Genome-wide association study of lipids in Greenlandic Inuit shows independent PCSK9 association and reveals a unique genetic architecture

doi:10.21203/rs.3.rs-2633131/v1

Perturbation of lipid homeostasis is a major risk factor for cardiovascular disease (CVD), the leading cause of death worldwide. We aimed to identify genetic variants affecting lipid levels, and thereby risk of CVD, in Greenlanders. Genome-wide association studies (GWAS) of six blood lipids, triglycerides, LDL-cholesterol, HDL-cholesterol, total cholesterol, as well as apolipoproteins A1 and B, were performed in up to 4473 Greenlanders. For genome-wide significant variants, we also tested for associations with additional traits, including CVD events. We identified 11 genome-wide significant loci associated with at least one of the lipid traits, and a potential causal variant near PCSK9 (rs12117661), which was independent of the known PCSK9 loss-of-function variant (rs11491147). rs12117661 was associated with lower LDL-cholesterol (β_SD(SE)=-0.22 (0.03), p=6.5x10^-12) and total cholesterol (-0.17 (0.03), p=1.1x10^-8) in the Greenlandic study population. Similar associations were observed in Europeans from the UK Biobank, where the variant was also associated with a lower risk of CVD outcomes. Moreover, rs12117661 was a top eQTL for PCSK9 across tissues in GTEx, and was located in a predicted regulatory element, supporting a possible causal impact on PCSK9 expression. Combined, the 11 GWAS hits explained up to 16.3% of the variance of the lipid traits. This suggests that the genetic architecture of lipid levels in Greenlanders is markedly different from Europeans, with fewer variants explaining the variance.

Biological sciences/Genetics/Genetic association study/Genome-wide association studies

Health sciences/Risk factors

Health sciences/Diseases/Cardiovascular diseases/Dyslipidaemias

genome-wide association study

PCSK9

lipids

LDL-cholesterol

genetic risk score

cardiovascular disease

Cardiovascular disease (CVD) is a major health burden worldwide, and ischemic heart disease and stroke have been estimated to account for more than one-fourth of all deaths in 2016, making these diseases the leading cause of death globally (1). Currently, one of the most successful strategies for preventing CVD is to treat dyslipidemia with lipid-lowering drugs (2). Circulating levels of blood lipids are affected by lifestyle, diet, and genetic variation (3). The genetics of circulating lipid levels has been studied widely, resulting in the identification of more than 940 loci in genome-wide association studies (GWAS) of levels of triglycerides, LDL-, HDL-, non-HDL-, and total cholesterol in more than 1.65 million individuals from different ancestry groups (4, 5). In Europeans, each of the identified variants only explains a small fraction of the trait variance, and combined in a genetic risk score (GRS) they explain up to 22% of the variance (4, 6, 7).

A key finding from a European GWAS of lipid levels is the discovery of a causal PCSK9 loss-of-function (LoF) variant (p.Arg46Leu; rs11591147) (8), which has led to the development of PCSK9-inhibitors to treat dyslipidemia (9). Besides this PCSK9 LoF variant, an independent signal in the PCSK9 region, driven by the rare PCSK9 p.Asn157Lys missense variant (rs143117125), has been reported in a cohort of more than 69,000 Norwegians (10).

Although the most recent genetic studies of circulating lipid levels include individuals of different ancestries, the majority are still of European ancestry (4, 5). Hence, additional genetic variants contributing to the variation in lipid levels could be identified by studying other ancestry groups. In this respect, the Greenlandic population could be of particular interest, because previous studies have shown that the genetic architecture of metabolic diseases in this population differs markedly from that in European populations (11, 12), and because an Arctic specific variant with a very high impact on LDL-cholesterol levels and CVD risk has been identified (13, 14). Moreover, the prevalence of high LDL cholesterol, as well as cardiometabolic diseases, in Greenland is high (15-18); In a recent population survey, more than 70% of the 45-54-year-old participants had LDL-cholesterol levels above 3 mmol/L, which is considered too high for individuals with no other CVD risk factors (18). Motivated by these observations, we aimed to investigate the genetics of triglycerides, LDL-cholesterol, HDL-cholesterol, total cholesterol, apolipoprotein A1 (apoA1), and apolipoprotein B (apoB) in Greenlanders. Furthermore, we aimed to describe the genetic architecture of these six lipid traits in Greenlanders compared to Europeans, and to assess if the lipid-associated genetic variants affected the risk of CVD.

Study population

We used data from three Greenlandic cohorts; The Inuit Health in Transition (IHIT, 2005-2010), comprising 3,115 individuals living in Greenland (19), the Greenland population study (B99) and BBH cohorts (1998-2001) comprising 1,401 individuals living in Greenland, and 503 Greenlanders living in Denmark, respectively (20). There was an overlap of individuals between IHIT and B99 (N = 295), and these were assigned to B99. By combining the three cohorts, we were able to include up to 4,473 individuals in the analyses.

