Pharmacogenomics of high-density lipoprotein-cholesterol-raising therapies

Abstract

High levels of HDL cholesterol (HDL-C) have traditionally been linked to lower incidence of cardiovascular disease, prompting the search for effective and safe HDL-C raising pharmaceutical agents. Although drugs such as niacin and fibrates represent established therapeutic approaches, HDL-C response to such therapies is variable and heritable, suggesting a role for pharmacogenomic determinants. Multiple genetic polymorphisms, located primarily in genes encoding lipoproteins, cholesteryl ester transfer protein, transporters and CYP450 genes have been shown to associate with HDL-C drug response in vitro and in epidemiologic studies. However, few of the pharmacogenomic findings have been independently validated, precluding the development of clinical tools that can be used to predict HDL-C response and leaving the goal of personalized medicine to future efforts.

Numerous epidemiologic studies have documented the inverse association between HDL cholesterol (HDL-C) and cardiovascular risk [1]. Notably, that paradigm was recently challenged by a large-scale Mendelian randomization analysis [2] that failed to find evidence of a causal link between HDL-C levels and the risk of myocardial infarction. However, because of decades of conflicting evidence and concerns surrounding the validity of the Mendelian randomization approach, current clinical practice continues advocating for increasing HDL-C levels, prompting the search for safe and effective interventions [1].

Although several classes of HDL-raising therapies exist, individual response to interventions aimed at increasing HDL-C is highly variable. In addition, the magnitude of change in HDL-C levels resulting from an intervention is moderately heritable, suggesting a role for genetic determinants. Data from the Health, Risk Factors, Exercise Training, and Genetics Family Study show that HDL-C response to lifestyle interventions such as exercise training clusters in families with heritability estimates ranging from 0.25 to 0.64 [3]. For pharmacological agents, heritability of the HDL-C response has not been explicitly estimated, but polymorphisms in several genes have been shown to contribute to the variability of treatment success [4]. This article provides a comprehensive review of the genetic mediators of HDL-C response to currently prescribed therapies, and delineates ways in which pharmacogenomics could inform future treatment approaches.

Dynamics of HDL metabolism

HDL participates in the reverse flux of cholesterol from peripheral tissues to the liver for eventual excretion from the body (Figure 1). The first steps of the process involve the addition of phospholipids to lipid-poor apolipoprotein A-I (apoA-I), synthesized in the liver and intestine, to form the HDL precursor. Addition of free cholesterol from peripheral cells to the HDL precursor, mediated by ABCA1 and LCAT forms mature HDL. Mature HDL particles are heterogeneous in size, density, composition and atherogenicity, and can be classified into two HDL₂ subfractions (HDL_2a and HDL_2b) and three HDL₃ subfractions (HDL_3a, HDL_3b and HDL_3c) [5,6]. Modification of mature HDL is multifaceted and involves the addition of lipids via ABCG1 and SR-BI1. CETP exchanges cholesteryl esters acquired from HDL for triglycerides with VLDL or LDL cholesterol, effecting depletion in cholesteryl esters and enrichment in triglycerides. The resulting HDL 3 fraction undergoes continuous conversion into HDL₂ and vice versa depending on the metabolic context [7]. The particles are then either taken up by the liver via SR-BI or modified by hepatic lipase and endothelial lipase. Metabolism by the latter releases lipid-poor apoA-I (LpA-I), completing the cycle. Most HDL-C directed interventions target these complex pathways, making several biological considerations germane to the pharmacogenomics of HDL-C raising therapies.

Figure 1.

HDL metabolism with known and possible targets of therapeutic intervention.

Black arrows indicate routes of transport and/or transformation. Colored lines indicate targets of HDL-directed pharmacotherapy.

apoB100: Apolipoprotein B100; EL: Endothelial lipase; FFA: Free fatty acid; HL: Hepatic lipase; LCAT: Lecithan-cholesterol acyltransferase; LDLR: LDL receptor; SR-B1: Scavenger receptor class B, type I.

