Genome Scans for Blood Pressure and Hypertension
The National Heart, Lung, and Blood Institute Family Heart Study
This study presents genome scans for hypertension and blood pressures from 2959 individuals in 500 white families from the National Heart, Lung, and Blood Institute Family Heart Study. Genome scans were performed with different methods of handling the 27% of individuals taking antihypertensive medications. Variance components LOD scores were estimated by assigning medicated hypertensive individuals (1) to have a blood pressure of 140/90; (2) to be missing; and (3) to have a randomly assigned systolic blood pressure between 140 and 160 (N[150,5] distribution) and diastolic blood pressure between 90 and 100 mm Hg (N[95,2.5] distribution). There were 5 regions with heterogeneity LOD scores ≥2.0 for hypertension (unadjusted for multiple models): 2.8 on chromosome 1 (192 cM), 2.6 on chromosome 7 (58 cM), 2.0 on chromosome 7 (127 cM), 2.4 on chromosome 12 (83 cM), and 2.4 on chromosome 15 (103 cM). Diastolic blood pressure had no LOD scores ≥2.0. Only chromosome 6 showed linkage for systolic blood pressure, with a LOD score of 3.3 at 88.7 cM from the initial randomization. Multiple randomizations of medicated subjects’ systolic blood pressures yielded a mean LOD score of 2.8±0.4, whereas setting medicated systolic blood pressures to 140 mm Hg yielded a LOD score of 3.3. Excluding the medicated individuals or using their treated blood pressures reduced the LOD scores to 0.8 and 1.3, respectively, indicating the importance of including the extremes of quantitative trait distributions in linkage analyses. These results overlap other published scans, particularly regions on chromosomes 1 and 6, which have been implicated in familial combined hyperlipidemia.
Anumber of studies of genome scans for blood pressure or hypertension have been published or reported.1–13⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓⇓ Surprisingly, there were few overlapping regions among the initial studies that appeared to be linked to blood pressure or hypertension. Although it is expected that with the addition of more genome search studies, some regions would begin to overlap by chance, concordance of multiple studies for a specific region may still provide the best evidence that the linkage is not a false-positive result. Part of the problem in comparing these studies is that the phenotype definition often differs across studies (hypertension versus blood pressure level). In addition, family ascertainment methods differ, and the sampling unit could include only sibpairs or extended pedigrees. However, it is unknown what effect these differences have on the resulting LOD scores. This study uses data from the National Heart, Lung, and Blood Institute (NHLBI) Family Heart Study to examine linkage to both hypertension and blood pressure in pedigrees. Although family studies of blood pressure will necessarily include persons taking antihypertensive medications because of the high prevalence of hypertension, the decision about how to analyze treated blood pressures may significantly affect the quantitative linkage results. Comparisons of results with different methods of handling medication effects on blood pressure and the degree of family aggregation of hypertension are made to provide insights into the resulting effects on the linkage results. The linkage of blood pressure and hypertension near loci linked to familial combined hyperlipidemia indicates that they also may be involved in the familial dyslipidemic hypertension syndrome14 and may suggest mechanisms of gene action.
The NHLBI Family Heart Study is a multicenter, population-based study of genetic and nongenetic determinants of coronary heart disease, atherosclerosis, and cardiovascular risk factors.15 Families were recruited by using probands identified from previously studied population-based cohorts from the Framingham Heart Study, the Forsyth County and Minneapolis Atherosclerosis Research in Communities (ARIC) study centers, and the Utah Health Family Tree Study. The study was approved by each institutional review board, and informed consent was obtained from all subjects. A total of 5975 subjects from 1280 families was recruited, and the Mammalian Genotyping Service16 genotyped the 401 largest pedigrees (402 markers, set 10). The average heterozygosity was 0.76 and average intermarker distance was 9.5 cM. A set of 240 markers was also genotyped at the University of Utah on a subset of these pedigrees, placed on the Marshfield map (http://research.marshfieldclinic.org/genetics), and the combined map was used for analysis. The number of white subjects with blood pressures and genotypes was 2959 from families of sizes ≤5, 58%; 6 to 10, 36%; and 11 to 18, 6%.
