Genome-wide Scan for Metabolic Syndrome and Related Quantitative Traits in Hong Kong Chinese and Confirmation of a Susceptibility Locus on Chromosome 1q21-q25

The details of ascertainment, exclusion criteria, and phenotyping for the HKFDS are described elsewhere (23). In brief, 179 families who were ascertained through a diabetic proband with available first-degree relatives are included in the present study, after exclusion of patients with clinical or autoimmune type 1 diabetes and families with known maturity-onset diabetes of the young or mitochondrial DNA nucleotide 3243 A>G mutations. Of these 179 families, 55 (137 affected subjects) with at least two MES-affected relatives (nonparent-offspring) are appropriate for linkage study of MES. Sixty-seven (49%) of these subjects in 36 families are included in our previous genome scan for type 2 diabetes genes (23). For assessing the relative contribution of families involved in our type 2 diabetes study on the linkage signal for MES, subset linkage analyses were performed in 36 families (82 affected relative pairs) involved in both the type 2 diabetes and MES studies and 19 families (48 affected relative pairs) involved in only the MES study. Studies on quantitative traits were performed in all 179 families (897 subjects) excluding subjects aged <16 years. The family relationships and the clinical characteristics of subjects involved in MES and QTL linkage studies are summarized in Tables 1 and 2, respectively. The average family size in the QTL studies is 5.0 ± 2.3. The average number of affected subjects per family in the MES study is 2.5 ± 1.0. Informed consent was obtained for each participating subject. This study was approved by the Clinical Research Ethics Committee of the Chinese University of Hong Kong.

All available family members were assessed for blood pressure (BP) and standard anthropometric parameters and completed a questionnaire on demographic data, family and medical histories of cardiovascular risk factors and complications, and lifestyle. Fasting blood samples were collected for measurement of plasma glucose (FPG), insulin, C-peptide, lipid profile, liver and renal function, and DNA. Subjects with no history of diabetes were tested with a 75-g oral glucose tolerance test. Insulin resistance was assessed by the homeostasis model assessment (HOMA%IR) as fasting insulin (μU/ml) × FPG (mmol/l)/22.5. MES was defined in accordance with the National Cholesterol Education Program Adult Treatment Panel III (NCEP III) guidelines (3). Patients who have at least three of the following five risk factors were classified as having MES: 1) hyperglycemia with known diabetes or FPG ≥6.1 mmol/l, 2) known hypertension or BP ≥130/85 mmHg, 3) hypertriglyceridemia with triglyceride (TG) ≥1.7 mmol/l, 4) HDL cholesterol <1.0 mmol/l in men or <1.3 mmol/l in women, and 5) central obesity with waist circumference (WC) >90 cm in men or >80 cm in women. The definition of central obesity was modified for Asian populations (24).

Detailed information on the genotype data and quality control have been described previously (23). In brief, 355 autosomal microsatellite markers (Research Genetics, Huntsville, AL) with an average intermarker distance of 10 cM and average heterozygosity of 71% were typed. Genetic relationships among family members were checked by RELPAIR (25) and PREST (26) using genome-scan data to assess consistency of allele-sharing pattern with the specified relationship. Misclassification of half siblings as full siblings and mislabeled DNA samples were identified and resolved. No families were removed as a result of relationship inconsistency. However, we removed nine subjects with no or unresolved relationship with other family members. Mendelian errors and potential genotyping errors were checked by PEDCHECK (v. 1.1) (27) and MERLIN (v. 0.9.12b) (28), respectively, and removed.

Marker allele frequencies were estimated from all subjects in the 179 families. Marker position was based on the sex-average maps from Marshfield Medical Research Foundation (http://research.marshfieldclinic.org/genetics). For the MES study, nonparametric multipoint linkage analysis was performed with MERLIN (v. 0.9.12b) using the score-pairs statistic (28). In chromosomal regions that showed suggestive linkage with MES, linkage analyses of the five respective factors (hyperglycemia, hypertension, hypertriglyceridemia, low HDL cholesterol, and central obesity) were also performed in the 55 MES families that consisted of family members with various degrees of MES factors.