It is important to note that HDL₂ and HDL₃ subfractions vary in their association with cardiovascular disease risk, and most notably, their response to HDL-C-raising interventions [3,8,9]. In addition to classifying subfractions by size and density, HDL particles can be distinguished by composition: LpA-I particles, which only contain apoA-I; and LpA-I:A-II particles, which contain both apoA-I and apoA-II [8]. However, it is still unclear if the most promising approach to improving lipid metabolism involves targeting the overall levels of HDL-C, the interconversion of the subfractions or HDL particle size [10]. Therefore, the complex and dynamic nature of HDL metabolism requires a nuanced approach to evaluating potential genetic predictors of therapeutic efficacy.

Genetic modifiers of HDL-C response to drug treatment

Several classes of drugs have been evaluated in clinical trials as means to raise HDL-C levels. Currently, the major pharmaceutical agents include niacin (nicotinic acid and vitamin B3), fibrates, statins and different forms of combination therapies. In addition, several other drugs prescribed for conditions other than dyslipidemia (e.g., thiazolidinedions) affect HDL-C levels, but their pharmacogenomic determinants have not been explicitly evaluated for lipid outcomes; therefore, the authors have excluded them from this review. Other approaches - that is, apoA1 mimetics, CETP inhibitors or recombinant HDL infusion have been proposed and are subjects of ongoing trials discussed in the following sections [10]. The evidence in support of genetic modifiers of drug efficacy and toxicity varies widely by class and is summarized in Table 1.

Table 1.

Summary of plausible genetic modifiers of HDL-cholesterol raising therapies.

Gene (SNPS if identified)	OMIM number	Lines of evidence	Ref
Niacin
ABCA1	600046	Cell experiments	[17,18]
CETP	118470	Animal models	[21]
HCAR2/GPR109A/HM74A	609163	Animal models	[13]
NR1H3/LXRA	602423	Cell experiments	[18]
PPARG	601487	Cell experiments	[18]
Fibrates
ABCA1	600046	Candidate gene study	[35]
APOA5 (rs3135506)	606368	Candidate gene study Cell experiments	[29–31] [28]
APOE (rs429358, rs7412)	107741	Candidate gene study	[32–34]
CETP	118470	Animal models Gene expression study	[24] [26]
CYP7A1	118455	Candidate gene study	[39]
PPARA (rs1800206)	170998	Candidate gene study	[33,36–38]
LPL (rs28935469)	609708	Candidate gene study	[33,40,41]
Statins
APOA1 (rs670)	107680	Candidate gene study	[50,56,57]
APOB	107730	Candidate gene study	[58]
CETP (rs708272)	118470	Candidate gene study	[50,63–66]
CYP3A4 (rs4986907,	124010	Candidate gene study	[50–54]
rs2740574, rs2242480)			[50,55]
CYP7A1 (rs8192871, rs3808607)	118455		[58] [60–62]
LPL	609708	Candidate gene study
SLCO1B1 (rs4149015, rs4149056)	604843	Candidate gene study Candidate gene study
CETP Inhibitors
ABCA1	600046	Cell experiments	[72]
ABCG1	603076	Cell experiments	[72]
APOE	107741	Cell experiments	[72]
CETP	118470	Animal models	[71]
LCAT	609967	Cell experiments	[72]

CETP, cholesterylester transfer protein; HDL-C, high-density lipoprotein cholesterol; LCAT, lecithin-cholesterol acyltransferase; LPL, lipoprotein lipase.