Sitting blood pressures were measured by random-zero sphygmomanometry; the mean of the last 2 of 3 measurements was used for analysis. All current medications were recorded for each subject and coded according to class of medication. Hypertension was defined as a blood pressure >140/90 mm Hg or the patient currently taking antihypertensive medications.
Blood pressures from unmedicated subjects were adjusted in a multiple regression model for field center, gender, age, age,2 age,3 and body mass index before the linkage analysis. Both quantitative and qualitative multipoint linkage analyses were performed, coding hypertensive status and medicated blood pressure levels in multiple ways to compare the consistency of the results. One way of handling medicated blood pressure levels in quantitative analyses is conservatively assuming that all persons had a minimum blood pressure of 140/90 mm Hg at the time of treatment onset, although for any individual either the systolic or diastolic measurement need not have been that high. Two concerns are the spike in the distribution that is created by fixing a large number of subjects to the same value and forcing hypertensive siblings to the same value, thereby increasing their phenotypic similarity. A second approach conservatively assumed that all subjects taking antihypertensive medications were stage I at diagnosis. Systolic and diastolic blood pressures were assigned for each medicated hypertensive subject by creating a random number from a normal distribution with a mean±SD value of 150±5 for systolic and 95±2.5 mm Hg for diastolic blood pressure (midpoints of stage I hypertension). Thus, 95% of all randomized values are expected to fall within a 140 to 160/90–100 mm Hg range (stage I). Any randomly generated blood pressure falling outside the range of 140 to 160 mm Hg systolic or 90 to 100 mm Hg diastolic was set to the closest boundary of these intervals to prevent extreme values from affecting the linkage analysis. Using a normal distribution rather than a log-normal or left-truncated normal distribution was a conservative approach for linkage analysis because extreme blood pressures were less likely to be generated. The assigned blood pressure values from the first randomization for the medicated individuals were used for a full genome search. Skewness and kurtosis after the first randomization was 0.28 and −0.89 for systolic and 0.26 and −1.0 for diastolic blood pressure. Multiple randomizations were used only on the primary linkage detected because the CPU time required for performing a large number of genome searches on each randomized set of values was prohibitive. Other approaches were to exclude all medicated individuals from the analysis or to use their actual medicated blood pressures. GENEHUNTER 2 was used for the quantitative trait analyses assuming no dominance effects and using all independent pairs in the variance components model.17
For the qualitative analyses, Allegro18 was used, which performs both model-based and nonparametric analyses. In the qualitative analyses, we analyzed the data two ways. First, we coded all hypertensive subjects in the pedigrees as affected. Second, we identified all sibships within each pedigree with two or more hypertensive siblings and coded these individuals as affected. Other hypertensive relatives who did not have a sibling with hypertension were coded as having unknown phenotypes (thus, uninformative for linkage) for the purpose of limiting possible sporadic cases of this common condition. Both dominant and recessive models were used in addition to the nonparametric model. Heterogeneity LOD scores are presented for the parametric models. Gene frequencies were arbitrarily selected as 0.01 for the dominant model and 0.14 for the recessive model, yielding a population prevalence of hypertension due to the genetic locus of 2% for both models if full penetrance is assumed. However, penetrance parameters of 70% for gene carriers and 5% for nongene carriers were used in the analysis. Allele frequencies were estimated from 200 family probands who were all randomly selected from each field center population. The various genome searches performed in this study are viewed as hypothesis-generating analyses. Therefore, significance of the LOD scores is not assessed, nor are the LOD scores adjusted for the multiple analysis models. Power of the data set was assessed by simulation of a trait with 30% total trait heritability (typical for blood pressure). At a 5% significance level, there was 80% power to detect a locus-specific heritability of 12% with the use of variance components methodology.
Table 1 shows the characteristics of the 2959 subjects included in this study. Most individuals who were hypertensive were currently prescribed antihypertensive medications (82%). The subjects, who averaged 52 years of age, were moderately overweight.
