A genome-wide linkage scan identifies multiple quantitative trait loci for HDL-cholesterol levels in families with premature CAD and MI.

Plasma HDL cholesterol levels (HDL-C) are an independent predictor of coronary artery disease (CAD). We have completed a genome-wide linkage scan for HDL-C in a US cohort consisting of 388 multiplex families with premature CAD (GeneQuest). The heritability of HDL-C in GeneQuest was 0.37 with gender and age as covariates (P = 5.1 × 10−4). Two major quantitative trait loci (QTL) for log-transformed HDL-C adjusted for age and gender were identified onto chromosomes 7p22 and 15q25 with maximum multipoint logarithm of odds (LOD) scores of 3.76 and 6.69, respectively. Fine mapping decreased the 7p22 LOD score to a nonsignificant level of 3.09 and split the 15q25 QTL into two loci, one minor QTL on 15q22 (LOD = 2.73) that spanned the LIPC gene, and the other at 15q25 (LOD = 5.63). A family-based quantitative transmission disequilibrium test (QTDT) revealed significant association between variant rs1800588 in LIPC and HDL-C in the GeneQuest population (P = 0.0067), which may account for the minor QTL on 15q22. The 15q25 QTL is the most significant locus identified for HDL-C to date, and these results provide a framework for the ultimate identification of the underlying HDL-C variant and gene on chromosomes 15q25, which will provide insights into novel regulatory mechanisms of HDL-C metabolism.

if the extreme phenotypes of interest are due to the segregation of many relatively rare alleles at that locus. Furthermore, whereas allelic associations can be due to spurious causes, especially heterogeneity/population stratifi cation, linkage analysis is not subject to such type 1 errors. To date, multiple quantitative trait loci (QTLs) have been identifi ed that show strong evidence of linkage for HDL-C levels (summarized in Table 1 ). Recent genome-wide association studies (GWAS) also identifi ed multiple loci represented by various single nucleotide polymorphisms (SNP) associated with HDL-C (13)(14)(15)(16)(17)(18). The previous work demonstrates the complex inheritance of genetic factors that infl uence this major CAD risk factor.
In the present study, we describe a whole genome linkage scan to identify chromosomal regions infl uencing HDL-C levels in families with premature CAD and MI. Five candidate QTLs were localized to chromosomes 3p25, 7p22, 13q12, 13q32, and 15q25 with maximum multipoint logarithm of odds (LOD) scores of 4.10, 4.21, 4.66, 3.95, independent predictor of atherosclerosis in both men and women ( 2,3 ). Families that have early onset, heritable CAD more frequently have low HDL-C than the general population. In men with CAD, low HDL-C is the most common lipid abnormality observed, affecting half the patients ( 4 ). Due to its anti-atherogenic properties, even small variations in HDL cholesterol levels are physiologically important. For each 1 mg/dl increase in HDL-C levels, there is a decrement of 2-3% in CAD risk ( 5 ). In addition to its prevalence in CAD patients, reduced HDL-C levels are a cornerstone of the metabolic syndrome because low HDL-C is associated with insulin resistance and abdominal obesity in humans ( 6 ). With the problem of obesity continuing to escalate in the United States, metabolic syndrome poses a major public health threat affecting 22% of the adult population ( 7 ). Because low HDL-C is prevalent in patients with both metabolic syndrome and CAD, the challenge of elucidating the causes of variation in HDL-C levels and discovering new drug treatments for the condition will continue to be critical.
Plasma HDL-C levels have a strong genetic component; approximately 50% of variation in human populations is due to genetic factors ( 8,9 ). While differences in plasma HDL-C have long been recognized to be controlled by genetic factors, our current understanding of the genetics of variation in HDL-C levels is largely based on studies of extreme monogenic HDL-C conditions. Although variation in genes caused by rare mutations may make some contribution to HDL-C levels ( 10,11 ), the majority of genetic variation in genes that control HDL-C levels in the general population has yet to be identifi ed ( 12 ). Identifi cation of the genes and genetic variants that control HDL-C concentrations are critical for preventative cardiology in reducing the public health burden of CAD and the metabolic syndrome.