Niacin

To date, niacin represents the most effective pharmaceutical approach to treat patients exhibiting low HDL-C levels [11,12]. Niacin improves the overall lipid profile primarily by activating GPR109A, the G-protein-coupled receptor expressed on adipocytes that reduces the release of free fatty acids to the liver leadsing to reduced triglyceride synthesis and circulating VLDL [13]. In turn, the reduction in VLDL-C inhibits CETP activity, resulting in increased HDL-C [14]. In addition to treating dyslipidemia, the activation of GPR109A is also responsible for the most common side effect of niacin therapy: the cutaneous vasodilatory response, or flushing [14]. However, niacin has also been shown to raise circulating HDL-C levels through several non-GPR109A dependent mechanisms. In hepatocytes, niacin downregulates cell surface expression of [beta]-chain adenosine triphosphate synthase, an enzyme that mediates endocytosis of HDL particles, which inhibits HDL liver uptake and catabolism [15]. The anticatabolic effect of niacin is specific to LpA-I particles [16] and enhances the reverse cholesterol transport function. Furthermore, it facilitates the conversion of smaller, denser HDL₃ particles to the larger, less dense HDL₂ particles as the smaller particles are allowed time to grow via uptake of peripheral tissue cholesterol [15]. Finally, niacin stimulates cholesterol efflux through activation of the PPAR[gamma]-liver X receptor (LXR)[alpha]-ABCA1 pathway, which further increases apoA-I lipidation and HDL biogenesis in the liver [17,18].

Although there is evidence demonstrating that response to niacin, in particular the side effect of flushing, clusters in families [19], no pharmacogenomic determinants of niacin efficacy or toxicity have been evaluated in humans [20]. However, evidence from animal and cell culture studies points to several plausible candidate genes. The mouse homolog of CETP has been identified as a potential mediator of niacin-induced HDL-C increase [21]. Based on the current understanding of niacin pharmacodynamics, it is possible that variation at the CETP locus affects the HDL-C response independently or via epistatic interactions with HCAR2, the gene encoding the GPR109A receptor in humans. In addition to considering genetic variation in HCAR2 itself, novel pharmacogenomic targets could also emerge from the evidence implicating PPAR[gamma] (encoded in humans by PPARG), LXR[alpha] (encoded by NR1H3) and ATP-binding cassette transporters, encoded by genes such as ABCA1, in niacin-mediated HDL efflux [17,18]. Overall, several candidate genes merit a closer evaluation as potential mediators of niacin response in humans, presenting fruitful opportunities for future studies.

Fibrates

In contrast to niacin, several of the pharmacogenomic determinants of HDL-C response to fibrates have been identified and characterized. Fibrates activate the intrahepatic nuclear receptor (PPAR[alpha]) by forming a heterodimer with retinoid X receptor, which in turn, binds to the peroxisome proliferator response element, thus modulating protein transcription of key drug target genes. Activation of PPAR ultimately increases plasma lipoprotein lipase (LPL) activity and raises A-I and A-II levels by stimulating the promoter region of APOA1 while reducing apoCIII, resulting in the observed changes in lipid fractions [22]. Similar to niacin, fibrates also increase the transcription of ABCA1, leading to increased synthesis of HDL in the liver in a PPAR-dependent manner [23]. In animal studies, fenofibrate also inhibited CETP expression, thus reducing cholesteryl ester transfer from HDL to VLDL [24]. It is important to note that human and rodent HDL-C responses to fibrate therapy vary substantially, which limits pharmacogenomic inference from animal models and increases the relative importance of epidemiologic evidence and/or human clinical studies [25]. However, in the case of CETP, animal model, data have been indirectly corroborated by a gene expression study in human subjects, although the effect of CETP variation on HDL-C response to fenofibrate has not been explicitly evaluated [26].

