- Email: [email protected]
Blood pressure genetics: time to focus
Journal of the American Society of Hypertension 3(4) (2009) 231–237
Review Article
Blood pressure genetics: time to focus Stephen B. Harrap, MBBS, PhD, FRACP Department of Physiology, University of Melbourne, Victoria, Australia Manuscript received and accepted June 9, 2009
Abstract This review briefly charts the recent history of the genetic study of high blood pressure (BP). After an inconsistent start it was hoped that very large genome-wide association studies (GWAS) might be able to provide some reliable answers. The two largest and most recent GWAS: CHARGE and GlobalBPgen were able to identify, despite some significant inconsistencies, genetic loci that accounted for only about 2% of the genetic factors believed to influence BP. The loci were associated with an estimated effect on BP of 1 mm Hg or less. No doubt many other loci exerting even smaller (<0.5 mm Hg) exist. This review contends that it is time to focus on the loci that can be viewed as confirmed, rather than extending the GWAS searches for less significant genetic influences. It is time to identify the precise deoxyribonucleic acid variants affecting BP, understand their mechanisms of action and think of ways in which such knowledge can be used to prevent and treat high BP in novel, effective, and possibly tailored ways. J Am Soc Hypertens 2009;3(4):231–237. Ó 2009 American Society of Hypertension. All rights reserved. Keywords: GWAS; CHARGE; GlobalBPgen; genetic loci.
Introduction A Difficult Problem The genetics of blood pressure (BP) has never been easy.1 With the ability to easily detect genetic variation, BP took its first genetic steps in the late 1980’s in and around genes considered worthy. After early enthusiasm generated by genes such as those encoding angiotensinogen2 and other components of the renin-angiotensin system (RAS), it soon became clear that for every positive study, there could be quoted an apparently credible negative result. This inconsistency wasn’t unique to the genetics of BP and casecontrol genetic association studies (comparing genotype frequencies between hypertensive and normotensive groups) that were trying to tackle the complex conditions were suffering similar problems.3 Although the ups and downs of the candidate gene analyses left many feeling deflated, there followed a period of intense activity to try and determine why things were Conflict of interest: none. Corresponding author: Stephen B. Harrap, MBBS, PhD, FRACP, Department of Physiology, University of Melbourne, Victoria 3010, Australia. Tel: þ613 8344 5836; fax: þ613 9349 4519. E-mail: [email protected]
difficult and how studies should be designed to respond.4 One approach was to abandon candidate genes, remove the blinkers and instead investigate the entire genome. This was first tackled using families and undertaking linkage studies (methods that had been proven in Mendelian disease). The great hope was that the big genomewide linkage scans from either side of the Atlantic known as British Genetics of Hypertension (BRIGHT)5 and the Family BP Program,6 would sort things out. However, much to the surprise of many, neither could find loci that achieved the Holy Grail of genome-wide significance. In the interim, methods were developed to undertake genome-wide association studies (GWAS), based not in families, but in cases and controls or in populations more broadly. These involved thousands of subjects and were expected to provide answers. Disappointingly, although there were odd hints, BP still seemed to be left without a consistent genetic story.7–10 Most acutely, in the Wellcome Trust Case Control Consortium, BP was left alone in the wings while other phenotypes such as diabetes and coronary artery disease took centre stage with their genetic partners.11
Increasing Power That genetic factors for BP must exist is predicated on analyses of BP variance that suggest that about 40 to
1933-1711/09/$ – see front matter Ó 2009 American Society of Hypertension. All rights reserved. doi:10.1016/j.jash.2009.06.001
232
S.B. Harrap / Journal of the American Society of Hypertension 3(4) (2009) 231–237
50% of BP variation is attributable to genes12 and that there indeed exist rare genetic variants having significant impact on BP.13 However, most of the evidence for the involvement of common genetic variants suggested that their phenotypic effects might be small. Part of the reason for the frustration in detecting such weak signals could have been the blurring effect of an inherently unstable phenotype for which a large proportion of variance is unique to each individual – being neither genetic nor shared environment. The logical approach was to undertake even bigger studies to augment statistical power. The two new transatlantic studies that rise to this challenge have been published recently in Nature Genetics.14,15 The 204 collective authors of these two papers undertook GWAS for BP in as many as 63,569 individuals. Each individual yielded about 500,000 genotypes, with additional genotyping in another 84,114 subjects, making a total dataset of around 32 billion genotypes. This is by any measure a veritable genetic sledgehammer. A sledgehammer, it transpired, to kill a flea. The flea in this case was 1 mm Hg, the estimated BP effect for each of the handful of genetic loci that achieved genome-wide significance. Most clinicians will be impressed by the enormity of the effort, but perhaps less so by the result, given that 1 mm Hg is imperceptible in their patients. They might be right, but it would be a mistake to think that this entire genetic endeavor is for naught. These juxtaposed (and intersecting) studies provide important insights, both general and specific, that deserve respect and scrutiny.
The Substrate The two consortia involved were Cohorts for Heart and Aging Research in Genome Epidemiology (CHARGE) and Global Blood Pressure Genetics (GlobalBPgen). Both were conglomerates of ‘‘smaller’’ samples that permitted the investigators to undertake meta-analyses. CHARGE comprised 6 groups, ranging in size from 1,760 to 8,096 subjects, that were from population-based cohorts, selected irrespective of BP. Within CHARGE, the Framingham Heart Study (FHS) and Athersclerosis Risk in Communities (ARIC) were the largest and majority shareholders. GlobalBPgen drew on a total of 32 samples, the smallest of which was 562 and the largest 15,858 (from the European Prospective Investigation of Cancer – Norfolk GWAS).
Blood Pressures There was a certain degree of heterogeneity between and within the studies in terms of design, phenotyping, and genotyping. Such heterogeneity is not ideal, but not necessarily fatal. It is also difficult to avoid, if one is to accrue sufficient numbers of subjects for studies of this kind. In terms of phenotypes, the BPs in CHARGE and GlobalBPgen were made over a long time frame, some as long
ago as the late 1940’s and through subsequent decades until this century. There have been changes in average pressures over this time and also changes in the environment.16 This might have seen alterations in the interactions between genes and environment that might bring certain loci in and out of focus. The impact of such protracted sampling is impossible to assess, but it adds a potential layer of variability that makes association studies harder. The methods of measurement also deserve consideration. Generally BP was measured in the seated position (sometimes supine) by the use of a variety of semi-automated oscillometric devices (particularly in GlobalBPgen) or mercury columns. In the CHARGE Study, 13,084 subjects from the ARIC, Cardiovascular Health Study, and the Rotterdam Study had BPs recorded many years ago with the Hawksley random zero sphygmomanometer – a device that was subsequently shown to introduce some inaccuracy.17,18 The impact of this is also difficult to judge, but it would certainly not strengthen the ability to identify loci associated with BP. All subjects were drawn from both Europe and North America and were of European origin (save for a replication analysis in 12,889 subjects of Indian Asian origin in GlobalBPgen).
