Genetics of Lipid Disorders

Abstract

Purpose of review

In this review, we will highlight recent advances in identifying genes and gene regions responsible for the variation in serum lipid levels. We will also consider the next directions for research based on these advances.

Recent findings

Large-scale genome-wide association studies have successfully screened common variants across the genome for association with serum lipids and have generated novel hypotheses about the causes of serum lipid variation.

Summary

Deep sequencing of GWA signals promises to expand the catalog of variants responsible for serum lipid variation and with a full catalog of variants, we may develop a panel of polymorphisms with clinical utility. In parallel, functional exploration of the GWA signals should expand our knowledge of lipoprotein metabolism and generate targets for pharmacologic intervention.

Introduction

Disorders of lipoprotein metabolism lead to atherosclerosis and cardiovascular disease (CVD), including myocardial infarction and stroke. Sixteen million adults in the United States alone live with coronary artery disease, which kills 450,000 people annually [1]. Because of the well-established relationship between serum lipid levels and CVD, serum lipids have long been of clinical interest. In particular, LDL-C is recognized not only as a biomarker of CVD risk, but also as a causal participant in the disease process.

Serum lipid levels including LDL-C, high-density lipoprotein cholesterol (HDL-C) and triglycerides (TG) are highly heritable: studies consistently estimate that over 50% of the total inter-individual variation in serum lipid levels can be explained by the genetic variation [2]. Therefore, there is substantial interest in discovering the genetic determinants of lipid levels, particularly genes that alter both serum lipids and CVD risk.

Mendelian lipoprotein disorders

Most of the well-characterized disorders of lipoprotein metabolism are monogenic, familial disorders with extreme phenotypes amenable to linkage analysis. For example, proprotein convertase subtilisin/kexin type 9 (PCKS9) has recently been implicated in a form of autosomal-dominant hypercholesterolemia (ADH). Its protein product normally associates with hepatic and extrahepatic LDL receptors and appears to target them for endosomal degradation [3]. Gain-of-function variants (such as D374Y) lead to a 10-fold reduction in LDL receptor levels, causing increased LDL-C levels and resulting in ADH [4,5].

Various Mendelian disorders of lipoprotein metabolism have now been described for each of the three lipoprotein traits, and up to now they have contributed the bulk of our knowledge about lipoprotein genetics [6]. Although those who harbor such traits experience profound phenotypes, the population-wide impact of the known Mendelian disorders is attenuated by their rarity. The monogenic lipid disorders discovered thus far can explain little of overall lipoprotein heritability. Indeed, the normal distribution of serum lipoprotein levels in the population suggests a polygenic model of inheritance: LDL-C, HDL-C, and TG levels are determined by the additive contribution of multiple loci.

Common variants

In contrast to the linkage studies that were well suited to identifying rare variants with profound effects, genome-wide association (GWA) studies have been designed to ascertain whether common genetic variants contribute to population-wide lipid variability. In such studies, hundreds of thousands of single-nucleotide polymorphisms (SNPs) are interrogated for association with serum lipid levels in large cohorts. GWA has become a viable technique for studying lipoprotein genetics on a large scale because serum lipids are accurately measured, lipid phenotypes are readily available in many cohorts, and genotyping hundreds of thousands of polymorphisms has become relatively inexpensive.

Each GWA study has shared similar approaches. Hundreds of thousands of SNPs are genotyped in a discovery cohort, with follow-up genotyping performed on a subset of those SNPs in a replication cohort. With linear regression, each allele is correlated to the serum level of each lipoprotein trait. Only genetic loci passing genome-wide significance (P < 5 × 10⁻⁸) are reported; the success of this stringent criterion in minimizing false associations is underscored by the fact that, in each successively larger GWA study, no previously-significant loci have dropped from significance.

The first large-scale lipids GWA study, published in 2007, involved 2,800 individuals genotyped at nearly 400,000 SNPs [7]. This study re-discovered associations known from prior genetic studies: that of HDL-C with CETP, and of LDL-C with APOE, which effectively served as positive controls that confirmed GWA as a valid approach to identifying lipid-associated loci. Progressively larger studies have been conducted; the most recent lipids GWA studies from 2009 involved up to 40,000 individuals of European descent and identified more than 30 chromosomal loci with common variants associated with lipid levels [8**,9**]. An in-progress study by the Global Lipids Genetics Consortium involving over 100,000 individuals will likely identify still more loci.