To date, most pharmacoepidemiologic studies of fibrates have focused on the lipoprotein genes, because upregulation of apoA-I (encoded by APOA1) is believed to be the central mechanism of action [10]. The APOA1/C3/A4 genes are located on human chromosome 11q23, extending to approximately 17 kb. The fourth member of that genetic cluster, APOA5, lies approximately 30 kb distal to APOA1/C3/A4 [27]. The focus on lipoproteins is justified by experimental evidence identifying APOA5 as an important PPAR[alpha] target gene and a potential mediator of fibrate effects on plasma lipids [28]. This evidence is further corroborated by epidemiologic studies. Data from the GOLDN study revealed that carriers of the minor G allele at the functional 56C→G (rs3135506) locus in APOA5 experienced a significantly larger increase in HDL-C following 3 weeks of daily 160 mg fenofibrate therapy when compared with noncarriers [29,30]. Another study, conducted in individuals with mixed dyslipidemia, linked common variants in the APOA5 region with response to combination therapy with fenofibrate and statins [31]. Studies investigating polymorphisms in APOE as potential mediators of fenofibrate response have yielded mixed results. In GOLDN, rs7412 was associated with changes in HDL-C among healthy subjects but not among participants with metabolic syndrome [32]. In the context of hypertriglyceridemia, the [epsilon]2 (T at rs429358 + T at rs7412) APOE genotype was associated with a more favorable lipid profile change than the [epsilon]3 (T at rs429358 + C at rs7412) or the [epsilon]4 (C at rs429358 + C at rs7412) genotypes [33]. However, in another study of dyslipidemic individuals, the association of the APOE genotypes with HDL-C response to fenofibrate was not significant, probably because of insufficient statistical power [34].

Fibrate therapy has been shown to increase the overall number of HDL particles and small HDL subclass particles (HDL₃), the latter due to the increase in cholesterol efflux triggered by enhanced ABCA1 expression [9,35]. Consistent with that hypothesis, GOLDN investigators reported a synergistic effect between ABCA1 variants and fenofibrate on HDL particle concentrations and small HDL [35]. The effect of fibrates on HDL subfraction profile also appears to be modulated by variation in PPARA itself, as evidenced by data from a small clinical trial of gemfibrozil conducted among abdominally obese men [36,37]. In the gemfibrozil trial, participants who were carriers of the variant allele of the functional L162V (rs1800206) polymorphism exhibited only a 6% increase in HDL₂ particles, compared with a 50% increase in wild-type participants [36]. The same polymorphism, however, was not a statistically significant predictor of HDL-C response to gemfibrozil in the Veterans Affairs HDL Intervention Trial, possibly because of baseline differences between populations [38]. On the other hand, the associations between common variation in CYP7A1, which encodes cholesterol 7[alpha]-hydroxylase and is crucial to bile acid synthesis, and the HDL-C response to fenofibrate were not specific to any subfraction but showed an overall gene-drug interaction [39].

Several other biologically plausible candidate genes, including LPL and PPARG, were investigated as potential mediators of fibrate response but were not found to be independently significant predictors of therapeutic outcomes [32,40]. Similarly, a recent analysis of the GOLDN data investigated whether previously validated genetic predictors of HDL particle number and size were associated with fenofibrate response, but did not yield any statistically significant results [41]. However, evidence suggests that the genetic determinants of fenofibrate response are nonlinear. A study conducted in hypertriglyceridemic subjects showed that the beneficial effect of [epsilon]2 APOE genotype on fenofibrate response is stronger among individuals with missense PPAR[alpha] −L162V (rs1800206) and LPL −P207L (rs28935469) variants [33]. Such synergy has clear biological plausibility, because all three gene products are in the fenofibrate pathway and have been shown to interact with each other. Specifically, the LPL enzyme, encoded by LPL, anchors to the endothelial matrix via an APOE -dependent mechanism and also interacts with apo E at the cell surface to recognize specific lipoproteins; Brisson et al. postulate that fenofibrate improves this interaction, resulting in a more favorable change in blood lipids [33]. As for the LPL/PPAR[alpha] interaction, LPL has been shown to induce generation of PPAR[alpha] ligands from circulating lipoproteins [42]. In individuals with the P207L (rs28935469) mutation, that process is compromised, but fenofibrate therapy can compensate by providing an exogenous mechanism of PPAR[alpha] activation, resulting in greater improvements in the lipid profile. These data show that such gene-gene interactions, although not extensively investigated, are critical to understanding complex traits such as lipid responses to fibrate therapy.

