosclerosis. Mendelian randomization was performed using three genetic instruments: a weighted genetic risk score using variants conditioned on KIV2-CN at a 4 Mb window around LPA (GRS), a KIV2-CN score, and a combined GRS + KIV2-CN score, and compared to the observational effects. The genetic instruments were all normalized such that 1 unit increase in the score is equal to 1 SD increase in Lp(a). a Associations (HR and 95% CI) of incident coronary heart disease (1056 cases; 21,207 controls) and myocardial infarction (580 cases; 21,377 controls) with the Lp(a) measurement and with genetic instruments among the genotyped and imputed FIN individuals (exact values in Supplementary Table 30). b Associations (Beta and 95% CI) of Lp(a) measurements and respective genetic instruments with standardized markers of subclinical atherosclerosis (CAC and AAC) among whole genome sequences of African Americans from 1701 JHS and 932 MESA participants, as well as European Americans from 1536 FHS and 1651 MESA participants (Supplementary Table 12). These data indicate that (1) a comprehensive Lp(a) genetic instrument (GRS + KIV2-CN) provides improved risk assessment compared to the Lp(a) phenotype, and (2) further stratifying this comprehensive instrument into separate Lp(a) variant classes provides additional risk stratification in that genomic sequence variants independent of KIV2-CN (i.e., GRS) have a stronger influence on clinical atherosclerosis compared to KIV2-CN. AAC Abdominal aortic calcium, CAC coronary artery calcium, CI confidence interval, FHS Framingham Heart Study, FIN FINRISK, GRS genetic risk score, HR hazard ratio, JHS Jackson Heart Study, KIV2-CN kringle IV-2 copy number, Lp(a) lipoprotein(a), Lp(a)-C lipoprotein(a) cholesterol, MESA Multi-Ethnic Study of Atherosclerosis