Biochemical measurements

Blood samples were collected from fasting participants in IHIT (19) as well as the majority of participants of B99, and in fasting and non-fasting participants of BBH. Serum concentrations of total cholesterol, HDL-cholesterol, triglycerides, apoA1, and apoB were measured and concentrations of LDL-cholesterol were calculated. Triglyceride measurement and the calculation of LDL-cholesterol was only done for fasting participants (21). Participants also had anthropometric measures and blood pressure taken, as well as an oral glucose tolerance test for those aged ≥35 years in B99 and BBH and for those aged ≥18 years in IHIT, from whom blood samples were also taken two hours after the glucose ingestion to measure plasma glucose, serum insulin, and serum C-peptide, as previously described (19, 20).

Cardiovascular disease outcomes

We obtained information on ischaemic stroke, myocardial infarction, and revascularization procedures from The Greenland National Patient register, the Danish National Patient Register, and the causes of death register from both Greenland and Denmark. We used the International Statistical Classification of Diseases and Related Health Problems- (ICD-) 8, ICD10, and International Classification of Primary Care 2 (ICPC-2) codes, as well as The Classification of Operations and Treatments and The Nordic Medico-Statistical Committee (NOMESCO) Classification of Surgical Procedures for procedures (Supplemental Table 1 and 2). Follow-up data in registries were only available for IHIT and B99, resulting in the inclusion of 4,110 individuals in the analyses of CVD events.

Genotyping

All samples were genotyped on the Multi-Ethnic Global Array (MEGA chip, Illumina), with ~2M variants. Genotypes were called jointly for all cohorts using the GenCall module of the GenomeStudio software (Illumina). The data set underwent quality control, removing duplicate samples, individuals missing >5% genotypes, and variants with >1% missing genotypes. In total, data from 4,607 individuals and 1,769,375 variants passed the quality control.

Imputation

To create a reference panel, we obtained short paired-end whole genome sequencing data (Illumina) for 447 Greenlanders. Genotyping was done using GATK best practices with their build 38 resource bundle. The variants were lifted to hg19 using Picard (http://broadinstitute.github.io/picard/) and sites with >3 % missingness or mendelian discordance >5 % based on the large number of trios and duos present in the data set were removed. We phased the remaining data using ShapeIt2 (22), using available duo/trio information for some of the individuals. These phased data were then combined with 1000 genomes data for 99 CEU, 91 GBR, 103 CHB, 105 CHS, and 93 CDX. For the 1000 genomes data, we included sites that overlapped with the whole genome data or had a minor allele frequency (MAF) > 1 %.

For the 4,182 individuals without WGS data, we phased the MegaChip data with more than 5 minor alleles using ShapIt2 also using the duo/trio family information as well as the constructed reference panel. The imputation was performed using IMPUTE2 (23). The recombination map for the reference variants was inferred with linear interpolation using the hg19 genomic map from IMPUTE2 as a template, and an effective population size of 1,500 (24). The imputation generated genotype data with an info score (r²) above 0.8 for 12,270,968 variants. Of these we removed 16 tri-allelic variants on chromosome 12 and applied a MAF filter of 0.5 %, resulting in a data set of 10,265,398 variants with a mean info score of >0.98, which we included in the analyses.

Statistical analysis

Analyses of the Greenlandic data

To perform association analyses, we used a linear mixed model implemented in the software GEMMA (25) to account for the relatedness and population structure. For each phenotype, we included all individuals across the three cohorts with information about that phenotype, and the general relatedness matrix used as additional input in GEMMA was estimated from the same individuals using only SNPs with MAF above 5% and at most 1% missingness. We assumed an additive effect and included sex, age, cohort, and whether an individual had imputed or genotype data as covariates. Lipid levels were quantile transformed to a standard normal distribution within each sex before association analyses to obtain p-values and effect size estimates. In GWAS analyses, we applied a significance threshold of p<5x10^-8. QQ plots showed no indications of inflation due to confounding in analyses of the six lipid traits (Supplementary Figure 1).

We estimated the partial variance explained (PVE) (26) for the identified variants in the Greenlanders and for the variants reported in Europeans (27).

CVD risk was assessed with Cox regression on the number of years lived until the first CVD event (survival time) using the R-package survival (28). These analyses were adjusted for sex, age, and the top 10 principal components (PC) to correct for the population structure. Years until an event was calculated as the years from birth until an event or end of follow-up (2018).