An additional avenue of pharmacogenomic investigations involves metabolic pathways responsible for the elimination or inactivation of the active therapeutic moiety of interest. In the case of fenofibrate, the elimination of fenofibric acid—the active moiety of the drug—occurs primarily through the glucuronidation pathway [43]. Assuming that access to the PPAR[alpha] site of action is indeed determined in part by circulating levels of fenofibric acid, as in vitro evidence suggests [43], genetic variation in uridine 5′-diphospho-glucuronosyltransferases emerges as a potentially important predictor of lipid response to fenofibrate. Data from the GOLDN study [44] have pursued associations between glucuronidation activity and fenofibrate response and established that uridine 5′-diphospho-glucuronosyltransferases are indeed important determinants of lipid changes, identifying a new class of markers that independently or in concert with other genes could eventually be used to predict HDL-C response to fenofibrate therapy.

HMG-CoA reductase inhibitors (statins)

Although the effect of statin therapy on HDL-C levels is subtle in comparison with the concomitant changes in LDL-C, it nevertheless contributes to the observed reduction in overall cardiovascular risk. Statins affect HDL profile via two complementary mechanisms: the downregulation of CETP and the reduction in particles containing apoB that can accept cholesterol esters [10]. Consequently, cholesterol transfer from HDL is impeded, resulting in an increase in the relative concentration of larger (HDL₂) particles and a more favorable HDL subfraction profile [45]. As expected, statin effects on HDL-C vary by dose and formulation, with pitavastatin showing the strongest HDL-C elevations in a meta-analysis of published trials [46]. Of note, the HDL-C-raising response observed from certain statins may not parallel their dose-dependent LDL-C-lowering response. For example, atorvastatin’s HDL-C-raising response appears to decline with higher doses that are typically associated with further LDL-C lowering; by contrast, rosuvastatin exhibits a more consistent dose response relationship for both LDL-C lowering and HDL-C raising [47]. Although this observation may bear clinical significance, the mechanisms underlying differential response remain poorly understood.

As with fibrates, lipid response to statin therapy is highly heterogeneous and in part mediated by genetic variation at multiple loci [48]. However, the modest increase in HDL-C attributed to statin therapy limits the ability to detect pharmacogenomic effects. The largest genome-wide association study of statin-mediated HDL-C changes to date, which combined three clinical trial populations involving almost 4000 individuals, did not yield any genome-wide Candidate gene studies, focusing primarily on CYP450 (CYP -) and drug transporter variants because of their importance in statin pharmacokinetics have yielded conflicting results. A pharmacogenetic study in newly diagnosed patients with coronary artery disease has linked common variation in CYP3A4 (CYP3A4*17, rs4986907) with significant alterations of HDL-C response to atorvastatin therapy [50]. The association with CYP3A4*17 (rs4986907) was evaluated in other atorvastatin-treated cohorts but no carriers of the variant allele were found among the studied subjects; as a result, this finding remains to be replicated [51]. The association with CYP3A4 was also not observed in analyses of other mutations in the same gene, CYP3A4*1G (rs2242480) and CYP3A4*1B (rs2740574), in hyperlipidemic patients treated with either atorvastatin or simvastatin [52,53]. The CYP3A4*1B (rs2740574) polymorphism, located in the promoter region was also associated with higher HDL-C variation in a study of Chilean hypercholesterolemic individuals [54]. Although insufficient statistical power remains a plausible explanation for the inconsistency, it is notable that CYP3A4 expression is affected by other coadministered therapeutic agents that could confound the genotype-drug response association; nevertheless, sensitivity analyses excluding patients who were administered other CYP3A4 substrates or inhibitors reported similar results [53].