For QTL studies, seven quantitative traits related to MES, including WC, systolic and diastolic BP, TGs, HDL cholesterol, FPG, and HOMA%IR, were examined. Data on BP were removed for subjects who were taking antihypertensive medications. Similarly, data on FPG and HOMA%IR were removed for the 68% of diabetic subjects who were taking oral antidiabetic drugs or insulin. Data were transformed using natural logarithm (WC, systolic BP, TGs, HDL cholesterol, and HOMA%IR), square root (diastolic BP), or double natural logarithm (FPG) to reduce skewness and kurtosis. Extreme values >4 SD from the mean were removed. The number of individuals who were removed for quantitative trait analyses ranged from zero to eight. No whole families were removed for analyses of any traits. Data were standardized to mean zero and unit variance and analyzed by the variance component method implemented in MERLIN with simultaneous adjustment of covariates including age and sex. Type 2 diabetes was also used as a covariate in regions that showed significant linkage. MERLIN-REGRESS (v. 0.9.12b) (29) was performed on selected traits and chromosomes for comparison with variance component-based results. Sex-specific linear regression of each trait for age was obtained in a Chinese cohort (573 subjects) recruited from a community-based health screening program in Hong Kong. Data of each family member were regressed on the linear model. Standardized residuals (using population mean and SD) were then analyzed with MERLIN-REGRESS using a population mean of zero and a variance of one and estimated trait heritability from the families. Logarithm of odds (LOD) scores with associated pointwise P values are reported (30). Two linked regions were defined as independent when their maximum LOD score positions differed by ≥40 cM.

Simulation studies were conducted to assess the significance of the observed linkage signals in both the MES and QTL studies. One thousand simulated datasets were generated by gene dropping with MERLIN under the null hypothesis of no linkage. The maps, marker allele frequencies, pedigree structures, and missing genotype patterns used in these simulations were identical to those in the actual data. Linkage analyses on MES and each of the seven quantitative traits were then conducted on the simulated datasets. In each simulation, the number of independent linked regions (ILRs) greater than a particular LOD score threshold was tallied. Using a locus-counting method (31), P_genome-wide for n-ILRs at a LOD score threshold was calculated as the proportion of simulations with ≥n ILRs at that particular LOD score. A significant excess of ILRs in the observed data set was defined as P_genome-wide < 0.05. Suggestive linkage was indicated by the LOD score that corresponds to an average of one ILR per genome scan expected by chance through simulation (32). Because of a priori hypothesis and the high correlations among the examined quantitative traits and MES, no adjustment for multiple comparisons was made on the P_genome-wide values.

The phenotypic correlation (ρ_P) between the pairwise combination of traits was calculated on the basis of genetic (ρ_G) and environmental (ρ_E) correlations and of heritabilities of the two traits (h1² and h2²), where

The MES study includes 55 families with 137 affected subjects (121 affected relative pairs; Table 1). The prevalence of metabolic abnormalities including hyperglycemia, hypertension, hypertriglyceridemia, low HDL cholesterol, and central obesity range from 64 to 87% (Table 2). The variance component-based QTL studies include 179 families with 897 subjects (2,127 relative pairs). These families also have a high prevalence of metabolic abnormalities (25–43%; Tables 1 and 2). The selection of quantitative traits for the QTL studies was based on their use in the definition of MES using the NCEP III criteria (3). Although families are selected through a diabetic proband, a large number of family members are unaffected. Thus, the distributions of transformed quantitative traits are approximately normal, and the mean values are similar to the general population except for FPG (Table 2). Heritabilities for WC, systolic and diastolic BP, TGs, and HDL cholesterol were moderate (h² = 0.45–0.63) but relatively low for FPG (h² = 0.28; J.K.Y.L. et al., unpublished data). We also assessed insulin resistance index (HOMA%IR; h² = 0.60) because it is a key factor linking various MES traits.