Genome-wide linkage and association scans provide comprehensive and unbiased approaches to identify HDL genes and may lead to the elucidation of unrecognized genetic pathways in HDL metabolism. Genome-wide association is more powerful than genome-wide linkage analysis to detect common alleles at a locus, but it is less powerful   and 7.57, respectively. After log-transformation and adjustment of age and gender, 15q25 and 7p22 QTLs remained signifi cant with maximum multipoint LOD scores of 6.69 and 3.76, respectively. Further fi ne mapping resulted in the drop of the LOD score of 7p22 QTL to 3.09 and of 15q25 to 5.63. Interestingly, the promoter SNP of LIPC rs1800588 showed signifi cant association with HDL in the family-based quantitative transmission disequilibrium test (QTDT) analysis. To the best of our knowledge, the 15q25 QTL is the most signifi cant locus identifi ed for HDL-C to date. Our study provides a framework for the ultimate cloning and identifi cation of genes that regulate plasma HDL-C levels.

Clinical data
The study population consists of 714 Caucasian individuals from 388 families with familial premature CAD and MI as de-scribed previously ( 19 ). Patients were recruited by cardiologists and data coordinators at the Cleveland Clinic Foundation over an approximately fi ve-year period. Institutional review boards approved protocols, and informed consent was obtained from every study participant. For recruitment, each proband in a family was required to have a living sibling meeting the same criteria. Participants answered a health questionnaire, had anthropomorphic measures taken, and had fasted blood drawn for measurement of serum markers and DNA extraction. HDL-C was measured by standard laboratory procedures.

Genotyping
DNA was extracted from whole blood using Puregene Kits (Gentra). Genome-wide genotyping of microsatellite markers was performed by the National Heart, Lung, and Blood Institute (NHLBI) Mammalian Genotyping Services directed by Dr. James L. Weber at Center for Medical Genetics, Marshfi eld Clinic, using Screening Set 11 with 408 markers that span the human genome at approximately every 10 cM (http://research. marshfi eldclinic.org/genetics/geneticResearch/screeningSets. asp). For fi ne mapping, additional markers were selected from posed by Lander and Kruglyak ( 25 ) specifi cally for sib pair linkage analysis in humans. LOD scores for suggestive, signifi cant, and highly signifi cant evidence of linkage are 2.2, 3.6, and 5.4, respectively.

RESULTS
We have completed a genome-wide linkage analysis to identify QTLs for plasma HDL-C levels in a well-characterized US cohort consisting of multiplex families (Gene-Quest). HDL-C values were available for 67% of the GeneQuest study population. A total of 714 Caucasian persons in 388 families with HDL-C data available were analyzed. The clinical and demographic features of the study population are shown in Table 2 . The mean value of HDL-C in the population was low (39.2 mg/dl). As the HDL-C levels did not present with a normal distribution ( Fig. 1A ), the values were also log-transformed prior to analysis. The transformed HDL-C values are shown in Fig.  1B . Age and sex were determined to be signifi cantly correlated with HDL-C values ( P = 0.0024 and P < 0.0001, respectively). Therefore, HDL-C values were also adjusted for age and gender using general linear regression analysis.
A residual heritability estimate of HDL-C in the study population was calculated as 0.37 with gender and age as covariates ( P = 5.1 × 10 Ϫ 4 ; SOLAR, http://solar.sfbrgenetics. org/). Genome-wide genotyping was carried out with 408 polymorphic markers that span the entire human genome at approximately every 10 cM. We performed both single-the Marshfi eld database, synthesized, tagged with 6 ′ -FAM (Sigma), and genotyped using an ABI 3100 genetic analyzer (Applied Biosystems) as previously described ( 20 ). The quality of genotyping for markers used for fi ne mapping was high.
For genotyping SNPs, the TaqMan PreDesigned SNP Genotyping Assays were performed on an ABI PRISM 7900HT Sequence Detection System as previously described ( 21,22 ).