Table 2 shows that only chromosome 6 had a maximum LOD score >2 for systolic blood pressure. The Figure shows 3 maximum LOD score plots of chromosome 6 for the quantitative analysis of systolic blood pressure. The line with the highest maximum LOD score (3.3) resulted from setting all treated hypertensive systolic blood pressures to 140 mm Hg. The line with the lowest maximum LOD score (0.8) resulted from excluding all subjects taking antihypertensive medications. If actual medicated blood pressures were analyzed, the maximum LOD score was 1.3 (not shown). The bold line in the middle was the mean LOD score of all randomized runs, which was 2.8±0.5, with a minimum of 2.0 and a maximum of 3.6. Each randomization showed that the peak maximum LOD score occurred at marker D6S1031 (ATA28B11), located at 88.7 cM. All randomizations also showed the smaller peak to the right of the main peak. The average percentage of the total phenotypic variance explained by the locus over all randomizations was 18.4%, with another 14.4% of the variance due to polygenes.
No chromosomal regions had maximum LOD scores >2.0 for diastolic blood pressure (Table 2). Regions with LOD scores between 1 and 2 overlapping similar LOD scores within a 50-cM interval for systolic blood pressure included chromosomes 1, 2, 6, 8, 12, and 13.
Qualitative Hypertension Results
The first nonparametric LOD-score column in Table 3 shows the results when all hypertensive individuals were coded as affected. The next column shows the results when only hypertensive individuals with siblings with hypertension were coded as affected. All relatives were included for this assessment even if they were not genotyped. Although no clear LOD-score pattern was found for the two methods of analysis, the most promising regions were chromosomes 7, 12, and 15. Parametric data analysis with the use of both a dominant and a recessive model also identified chromosomes 7 and 12 as promising regions (Table 4). The dominant model appeared best for chromosome 7, whereas the recessive model appeared better for chromosome 12. In addition, chromosome 1 had a LOD score >2 for the dominant model. This occurred at a location that was 28 cM telomeric from the nonparametric results. Chromosome 15 became weaker and the LOD score maximized at a location different than with the nonparametric model.
Various studies have now been published examining linkage to either blood pressure as a quantitative trait or to hypertension as a discrete trait. An important unresolved question is whether there are genes that influence normal variation in blood pressure but may be distinct from genes that lead to the clinical definition of hypertension. Although this question is difficult to address, this study, with a large sample size of genotyped subjects and a large proportion of medicated hypertensive subjects, has performed genome searches for both phenotypes, allowing a comparison of the results in the same set of pedigrees. The largest maximum LOD score in this study was from the quantitative analysis of systolic blood pressure. There were several regions >2.0 for hypertension. It is possible that many of the LOD scores >2.0 for hypertension are merely false-positive results and do not occur with a greater frequency than for blood pressure. The likelihood that some of these regions may be false-positive regions may be indirectly assessed, albeit in a very crude way, by comparing the LOD scores from different subsets of patients to look for consistency. Consistency by itself does not rule out false-positive results, since a shared genomic region in all families may be linked to a phenotype merely by chance. The results of Table 3 suggest that many of these regions are consistent across different definitions of hypertension aggregation. There is only somewhat more variability in the results when nonparametric and parametric models were used for linkage of hypertension (Table 4). It should be remembered that the LOD scores presented were not adjusted for the multiple models and phenotypes analyzed. For example, if 8 independent models are run, LOD scores should be reduced by log8, or 0.9 LOD units.
It does not appear that using a stronger definition of familiality, such as two siblings with hypertension as opposed to any relative with hypertension, increased the strength of the linkages seen, although fewer affected individuals analyzed in the former analysis may have offset a possible increased effect size in the smaller subset, resulting in equivalent power. Results for chromosomes 1 and 7 were increased to >2.0 when parametric models were used, whereas chromosome 12 remained about the same and chromosome 15 showed decreased linkage. Therefore, despite the multiple methods of analysis, the main linkage results can be summarized without strict attention to the underlying model.