Table 3 summarizes regions with LOD scores >1.18 (P_pointwise < 0.01). The region that showed the strongest evidence for linkage is on chromosome 1q21–25 at 169.5 cM (LOD = 4.50, one LOD confidence interval = 163–185 cM, P_genome-wide = 0.002). Two other regions, chromosome 2 at 44.1 cM (LOD = 2.22) and chromosome 16 at 65.4 cM (LOD = 1.75), have suggestive evidence for linkage based on observing an average of one ILR per genome scan (32) at a LOD score of 1.57 (Tables 3 and 4) in our simulation. ILRs are significantly more frequent in the actual data than expected at LOD >2.08 (Table 4). Analyses on individual components of MES in these 55 families show clustering of linkage signals for most traits at the MES loci (Fig. 1). The strongest signals are for central obesity (LOD = 4.97) on chromosome 1, hypertension on chromosome 2 (LOD = 3.16), and low HDL cholesterol on chromosome 16 (LOD = 1.91). In a subset analyses of chromosome 1, the respective LOD scores for MES at 169.5 cM are 3.82 for the 36 families involved in the previous type 2 diabetes study and 0.90 for 19 families involved in only the MES study. The obesity indexes of diabetic subjects in the former study (BMI, 27.0 ± 4.6 kg/m²; WC, 93 ± 10 cm for men and 83 ± 10 cm for women, respectively) were similar to those in the present study (Table 2).

In the QTL analyses, we observed genome-wide evidence of linkage for WC on chromosome 1q21–25 at 180.3 cM (LOD = 3.71, one LOD confidence interval = 172–190 cM, P_genome-wide = 0.019; Tables 3 and 4, Fig. 2). The respective LOD score decreases to 2.87 after adjustment for type 2 diabetes status. We also observed suggestive evidence of linkage (LOD = 1.69–2.99) for several metabolic traits. These include diastolic BP on chromosome 5q; TGs on chromosomes 2q, 3q, 6q, 9q, 10q, and 17q; HDL cholesterol on chromosomes 12p, 12q, and 22q; FPG on chromosome 2p; and HOMA%IR on chromosome 6q (Tables 3 and 4, Fig. 2). Most traits except for BP show more ILRs than expected at several LOD score thresholds, especially for TGs and HDL cholesterol (Table 4). We also analyzed two regions using a regression-based method. These include WC on chromosome 1, showing significant linkage, and FPG on chromosome 2, showing suggestive linkage and with different mean values in family versus population data. We observed a higher LOD score for WC (5.97 vs. 3.71) but a similar LOD score for FPG (1.90 vs. 2.07) using the latter method.

Several regions show nominal or suggestive linkages with two or more MES-related traits: chromosome 1 (169.5–181.5 cM) for MES, WC, and diastolic BP; chromosomes 2 (44.1–57.3 cM) for MES, FPG, and diastolic BP; and chromosome 16 (45.2–65.4 cM) for MES, HOMA%IR, and HDL cholesterol (Table 3).

We have carried out both qualitative and quantitative trait linkage analyses to localize genes that contribute to susceptibility to MES. We used simulation to assess the significance of linkage results, which took into account the incomplete marker information content, sample size, and ascertainment scheme of our families. We found significant linkage on chromosome 1 (169.5–181.5 cM) for MES and WC and suggestive linkages on chromosomes 2 (44.1–57.3 cM) and 16 (45.2–65.4 cM) for MES and various metabolic traits. Because only a small proportion of subjects (n = 137) are included in both studies and two different analytical methods are used, the overlapping linkage signals for MES and quantitative traits in these chromosomal regions provide complementary evidence for susceptibility gene(s) for multiple metabolic abnormalities and/or multiple genes at the same loci for different metabolic abnormalities.

We and others previously reported significant evidence of linkage of chromosome 1q21–25 to type 2 diabetes (24,34). Here, we show that chromosome 1q21–25 is also significantly linked to MES and central obesity (LOD = 4.50 at 169.5 cM and LOD = 3.71 at 180.3 cM, respectively). Sixty-seven (49%) individuals from 36 families in our MES study were part of the previous type 2 diabetes study. Subset analyses show that most but not all of the linkage signal for MES is derived from families involved in the previous type 2 diabetes study rather than from the 19 families involved only in the MES study. Of the analyses of individual phenotypes and quantitative traits, the highest LOD scores are observed for obesity indexes. Significant linkage signals for WC are observed in both variance-component and regression-based analyses; the latter are robust to ascertainment bias but require correct specification of population mean, variance, and heritability (29). This signal is partly explained by type 2 diabetes status. In bivariate analyses of our families, we observed higher correlations between obesity indexes and other metabolic indexes (BP, lipid and glucose profiles; phenotypic correlation, 0.28–0.48; genetic correlation, 0.16–0.60) compared with correlations among other metabolic indexes (phenotypic correlation, 0.09–0.41; genetic correlation, 0.01–0.43; J.K.Y.L. et al., unpublished data). Collectively, the present and previous results strongly suggest the presence of susceptibility gene(s) on chromosome 1q21–25 for clustered phenotypes of MES and type 2 diabetes. Obesity may be the key factor linking the pathogenesis of various metabolic abnormalities. These results also support our previous structural equation modeling analysis that demonstrated that obesity, rather than insulin resistance, was the key factor linking MES traits in our population (9). However, the weaker signal of hypertension in the MES linkage study (this study) and the weaker phenotypic correlations between BP and other metabolic traits (0.09–0.27) suggest that other genetic and environmental factors play a more important role in the pathogenesis of hypertension in these families.