Quantitative Approaches Both studies undertook primary analyses based on quantitative measures of systolic BP (SBP) and diastolic BP (DBP) across the full population range, including hypertensive subjects. The goal was to identify loci that might influence SBP or DBP within populations, not just in hypertensive subjects. This is a welcome and important approach, as it recognizes the fact that, as a result of their numbers, those with average levels of BP, who are exposed to modest cardiovascular risk, contribute more cases of BP-related cardiovascular death than the smaller number of those with the highest BPs and greatest individual risk.19 This means that this sort of genetic discovery is relevant to populations and not just individuals. The quantitative approach also partially avoids some potential pitfalls of the case (hypertensive) vs. control (normotensive) paradigm. Such designs are at risk of bias, such as population stratification, where unrecognized and irrelevant (for BP at least) genetic or environmental differences become entangled in the analyses and create spurious associations. It also minimizes the impact of unrecognized cases among the control subjects. This is especially important for conditions of high prevalence such as hypertension where younger ‘‘controls’’ might in fact be genetically predisposed to later high BP.
Collaboration Although CHARGE and GlobalBPgen were published as separate papers, they collaborated very closely. Both
S.B. Harrap / Journal of the American Society of Hypertension 3(4) (2009) 231–237
233
Figure. The summarized results regarding the top 10 genetic SBP markers from CHARGE and GlobalBPgen. The columns show for each marker the chromosome, whether it was in the CHARGE (C) or GlobalBPgen (G) top 10, its chromosomal location, the nearest candidate gene, the estimated effect on SBP associated with the marker and whether the marker reached threshold (dark gray shading, P< 1 105) or genome-wide (black shading, P< 5 108) significance in the CHARGE, GlobalBPgen, and their pooled data sets in the last 3 columns, respectively. CHARGE, Cohorts for Heart and Aging Research in Genome Epidemiology; GlobalBPgen, Global Blood Pressure Genetics; SBP, systolic blood pressure.
undertook their own primary analyses and each derived their top 10 markers for SBP and DBP. These were tested in the reciprocal data set separately and then in the pooled dataset of nearly 64,000 subjects. This collaborative arrangement was designed to augment power and provide an opportunity to replicate putative genetic loci. However, this arrangement was not only a test of genetic loci, it was a test of the approach itself. Here we have two studies that independently might have been expected to provide sufficient power to detect loci of modest effect. How do they compare and by increasing study size to this degree, can we at last find some consensus? It is fascinating to see how the results actually unfold.
Loci for Systolic Pressure For simplicity, the focus will be on SBP, but the story is similar for DBP. The Figure shows the findings for the 20 top loci ordered by chromosome (Chr) and chromosomal position from both studies (10 from CHARGE — C and
10 from GlobalBPgen — G). The nearest gene is shown for each of the loci. On chromosome 10 the loci from CHARGE and GlobalBPgen (indicated by *) were relatively close and although closer to the gene NT5C2 the GlobalBPgen investigators referred to this as a marker of CYP17A1. As shown, the estimated effect size per allele for each of these markers was sometime around, but more often less than 1 mm Hg. The columns CHARGE and GlobalBPgen show whether the individual loci achieved threshold or genome-wide significance in either study. Where loci achieved threshold significance (P < 1 105) the cells are dark gray, when they achieved genome-wide level of significance (P < 5 108) the cells are black. It is clear that there is little overlap in the top 10 loci from either study, with none of the top 10 loci achieving threshold significance in both studies. Some of the discrepancies were large, with 6 of the top 10 loci from GlobalBPgen not even achieving nominal significance of P < .05 in CHARGE
234
S.B. Harrap / Journal of the American Society of Hypertension 3(4) (2009) 231–237
and 4 of the top 10 from CHARGE suffering the same fate in GlobalBPgen. In the results of the pooled analyses, 6 loci achieved genome-wide significance. Four of these were originally from CHARGE and two were from GlobalBPgen. These were the markers in or near the genes MTHFR, CYP17A1, NT5C2, PLEKHA7, ATP2B1, and SH2B3. The GlobalBPGen decided to go one further step and attempted to replicate their results for three of their SBP markers from the pooled analyses that had reached a significance of P < 5 107 (the level of significance used in the Wellcome Trust Case Control Consortium studies), being those near NT5C2, MTHFR, and PLCD3. For this purpose as many as 71,225 subjects of European ancestry and 12,889 of Indian Asian ancestry were genotyped. Given what had already been seen, the Indian Asian group was manifestly underpowered statistically by its relatively small size. From the European group, the NT5C2 marker stood the test (P < 9 1015) in 71,225 subjects and MTHFR also gave a fair result (P < 2 105) in 19,751 subjects. However, the PLCD3 marker fell at the hurdle in 17,887 subjects with a P value of only .045.
Many Loci, Small Effects What are we to make of all this? Firstly, these studies confirm that the estimated effects for the genetic factors relevant to BP are small. This observation, combined with the inherent variability of BP, explain, why it has been so difficult in GWAS to identify loci with genome-wide significance. Secondly, the alleles associated with BP were reasonably common, with average minor allele frequencies around 20% to 30%. Thirdly, the loci reported seem to interact in an additive fashion. The CHARGE investigators plotted average pressure levels as a weighted function of the number of BP alleles carried. This showed a linear increase of about 8 mm Hg across the range as alleles accumulated. Therefore, on average, there did not seem to be multiplicative interactions between loci that might have caused a more exponential increase in pressure for increasing alleles. There must be some caution in interpreting these analyses, as they were made in the same subjects from which the markers were discovered and results in independent populations might differ. However, the patterns seen in three studies are entirely consistent with the polygenic hypothesis of common conditions such as hypertension, which can be considered (at least in genetic terms) the result of inheriting a number of predisposing alleles, each of which is common and each of which exerts small BP effects individually. BP is truly the sum of many genetic parts.