Analyses of UK Biobank data

Variants of interest were analysed in the UK Biobank version 3 (29) (UK Biobank project ID: 32683). LDL-cholesterol (Field ID: 30780 ‘LDL direct’) and total cholesterol (Field ID: 30690 ‘Cholesterol’) were rank-based inverse normal transformed separately for each sex and analysed in a linear model using plink2 (30). CVD risk was assessed using Cox regression, defining CVD events by the diagnosis codes in supplementary tables 1 and 2 extracted using the field IDs: 41270 (ICD10), 41271 (ICD9), 41272 (OPCS), 20002 (self-reported), and 20004 (self-reported operations) and their related data fields for date of diagnosis/operation. All analyses were adjusted for birth year and month, sex, and the first 10 PCs, and conditioned on the PCSK9 loss-of-function variant (rs11591147) to obtain the independent effect estimates for the PCSK9 rs12117661 variant.

Genetic risk scores

Genetic risk scores (GRS) were calculated for each of the six lipid traits, by including variants identified by the GWAS (p<5x10^-8) and using the estimated effect size from the GWAS as weight. We did 10-fold cross-validation, dividing all the individuals randomly into 10 groups of equal size and estimating the GRS of the individuals in each group by using variants and effect sizes from a GWAS performed on the remaining 9 groups combined. Each of these GWAS was performed as the main lipid GWAS performed in this study. To avoid including multiple variants that associate due to LD, we sorted all the variants by p-value, kept the top variant, and discarded all other SNPs in a region +/- 5 Mb region around it. We repeated this procedure until there were no more variants left with p<5x10^-8.

We calculated the variance explained by each GRS by the partial r² of the GRS in a model adjusted for age, sex, and the first 10 PCs using quantile-transformed lipid levels.

In-silico variant analyses

The Ensembl database (31) and Encode 3 data (32) were queried to assess the co-localization of variants with regulatory elements, such as transcription factor binding sites, promoter regions, and regions of DNase hypersensitivity. Moreover, RNA expression data from 48 tissues (with >70 samples, range: 80-399) were queried through the GTEx Portal to assess possible effects of the genetic variants on the expression of nearby genes (33).

We identified 11 loci that were genome-wide significantly (p<5x10^-8) associated with at least one of the six traits (Figure 1A, Table 1) in up to 4,473 Greenlanders (Supplemental Table 3 shows participant characteristics). Lead variants from ten of the 11 loci were at least nominally associated with more than one of the lipid traits at p<0.05 (Figure 1B and Supplementary Table 4).

Table 1. Lead variants of the 11 genome-wide significant loci. Effect sizes (β_SD) and p-values were calculated based on quantile transformed trait values and shown for the lipid trait with the strongest association. Loci number 1 and 2 are potentially novel and the remaining loci have been associated with lipid levels previously (13, 37). Minor allele frequencies (MAF) are shown for Greenlanders (GR) and for Europeans (EU). The European frequencies were obtained for the non-Finnish Europeans from the GnomAD data set v.2 (https://gnomad.broadinstitute.org/). PVE, partial variance explained.

Locus	Variant	Chr:pos (effect allele)	Variant type (Nearby genes)	MAF (%) GR	MAF (%) EU	Lipid trait	N	β_SD (SE)	p-value	PVE (%)
1	rs78168195	2:228480653 (G)	Intron (C2orf83)	6.1	8.1	LDL-cholesterol	3809	0.27 (0.05)	7.9x10^-9	0.87
2	rs148015058	7:25766657 (T)	Intergenic (NFE2L3, NPVF)	3.9	14.4	Triglycerides	3970	0.35 (0.06)	2.6x10^-8	0.77
3	rs4927191	1:55491702 (C)	Intergenic (PCSK9)	16.4	24.3	LDL-cholesterol	3809	-0.23 (0.03)	1.2x10^-12	1.29
4	rs312029	2:21461730 (C)	Intergenic (TDRD15, APOB)	14.0	12.9	Total cholesterol	4341	-0.19 (0.03)	6.2x10^-10	0.87
5	rs2169387 (37)	8:9181395 (A)	Intergenic (PPP1R3B)	18.6	9.7	Total cholesterol	4342	-0.15 (0.03)	1.5x10^-8	0.74
6	rs287 (37)	8:19815556 (G)	Intron (LPL)	29.3	25.2	Triglycerides	3970	-0.16 (0.03)	3.3x10^-9	0.90
7	rs2519093	9:136141870 (T)	Intron (ABO)	25.6	19.8	LDL-cholesterol	3809	0.16 (0.03)	1.0x10^-9	0.95
8	rs66505542	11:116623213 (TA)	Intron (BUD13)	28.9	15.1	Triglycerides	3970	0.25 (0.03)	1.3x10^-22	2.46
9	rs8045855	16:57000696 (A)	Intron (CETP)	48.9	18.4	HDL-cholesterol	4473	-0.20 (0.02)	6.7x10^-20	1.81
10	rs730882082	19:11215991 (A)	Missense (LDLR)	15.8	0	LDL-cholesterol	3809	0.66 (0.03)	7.2x10^-99	10.90
11	rs429358	19:45411941 (C)	Missense (APOE)	21.3	14.9	LDL-cholesterol	3809	0.20 (0.03)	2.8x10^-12	1.26