Of more importance to the analysis, was how the large proportion of subjects taking antihypertensive medications was analyzed. It is clear that deleting medicated individuals from the analysis greatly diminishes power. The maximum LOD score for systolic blood pressure on chromosome 6 was dramatically reduced from the range of 2.8 to 3.3 to 0.8 after exclusion of these subjects. This indicates that variation contributed from the upper extremes of systolic blood pressure is essential in detecting linkage in this data set. It also suggests that any underlying gene in this region that explains the variation in systolic blood pressure is also related to the development of hypertension since the result requires the hypertensive subjects to be included. The maximum LOD scores from qualitative analyses of this region varied around 1.0, suggesting greater information in the quantitative analysis as long as the medicated subjects were included.
The randomization of systolic blood pressure was a conservative measure to estimate linkage because it introduced random variation within families with hypertensive individuals. Assuming that all subjects were stage I instead of allowing some to be stage II may have reduced the variance between affected and unaffected relatives, also reducing the LOD score. The LOD score obtained when all subjects taking antihypertensive treatment were fixed to have a systolic blood pressure of 140 mm Hg was well inside the LOD score distribution obtained from multiple randomizations, suggesting that the presence of a large spike in the expected normal distribution did not greatly affect the results in this instance. Rather than randomize medicated blood pressures, the Framingham study used a nonparametric censored distribution approach that may better adjust the medicated blood pressures, especially if the stratum-specific unmedicated sample size is very large for proper adjustment.4
Overlap With Published Linkage Regions
Assessing the concordance of linkage regions across published studies is essential prior to investing the extremely intensive and costly resources to isolate mutations in genes responsible for the linkage results. Obviously, the more studies that are published, the more concordance one would expect to see across the genome. However, the agreement of a large number of studies on the same interval tends to supply confidence in that region. The results found in this study overlap some important regions previously seen in other studies.
There were two regions on chromosome 1 that are of interest. First, a LOD score >1.0 was found for all three models for hypertension and for systolic blood pressure in the range of 164 to 199 cM. The highest LOD score of 2.8 for hypertension occurred with the dominant parametric model at 192 cM (ATA4E02). Another study has shown maximum LOD scores for hypertension also occurring on chromosome 1 at 192 cM.6 This region is close to a region (176 to 178 cM, D1S104, D1S1677) implicated in familial combined hyperlipidemia by three other studies.19–21⇓⇓ One of these studies was an analysis of 71 familial combined hyperlipidemia (FCHL) families (170 affected subjects) from the NHLBI Family Heart Study,20 indicating that within a subset of the same data set both FCHL and hypertension/blood pressure are linked to this region. A mouse gene in the orthologous region has now been cloned that alters the cellular redox state and elevates triglyceride and free fatty acid levels.22 If the gene(s) underlying these regions are the same, it implies that the genes have pleiotropic effects on both blood pressure and lipids leading to familial dyslipidemic hypertension, a common syndrome.14 However, there is a large number of genes in this wide interval, and it may be unlikely that the same gene is responsible for both conditions. A study of type II diabetes (170 cM, near Apo A-II) and HDL-C (198 cM, near D1S408) also have found linkage to this region.23,24⇓ Because of previous experience that LOD score peaks can move significantly along the chromosome as data are added to an analysis or different subsets are analyzed, one cannot exclude the possibility that the FCHL and blood pressure linkage regions are different regions harboring genes specific to each phenotype and are not pleiotropic. There was little evidence in the Family Heart Study data set for linkage to triglyceride or LDL-C levels (unpublished results). The second region on chromosome 1 occurred over the angiotensinogen gene for both systolic and diastolic blood pressure with LOD scores >1.0 (Table 2).
We found a LOD score of 2.0 in all hypertensive subjects and 1.4 in only hypertensive sibpairs on chromosome 7 at 115 cM. This is near another familial dyslipidemia locus recently identified for blood pressure on chromosome 7 near 112 cM (LOD 2.5), as both fasting insulin and systolic blood pressure both appear to link to this region in Hispanic families.10 Mild LOD scores on chromosome 8 replicate linkage of blood pressure to the area around the LPL gene, which shows more significant results when hyperinsulinemic families are studied.13,25⇓ Other chromosomal areas of interest are chromosomes 12 and 15, although there is currently no compelling information that would also link these regions to dyslipidemia.