The pattern of mixed and generally insufficient evidence holds true for other genes encoding CYP450 proteins. The candidate gene analysis of atorvastatin response determinants in newly-diagnosed coronary artery disease patients by Poduri et al. also found significant associations between changes in HDL-C and the rs8192871 variant in CYP7A1, which tags a promoter variant known as −204A→C (rs3808607) [50]. However, that finding was not replicated in an independent analysis [55], probably because of either different linkage disequilibrium patterns between populations, insufficient power to detect an effect, or gene-gene/gene-environment interactions. It is also possible that variation in CYP7A1, a gene encoding the rate-limiting enzyme in the biosynthesis of bile acids from cholesterol, explains more of the variation in LDL-C response to statins [55] and any concurrent changes in HDL-C response are spurious.

Despite the relevance of apoA-I to HDL metabolism, data on APOA1 variation and the HDL-C raising effects of statin therapy are similarly inconsistent. Candidate gene studies identified an APOA1 promoter polymorphism (−75G→A, rs670) as a robust predictor of HDL-C response to pravastatin [56] and atorvastatin [50], but the association was not replicated in another atorvastatin-treated population [57]. However, variation in APOB was found to be strongly associated with fluvastatin-mediated HDL-C increases in the Malmo Diet and Cancer Cardiovascular Cohort [58]. In the same population, it was also found that genetic mediators of lipid response to statins may vary by gender, with an LPL variant predicting treatment outcomes in women but not men [58]. Haplotypes in LPL were also associated with changes in HDL-C following lovastatin treatment in individuals who underwent coronary artery bypass surgery, providing indirect validation for the Malmo findings and supporting the hypothesis of pleiotropic effects of LPL [4].

Among drug transporters, OATP1B1 (encoded by SLCO1B1) has received considerable attention as it enables statins to enter hepatocytes, a required step for both action and subsequent clearance via bile [59]. In children diagnosed with familial hypercholesterolemia, the −11187G→A (rs4149015) mutation in SLCO1B1 was associated with lower peak plasma pravastatin C_max and a greater increase in HDL-C following 2-month statin therapy [60]. In another pediatric population described in the study by Hedman et al. that included heart transplant recipients treated with statins, similar increases were associated with the C allele at the 521T→C (rs4149056) locus, also in SLCO1B1 [60]. Fitting the pattern of inconsistency in pharmacogenetic studies of statins, a candidate gene study of the SLCO1B1 521T→C (rs4149056) mutation in adults failed to replicate the differential changes in HDL-C [61]. In addition, the rs4149056 variant was shown to be unrelated to pitavastatin-mediated HDL-C changes in a Chinese population [62]. It is likely that SLCO1B1 polymorphisms are less relevant to HDL-C response than to the risk of adverse events (i.e., statin-induced myopathy), where the observed associations are more consistent [48].

Because of the postulated CETP-dependent mechanism of statin action on HDL-C levels [10], variation in CETP has been considered as a potential predictor of treatment efficacy. Both CETP haplotypes [63,64] and individual variants [48,64,65], including the well-studied TaqIB polymorphism (rs708272) [50], have been shown to mediate HDL-C response to statin therapy. However, a large-scale meta-analysis of patients treated with pravastatin failed to detect any pharmacogenetic interactions with the TaqIB (rs708272) genotype and HDL-C changes, although the null findings may, in part, have been explained by the lack of statistical power as two of the three trials in the meta-analysis were aimed at secondary prevention [66].

On balance, the evidence of pharmacogenomic determinants of HDL-C response to statins is mixed. As most studies are designed to detect statin-mediated changes in LDL-C that are considerably greater than in HDL-C, lack of statistical power remains a pivotal concern. Future insights into the genetic predictors of statin-induced HDL-C changes are likely to come from the newly formed Genomic Investigation of Statin Therapy consortium, which includes genome-wide association study data on approximately 20,000 statin-treated individuals and plans to collect exome sequencing and epigenetic data on their subset [48].