The Missing Genes However, the loci reported in CHARGE and GlobalBPgen explain only about 1% of total variation in BP, or about 2% of
Table A list of candidate genes that have been associated at various times with blood pressure 11BHSD2 ADD2 ATP1A1 CAV1 CLCNKB CYP11B1 CYP4A11 ESR1 FYN GRK-4 ITGB2 NCOA1 PTGIS SCNN1B SLC12A1 SPON1 WNK4
5HT2A ADRB2 ATP2A2 CCR2 CNR2 CYP11B2 DRD-1 ESR2 GCH1 HNF1 KCNJ1 NEDD4L QPCT SCNN1G SLC12A3 TH
ABCA1 AGT BMAL1 CCR5 COL1A2 CYP11B2 EDNRA ET1 GCK IL6 KIAA0789 NR3C2 REN SGK1 SLC4A5 TSC
ACE AGTR1 CAPN10 CD31 COMT CYP19A1 ELN FAS GNB3 INPPL1 M6PR PDE4D ROS1 SHIPS2 SLC7A1 UMOD
ADD1 AMPD1 CASR CHGB CORIN CYP3A5 NOS3 FGF1 GPX1 ITGA2 MTP PNMT SAH SLC SLC9A2 WNK1
the genetic variance. Where then are the missing genetic factors? One possibility is that there exist other less potent loci that have escaped the net cast by the strict statistical definitions of significance. These could be loci for which BP effects in these populations are smaller than 0.5 mm Hg. If this is the case, it implies that there must be a multitude of such loci to make up the missing variance. The CHARGE investigators make the point that if one lowers one’s statistical sights to P < 1 103, one observes 890 more markers associated with BP than expected by chance. Perhaps some of the candidate genes that have been associated in smaller studies previously (Table) are in this group – a glimmer of hope for those candidate gene investigators who felt themselves shunned by the advent of GWAS.
The Discrepancies The next observation is that, just as in the early days of candidate genes, inconsistency and discrepancies still plague the genetics of BP. The markers that seriously failed the tests of replication cannot be simply dismissed as a consequence of moderate effect size, as suggested by the GlobalBPgen investigators. There was no clear relationship between effect size estimated in one study and the likelihood that a marker would fail to achieve even P < .05 in the other. Such divergence raises important questions.20 Among the possible causes is heterogeneity in a variety of factors such as differences in genotyping platforms, different marker sets, and a whole range of factors associated with underlying population heterogeneity. These could include different patterns of linkage disequilibrium, which means that as result of different genetic architecture, markers in a particular region are linked to the causative variant in one population, but not in another. There could
S.B. Harrap / Journal of the American Society of Hypertension 3(4) (2009) 231–237
also be population-specific differences in genetic and environmental interactions that might obscure or highlight a particular locus in one or other population. Given the fact that both studies were set in subjects of European descent in both Europe and the United States, it is hard to invoke population differences as an explanation for the divergence. One last possibility that can’t be excluded is that even studies of 30,000 subjects can be misleading.
Genetic Screening? For those dreaming of the day when the routine health screening will involve tests of markers for high BP, these studies are not encouraging. It is difficult to envisage that markers that account for 1 mm Hg or less of BP and can be difficult to replicate between populations and even within the constituent studies of meta-analyses might be useful and reliable for personal prediction. The gold standard remains careful measurement with a BP cuff and taking a good family history (or better still measurement of the parents’ BP).
Where Next? In some ways, the genetics of BP is at a fork in the road. One might choose to continue down the path of pinning down more and more genetic loci. However, CHARGE and GlobalBPgen are telling us that we would need even bigger studies to pick up loci with BP effects of less than 0.5 mm Hg. This becomes a case of diminishing returns, particularly as the numbers of subjects required for such studies rise exponentially as the phenotypic effects gets smaller.
Time to Reflect and Focus But genetics should be about understanding, not about cataloguing. It is time to pause and expend effort mining the discoveries to date. This begins with identifying the precise location and nature of the deoxyribonucleic acid (DNA) variation that explains these BP associations. This is critical to revealing and understanding mechanisms against which, hopefully novel, preventive, and treatment strategies might be targeted. For example, the first priority would no doubt be on chromosome 10. Both CHARGE and GlobalBPgen highlighted this locus, but had their own top markers at either end of an associated stretch of DNA about 430 kb long. The top CHARGE marker was in the gene CYP17A1 that encodes the enzyme steroid 17a-hydroxylase – a gene with good credentials in the Mendelian BP world. The top GlobalBPgen marker was closer to the gene NT5C2 that encodes a 5’-nucleotidase acting as a cytoplasmic enzyme involved in the maintenance of intracellular purine/pyrimidine nucleotides.
235
Think Laterally Given the experiences with recent discoveries, most notably that for myocardial infarction, one should not assume that causative variants would be in a gene of physiological relevance or even close to any gene per se. There is a good deal of variation in the DNA sequence between CYP17A1 and NT5C2 and careful thought will need to be given to the best strategy for recognizing the highest priority candidates and the most appropriate means to validate their status. Searching across intergenic regions for changes in DNA sequence that might alter gene expression to produce quantitative effects (as opposed to changing protein sequence and causing qualitative effects) is extremely difficult, especially as we cannot recognize many such sequences at this time. Another marker worthy of detailed investigation might be that on chromosome 12 close to the gene ATP2B1. This was the most significant marker in CHARGE, which had also recently been revealed as associated with BP in a reasonably large study in Korea.21 This gene encodes a plasma membrane calcium/calmodulin-dependent ATPase found in vascular endothelium and implicated in animal models of hypertension. The validation of causative DNA variants will demand a careful and strategic combination of molecular genetics and cellular biology, biochemistry, and physiology to trace the path of influence from DNA through ribonucleic acid proteins, cells, tissues, organs to man. And one needs to be careful in piecing these elements together. For example, the locus on chromosome 1 identified in GlobalBPgen for SBP is closest to the methylenetetrahydrofolate reductase (MTHFR) gene. It is also in the vicinity of the NPPA and NPPB genes that had been associated very strongly with differences in levels of plasma atrial natriuretic peptide and less so with BP.22 However, results from GlobalBPgen raise the possibility that the NPPA and NPPB genes are not the outstanding contributors for BP in this region. Therefore, although it might be tempting to associate the BP and peptide changes as logical consequences of variation in NPPA and NPPB genes, such effects might be coincidental, with the BP being the result of variation towards MTHFR.