Two of the 11 loci, on chromosome 2 (rs78168195) and chromosome 7 (rs148015058), were potentially novel (Table 1 and Supplementary Figure 2A-B). The locus on chromosome 2, was primarily associated with higher levels of LDL-cholesterol and the locus on chromosome 7, was associated with higher levels of triglycerides (Table 1 and Supplementary Table 4). Notably, the lead variants from each of these loci were either not significant or borderline significant, when considering that six traits were analysed. In addition, the two lead variants were more common in Europeans than in Greenlanders (Table 1), and were not identified in a much larger GWAS in Europeans (4). This makes them very unlikely to be true signals.

The lead variants from the remaining nine of the 11 loci were located in previously reported lipid loci (Table 1). However, in the locus on chromosome 1, near PCSK9, we identified a signal, which was not explained by previously reported lead variants. We explored this signal in more detail.

A possible causal variant near PCSK9

The minor allele of the lead variant (rs4927191) in the locus near PCSK9 on chromosome 1 was associated with lower levels of LDL-cholesterol (Table 1), total cholesterol (-0.17 (0.03), p=2.3x10^-9), and ApoB (-0.20 (0.05), p=1.5x10^-4), as well as higher levels of HDL-cholesterol (0.08 (0.03), p=0.0082) (Supplementary Table 4).

Importantly, these associations were independent of the well-established, causal PCSK9 LoF variant (rs11591147); conditioning on this LoF variant did not attenuate our signal for any of the lipid traits (Figure 2A-B) and the LoF variant had an estimated allele frequency of 0% in the Inuit ancestry and 0.6% in the Greenlandic population. Also, another coding PCSK9 variant (rs143117125), suggested to be causal in Europeans, was too rare to be found in our whole genome sequencing data of 447 individuals, and could therefore not explain the observed associations either.

Our PCSK9-association signal comprised several variants in high LD. Particularly eight variants separated from the rest of the association signal with p-values below 10^-9 for association with the primary trait, LDL-cholesterol (Figure 2A and Supplementary Table 5). These eight top variants were all intergenic and located between BSND and PCSK9 (Figure 2C). In the GTEx Portal, we could assess if six of the top variants were associated with PCSK9 expression across 48 tissues, rs200159426 and rs187607506 were not available. Interestingly, five of the top variants were the five variants most significantly associated with expression of PCSK9 across tissues in GTEx (Figure 2D, Supplementary Table 6), indicating that the causal variant could be one of these five variants. PCSK9 expression was significantly affected in the tibial nerve, visceral and subcutaneous adipose tissue, lung, whole blood, and the tibial artery (Supplemental Table 6). Next, we queried the Ensembl database and Encode 3 data, and found that rs12117661 was the only of the eight potential causal variants, which was located in a predicted regulatory region, namely a CTCF binding site (ENSR00000251959) in a region with open chromatin (Figure 2C).

The PCSK9 rs12117661 variant was the strongest candidate to be the causal variant and we, therefore, assessed associations with additional cardiometabolic traits (results for the other seven PCSK9 top variants are presented in Supplementary Table 5 and Supplementary Table 7). The minor allele of rs12117661 was associated with lower levels of LDL-cholesterol, total cholesterol, and ApoB, as well as higher levels of HDL-cholesterol (Figure 3A and Supplementary Table 8). We observed no associations between rs12117661 and measures of glucose homeostasis, blood pressure, or measures of body composition (Supplemental Table 8). In data from 3,716 Greenlandic individuals with a median time lived until first event (or end of follow-up) of 56.2 years, rs12117661 was not significantly associated with risk of ischaemic stroke, myocardial infarction, revascularization procedures, or a combined measure of CVD events (Figure 3D and Supplementary Table 7). However, the point estimates indicated a potential markedly lower risk of CVD events, hence, we tested for similar association results in the much larger European UK Biobank cohort. Here, the minor allele of rs12117661 had a frequency of 24.6% and was associated with lower levels of LDL-cholesterol and total cholesterol when conditioning on the PCSK9 LoF variant (Figure 3B, Supplementary Table 5), and with a lower risk of revascularization procedures and any CVD event (Figure 3D, Supplementary Table 7).

Impact of the identified variants

We also assessed the impact on CVD of the lead variants from the remaining 10 loci that were significantly associated with one of the six lipid traits in the initial GWAS analyses. The derived allele of the LDLR rs730882082 variant was associated with a higher risk of revascularization procedures and any CVD event (Supplementary Table 9), in line with what has been shown previously in this study population (14). Furthermore, the lead variant from the APOB association signal on chromosome 2 (rs312029) was nominally associated with a higher risk of revascularization procedures (Supplementary Table 9).