The lipids GWA studies have led to the observation that genes with rare Mendelian mutations of large effect size also harbor common variants of more modest effect sizes [8**,9**] (Table 1). These findings prompt the reciprocal hypothesis: many lipid-associated loci mapped using common variants will also harbor novel low-frequency (1% to 5% frequency) and rare (<1% frequency) variants with more pronounced effects.

Table 1.

Loci associated with common SNPs in a 40,000-individual GWA study (adapted from [8**])

Trait	Chr.	SNP	P value	Locus	Associated Mendelian Disorders
LDL	1p13	rs12740374	2 × 10−42	CELSR2, PSRC1, SORT1
LDL	2p24	rs515135	5 × 10−29	APOB	Autosomal dominant hypercholesterolemia
LDL	19q13	rs4420638	4 × 10−27	APOE-APOC1- APOC4- APOC2	Type IB hyperlipoproteinemia (APOC2); Type III hyperlipoproteinemia (APOE)
LDL	19p13	rs6511720	2 × 10−26	LDLR	Autosomal dominant hypercholesterolemia
LDL	2p21	rs6544713	2 × 10−20	ABCG8	Sitosterolemia
LDL	5q13	rs3846663	8 × 10−12	HMGCR
LDL	5q23	rs1501908	1 × 10−11	TIMD4-HAVCR1
LDL	20q12	rs6102059	4 × 10−9	MAFB
LDL	12q24	rs2650000	2 × 10−8	HNF1A
LDL	19p13	rs10401969	2 × 10−8	NCAN, CILP2, PBX4
LDL	1p32	rs11206510	4 × 10−8	PCSK9	Autosomal dominant hypercholesterolemia
HDL	16q13	rs173539	4 × 10−75	CETP	Cholesteryl ester transfer protein deficiency
HDL	8p21	rs12678919	2 × 10−34	LPL
HDL	15q22	rs10468017	8 × 10−23	LIPC
HDL	18q21	rs4939883	7 × 10−15	LIPG
HDL	16q22	rs2271293	9 × 10−13	LCAT	LCAT deficiency (fish-eye disease)
HDL	11q23	rs964184	1 × 10−12	APOA1-APOC3- APOA4- APOA5	Primary hypoalphalipoproteinemia (APOA1)
HDL	11q12	rs174547	2 × 10−12	FADS1-FADS2- FADS3
HDL	12q24	rs2338104	1 × 10−10	MMAB, MVK
HDL	9p22	rs471364	3 × 10−10	TTC39B
HDL	20q13	rs1800961	8 × 10−10	HNF4A
HDL	9q31	rs1883025	1 × 10−9	ABCA1	Tangier disease
HDL	20q13	rs7679	4 × 10−9	PLTP
HDL	19p13	rs2967605	1 × 10−8	ANGPTL4
HDL	1q42	rs4846914	4 × 10−8	GALNT2
TG	11q23	rs964184	4 × 10−62	APOA1-APOC3-APOA4- APOA5
TG	8p21	rs12678919	2 × 10−41	LPL	Type IA hyperlipoproteinemia
TG	2p23	rs1260326	2 × 10−31	GCKR
TG	8q24	rs2954029	3 × 10−19	TRIB1
TG	7q11	rs714052	3 × 10−15	MLXIPL
TG	11q12	rs174547	2 × 10−14	FADS1-FADS2- FADS3
TG	2p24	rs7557067	9 × 10−12	APOB
TG	19p13	rs17216525	4 × 10−11	NCAN, CILP2, PBX4
TG	20q13	rs7679	7 × 10−11	PLTP
TG	8p23	rs7819412	3 × 10−8	XKR6-AMAC1L2

Already, 30 common lipid-associated variants have explained 9.3%, 7.7%, and 7.4% of inter-individual variability for HDL-C, LDL-C, and TG, respectively [8**]. These variants, therefore, explain ~15–18% of heritability – a proportion that likely will increase with publication of the 100,000-individual lipids GWA study. Much of the remaining heritability may ultimately be attributable to low-frequency and rare variants.