Is It Worth the Effort? Some might argue that we are better off putting money into other aspects of cardiovascular disease prevention, as the genetic loci are associated with such small effects. However, it would be a mistake to relegate genetics on these grounds. Firstly, until the precise DNA variants are identified, it is impossible to know the real impact on BP. Even a GWAS using 500,000 markers is pretty sparse considering the 3 billion base pairs in the human chromosome. It is likely that most markers will be at a distance
236
S.B. Harrap / Journal of the American Society of Hypertension 3(4) (2009) 231–237
from the causative variants and that physical separation can weaken the apparent BP effect, as judged by the marker alone. Secondly, it needs to be emphasized that the magnitude of BP effect resulting from common DNA variation does not necessarily reflect the importance of the physiological control mechanism influenced by the variant. A locus associated with 0.5 to 1 mm Hg of SBP says nothing about the robustness of the system with which the locus interacts. These small clues might be just the tip of the iceberg. More important is the nature of the iceberg. Indeed, the excitement of GWAS is not the discovery of loci, it is the potential to define undiscovered physiological pathways that will offer new approaches to prevention and treatment and augment existing therapeutic and health promotion strategies. Finally, there is a real possibility that a diversity of DNA variants will have effects that converge on a common physiological mechanism.23 Therefore, if several loci, each with an effect of 0.5 to 1 mm Hg mediate their effects through the same physiological focal point, such a point becomes a prime target for intervention.
7.
8.
9.
10.
11.
Conclusion This is a fascinating time for the genetics of BP. We need to recognize that BP is a unique phenotype and more difficult than most when it comes genetic discovery. Perhaps we can do both, but if we need to choose between accumulating more markers in the BP catalogue vs. learning exactly how the existing markers are implicated, the latter focused approach is the one that will bring us nearer to the ultimate goal of making genetics of use to our communities and our patients.
References 1. Harrap SB. Where are all the blood pressure genes? Lancet 2003;361:2149–51. 2. Jeunemaitre X, Soubrier F, Kotelevtsev YV, Lifton RP, Williams CS, Charru A, et al. Molecular basis of human hypertension: role of angiotensinogen. Cell 1992;71: 169–80. 3. Weiss KM, Terwilliger JD. How many diseases does it take to map a gene with SNPs? Nat Genet 2000;26: 151–7. 4. Cardon LR, Bell JI. Association study designs for complex diseases. Nat Rev Genet 2001;2:91–9. 5. Caulfield M, Munroe P, Pembroke J, Samani N, Dominiczak A, Brown M, et al. Genome-wide mapping of human loci for essential hypertension. Lancet 2003; 361:2118–23. 6. Province MA, Kardia SLR, Ranade K, Rao DC, Thiel BA, Cooper RS, et al. A meta-analysis of genomewide linkage scans for hypertension: The National
12.
13.
14.
15.
16.
17.
18.
19.
Heart Lung and Blood Institute Family Blood Pressure Program. Am J Hypertens 2003;16:144–7. Levy D, Larson MG, Benjamin EJ, Newton-Cheh C, Wang TJ, Hwang SJ, et al. Framingham Heart Study 100K Project: genome-wide associations for blood pressure and arterial stiffness. BMC Med Genet 2007;8:S3. Wang Y, O’Connell JR, McArdle PF, Wade JB, Dorff SE, Shah SJ, et al. Whole-genome association study identifies STK39 as a hypertension susceptibility gene. PNAS 2009;106:226–31. Sabatti C, Service SK, Hartikainen AL, Pouta A, Ripatti S, Brodsky J, et al. Genome-wide association analysis of metabolic traits in a birth cohort from a founder population. Nat Genet 2009;41:35–46. Gu CC, Hunt SC, Kardia S, Turner ST, Chakravarti A, Schork N, et al. An investigation of genome-wide associations of hypertension with microsatellite markers in the Family Blood Pressure Program. Hum Genet 2007; 121:577–90. Burton PR, Clayton DG, Cardon LR, Craddock N, Deloukas P, Duncanson A, et al. Wellcome Trust Case Control Consortium. Nature 2007;447:661–78. Harrap SB, Stebbing M, Hopper JL, Hoang HN, Giles GG. Familial patterns of covariation for cardiovascular risk factors in adults: The Victorian Family Heart Study. Am J Epidemiol 2000;152:704–15. Ji W, Foo JN, O’Roak BJ, Zhao H, Larson MG, Simon DB, et al. Rare independent mutations in renal salt handling genes contribute to blood pressure variation. Nat Genet 2008;40:592–9. Levy D, Ehret GB, Rice K, Verwoert GC, Launer LJ, Dehghan A, et al. Genome-wide association study of blood pressure and hypertension. Nat Genet 2009;41: 677–87. Newton-Cheh C, Johnson T, Gateva V, Tobin MD, Bochud M, Coin L, et al. Genome-wide association study identifies 8 loci associated with blood pressure. Nat Genet 2009;41:666–76. Tunstall-Pedoe H, Connaghan J, Woodward M, Tolonen H, Kuulasmaa K. Pattern of declining blood pressure across replicate population surveys of the WHO MONICA project, mid-1980s to mid1990s, and the role of medication. BMJ 2006;332: 629–32. McGurk C, Nugent A, McAuley D, Silke B. Sources of inaccuracy in the use of the Hawksley random-zero sphygmomanometer. J Hypertens 1997;15:1379–84. Brown WCB, Kennedy S, Inglis GC, Murray LS, Lever AF. Mechanisms by which the Hawksley random zero sphygmomanometer underestimates blood pressure and produces a non-random distribution of RZ values. J Human Hypertens 1997;11:75–93. Rose G. Strategy of prevention: lessons from cardiovascular disease. Br Med J Clin Res Ed 1981;282:1847–51.
S.B. Harrap / Journal of the American Society of Hypertension 3(4) (2009) 231–237
20. McCarthy MI, Abecasis GR, Cardon LR, Golstein DB, Little J, Ioannidis JPA, et al. Genome-wide association studies for complex traits: consensus, uncertainty, and challenges. Nat Rev Genet 2008;9:356–69. 21. Cho YS, Go MJ, Kim YJ, Heo JY, Oh JH, Ban HJ, et al. A large-scale genome-wide association study of Asian populations uncovers genetic factors influencing 8 quantitative traits. Nat Genet 2009;41:527–34.
237
22. Newton-Chen C, Larson MG, Vasan RS, Levy D, Bloch KD, Surti A, et al. Association of common variants in NPPA and NPPB with circulating natriuretic peptides and blood pressure. Nat Genet 2009;41: 348–53. 23. Harrap SB, Petrou S. Utility of genetic approaches to common cardiovascular diseases. Am J Physiol 2001; 281:H1–6.