Emerging HDL-C-raising therapies

The process of CETP inhibition, which promotes cholesterol efflux and raises the levels of large HDL particles, is the common mechanism underlying the HDL-C effects of niacin, fibrates and statins [67]. The new emerging class of HDL-C-raising therapies, currently undergoing trials for safety and efficacy, targets the CETP molecule directly. Although torcetrapib—the first CETP inhibitor assessed in clinical trials—showed an association with increased cardiovascular morbidity and mortality in statin-treated patients, largely attributed to poorly understood off-target effects on blood pressure, new agents such as dalcetrapib and anacetrapib do not seem to confer the same hazard [68]. Unfortunately, a Phase III study of dalcetrapib was recently stopped due to the lack of effectiveness in reducing adverse coronary events, echoing the fate of the AIM-HIGH trial [69], which similarly failed to demonstrate that addition of an HDL-raising component (niacin) to LDL-lowering therapy (simvastatin) can further the risk of myocardial infarction. However, in an on-treatment analysis of AIM-HIGH data, the difference in median HDL-C between groups was hardly clinically meaningful (4 mg/dl), suggesting that these results should not be extrapolated to other HDL-raising agents and indicating possible drug interactions [70].

Until the burden of proof to break the causal link between HDL-C and cardiovascular disease is met, however, the search for HDL-C raising agents continues. In addition to CETP, promising HDL-raising candidates identified by cell culture studies include ABCA1 and ABCG1, which encode transporters, LCAT, which encodes the lecithin-cholesterol acyltransferase enzyme, and APOE [71,72]. Direct LCAT activators and LXR agonists represent other potential venues of raising HDL-C by augmenting reverse cholesterol transport, but both classes of drugs and agents such as apoA-I mimetics, are in early stages of development and have not undergone pharmacogenomic testing [73].

From bench to bedside: challenges to clinical application of pharmacogenomic findings

Despite the plethora of replicated genetic loci that have been shown to be associated with HDL-C levels in cross-sectional studies, few such predictors have been validated for longitudinal changes in HDL-C following treatment. Reasons for failure to replicate positive findings are manifold and include the challenges of finding independent study populations with comparable drug interventions and phenotype ascertainment, small effect sizes, inadequate statistical power, low frequency of functional alleles and chance [74–76]. As a result, some investigators have pursued alternative validation approaches, for example, functional assays in cell lines and/or animal knockout models. However, the vast majority of pharmacogenomic findings in general, and of HDL-C raising therapies in particular, are yet to be validated.

In addition to the challenge of replication, it is important to consider potential gene-environment interactions between pharmacogenetic variants and patients’ dietary and exercise habits. Although evidence on the effectiveness of lifestyle changes on HDL-C levels is mixed [77,78], it is plausible that some putative genetic predictors on HDL-C response synergistically interact with nutritional factors, as was shown in the context of xenobiotic-metabolizing enzymes [79]. That could lead to the main effects of the genetic variant to be obscured unless evaluated in a stratified manner. Although investigations to detect such interactions will require a large number of participants, future studies should consider possible interactions between pharmaceutical agents, diet/exercise and genetic variants in determining success of HDL-C-raising therapies.

Considering that most variants explain only a small fraction of response variability, another challenge to the concept of personalized treatment of low HDL-C levels involves the cost-effectiveness of genetic tests. Even in the case of the lipid response to statins, for which pharmacogenomic determinants have been thoroughly investigated, routine genetic testing is not currently recommended [80]. In addition, the field is yet to reach agreement on the level of evidence required to justify clinical utility of such testing [81]. However, rapid advancements in next-generation sequencing assays stand to improve the cost-effectiveness of screening for pharmacogenomic markers by making the availability of sequence data possible for each patient [82]. The realization of comprehensive DNA sequence-based pharmacogenomic studies holds much promise to identify higher impact variants and validate novel predictors of drug response for a wide range of therapies. In summary, discovery of pharmacogenomic determinants of HDL-C raising therapies has been limited, but previous studies will ultimately guide ongoing efforts with the fundamental goal to advance personalization of lipid normalizing agents.

