The FOXC2 C-512T Polymorphism Is Associated with Obesity and Dyslipidemia
The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Abstract
The transcription factor FOXC2 has been implicated in resistance to diet-induced obesity and insulin resistance. To investigate the possible role for FOXC2 in obesity and related phenotypes, we performed two association studies for obesity using unrelated case-control materials by genotyping the FOXC2 C-512T polymorphism. In the first study (127 obese and 127 normal-weight nondiabetic subjects matched for age and sex), the C-allele showed association with obesity, odds ratio 1.74 (1.12 to 2.73; p < 0.01) for the C- vs. T-allele and 1.81 (1.04 to 3.25; p < 0.05) for the C/C and C/T vs. T/T genotype. BMI was higher in carriers of the C/C and C/T genotype in normal weight [adjusted p value (padj) = 0.02] but not in obese subjects (padj = 0.1). In the replication study (223 obese and 231 nonobese subjects), subjects with the C/C genotype exhibited an increased risk for obesity, odds ratio 2.01 (1.15 to 3.52; p = 0.01). Obese carriers of the C-allele had lower high-density lipoprotein-cholesterol [1.1 (0.9 to 1.3) vs. 1.2 (1.0 to 1.4) mM, padj = 0.006] and increased triglyceride levels (1.95 [1.30 to 2.68] vs. 1.60 [1.10 to 2.40] mM, padj = 0.02) compared with obese carriers of the T/T genotype. Our data suggest that FOXC2 is a weak but consistent candidate gene for obesity and dyslipidemia.
Transgenic overexpression of the human winged helix family transcription factor FOXC2 in white adipose tissue of mice results in a phenotype resistant to diet-induced obesity and insulin resistance (1). The T-allele of the FOXC2 C-512T polymorphism has been associated with enhanced insulin sensitivity, as judged by homeostasis model assessment of insulin resistance (HOMA-IR) and plasma triglyceride (TG)1 levels in a genotype discordant sibling pair analysis in whites, and by basal glucose uptake during a euglycemic hyperinsulinemic clamp and decreased plasma TGs in Pima Indians (2,3). Interestingly, these observations have only been found or are only significant in women (2,3). Individuals homozygous for the−512T-allele have a lower BMI (p = 0.009) and body fat percentage (p = 0.02) than carriers of the C-allele (3). Despite this putative association with insulin resistance, three separate studies, performed in Pima Indians, Japanese, and Danish whites, respectively, have failed to identify an association between FOXC2 C-512T and type 2 diabetes (3,4,5). By contrast, the Danish study has suggested that the T-allele is associated with increased plasma TGs (p = 0.03), fasting C-peptide (p = 0.009), and insulinogenic index (p = 0.04).
Because obesity is a common factor behind insulin resistance, we performed a case-control association study to investigate the possible role for FOXC2 C-512T in obesity and related phenotypes using two unrelated case-control materials (Table 1). In contrast to the Genbank FOXC2 sequence (GenBank accession no. NM 005251), in Pima Indians (3) and in Japanese (4), the more common allele of this polymorphism in Scandinavian populations is not C but T (2,5) (Table 2). In the initial case-control association study, the C-allele was more common among the obese subjects (p = 0.0095), as were the C/C and C/T genotypes (p = 0.027). No difference was observed when grouping C/T and T/T vs. C/C (p = 0.2). There was an increased risk for obesity; odds ratio (OR) were 1.74 (1.12 to 2.73) and 1.81 (1.04 to 3.25) for the C/C and C/T vs. T/T genotype. Normal weight C-allele carriers had a higher BMI compared with subjects homozygous for the T-allele [adjusted p value (padj) = 0.02] (Table 3). This was not the case in the obese subjects (padj = 0.1). In contrast to earlier results in nonobese populations (2,3), we did not observe any differences in measures of insulin resistance, but the amount of available data were limited (Table 1).