Review Article
Blood pressure genetics: time to focus Stephen B. Harrap, MBBS, PhD, FRACP Department of Physiology, University of Melbourne, Victoria, Australia Manuscript received and accepted June 9, 2009
Abstract This review briefly charts the recent history of the genetic study of high blood pressure (BP). After an inconsistent start it was hoped that very large genome-wide association studies (GWAS) might be able to provide some reliable answers. The two largest and most recent GWAS: CHARGE and GlobalBPgen were able to identify, despite some significant inconsistencies, genetic loci that accounted for only about 2% of the genetic factors believed to influence BP. The loci were associated with an estimated effect on BP of 1 mm Hg or less. No doubt many other loci exerting even smaller (<0.5 mm Hg) exist. This review contends that it is time to focus on the loci that can be viewed as confirmed, rather than extending the GWAS searches for less significant genetic influences. It is time to identify the precise deoxyribonucleic acid variants affecting BP, understand their mechanisms of action and think of ways in which such knowledge can be used to prevent and treat high BP in novel, effective, and possibly tailored ways. J Am Soc Hypertens 2009;3(4):231–237. Ó 2009 American Society of Hypertension. All rights reserved. Keywords: GWAS; CHARGE; GlobalBPgen; genetic loci.
Introduction A Difficult Problem The genetics of blood pressure (BP) has never been easy.1 With the ability to easily detect genetic variation, BP took its first genetic steps in the late 1980’s in and around genes considered worthy. After early enthusiasm generated by genes such as those encoding angiotensinogen2 and other components of the renin-angiotensin system (RAS), it soon became clear that for every positive study, there could be quoted an apparently credible negative result. This inconsistency wasn’t unique to the genetics of BP and casecontrol genetic association studies (comparing genotype frequencies between hypertensive and normotensive groups) that were trying to tackle the complex conditions were suffering similar problems.3 Although the ups and downs of the candidate gene analyses left many feeling deflated, there followed a period of intense activity to try and determine why things were Conflict of interest: none. Corresponding author: Stephen B. Harrap, MBBS, PhD, FRACP, Department of Physiology, University of Melbourne, Victoria 3010, Australia. Tel: þ613 8344 5836; fax: þ613 9349 4519. E-mail: [email protected]
difficult and how studies should be designed to respond.4 One approach was to abandon candidate genes, remove the blinkers and instead investigate the entire genome. This was first tackled using families and undertaking linkage studies (methods that had been proven in Mendelian disease). The great hope was that the big genomewide linkage scans from either side of the Atlantic known as British Genetics of Hypertension (BRIGHT)5 and the Family BP Program,6 would sort things out. However, much to the surprise of many, neither could find loci that achieved the Holy Grail of genome-wide significance. In the interim, methods were developed to undertake genome-wide association studies (GWAS), based not in families, but in cases and controls or in populations more broadly. These involved thousands of subjects and were expected to provide answers. Disappointingly, although there were odd hints, BP still seemed to be left without a consistent genetic story.7–10 Most acutely, in the Wellcome Trust Case Control Consortium, BP was left alone in the wings while other phenotypes such as diabetes and coronary artery disease took centre stage with their genetic partners.11
Increasing Power That genetic factors for BP must exist is predicated on analyses of BP variance that suggest that about 40 to
1933-1711/09/$ – see front matter Ó 2009 American Society of Hypertension. All rights reserved. doi:10.1016/j.jash.2009.06.001
232
S.B. Harrap / Journal of the American Society of Hypertension 3(4) (2009) 231–237
50% of BP variation is attributable to genes12 and that there indeed exist rare genetic variants having significant impact on BP.13 However, most of the evidence for the involvement of common genetic variants suggested that their phenotypic effects might be small. Part of the reason for the frustration in detecting such weak signals could have been the blurring effect of an inherently unstable phenotype for which a large proportion of variance is unique to each individual – being neither genetic nor shared environment. The logical approach was to undertake even bigger studies to augment statistical power. The two new transatlantic studies that rise to this challenge have been published recently in Nature Genetics.14,15 The 204 collective authors of these two papers undertook GWAS for BP in as many as 63,569 individuals. Each individual yielded about 500,000 genotypes, with additional genotyping in another 84,114 subjects, making a total dataset of around 32 billion genotypes. This is by any measure a veritable genetic sledgehammer. A sledgehammer, it transpired, to kill a flea. The flea in this case was 1 mm Hg, the estimated BP effect for each of the handful of genetic loci that achieved genome-wide significance. Most clinicians will be impressed by the enormity of the effort, but perhaps less so by the result, given that 1 mm Hg is imperceptible in their patients. They might be right, but it would be a mistake to think that this entire genetic endeavor is for naught. These juxtaposed (and intersecting) studies provide important insights, both general and specific, that deserve respect and scrutiny.
The Substrate The two consortia involved were Cohorts for Heart and Aging Research in Genome Epidemiology (CHARGE) and Global Blood Pressure Genetics (GlobalBPgen). Both were conglomerates of ‘‘smaller’’ samples that permitted the investigators to undertake meta-analyses. CHARGE comprised 6 groups, ranging in size from 1,760 to 8,096 subjects, that were from population-based cohorts, selected irrespective of BP. Within CHARGE, the Framingham Heart Study (FHS) and Athersclerosis Risk in Communities (ARIC) were the largest and majority shareholders. GlobalBPgen drew on a total of 32 samples, the smallest of which was 562 and the largest 15,858 (from the European Prospective Investigation of Cancer – Norfolk GWAS).
Blood Pressures There was a certain degree of heterogeneity between and within the studies in terms of design, phenotyping, and genotyping. Such heterogeneity is not ideal, but not necessarily fatal. It is also difficult to avoid, if one is to accrue sufficient numbers of subjects for studies of this kind. In terms of phenotypes, the BPs in CHARGE and GlobalBPgen were made over a long time frame, some as long
ago as the late 1940’s and through subsequent decades until this century. There have been changes in average pressures over this time and also changes in the environment.16 This might have seen alterations in the interactions between genes and environment that might bring certain loci in and out of focus. The impact of such protracted sampling is impossible to assess, but it adds a potential layer of variability that makes association studies harder. The methods of measurement also deserve consideration. Generally BP was measured in the seated position (sometimes supine) by the use of a variety of semi-automated oscillometric devices (particularly in GlobalBPgen) or mercury columns. In the CHARGE Study, 13,084 subjects from the ARIC, Cardiovascular Health Study, and the Rotterdam Study had BPs recorded many years ago with the Hawksley random zero sphygmomanometer – a device that was subsequently shown to introduce some inaccuracy.17,18 The impact of this is also difficult to judge, but it would certainly not strengthen the ability to identify loci associated with BP. All subjects were drawn from both Europe and North America and were of European origin (save for a replication analysis in 12,889 subjects of Indian Asian origin in GlobalBPgen).