We also assessed how much of the trait variance was explained by the lead variant from each of the 11 loci, as well as the potential novel causal PCSK9 variant. This varied between 0.74% and 10.9% for the lead variants (Table 1), and the PCSK9 rs12117661 variant explained 1.25% of the LDL-cholesterol variance (Supplementary Table 5). Compared to variants identified in European GWAS of the same traits, our lead variants were relatively common and explained a larger proportion of the variance (Supplementary Figure 3), suggesting that the genetic architecture of these traits in Greenlanders comprises a lower number of variants of either high frequency, high effect, or both.

Finally, to further investigate the impact of the lead variants from the 11 loci, we combined them in a GRS for each of the six lipid traits. There was a correlation between the GRS for LDL-cholesterol, total cholesterol, and ApoB, as well as between the GRS for HDL-cholesterol and ApoA (Supplemental Figure 4). Interestingly, the GRS explained from 0.53% for ApoA1 to as much as 16.3% for LDL-cholesterol when adjusting for BMI, age, and PC1-10 (Figure 4, Supplementary Table 10) and from 0.73% for apoA1 to 13.7% for LDL-cholesterol when only adjusting for PC1-10 (Supplementary Table 10). For triglycerides, total cholesterol, LDL-cholesterol, and ApoB, the GRS explained more of the variance than did age, and for total cholesterol, LDL-cholesterol, and ApoB, the GRS also explained more than BMI (Figure 4, Supplementary Table 10). Notably, the LDLR rs730882082 variant explained far more of the variance in LDL-cholesterol, total cholesterol, and ApoB than the remaining variants in the GRS, namely 10.9%, 7.6%, and 6.18% (Supplementary Table 4).

In our GWAS analyses of six lipid traits in Greenlanders, we identified 11 distinct loci harbouring variants that were genome-wide significantly associated with at least one of these traits. Nine of the 11 signals mapped to loci previously reported to be associated with circulating lipid levels (4, 5). The remaining two loci were potentially novel, however, not convincing. Notably, we identified a potential novel causal variant (rs12117661) near PCSK9 on chromosome 1.

Importantly, the PCSK9 LoF variant (rs11591147), which is strongly associated with LDL-cholesterol in Europeans, and has been hypothesized to be the causal variant in this locus (34, 35) was absent from the Inuit ancestry and had a low MAF in the general Greenlandic population. Furthermore, the LoF variant was not associated with levels of circulating lipids in our Greenlandic study population, and conditioning on this variant did not attenuate the association signal for our lead variant in the PCSK9 signal. Therefore, our signal near PCSK9 was independent of the LoF variant. A lipid-associated PCSK9 signal independent of the LoF variant has previously been reported in Europeans, and it was suggested that the rare missense variant rs143117125 (p.N157K) was the causal variant in this signal (10). This variant was too rare to be detected in the Greenlandic whole genome sequencing data and could therefore not explain our association signal either.

We identified the rs12117661 variant as a potential causal variant in the PCSK9 association signal. This variant was among the variants showing the strongest association with lower levels of LDL-cholesterol and total cholesterol in the Greenlanders. Analyses of data from UK Biobank showed strong associations as well, even when conditioning on the PCSK9 LoF variant. Of note, the observed effect sizes for rs12117661 in Europeans were markedly lower than in the Greenlandic cohort. This could potentially indicate the presence of a variant in the Greenlanders, which could not be detected with the current high depth short read sequencing data.

In Greenlanders, rs12117661 also showed a non-significant association with a lower risk of peripheral artery disease, and in Europeans, it was very significantly associated with a lower risk of revascularization procedures and any CVD event, even when conditioning on the PCSK9 LoF variant. In silico analyses showed that the rs12117661 variant was located in a predicted transcription factor binding site, indicating that this variant might affect the transcript of nearby genes, including PCSK9. Moreover, rs12117661 was strongly associated with lower PCSK9 expression levels in multiple tissues in data from the GTEx portal (33). Lower PCSK9 expression is consistent with the observed association with lower levels of circulating LDL-cholesterol and total cholesterol, as lower PCSK9 activity will increase LDL-receptor recycling, and thereby reduce hepatic LDL-cholesterol uptake and degradation (36). Further studies of gene expression are needed to identify the specific tissues where the variant exerts its effect to cause lower lipid levels.

The PCSK9 rs12117661 variant had an effect size for LDL-cholesterol of -0.22 (0.03) SD and explained 1.3% of the variance in LDL-cholesterol levels in Greenland. Combined with an effect allele frequency of 15.6%, the impact of this variant is relatively large in the Greenlandic population. In comparison, the effect size of the PCSK9 LoF variant (rs11591147) on LDL-cholesterol of up to -0.53 SD in Europeans is larger (37, 38), but the LoF variant only has a frequency of 1-2% in Europeans (10, 37, 38), and therefore only explains 0.64% of the variance in LDL-cholesterol in Europeans (27). The population-level impact is therefore much larger for the identified PCSK9 rs12117661 variant in Greenlanders, than for the LoF variant in Europeans.