| Case-control association study | |||
|---|---|---|---|
| Obese subjects | Control subjects | p | |
| n (men/women) | 127 (34/93) | 127 (34/93) | 1.0 |
| Age (years) | 39.0 (34.0 to 45.0) | 40.0 (34.9 to 44.0) | 0.3 |
| BMI (kg/m2) | 39.8 (35.8 to 44.5) | 22.4 (21.0 to 23.7) | <0.0001 |
| Body fat (%) | 43.6 (38.2 to 58.5) (n = 101) | 27.5 (23.2 to 29.4) (n = 87) | <0.0001 |
| WHR | 0.90 (0.83 to 0.98) (n = 99) | 0.80 (0.76 to 0.86) (n = 117) | <0.0001 |
| fP-glucose (mM) | 5.4 (4.7 to 6.0) (n = 121) | 4.8 (4.6 to 5.0) (n = 125) | <0.0001 |
| fS-insulin (μU/mL) | 14.2 (9.9 to 20.8) (n = 47) | 4.7 (3.9 to 7.1) (n = 94) | <0.0001 |
| HOMA-IR | 3.1 (2.3 to 4.3) (n = 43) | 1.0 (0.8 to 1.5) (n = 94) | <0.0001 |
| fP-HDL (mM) | 1.16 (0.95 to 1.37) (n = 76) | 1.50 (1.31 to 1.73) (n = 110) | <0.0001 |
| fP-TG (mM) | 1.52 (1.00 to 2.12) (n = 120) | 0.82 (0.66 to 1.00) (n = 110) | <0.0001 |
| Replication study | |||
| Obese subjects | Control subjects | p | |
| n (men/women) | 223 (77/146) | 231 (99/132) | 0.08 |
| Age (years) | 39.0 (34.0 to 45.0) | 37.0 (32.0 to 43.0) | 0.05 |
| BMI (kg/m2) | 39.0 (36.0 to 42.7) | 24.0 (22.0 to 25.4) | <0.0001 |
| Body fat (%) | 48.7 (42.6 to 55.7) (n = 220) | 24.5 (19.2 to 29.1) (n = 227) | <0.0001 |
| WHR | 0.98 (0.93 to 1.03) (n = 203) | 0.91 (0.84 to 0.96) (n = 203) | <0.0001 |
| fP-glucose (mM) | 5.0 (4.5 to 5.6) (n = 215) | 4.9 (4.6 to 5.2) (n = 208) | 0.08 |
| fS-insulin (μU/mL) | 16.0 (10.5 to 22.0) (n = 213) | 6.6 (4.8 to 8.7) (n = 192) | <0.0001 |
| HOMA-IR | 3.5 (2.3 to 5.3) (n = 212) | 1.4 (1.1 to 2.0) (n = 190) | <0.0001 |
| fP-HDL (mM) | 1.1 (1.0 to 1.3) (n = 197) | 1.3 (1.1 to 1.6) (n = 176) | <0.0001 |
| fP-TG (mM) | 1.70 (1.20 to 2.55) (n = 213) | 1.1 (0.8 to 1.8) (n = 195) | <0.0001 |
- Data are median (interquartile range) followed by number of subjects with available data. p values refer to Mann-Whitney statistics except for the gender comparison where it refers to Fisher's exact test.
| Case-control association study | |||
|---|---|---|---|
| Obese subjects | Control subjects | p | |
| Allele | |||
| C | 102 (40.2) | 77 (30.3) | 0.0095 |
| T | 152 (59.8) | 177 (69.7) | |
| Genotype | |||
| C/C | 21 (16.5) | 14 (11.0) | 0.027 |
| C/T | 60 (47.2) | 49 (38.6) | |
| T/T | 46 (36.2) | 64 (50.4) | |
| Replication study | |||
| Obese subjects | Control subjects | p | |
| Allele | |||
| C | 166 (37.2) | 147 (31.8) | 0.094 |
| T | 280 (62.8) | 315 (68.2) | |
| Genotype | |||
| C/C | 39 (17.5) | 22 (9.5) | |
| C/T | 88 (39.5) | 103 (44.6) | 0.043* |
| T/T | 96 (43.0) | 106 (45.9) | 0.013† |
- Data are n (%). In the case-control association study, allele and genotype frequencies (grouping the rare C/C with C/T vs. T/T genotype) differed significantly between the two populations based on McNemar's test for matched samples. All genotype frequencies were in Hardy-Weinberg equilibrium.
- * In the replication study, genotype frequencies differed significantly between the two populations based on χ2 test when comparing C/C vs. C/T and T/T.