Quantitative Approaches Both studies undertook primary analyses based on quantitative measures of systolic BP (SBP) and diastolic BP (DBP) across the full population range, including hypertensive subjects. The goal was to identify loci that might influence SBP or DBP within populations, not just in hypertensive subjects. This is a welcome and important approach, as it recognizes the fact that, as a result of their numbers, those with average levels of BP, who are exposed to modest cardiovascular risk, contribute more cases of BP-related cardiovascular death than the smaller number of those with the highest BPs and greatest individual risk.19 This means that this sort of genetic discovery is relevant to populations and not just individuals. The quantitative approach also partially avoids some potential pitfalls of the case (hypertensive) vs. control (normotensive) paradigm. Such designs are at risk of bias, such as population stratification, where unrecognized and irrelevant (for BP at least) genetic or environmental differences become entangled in the analyses and create spurious associations. It also minimizes the impact of unrecognized cases among the control subjects. This is especially important for conditions of high prevalence such as hypertension where younger ‘‘controls’’ might in fact be genetically predisposed to later high BP.
Collaboration Although CHARGE and GlobalBPgen were published as separate papers, they collaborated very closely. Both
S.B. Harrap / Journal of the American Society of Hypertension 3(4) (2009) 231–237
233
Figure. The summarized results regarding the top 10 genetic SBP markers from CHARGE and GlobalBPgen. The columns show for each marker the chromosome, whether it was in the CHARGE (C) or GlobalBPgen (G) top 10, its chromosomal location, the nearest candidate gene, the estimated effect on SBP associated with the marker and whether the marker reached threshold (dark gray shading, P< 1 105) or genome-wide (black shading, P< 5 108) significance in the CHARGE, GlobalBPgen, and their pooled data sets in the last 3 columns, respectively. CHARGE, Cohorts for Heart and Aging Research in Genome Epidemiology; GlobalBPgen, Global Blood Pressure Genetics; SBP, systolic blood pressure.
undertook their own primary analyses and each derived their top 10 markers for SBP and DBP. These were tested in the reciprocal data set separately and then in the pooled dataset of nearly 64,000 subjects. This collaborative arrangement was designed to augment power and provide an opportunity to replicate putative genetic loci. However, this arrangement was not only a test of genetic loci, it was a test of the approach itself. Here we have two studies that independently might have been expected to provide sufficient power to detect loci of modest effect. How do they compare and by increasing study size to this degree, can we at last find some consensus? It is fascinating to see how the results actually unfold.
Loci for Systolic Pressure For simplicity, the focus will be on SBP, but the story is similar for DBP. The Figure shows the findings for the 20 top loci ordered by chromosome (Chr) and chromosomal position from both studies (10 from CHARGE — C and
10 from GlobalBPgen — G). The nearest gene is shown for each of the loci. On chromosome 10 the loci from CHARGE and GlobalBPgen (indicated by *) were relatively close and although closer to the gene NT5C2 the GlobalBPgen investigators referred to this as a marker of CYP17A1. As shown, the estimated effect size per allele for each of these markers was sometime around, but more often less than 1 mm Hg. The columns CHARGE and GlobalBPgen show whether the individual loci achieved threshold or genome-wide significance in either study. Where loci achieved threshold significance (P < 1 105) the cells are dark gray, when they achieved genome-wide level of significance (P < 5 108) the cells are black. It is clear that there is little overlap in the top 10 loci from either study, with none of the top 10 loci achieving threshold significance in both studies. Some of the discrepancies were large, with 6 of the top 10 loci from GlobalBPgen not even achieving nominal significance of P < .05 in CHARGE
234
S.B. Harrap / Journal of the American Society of Hypertension 3(4) (2009) 231–237
and 4 of the top 10 from CHARGE suffering the same fate in GlobalBPgen. In the results of the pooled analyses, 6 loci achieved genome-wide significance. Four of these were originally from CHARGE and two were from GlobalBPgen. These were the markers in or near the genes MTHFR, CYP17A1, NT5C2, PLEKHA7, ATP2B1, and SH2B3. The GlobalBPGen decided to go one further step and attempted to replicate their results for three of their SBP markers from the pooled analyses that had reached a significance of P < 5 107 (the level of significance used in the Wellcome Trust Case Control Consortium studies), being those near NT5C2, MTHFR, and PLCD3. For this purpose as many as 71,225 subjects of European ancestry and 12,889 of Indian Asian ancestry were genotyped. Given what had already been seen, the Indian Asian group was manifestly underpowered statistically by its relatively small size. From the European group, the NT5C2 marker stood the test (P < 9 1015) in 71,225 subjects and MTHFR also gave a fair result (P < 2 105) in 19,751 subjects. However, the PLCD3 marker fell at the hurdle in 17,887 subjects with a P value of only .045.
Many Loci, Small Effects What are we to make of all this? Firstly, these studies confirm that the estimated effects for the genetic factors relevant to BP are small. This observation, combined with the inherent variability of BP, explain, why it has been so difficult in GWAS to identify loci with genome-wide significance. Secondly, the alleles associated with BP were reasonably common, with average minor allele frequencies around 20% to 30%. Thirdly, the loci reported seem to interact in an additive fashion. The CHARGE investigators plotted average pressure levels as a weighted function of the number of BP alleles carried. This showed a linear increase of about 8 mm Hg across the range as alleles accumulated. Therefore, on average, there did not seem to be multiplicative interactions between loci that might have caused a more exponential increase in pressure for increasing alleles. There must be some caution in interpreting these analyses, as they were made in the same subjects from which the markers were discovered and results in independent populations might differ. However, the patterns seen in three studies are entirely consistent with the polygenic hypothesis of common conditions such as hypertension, which can be considered (at least in genetic terms) the result of inheriting a number of predisposing alleles, each of which is common and each of which exerts small BP effects individually. BP is truly the sum of many genetic parts.