The combination of a relatively large effect size and high frequency makes the PCSK9 rs12117661 variant useful for assessing the long-term effects of using PCSK9 inhibitors to treat hypercholesterolemia. This is of great interest since most trials so far have only assessed the effects of the PCSK9 inhibitors with no more than 3 years of follow-up (39, 40). Assessing possible long-term effects is particularly relevant, as there are concerns about PCSK9 inhibitors exhibiting adverse neurocognitive effects as well as increasing the risk of type 2 diabetes and cancer (40-42). The link to type 2 diabetes is further supported by studies of the PCSK9 rs11591147 LoF variant (43). Contrary to these findings, we did not observe any indications of adverse glucose homeostasis in carriers of the rs12117661 variant in the Greenlandic data. This is consistent with findings from the UK Biobank (44) and the FinnGen study (http://r7.finngen.fi/), where we observed no association between the rs12117661 variant and type 2 diabetes. Moreover, in the same datasets, no association with any type of cancer was observed. Instead, we showed that the rs12117661 variant was associated with a beneficial effect on CVD outcomes in Europeans from the UK Biobank, and the variant was associated with lower risk of similar outcomes in the FinnGen study (http://r7.finngen.fi/). We were unable to show any significant effect of the rs12117661 variant on CVD outcomes in the Greenlanders, likely due to the low number of CVD events. Additional analyses in larger study populations are needed to address these adverse and beneficial effects of the variant further.

Genetic architecture of lipid traits in Greenlanders

We found that the Greenlandic population has a unique genetic architecture of circulating lipid levels comprising relatively common variants with large effect sizes explaining more variance than variants in Europeans (4, 27). Combined, the 11 genome-wide significant variants explained a large amount of the variance in Greenlanders, particularly for LDL-cholesterol with PVE of 16.3%. This was mainly driven by the LDLR rs730882082 variant, which has previously been found to have a very large effect on LDL- and total cholesterol in Greenlanders (14). In Europeans, up to 22% of the variance in LDL-cholesterol has been explained by genetic risk scores, but these scores comprised up to 2 million variants (4-7). Taken together, the genetic architecture of lipid traits in Greenlanders is similar to what has been reported for other metabolic traits, where few variants explain a large proportion of the variance (11, 12).

In conclusion, we identified 11 independent loci associated with levels of at least one of six lipid traits in Greenlanders, including an independent variant associated with LDL- and total cholesterol levels through regulation of PCSK9 expression. The suggested causal variant (rs12117661) in the PCSK9 locus was located in a predicted regulatory site and was strongly associated with lipid levels, expression of PCSK9 in Greenlanders, and with lower risk of certain CVD outcomes in Europeans. Our results revealed a unique genetic architecture of lipid traits among Greenlanders, where few variants explained a large amount of the trait variance.

Data Availability Statement

The data generated for this paper can be found within the published paper and its supplementary file.

The accession numbers for the genotype data are EGAD00010001427, EGAD00010001428, and EGAD00010002057 (https://ega-archive.org/datasets).

Acknowledgments

We gratefully acknowledge both the staff and the participants of the Greenlandic population health surveys. This research has been conducted using data from UK Biobank, a major biomedical database.

Author Contribution Statement

NKS, MKA, IM, TH, and AA were involved in the conceptualization, design, and conduct of the study. PB, CVLL, NG, MEJ, EZ, IM, TH, and AA curated the data and supervised the study. NKS, MKA, EJ, FFS, AKN, GGE, JM, CGS, RFB, and AG did the formal analyses and interpreted the results. NKS and MKA wrote the first draft of the manuscript, and all authors edited, reviewed, and approved the final version of the manuscript.

Ethical Approval

The study was approved by the Commission for Scientific Research in Greenland (project 2011–13, ref. no. 2011–056978; project 2013–13, ref.no. 2013–090702; and project 2012-16/17, ref.no. 2017-12997) and was conducted in accordance with the ethical standards of the Declaration of Helsinki, second revision. All participants gave informed consent.

Competing Interests

We have no competing interests to report.