Next steps after genetic mapping using common variants

GWA has provided a flood of common genetic variants that are strongly associated with lipoprotein levels and have begun to explain substantial fractions of heritability. The tasks now at hand are to narrow the scope of localization from the chromosomal region to the level of the functional element (gene or regulatory element) and to explain the molecular mechanisms that ultimately lead to the observed lipid phenotypes. DNA sequencing and functional biology will be instrumental to these ends.

Sequencing

Sequencing offers the promise of identifying the variants driving the GWA signals. Because of the underlying patterns of linkage disequilibrium, a GWA SNP may only narrow the search to a several-hundred-kilobase region. Targeted, large-scale sequencing in unrelated individuals should ascertain all of the variants in the region and, therefore, to reduce the scope of the functional search down to a focused assortment of dozens of specific variants, one or more of which is likely to have functional consequences.

In some cases, sequencing may even identify smoking-gun variants, such as early truncations, that strongly indicate causation. Studies of PCSK9 are suggestive in this regard. The gene harbors a common variant (19% minor allele frequency) identified in a 40,000-person lipids GWA study [10]. PCSK9 has also been shown to contain low-frequency nonsense variants that lead to a 15–28 percent reduction in LDL-C and a 47–88 percent decrease in lifetime risk of CVD [11]. Additionally, the rare gain of function variants, discussed above, implicate PCSK9 in a form of ADH. If this pattern of common, low-frequency, and rare variants being present at a lipid locus is typical, then sequencing near GWA loci may not only identify the source of the GWA hit, but also a trove of causal low-frequency and rare variants. Nevertheless, even variants that appear likely to be causal (truncations, nonconservative substitutions) will require functional validation.

As the efficiency of sequencing increases and the cost falls, targeted sequencing will give way to whole-exome sequencing, and, in turn, to whole-genome sequencing – far exceeding the association boundaries of the regions initially highlighted by GWA studies. Such enhancement of our catalogue of lipid variants by sequencing may also prove fruitful for improved risk prediction and for the identification of a set of genes enriched for plausible pharmacological targets.

Functional validation

GWA and deep resequencing are highly informative, but fall short of elucidating a functional link between DNA sequence variant and lipoprotein phenotype. To provide convincing evidence of causality, functional studies will be required. Many GWA-identified loci are in linkage disequilibrium blocks spanning one or more genes; such loci immediately suggest protein targets to interrogate, whether through animal models, molecular biology, or emerging techniques. More difficult to study will be the loci that harbor no known genes, though many will ultimately be discovered to harbor regulatory elements important for modulating the expression of genes in lipoprotein metabolic pathways.

Ultimately, the fulfillment of a genetic Koch’s postulates would provide the most convincing evidence of causality, requiring: (a) identification of the SNP that, presumptively, causes variation in a lipid trait; (b) development of a valid model, such as that from induced pluripotent stem cells (iPS) from individuals with and without the variant; (c) highly targeted manipulation of the DNA to replace the common allele with the variant; and (d) observation of a change in phenotype consistent with the proposed directionality and degree of variation (i.e., substituting the variant allele for the common allele and observing a phenotype consistent with that found in those natively harboring the variant allele). Such an approach would be particularly useful in confirming the causality of variants that appear to reside in intergenic “deserts.”

Clinical applications

In parallel to the work being done on elucidating the biology underlying the GWA signals, efforts are underway to take advantage of the presently available data for clinical risk prediction and personalized pharmacology.

Genetic risk stratification

There are advantageous theoretical properties to a genetic screen for plasma lipoproteins. Genetic scores are immutable within an individual, which may reduce costs due to decreased repeat testing, and allow for very early detection of individuals at increased risk for CVD. Such scores may better represent the cumulative lifetime burden of exposure to lipids than point-estimates of fasting lipid levels [12].