The Missing Genes However, the loci reported in CHARGE and GlobalBPgen explain only about 1% of total variation in BP, or about 2% of
Table A list of candidate genes that have been associated at various times with blood pressure 11BHSD2 ADD2 ATP1A1 CAV1 CLCNKB CYP11B1 CYP4A11 ESR1 FYN GRK-4 ITGB2 NCOA1 PTGIS SCNN1B SLC12A1 SPON1 WNK4
5HT2A ADRB2 ATP2A2 CCR2 CNR2 CYP11B2 DRD-1 ESR2 GCH1 HNF1 KCNJ1 NEDD4L QPCT SCNN1G SLC12A3 TH
ABCA1 AGT BMAL1 CCR5 COL1A2 CYP11B2 EDNRA ET1 GCK IL6 KIAA0789 NR3C2 REN SGK1 SLC4A5 TSC
ACE AGTR1 CAPN10 CD31 COMT CYP19A1 ELN FAS GNB3 INPPL1 M6PR PDE4D ROS1 SHIPS2 SLC7A1 UMOD
ADD1 AMPD1 CASR CHGB CORIN CYP3A5 NOS3 FGF1 GPX1 ITGA2 MTP PNMT SAH SLC SLC9A2 WNK1
the genetic variance. Where then are the missing genetic factors? One possibility is that there exist other less potent loci that have escaped the net cast by the strict statistical definitions of significance. These could be loci for which BP effects in these populations are smaller than 0.5 mm Hg. If this is the case, it implies that there must be a multitude of such loci to make up the missing variance. The CHARGE investigators make the point that if one lowers one’s statistical sights to P < 1 103, one observes 890 more markers associated with BP than expected by chance. Perhaps some of the candidate genes that have been associated in smaller studies previously (Table) are in this group – a glimmer of hope for those candidate gene investigators who felt themselves shunned by the advent of GWAS.
The Discrepancies The next observation is that, just as in the early days of candidate genes, inconsistency and discrepancies still plague the genetics of BP. The markers that seriously failed the tests of replication cannot be simply dismissed as a consequence of moderate effect size, as suggested by the GlobalBPgen investigators. There was no clear relationship between effect size estimated in one study and the likelihood that a marker would fail to achieve even P < .05 in the other. Such divergence raises important questions.20 Among the possible causes is heterogeneity in a variety of factors such as differences in genotyping platforms, different marker sets, and a whole range of factors associated with underlying population heterogeneity. These could include different patterns of linkage disequilibrium, which means that as result of different genetic architecture, markers in a particular region are linked to the causative variant in one population, but not in another. There could
S.B. Harrap / Journal of the American Society of Hypertension 3(4) (2009) 231–237
also be population-specific differences in genetic and environmental interactions that might obscure or highlight a particular locus in one or other population. Given the fact that both studies were set in subjects of European descent in both Europe and the United States, it is hard to invoke population differences as an explanation for the divergence. One last possibility that can’t be excluded is that even studies of 30,000 subjects can be misleading.
Genetic Screening? For those dreaming of the day when the routine health screening will involve tests of markers for high BP, these studies are not encouraging. It is difficult to envisage that markers that account for 1 mm Hg or less of BP and can be difficult to replicate between populations and even within the constituent studies of meta-analyses might be useful and reliable for personal prediction. The gold standard remains careful measurement with a BP cuff and taking a good family history (or better still measurement of the parents’ BP).
Where Next? In some ways, the genetics of BP is at a fork in the road. One might choose to continue down the path of pinning down more and more genetic loci. However, CHARGE and GlobalBPgen are telling us that we would need even bigger studies to pick up loci with BP effects of less than 0.5 mm Hg. This becomes a case of diminishing returns, particularly as the numbers of subjects required for such studies rise exponentially as the phenotypic effects gets smaller.
Time to Reflect and Focus But genetics should be about understanding, not about cataloguing. It is time to pause and expend effort mining the discoveries to date. This begins with identifying the precise location and nature of the deoxyribonucleic acid (DNA) variation that explains these BP associations. This is critical to revealing and understanding mechanisms against which, hopefully novel, preventive, and treatment strategies might be targeted. For example, the first priority would no doubt be on chromosome 10. Both CHARGE and GlobalBPgen highlighted this locus, but had their own top markers at either end of an associated stretch of DNA about 430 kb long. The top CHARGE marker was in the gene CYP17A1 that encodes the enzyme steroid 17a-hydroxylase – a gene with good credentials in the Mendelian BP world. The top GlobalBPgen marker was closer to the gene NT5C2 that encodes a 5’-nucleotidase acting as a cytoplasmic enzyme involved in the maintenance of intracellular purine/pyrimidine nucleotides.
235
Think Laterally Given the experiences with recent discoveries, most notably that for myocardial infarction, one should not assume that causative variants would be in a gene of physiological relevance or even close to any gene per se. There is a good deal of variation in the DNA sequence between CYP17A1 and NT5C2 and careful thought will need to be given to the best strategy for recognizing the highest priority candidates and the most appropriate means to validate their status. Searching across intergenic regions for changes in DNA sequence that might alter gene expression to produce quantitative effects (as opposed to changing protein sequence and causing qualitative effects) is extremely difficult, especially as we cannot recognize many such sequences at this time. Another marker worthy of detailed investigation might be that on chromosome 12 close to the gene ATP2B1. This was the most significant marker in CHARGE, which had also recently been revealed as associated with BP in a reasonably large study in Korea.21 This gene encodes a plasma membrane calcium/calmodulin-dependent ATPase found in vascular endothelium and implicated in animal models of hypertension. The validation of causative DNA variants will demand a careful and strategic combination of molecular genetics and cellular biology, biochemistry, and physiology to trace the path of influence from DNA through ribonucleic acid proteins, cells, tissues, organs to man. And one needs to be careful in piecing these elements together. For example, the locus on chromosome 1 identified in GlobalBPgen for SBP is closest to the methylenetetrahydrofolate reductase (MTHFR) gene. It is also in the vicinity of the NPPA and NPPB genes that had been associated very strongly with differences in levels of plasma atrial natriuretic peptide and less so with BP.22 However, results from GlobalBPgen raise the possibility that the NPPA and NPPB genes are not the outstanding contributors for BP in this region. Therefore, although it might be tempting to associate the BP and peptide changes as logical consequences of variation in NPPA and NPPB genes, such effects might be coincidental, with the BP being the result of variation towards MTHFR.