Our linkage results in this region confirm the findings of studies in other populations on disease phenotypes and quantitative traits related to MES (18,35), obesity (36,37), type 2 diabetes or glucose intolerance (23,34), and dyslipidemia or familial combined hyperlipidemia (FCHL) (38–42). To our knowledge, no linkage of chromosome 1q21–25 with hypertension has been reported so far, consistent with our observations that hypertension is less correlated with other MES traits. Recently, the upstream transcription factor 1 gene located on chromosome 1q21 was reported to be linked to and associated with FCHL (43). The contributions of this and other candidate genes associated with type 2 diabetes, MES, and related traits remain to be determined.

We observed suggestive linkage with MES on chromosomes 2 (LOD = 2.07) and 16 (LOD = 1.75). These regions also show nominal or suggestive linkage with MES-related traits. On chromosome 2, the strongest signals are for hypertension, FPG, and diastolic BP. On chromosome 16, the strongest signals are for low HDL cholesterol, HDL cholesterol level, and HOMA%IR. The correspondence in linkage signals for MES and QTLs for BP on chromosome 2 and HDL cholesterol on chromosome 16 in two largely independent datasets using different analytical methods suggest that they may be true susceptibility loci. In contrast to the chromosome 1 signal, the linkage signals on these chromosomes were not observed in our scan for type 2 diabetes genes. This suggests that these loci on chromosomes 2 and 16 may have a primary effect on individual traits rather than on MES as a whole. The chromosome 2 region was previously reported to be linked to MES (19); hypertension (44); obesity (45); and TG (46), leptin (47), and adiponectin levels (48). The HDL cholesterol linkage result on chromosome 16 replicates the finding of a coronary heart disease genome scan (21).

In the QTL analyses, nominal evidence for linkage (P < 0.01) for various metabolic traits is reached for many chromosomal regions. Many of these are expected to be false positive but cannot be distinguished from true signals. Several metabolic traits show linkage signals in more ILRs than expected, and some regions reach suggestive linkage (e.g., TG with LOD score of 2.58 on chromosome 6q), suggesting the presence of susceptibility loci, which may warrant further investigation.

The collection of extensively phenotyped families allows us to analyze complementary disease and quantitative phenotypes. We demonstrate that using different analytical methods on two largely independent datasets is a powerful approach to mapping loci for diseases with complex phenotypes. The identification of the genes that affect risk to MES will lead to a better understanding of the pathogenesis and improve diagnosis of MES in Chinese and other populations.

This work was supported by grants from the Hong Kong Research Grants Committee; the Chinese University of Hong Kong Strategic Grant Program; the Innovation and Technology Support Fund, Hong Kong (ITS/33/00); and U.S. Public Health Service Grants DK-20595, DK-47486, DK-55889, and DK-58026. G.I.B. is an investigator of the Howard Hughes Medical Institute.

We thank Emily Poon and Patty Tse for technical support and Dr. June Li and Cherry Chiu for help with the recruitment of patients and their family members. We thank William Wen for computational support. We thank all nursing and medical staff at the PWH Diabetes and Endocrine Centre for dedication and professionalism. We thank all patients and their relatives for participating in this study. We are also indebted to the late Professors Robert Turner, Oxford University, and Julian Critchley, The Chinese University of Hong Kong, for inspiration and support.