World Health Organization. The top 10 causes of death 2018 [cited 2020 January 24]. Available from: https://www.who.int/news-room/fact-sheets/detail/the-top-10-causes-of-death.
Ciffone NA, Copple T. Managing dyslipidemia for CVD prevention: A review of recent clinical practice guidelines. The Nurse practitioner. 2019;44(1):8-16.
Chait A, Goldberg I. Treatment of Dyslipidemia in Diabetes: Recent Advances and Remaining Questions. Curr Diabetes Rep. 2017;17(11):10.
Graham SE, Clarke SL, Wu KH, Kanoni S, Zajac GJM, Ramdas S, et al. The power of genetic diversity in genome-wide association studies of lipids. Nature. 2021;600(7890):675-9.
Klarin D, Damrauer SM, Cho K, Sun YV, Teslovich TM, Honerlaw J, et al. Genetics of blood lipids among similar to 300,000 multi-ethnic participants of the Million Veteran Program. Nature Genet. 2018;50(11):1514-+.
Natarajan P, Peloso GM, Zekavat SM, Montasser M, Ganna A, Chaffin M, et al. Deep-coverage whole genome sequences and blood lipids among 16,324 individuals. Nat Commun. 2018;9(1):3391.
Wu H, Forgetta V, Zhou S, Bhatnagar SR, Paré G, Richards JB. Polygenic Risk Score for Low-Density Lipoprotein Cholesterol Is Associated With Risk of Ischemic Heart Disease and Enriches for Individuals With Familial Hypercholesterolemia. Circulation Genomic and precision medicine. 2021;14(1):e003106.
Cohen J, Pertsemlidis A, Kotowski IK, Graham R, Garcia CK, Hobbs HH. Low LDL cholesterol in individuals of African descent resulting from frequent nonsense mutations in PCSK9. Nature Genet. 2005;37(2):161-5.
Shapiro MD, Tavori H, Fazio S. PCSK9: From Basic Science Discoveries to Clinical Trials. Circulation research. 2018;122(10):1420-38.
Nielsen JB, Rom O, Surakka I, Graham SE, Zhou W, Roychowdhury T, et al. Loss-of-function genomic variants highlight potential therapeutic targets for cardiovascular disease. Nat Commun. 2020;11(1):6417.
Moltke I, Grarup N, Jorgensen ME, Bjerregaard P, Treebak JT, Fumagalli M, et al. A common Greenlandic TBC1D4 variant confers muscle insulin resistance and type 2 diabetes. Nature. 2014;512(7513):190-+.
Thuesen ACB, Stæger FF, Kaci A, Solheim MH, Aukrust I, Jørsboe E, et al. A novel splice-affecting HNF1A variant with large population impact on diabetes in Greenland. Lancet Reg Health Eur. 2023;24:100529.
Dube JB, Wang J, Cao HN, McIntyre AD, Johansen CT, Hopkins SE, et al. Common Low-Density Lipoprotein Receptor p.G116S Variant Has a Large Effect on Plasma Low-Density Lipoprotein Cholesterol in Circumpolar Inuit Populations. Circ-Cardiovasc Genet. 2015;8(1):100-U90.
Jørsboe E, Andersen MK, Skotte L, Stæger FF, Færgeman NJ, Hanghøj K, et al. An LDLR missense variant poses high risk of familial hypercholesterolemia in 30% of Greenlanders and offers potential of early cardiovascular disease intervention. HGG Adv. 2022;3(4):100118.
Bjerregaard P, Jørgensen ME. Prevalence of obesity among Inuit in Greenland and temporal trend by social position. Am J Hum Biol. 2013;25(3):335-40.
Bjerregaard P, Young TK, Hegele RA. Low incidence of cardiovascular disease among the Inuit - what is the evidence? Atherosclerosis. 2003;166(2):351-7.
Jørgensen ME, Borch-Johnsen K, Witte DR, Bjerregaard P. Diabetes in Greenland and its relationship with urbanization. Diabetic Med. 2012;29(6):755-60.
Larsen CVL, Hansen CB, Ingemann C, Jørgensen ME, Olesen I, Sørensen IK, et al. Befolkningsundersøgelsen i Grønland 2018. Levevilkår, livsstil og helbred. [Population Health Survey in Greenland 2018. Living conditions, lifestyle, and health. In Danish]. Statens Institut for Folkesundhed, Studiestræde 6, 1455 København K: Statens Institut for Folkesundhed, SDU; 2019.
Bjerregaard P. Inuit Health in Transition – Greenland survey 2005-2010. Population sample and survey methods. SIF writings on Greenland [Internet]. 2011 May 07, 2019 [cited 2019 May 07]; 19. Available from: https://www.sdu.dk/da/sif/rapporter/2011/inuit_health_in_transition.
Bjerregaard P, Curtis T, Borch-Johnsen K, Mulvad G, Becker U, Andersen S, et al. Inuit health in Greenland: a population survey of life style and disease in Greenland and among Inuit living in Denmark. Int J Circumpolar Health. 2003;62 Suppl 1:3-79.