Genetic statistical analyses
Before linkage scanning, obvious pedigree errors, data errors, genotyping errors, and locus-order errors that commonly occur with a large-scale linkage analysis were corrected. Allele frequencies for all markers genotyped for the cohort were estimated by maximum likelihood methods using the S.A.G.E. program FREQ ( 23 ). Pedigree relationships were checked using RELTEST, which uses a Markov process model of allele sharing along the chromosome and classifi es pairs of pedigree members according to their true relationship by use of genome-scan data ( 24 ). Twenty-seven of 428 pedigrees with uncorrectable errors were excluded from further linkage analysis. The S.A.G.E. program MARKERINFO was used to detect any Mendelian inheritance inconsistencies. Three families with inconsistent Mendelian inheritance were eliminated from the study. Three pairs of monozygotic twins were identifi ed in GeneQuest and excluded from further statistical analysis. Only 388 Caucasian families with HDL-C data were analyzed for linkage.
A genome-wide linkage analysis was performed using the program GENEHUNTER (GENEHUNTER 2.1 package, Whitehead Institute, Cambridge, MA) using the sibs quantitative trait mapping function, maximum likelihood QTL variance estimation. All sib pairs were used for analysis. Positions of markers were from Center for Medical Genetics (see http://research.marshfi eldclinic. org/genetics/ for Marshfi eld Genetic Database marker information). Age and sex were determined to be important covariates for HDL-C levels and were modeled in the linkage analysis using general linear regression with SAS Version 9.00. To evaluate the signifi cance of the linkage results, we followed the criteria pro- mosome 7p locus followed with maximum multipoint evidence of linkage equal to a LOD of 4.21 at 6.1 cM ( Fig. 3 ). Single-point analysis yielded a signifi cant result at the closest marker D7S3056 (LOD = 4.11; 7.0 cM) ( Table 3 ). Finally, multipoint analysis for the 13q31-32 locus reached an LOD score of 3.95 at 68.8 cM between markers D13S317 (64 cM) and D13S793 (76.0 cM) ( Fig. 2 ).
For the two strongest linkage loci on 7p22 and 15q25, fi ne mapping was carried out with additional microsatellite markers and di-allelic SNP markers. For the 7p22 QTL, we genotyped the GeneQuest families with D7S1532 and two candidate SNPs, rs10499320 and rs10486788, which in the Framingham Heart Study showed potential association with HDL-C ( 13 ). Linkage analysis for HDL-C after log-transformation and adjustment for age and gender showed a major, complete linkage peak (half point and multipoint linkage analyses using the Gene-Hunter sibs quantitative trait mapping function and maximum likelihood QTL variance estimation, the results of which are shown in Table 3 and Figs. 2-5 . Of the loci identifi ed for HDL-C adjusted for age and gender, the 15q25 region displayed the strongest evidence for linkage to HDL-C. Model-free multipoint linkage analysis revealed high signifi cance at 86 cM in a region between markers D15S655 (83 cM) and D15S652 (90cM) as shown in Fig. 3 . Additionally, single-point linkage analysis confi rmed that high signifi cance was reached at both markers: D15S655 (LOD 6.76) and D15S652 (LOD 5.72) ( Table 3 ). Therefore, for the chromosome 15 locus, the maximum multipoint and single-point LOD scores were 7.57 and 6.76, respectively.
Four additional loci were detected with signifi cant evidence of multipoint linkage: 3p25, 7p22, 13q12, and 13q31-32. The second strongest locus for HDL-C was at 13q12 (20 cM) with maximum multipoint evidence of linkage equal to a LOD of 4.66 ( Fig. 3 ). This locus is in close proximity to marker ATA5A09N (20 cM), and single-point analysis confi rmed signifi cant evidence of linkage at ATA5A09N with a LOD of 6.17 ( Table 3 ). Maximum multipoint linkage for chromosome 3p25 (LOD = 4.10) was reached at 23 cM between markers GATA131D09 (19.3 cM) and D3S4545 (26 cM) ( Table 3 and Fig. 3 ). The chro- Fig. 2. Likelihood plots for QTLs for HDL-C adjusted for age and gender. The Y-axis of each plot is the LOD score; the X-axis is the marker map position. Solid lines represent the multipoint linkage analysis; horizontal dashed lines indicate the signifi cance threshold which is equal to a LOD score value of 3.6. Signifi cant linkages to chromosomes 3p25, 7p22, 13q12, 13q32, and 15q25 were detected with multipoint allele sharing LOD scores of 4.10, 4.21, 4.66, 3.95, and 7.57, respectively. HDL-C, high density lipoprotein cholesterol; LOD, logarithm of odds; QTL, quantitative trait locus. families using a family-based QTDT, focusing on the promoter and exonic SNPs. For the LIPC promoter, we selected SNP rs1800588 because this promoter SNP has been reported to be associated with plasma HDL-C levels ( 14 ). Furthermore, two exonic SNPs, rs690 and rs6083, were selected among tagging SNPs for LIPC identifi ed by the Tagger program and Haploview 4.1 using the threshold minor allele frequency of 0.3 and R 2 of 0.6. As shown in Table 6 , signifi cant association with HDL-C levels was identifi ed for SNP rs1800588 ( P = 0.0067), but not with rs690 or rs6083.