Is It Worth the Effort? Some might argue that we are better off putting money into other aspects of cardiovascular disease prevention, as the genetic loci are associated with such small effects. However, it would be a mistake to relegate genetics on these grounds. Firstly, until the precise DNA variants are identified, it is impossible to know the real impact on BP. Even a GWAS using 500,000 markers is pretty sparse considering the 3 billion base pairs in the human chromosome. It is likely that most markers will be at a distance
236
S.B. Harrap / Journal of the American Society of Hypertension 3(4) (2009) 231–237
from the causative variants and that physical separation can weaken the apparent BP effect, as judged by the marker alone. Secondly, it needs to be emphasized that the magnitude of BP effect resulting from common DNA variation does not necessarily reflect the importance of the physiological control mechanism influenced by the variant. A locus associated with 0.5 to 1 mm Hg of SBP says nothing about the robustness of the system with which the locus interacts. These small clues might be just the tip of the iceberg. More important is the nature of the iceberg. Indeed, the excitement of GWAS is not the discovery of loci, it is the potential to define undiscovered physiological pathways that will offer new approaches to prevention and treatment and augment existing therapeutic and health promotion strategies. Finally, there is a real possibility that a diversity of DNA variants will have effects that converge on a common physiological mechanism.23 Therefore, if several loci, each with an effect of 0.5 to 1 mm Hg mediate their effects through the same physiological focal point, such a point becomes a prime target for intervention.
7.
8.
9.
10.
11.
Conclusion This is a fascinating time for the genetics of BP. We need to recognize that BP is a unique phenotype and more difficult than most when it comes genetic discovery. Perhaps we can do both, but if we need to choose between accumulating more markers in the BP catalogue vs. learning exactly how the existing markers are implicated, the latter focused approach is the one that will bring us nearer to the ultimate goal of making genetics of use to our communities and our patients.
References 1. Harrap SB. Where are all the blood pressure genes? Lancet 2003;361:2149–51. 2. Jeunemaitre X, Soubrier F, Kotelevtsev YV, Lifton RP, Williams CS, Charru A, et al. Molecular basis of human hypertension: role of angiotensinogen. Cell 1992;71: 169–80. 3. Weiss KM, Terwilliger JD. How many diseases does it take to map a gene with SNPs? Nat Genet 2000;26: 151–7. 4. Cardon LR, Bell JI. Association study designs for complex diseases. Nat Rev Genet 2001;2:91–9. 5. Caulfield M, Munroe P, Pembroke J, Samani N, Dominiczak A, Brown M, et al. Genome-wide mapping of human loci for essential hypertension. Lancet 2003; 361:2118–23. 6. Province MA, Kardia SLR, Ranade K, Rao DC, Thiel BA, Cooper RS, et al. A meta-analysis of genomewide linkage scans for hypertension: The National
12.
13.
14.
15.
16.
17.
18.
19.
Heart Lung and Blood Institute Family Blood Pressure Program. Am J Hypertens 2003;16:144–7. Levy D, Larson MG, Benjamin EJ, Newton-Cheh C, Wang TJ, Hwang SJ, et al. Framingham Heart Study 100K Project: genome-wide associations for blood pressure and arterial stiffness. BMC Med Genet 2007;8:S3. Wang Y, O’Connell JR, McArdle PF, Wade JB, Dorff SE, Shah SJ, et al. Whole-genome association study identifies STK39 as a hypertension susceptibility gene. PNAS 2009;106:226–31. Sabatti C, Service SK, Hartikainen AL, Pouta A, Ripatti S, Brodsky J, et al. Genome-wide association analysis of metabolic traits in a birth cohort from a founder population. Nat Genet 2009;41:35–46. Gu CC, Hunt SC, Kardia S, Turner ST, Chakravarti A, Schork N, et al. An investigation of genome-wide associations of hypertension with microsatellite markers in the Family Blood Pressure Program. Hum Genet 2007; 121:577–90. Burton PR, Clayton DG, Cardon LR, Craddock N, Deloukas P, Duncanson A, et al. Wellcome Trust Case Control Consortium. Nature 2007;447:661–78. Harrap SB, Stebbing M, Hopper JL, Hoang HN, Giles GG. Familial patterns of covariation for cardiovascular risk factors in adults: The Victorian Family Heart Study. Am J Epidemiol 2000;152:704–15. Ji W, Foo JN, O’Roak BJ, Zhao H, Larson MG, Simon DB, et al. Rare independent mutations in renal salt handling genes contribute to blood pressure variation. Nat Genet 2008;40:592–9. Levy D, Ehret GB, Rice K, Verwoert GC, Launer LJ, Dehghan A, et al. Genome-wide association study of blood pressure and hypertension. Nat Genet 2009;41: 677–87. Newton-Cheh C, Johnson T, Gateva V, Tobin MD, Bochud M, Coin L, et al. Genome-wide association study identifies 8 loci associated with blood pressure. Nat Genet 2009;41:666–76. Tunstall-Pedoe H, Connaghan J, Woodward M, Tolonen H, Kuulasmaa K. Pattern of declining blood pressure across replicate population surveys of the WHO MONICA project, mid-1980s to mid1990s, and the role of medication. BMJ 2006;332: 629–32. McGurk C, Nugent A, McAuley D, Silke B. Sources of inaccuracy in the use of the Hawksley random-zero sphygmomanometer. J Hypertens 1997;15:1379–84. Brown WCB, Kennedy S, Inglis GC, Murray LS, Lever AF. Mechanisms by which the Hawksley random zero sphygmomanometer underestimates blood pressure and produces a non-random distribution of RZ values. J Human Hypertens 1997;11:75–93. Rose G. Strategy of prevention: lessons from cardiovascular disease. Br Med J Clin Res Ed 1981;282:1847–51.
S.B. Harrap / Journal of the American Society of Hypertension 3(4) (2009) 231–237
20. McCarthy MI, Abecasis GR, Cardon LR, Golstein DB, Little J, Ioannidis JPA, et al. Genome-wide association studies for complex traits: consensus, uncertainty, and challenges. Nat Rev Genet 2008;9:356–69. 21. Cho YS, Go MJ, Kim YJ, Heo JY, Oh JH, Ban HJ, et al. A large-scale genome-wide association study of Asian populations uncovers genetic factors influencing 8 quantitative traits. Nat Genet 2009;41:527–34.
237
22. Newton-Chen C, Larson MG, Vasan RS, Levy D, Bloch KD, Surti A, et al. Association of common variants in NPPA and NPPB with circulating natriuretic peptides and blood pressure. Nat Genet 2009;41: 348–53. 23. Harrap SB, Petrou S. Utility of genetic approaches to common cardiovascular diseases. Am J Physiol 2001; 281:H1–6.