Bjerregaard P, Jorgensen ME, Borch-Johnsen K, Greenland Population S. Serum lipids of Greenland Inuit in relation to Inuit genetic heritage, westernisation and migration. Atherosclerosis. 2004;174(2):391-8.
Delaneau O, Zagury J-F. Data Production and Analysis in Population Genomics: Human Press; 2012.
Howie BN, Donnelly P, Marchini J. A Flexible and Accurate Genotype Imputation Method for the Next Generation of Genome-Wide Association Studies. PLoS Genet. 2009;5(6):15.
Fumagalli M, Moltke I, Grarup N, Racimo F, Bjerregaard P, Jorgensen ME, et al. Greenlandic Inuit show genetic signatures of diet and climate adaptation. Science. 2015;349(6254):1343-7.
Zhou X, Stephens M. Genome-wide efficient mixed-model analysis for association studies. Nature Genet. 2012;44(7):821-U136.
Shim H, Chasman DI, Smith JD, Mora S, Ridker PM, Nickerson DA, et al. A Multivariate Genome-Wide Association Analysis of 10 LDL Subfractions, and Their Response to Statin Treatment, in 1868 Caucasians. PLoS One. 2015;10(4):20.
Liu DJ, Peloso GM, Yu H, Butterworth AS, Wang X, Mahajan A, et al. Exome-wide association study of plasma lipids in > 300,000 individuals. Nature Genet. 2017;49(12):1758-+.
Therneau T. A Package for Survival Analysis in S. Version 2.41.3. 2015.
Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, et al. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018;562(7726):203-9.
Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. Second-generation PLINK: rising to the challenge of larger and richer datasets. GigaScience. 2015;4(1).
Cunningham F, Achuthan P, Akanni W, Allen J, Amode M R, Armean IM, et al. Ensembl 2019. Nucleic Acids Res. 2018;47(D1):D745-D51.
ENCODE Project Consortium. An integrated encyclopedia of DNA elements in the human genome. Nature. 2012;489(7414):57-74.
GTEx Consortium. The Genotype-Tissue Expression (GTEx) pilot analysis: Multitissue gene regulation in humans. Science. 2015;348(6235):648-60.
Benn M, Nordestgaard BG, Grande P, Schnohr P, Tybjaerg-Hansen A. PCSK9 R46L, Low-Density Lipoprotein Cholesterol Levels, and Risk of Ischemic Heart Disease 3 Independent Studies and Meta-Analyses. J Am Coll Cardiol. 2010;55(25):2833-42.
Kotowski IK, Pertsemlidis A, Luke A, Cooper RS, Vega GL, Cohen JC, et al. A spectrum of PCSK9 Alleles contributes to plasma levels of low-density lipoprotein cholesterol. Am J Hum Genet. 2006;78(3):410-22.
Horton JD, Cohen JC, Hobbs HH. PCSK9: a convertase that coordinates LDL catabolism. J Lipid Res. 2009;50:S172-S7.
Surakka I, Horikoshi M, Magi R, Sarin AP, Mahajan A, Lagou V, et al. The impact of low-frequency and rare variants on lipid levels. Nature Genet. 2015;47(6):589-97.
The Neale Lab. UKBiobank GWAS results [Available from: http://www.nealelab.is/uk-biobank/.
AlTurki A, Marafi M, Dawas A, Dube MP, Vieira L, Sherman MH, et al. Meta-analysis of Randomized Controlled Trials Assessing the Impact of Proprotein Convertase Subtilisin/Kexin Type 9 Antibodies on Mortality and Cardiovascular Outcomes. Am J Cardiol. 2019;124(12):1869-75.
Schmidt AF, Pearce LS, Wilkins JT, Overington JP, Hingorani AD, Casas JP. PCSK9 monoclonal antibodies for the primary and secondary prevention of cardiovascular disease. Cochrane Database of Systematic Reviews. 2017(4):111.
Robinson JG, Farnier M, Krempf M, Bergeron J, Luc G, Averna M, et al. Efficacy and Safety of Alirocumab in Reducing Lipids and Cardiovascular Events. N Engl J Med. 2015;372(16):1489-99.
Sabatine MS, Giugliano RP, Wiviott SD, Raal FJ, Blom DJ, Robinson J, et al. Efficacy and Safety of Evolocumab in Reducing Lipids and Cardiovascular Events. N Engl J Med. 2015;372(16):1500-9.
Nelson CP, Lai FY, Nath M, Ye S, Webb TR, Schunkert H, et al. Genetic Assessment of Potential Long-Term On-Target Side Effects of PCSK9 (Proprotein Convertase Subtilisin/Kexin Type 9) Inhibitors. Circ-Genom Precis Med. 2019;12(1):9.
UKBiobank PheWeb [Internet]. Available from: https://pheweb.org/UKB-TOPMed/.

There is no duality of interest

Supplementalmaterial.docx

Genome-wide association study of lipids in Greenlandic Inuit shows independent PCSK9 association and reveals a unique genetic architecture

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Materials And Methods

Results

Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1