DISCUSSION
The present study reports evidence from genome-wide linkage analysis that multiple QTLs infl uence HDL-C levels in a cohort of premature CAD and MI families (Gene-Quest). In particular, we identifi ed one locus on chromosome 15q25 with highly signifi cant evidence of linkage to date, displaying a LOD score that is greater than the cutoff LOD score of 5.40 for highly signifi cant evidence of linkage at 86 cM. Four regions were also detected with evidence of linkage to HDL-C levels on chromosomes 3p25, 7p22, 13q12, and 13q32.
Supporting the current fi ndings, several independent studies provide evidence for the candidate QTLs on chromosomes 15q25, 13q12, 3p25, and 13q32. Linkage for HDL-C was observed on chromosome 15 in Turkish families with dyslipidemia (LOD = 3.05 at 15q23 at 66.5 Mb) ( 28 ); French Canadian families (LOD = 1.6 at 15q25.1 at 78.1 Mb) ( 29 ); and for unesterifi ed HDL 2b -C in Mexican-peak before fi ne mapping). However, the maximum LOD score of 7p22 dropped to 3.09, which did not exceed the signifi cance threshold of a LOD score of 3.6 ( Fig. 6 and Table 4 ). The size of the one-LOD drop interval was not changed by fi ne mapping. A QTDT did not identify any association between rs10499320 and rs10486788 and HDL-C ( P > 0.05) ( Table 5 ). Haplotypes formed by these two SNPs were predicted by PHASE software (http:// stat.washington.edu/stephens/software.html), and none of the haplotypes showed any association with HDL-C levels (data not shown).
For fi ne mapping of the 15q25 QTL, we studied D15S983 and two SNPs, rs1491579 and rs1638634, adjacent to marker D15S655 with the highest LOD score. The fi ne mapping study splits the QTL into two linkage peaks, one with a maximum multipoint LOD score of 2.73 at chromosome 15q22 covering the LIPC gene (encoding hepatic lipase) and the other with a maximum multipoint LOD score of 5.63 remaining at chromosome 15q25 ( Fig. 6 and Table 4 ). The fi ne mapping sharpened the major linkage peak by narrowing the one-LOD drop interval from 14.8 cM to 9.1 cM. A QTDT did not identify any association between rs1491579 and rs1638634 and HDL-C ( P > 0.05) ( Table 5 ). None of the haplotypes formed by these two SNPs showed any association with HDL-C levels (data not shown).
Recent genome-wide SNP association studies and earlier candidate gene analysis revealed association of SNPs in the LIPC gene with HDL-C levels ( 14,27 ). Because LIPC is located within the small QTL for HDL-C on 15q22 ( Fig. 6 ), we assessed its association with HDL-C in the GeneQuest Fig. 3. Detailed likelihood plots for signifi cant QTLs for HDL-C on chromosomes 3p25, 7p21, 13q12, 13q32, and 15q25. The Y-axis of each plot is the LOD score; the X-axis is the marker map position. Solid lines represent the multipoint linkage analysis; horizontal dashed lines indicate the signifi cance threshold which is equal to a LOD score value of 3.6. HDL-C, high density lipoprotein cholesterol; LOD, logarithm of odds; QTL, quantitative trait locus. studies using transgenic and gene-targeted mice have revealed >100 genes infl uencing the development of atherosclerotic lesions ( 37,38 ), and a larger number of genes infl uencing HDL metabolism is expected to be identifi ed. It is interesting to note that a recent study ( 39 ) using bivariate linkage analysis of coronary artery calcifi cation (CAC), a measure of atherosclerosis determined by electron beam-computed tomography, provided evidence of two regions with pleiotropic effects on CAC and HDL-C on chromosomes 4p16 (MLS = 3.03, P = 0.00084) and 9p12 (MLS = 3.21, P = 0.00056), which may suggest an underlying genomic mechanism for pleiotropism. In future studies we can explore gene-gene interactions for HDL-C (including related plasma parameters) and clinical cardiovascular phenotypes as both data are available, and significant genetic loci were identifi ed for both phenotypes.