Chromosome 6 showed the best evidence in our study for a linked gene to systolic blood pressure with a mean LOD score from the randomized analyses of 2.8 at marker D6S1031 at 88.7 cM on the Marshfield map. This LOD score is only suggestive of linkage, especially after adjustment for multiple models analyzed. A recent genome scan analysis of blood pressure in FCHL families showed a LOD score of 2.5 for diastolic blood pressure, with a peak nearest to the same marker that showed linkage in our study, D6S1031.13 They also found concordant linkage of systolic blood pressure and free fatty acids to a region on chromosome 4p, with a LOD score of 4.6 for systolic blood pressure (≈35 to 40 cM). These results again point to the genetic concordance of dyslipidemia and hypertension within the same data set. Our data showed only weak LOD scores (0.5 to 0.7) ≈60 to 66 cM.
We conclude that the two best linkage signals for hypertension or blood pressure in the large NHLBI Family Heart Study, chromosomes 1 and 6, localize near regions linked to FCHL. Results on chromosomes 4 and 8 may support other pleiotropic genes for hypertension and dyslipidemia. Although different statistical analysis methods and family selection criteria showed only minor differences in results, how one handles medicated individuals is extremely important.
This study is presented on behalf of the investigators of the NHLBI Family Heart Study. Participating institutions and principal staff of the study are as follows: Forsyth County/University of North Carolina/Wake Forest University: Gerardo Heiss, Stephen Rich, Greg Evans, James Pankow, H.A. Tyroler, Jeannette T. Bensen, Catherine Paton, Delilah Posey, and Amy Haire; University of Minnesota Field Center: Donna K. Arnett, Aaron R. Folsom, Larry Atwood, James Peacock, and Greg Feitl; Boston University/ Framingham Field Center: R. Curtis Ellison, Richard H. Myers, Yuqing Zhang, Andrew G. Bostom, Luc Djoussé, Jemma B. Wilk, and Greta Lee Splansky; University of Utah Field Center: Steven C. Hunt, Roger R. Williams (deceased), Paul N. Hopkins, Hilary Coon, and Jan Skuppin; Coordinating Center, Washington University, St. Louis: Michael A. Province, D.C. Rao, Ingrid B. Borecki, Yuling Hong, Mary Feitosa, Jeanne Cashman, and Avril Adelman; Central Biochemistry Laboratory, University of Minnesota: John H. Eckfeldt, Catherine Leiendecker-Foster, Michael Y. Tsai, and Greg Rynders; Central Molecular Laboratory, University of Utah: Mark F. Leppert, Jean-Marc Lalouel, Tena Varvil, Lisa Baird; National Heart, Lung, and Blood Institute–Project Office: Phyliss Sholinsky, Millicent Higgins (retired), Jacob Keller (retired), Sarah Knox, and Lorraine Silsbee.
This work was supported by National Heart, Lung, and Blood Institute cooperative agreement grants U01 HL56563, U01 HL56564, U01 HL56565, U01 HL56566, U01 HL56567, U01 HL56568, and U01 HL56569.
↵*A list of the NHLBI Family Heart Study participating institutions and principal staff is given in the Appendix.
- Received February 14, 2002.
- Accepted May 8, 2002.
Krushkal J, Ferrell R, Mockrin SC, Turner ST, Sing CF, Boerwinkle E. Genome-wide linkage analyses of systolic blood pressure using highly discordant siblings. Circulation. 1999; 99: 1407–1410.
Rice T, Rankinen T, Province MA, Chagnon YC, Perusse L, Borecki IB, Bouchard C, Rao DC. Genome-wide linkage analysis of systolic and diastolic blood pressure: the Quebec Family Study. Circulation. 2000; 102: 1956–1963.
Levy D, DeStefano AL, Larson MG, O’Donnell CJ, Lifton RP, Gavras H, Cupples LA, Myers RH. Evidence for a gene influencing blood pressure on chromosome 17: genome scan linkage results for longitudinal blood pressure phenotypes in subjects from the Framingham heart study. Hypertension. 2000; 36: 477–483.