Expert commentary

The pursuit of pharmacogenomic determinants of HDL-C-raising therapies has the dual purpose of advancing the understanding of systems biology underlying the variability of response xenobiotics, and of guiding clinicians in making treatment decisions on the basis of individual likelihood of drug effectiveness and/or adverse reactions.

Knowledge gained through pharmacogenomic studies must ultimately bring us closer to distinguishing superior or inferior candidates to receive a given drug therapy on the basis of genomic markers. To date, advances in genomic techniques have made possible the systematic pursuit and discovery of numerous genetic markers. Nonetheless, the translational value of the data forthcoming from such pursuits to personalized medicine has been limited.

For currently used therapies targeting HDL-C, there is either a paucity of data regarding pharmacogenomic sources of variability in response (as is the case for niacin), or, where such data are available (as is the case for fibrates), currently identified genomic markers explain only a small percentage of drug response variability. Such modest impact can be attributed to either small absolute effect of the polymorphism and/or modest effect allele frequency in the relevant population. At this point, it appears that actionable pharmacogenomics-based guidance for clinicians seeking to raise HDL-C will have to include a broader array of both genomic and nongenomic factors contributing to drug response variability, particularly in the case of the emerging HDL-C-raising methods such as CETP inhibitors.

Finally, to realize the translational value of genetic determinants of HDL-C response, prospective clinical trials must test the practical utility and cost-effectiveness of population-scale pharmacogenomics. Although much has been learned to date, further work is clearly needed before genetic tools are widely used to optimize HDL-C raising therapy regimens and thus clinical outcomes.

Five-year view

The search for pharmacogenomic predictors of HDL-C response to therapy will continue and expand beyond candidate gene studies. Studies such as GOLDN and the GIST consortium are currently investigating lipid response to fenofibrate and statins, respectively, via genome-wide association scans, exome sequencing and epigenetic approaches. In addition, studies that follow currently ongoing trials of novel HDL-C-raising therapies will yield new information on genetic predictors of response heterogeneity for several other classes of drugs. These findings will augment the initial set of putative genetic determinants of HDL-C response, which should subsequently be validated in independent cohorts. Upon successful replication, future studies could develop algorithms that would enable clinicians to use DNA testing to predict response to HDL-C-raising therapies, leading to truly personalized prescription approaches. Concurrently, studies will continue re-evaluating the causal link between HDL-C and heart disease, which could ultimately alter the nature of clinical guidelines and therapeutic approaches, challenging the current priorities of pharmaceutical companies and pharmacogenomics researchers.

Key issues.

• HDL metabolism is dynamic and multifaceted, requiring a nuanced evaluation of putative genetic determinants of response to HDL-cholesterol (HDL-C)-raising therapy.

• HDL-C response to therapeutic interventions is characterized by high variability and moderate heritability, suggesting a role for pharmacogenomic factors.

• Although niacin is currently the most effective HDL-C-raising agent, no pharmacogenomic determinants of niacin efficacy have been described in humans.

• HDL-C response to fenofibrate is associated with genetic variation in the APOA1/C3/A4/A5 cluster, ABCA1, APOE, CETP, CYP7A1 and PPARa.

• Potential genetic predictors of HDL-C response to statin therapy are dependent on the type of statin and include variants in genes encoding lipoproteins, CYP450 proteins, CETP, LPL and OATP1B1.

• The effect of novel HDL-C-raising therapies, such as CETP inhibitors is currently under investigation and is likely to be modified by polymorphisms in relevant biologic pathways.

• Most of the currently available pharmacogenomic findings regarding HDL-C response have not been replicated in independent cohorts, leaving open the question of their validity.

• Until the findings are replicated and genetic risk prediction algorithms are developed and validated, the promise of personalized HDL-C therapy remains elusive.