Because CVD is the ultimate outcome of interest, not lipoprotein levels per se, genetic screens may be particularly helpful if not all causes of lipid variation confer equal CVD risk (for example, if certain causes of increased HDL-C are protective, whereas others are not). Such a possibility was brought to the fore by the failure of torcetrapib, the CETP inhibitor, to reduce cardiovascular mortality or intermediate endpoints, despite increasing HDL (though this result should not be overinterpreted, as torcetrapib has molecule-specific off-target effects that trigger mineralocorticoid excess) [13–15]. On the other hand, persons with null mutations in APOC3 experience increased HDL-C, reduced TG, and a reduction in coronary calcification [16*]. This suggests that not all causes of elevated HDL-C carry the same consequences for CVD. Specific variants contributing to the observed serum lipoprotein levels may convey substantial information beyond that encoded by the scalar lipid levels themselves.

Nevertheless, genetic scores have not proven superior to lipoprotein levels and family history for CVD risk discrimination. For example, the use of one risk allele in the well-known 9p21 locus in white women did not improve discrimination or reclassification beyond that achievable using serum lipids, C-reactive protein (CRP), and family history [17*]. In contrast, we have demonstrated a lipid genetic panel using SNPs identified in the pre-GWA era; the panel did not improve risk discrimination but did improve risk reclassification [12]. As more variants are identified through GWA studies and sequencing, expanded panels will be developed that may achieve the goal of improved discrimination. At that point, the clinical utility of such panel should be examined in the context of a clinical trial.

Pharmacogenetics

Whereas genetic risk profiling is used to rule-in those who would ordinarily not be considered for therapy, pharmacogenetics may be used to exclude patients from what would otherwise be the default treatment. Here we will split the subject of lipid pharmacogenetics into two topics: efficacy-oriented pharmacogenetics and toxicity-oriented pharmacogenetics.

Trials have now demonstrated proof-of-principle for efficacy-oriented pharmacogenetic approaches to predicting statin response [18*,19*]. For example, the PROVE-IT TIMI 22 trial examined the effect of SNPs from genes in LDL-C metabolism pathways and statin pharmacokinetics pathways on the response to atorvastatin or pravastatin [18*]. Carriers with apolipoprotein E (APOE) isoform ε2 demonstrated greater reduction in LDL-C did than ε4 carriers, and were more likely to reach target LDL-C levels [18*]. In this study, APOE isoforms explained 3.8% of the residual variance in response to statin therapy. Nevertheless, statin therapy can already be titrated based on lipid levels at follow-up. Perhaps a pharmacogenetic panel could be built to address whether a failure to reduce LDL-C despite statin therapy was due to nonadherence or, instead, to genetically-encoded resistance, but the clinical utility of such a tool is unclear. Efficacy-oriented pharmacogenetic approaches are unlikely to be clinically compelling in the absence of a trial that demonstrates that an interaction between statin choice and genotype can lead to differential outcomes, even after controlling for lipoprotein levels.

Toxicity-oriented pharmacogenetic approaches may be of clinical use if the consequences of an adverse event are severe – e.g., cause harm per se or, indirectly, by inducing reluctance to try any drug in the same class, even one with reduced likelihood of causing a future adverse event. Clinical trials have now examined the association between genotype and statin-induced myopathy. In the SEARCH trial, a SLCO1B1 allele (SLCO1B1*5) raised the risk for myopathy in an allelic dose-dependent fashion (OR = 4.5 for CT heterozygotes vs TT homozygotes, and OR = 16.9 for CC homozygotes vs TT homozygotes) [20**]. Roughly 60% of the population risk of myopathy from taking 80mg simvastatin daily was attributed to this allele; nevertheless, the absolute risk of simvastatin-induced myopathy was modest (0.6% in TT homozygotes, 3% in CT heterozygotes, and 18% in CC homozygotes). Another group found similar relative effect sizes of allelic dose on risk of myopathy in simvastatin users, but not in those taking pravastatin [21*]. Therefore, knowledge of such allelic variants may help guide initial statin therapy in order to reduce the risk of harm and to increase the likelihood of adherence.

Conclusion

GWA studies have successfully screened common variants across the genome for association with serum lipids and have generated novel hypotheses about the causes of lipid variation. Deep sequencing of GWA signals promises to expand the catalog of variants responsible for serum lipid variation and with a full catalog of variants, we may develop a panel of polymorphisms with clinical utility. In parallel, functional exploration of the GWA signals should expand our knowledge of lipoprotein metabolism and generate targets for pharmacologic intervention.