Sharma P, Fatibene J, Ferraro F, Jia H, Monteith S, Brown C, Clayton D, O’Shaughnessy K, Brown MJ. A genome-wide search for susceptibility loci to human essential hypertension. Hypertension. 2000; 35: 1291–1296.
Hsueh WC, Mitchell BD, Schneider JL, Wagner MJ, Bell CJ, Nanthakumar E, Shuldiner AR. QTL influencing blood pressure maps to the region of PPH1 on chromosome 2q31–34 in Old Order Amish. Circulation. 2000; 101: 2810–2816.
Pankow JS, Rose KM, Oberman A, Hunt SC, Atwood LD, Djousse L, Province MA, Rao DC. Possible locus on chromosome 18q influencing postural systolic blood pressure changes. Hypertension. 2000; 36: 471–476.
Rankinen T, An P, Rice T, Sun G, Chagnon YC, Gagnon J, Leon AS, Skinner JS, Wilmore JH, Rao DC, Bouchard C. Genomic scan for exercise blood pressure in the Health, Risk Factors, Exercise Training and Genetics (HERITAGE) Family Study. Hypertension. 2001; 38: 30–37.
Cheng LS, Davis RC, Raffel LJ, Xiang AH, Wang N, Quinones M, Wen PZ, Toscano E, Diaz J, Pressman S, Henderson PC, Azen SP, Hsueh WA, Buchanan TA, Rotter JI. Coincident linkage of fasting plasma insulin and blood pressure to chromosome 7q in hypertensive Hispanic families. Circulation. 2001; 104: 1255–1260.
Allayee H, de Bruin TW, Michelle Dominguez K, Cheng LS, Ipp E, Cantor RM, Krass KL, Keulen ET, Aouizerat BE, Lusis AJ, Rotter JI. Genome scan for blood pressure in Dutch dyslipidemic families reveals linkage to a locus on chromosome 4p. Hypertension. 2001; 38: 773–778.
Williams RR, Hunt SC, Hopkins PN, Stults BM, Wu LL, Hasstedt SJ, Barlow GK, Stephenson SH, Lalouel JM, Kuida H. Familial dyslipidemic hypertension: evidence from 58 Utah families for a syndrome present in approximately 12% of patients with essential hypertension. JAMA. 1988; 259: 3579–3586.
Higgins M, Province M, Heiss G, Eckfeldt J, Ellison RC, Folsom AR, Rao DC, Sprafka JM, Williams R. NHLBI Family Heart Study: objectives and design. Am J Epidemiol. 1996; 143: 1219–1228.
Weber JL, Broman KW. Genotyping for human whole-genome scans: past, present, and future.In: Advances in Genetics. New York, NY: Academic Press, Inc; 2001: 77–96.
Coon H, Myers RH, Borecki IB, Arnett DK, Hunt SC, Province MA, Djousse L, Leppert MF. Replication of linkage of familial combined hyperlipidemia to chromosome 1q with additional heterogeneous effect of apolipoprotein A-I/C-III/A-IV locus. The NHLBI Family Heart Study Arterioscler Thromb Vasc Biol. 2000; 20: 2275–2280.
Elbein SC, Hoffman MD, Teng K, Leppert MF, Hasstedt SJ. A genome-wide search for type 2 diabetes susceptibility genes in Utah Caucasians. Diabetes. 1999; 48: 1175–1182.
Coon H, Leppert MF, Eckfeldt JH, Oberman A, Myers RH, Peacock JM, Province MA, Hopkins PN, Heiss G. Genome-wide linkage analysis of lipids in the hypertension genetic epidemiology network (HyperGEN) blood pressure study. Arterioscler Thromb Vasc Biol. 2001; 21: 1969–1976.
Wu D-A, Bu X, Warden CH, Shen DDC, Jeng C-Y, Sheu WHH, Fuh MMT, Katsuya T, Dzau VJ, Reaven GM, Lusis AJ, Rotter JI, Chen Y-DI. Quantitative trait locus mapping of human blood pressure to a genetic region at or near the lipoprotein lipase gene on chromosome 8p22. J Clin Invest. 1996; 97: 2111–2118.