Our QTDT did not detect any association between HDL-C and SNP rs10499320 or between HDL-C and SNP rs10486788 in the GeneQuest families.  ( 34 ). The chromosome 13q32 region has also been identifi ed for linkage to total cholesterol, LDL cholesterol, and various lipid-related traits in two studies involving families with familial hypercholesterolemia ( 35,36 ). In total, these independent reports provide some support that the genetic regions identifi ed in the current study harbor genetic variants that regulate HDL-C levels.
Our genome-wide linkage scan and heritability analysis of a US Caucasian population clearly indicate that HDL-C and CAD/MI are complex traits with mixed contributions from multiple genetic and environmental factors. Various signifi cant association was detected. For the minor QTL at 15q22, we analyzed the LIPC gene for its association with HDL-C using a family-based QTDT in the Gene-Quest families. Interestingly, a LIPC promoter SNP, rs1800588, showed signifi cant association with HDL-C, but two exonic SNPs, rs690 and rs6083, were not associated with HDL-C in the GeneQuest population ( Table  6 ). The rs1800588 association may account for the minor QTL on 15q22 identifi ed by fi ne mapping. These results are identical to those generated by a population-For chromosome 15q25 QTL for HDL-C, fi ne mapping studies split the QTL into two separate QTLs, one major QTL on 15q25 with a maximum LOD score of 5.63 and the other minor QTL on 15q22 that showed a maximum LOD score of 2.73 and covered the LPIC gene. Two SNPs, including rs1491579 and rs11638634 close to marker D15S655 showing the maximum LOD score, were analyzed for association with HDL-C, but no   based association study reported in a Turkish population ( 27 ). To the best of our knowledge, this is the fi rst family-based QTDT study to demonstrate the association between LIPC and HDL-C. On the other hand, the specifi c gene responsible for the major 15q25 HDL-C QTL remains to be identifi ed. Due to its highly signifi cant linkage to HDL-C, the chromosome 15 locus provides the most promise for gene identifi cation. Several interesting candidate genes reside within this genetic interval. The most promising candidate genes for HDL-C at this locus include START domain containing 5 ( STARD5 ), PL1N1, ADANTSL3, and endonuclease VIII-like 1 ( NEIL1 ). STARD5 is a member of the steroidogenic acute regulatory lipid transfer (START) domain superfamily of proteins involved in several pathways of intracellular traffi cking and metabolism of cholesterol ( 40 ). PLIN1 is a cAMP-dependent protein kinase substrate in adipocytes that plays a role in lipolysis and has been shown to be associated with metabolic variables in Caucasian women ( 41 ). ADAMTSL3 belongs to the ADAMTS metalloprotease family ( 42 ). NEIL1 knockout mice developed metabolic syndrome with severe obesity, dyslipidemia, and fatty liver disease ( 43 ). It would be interesting to determine whether SNPs in these genes may account for the major 15q25 HDL-C QTL.
One limitation of the current study is that empirical P values for each QTL based on trait or marker resimulation data could not be estimated because such a program was not implemented in the GeneHunter package. The other limitation is that the HDL-C QTLs identifi ed in this study were mostly derived from patients and families with CAD and MI (97.2% of the GeneQuest population) ( Table 2 ), which may be enriched for low HDL-C levels. These QTLs may be different from those modulating HDL-C in general healthy populations.

CONCLUSION
The genome-wide linkage analysis for HDL-C QTLs described here provides highly signifi cant evidence for the presence of a locus on chromosome 15q25 controlling HDL-C values. The 15q25 QTL is the most signifi cant locus identifi ed for HDL-C and represents a novel candidate genetic region that infl uences plasma HDL-C levels. These results will lay the foundation for the successful identifi cation of genetic variants that infl uence HDL-C at this QTL.