- † In the replication study, genotype frequencies differed significantly between the two populations based on Fisher's exact test when comparing C/C vs. C/T and T/T.
| Case-control association study | ||||||||
|---|---|---|---|---|---|---|---|---|
| Obese subjects | Control subjects | |||||||
| C/C and C/T | T/T | p | padj | C/C and C/T | T/T | p | padj | |
| Age (years) | 39.0 (34.5 to 45.0) | 38.5 (33.0 to 43.0) | 0.4 | NA | 40.1 (34.4 to 44.0) | 40.4 (35.5 to 44.8) | 0.4 | NA |
| Sex (men/women) | 21/60 | 13/33 | 0.8 | NA | 16/47 | 18/46 | 0.8 | NA |
| BMI (kg/m2) | 40.5 (35.2 to 46.3) | 41.3 (38.0 to 45.0) | 0.9 | 0.1 | 22.7 (21.9 to 24.0) | 21.7 (20.7 to 23.3) | 0.01 | 0.02 |
| Body fat (%) | 43.5 (38.0 to 56.0) | 44.9 (39.9 to 61.0) | 0.3 | 0.5 | 27.6 (23.5 to 30.7) | 27.5 (23.0 to 28.8) | 0.3 | 0.4 |
| WHR | 0.91 (0.83 to 0.99) | 0.88 (0.84 to 0.96) | 0.9 | 0.9 | 0.80 (0.74 to 0.86) | 0.80 (0.77 to 0.86) | 0.2 | 0.2 |
| fP-glucose (mM) | 5.4 (4.7 to 6.2) | 5.2 (4.6 to 5.6) | 0.2 | 0.2 | 4.8 (4.4 to 5.0) | 4.7 (4.6 to 5.0) | 0.7 | 0.6 |
| fS-insulin (μU/mL) | 15.4 (11.8 to 24.8) | 12.8 (9.6 to 16.5) | 0.2 | 0.4 | 4.7 (3.9 to 6.8) | 5.1 (3.9 to 7.7) | 0.5 | 0.9 |
| HOMA-IR | 3.41 (2.47 to 4.54) | 2.68 (2.00 to 3.93) | 0.2 | 0.5 | 1.01 (0.81 to 1.52) | 1.10 (0.82 to 1.49) | 0.5 | 0.9 |
| fP-HDL (mM) | 1.17 (0.97 to 1.39) | 1.14 (0.94 to 1.28) | 0.4 | 0.5 | 1.51 (1.30 to 1.72) | 1.46 (1.30 to 1.75) | 0.7 | 0.6 |
| fS-TG (mM) | 1.61 (1.08 to 2.34) | 1.23 (0.99 to 2.00) | 0.1 | 0.2 | 0.86 (0.64 to 1.02) | 0.81 (0.66 to 0.94) | 0.7 | 0.6 |
| Replication study | ||||||||
| Obese subjects | Control subjects | |||||||
| C/C and C/T | T/T | p | padj | C/C and C/T | T/T | p | padj | |
| Age (years) | 39.0 (33.0 to 44.0) | 38.5 (34.0 to 46.0) | 1.0 | NA | 37.0 (31.5 to 43.0) | 35.0 (32.0 to 42.2) | 0.5 | NA |
| Sex (men/women) | 47/80 | 30/66 | 0.4 | NA | 57/68 | 42/64 | 0.4 | NA |
| BMI (kg/m2) | 38.8 (35.6 to 42.6) | 39.2 (36.3 to 42.8) | 0.3 | 0.3 | 23.7 (21.6 to 25.1) | 24.2 (22.5 to 25.7) | 0.2 | 0.2 |
| Body fat (%) | 48.3 (41.5 to 54.4) | 49.1 (43.3 to 57.4) | 0.2 | 0.2 | 23.8 (18.9 to 28.4) | 25.2 (19.6 to 29.6) | 0.1 | 0.2 |
| WHR | 0.98 (0.93 to 1.03) | 0.98 (0.91 to 1.03) | 0.2 | 0.1 | 0.91 (0.84 to 0.96) | 0.90 (0.84 to 0.91) | 0.4 | 0.2 |
| FP-glucose (mM) | 5.1 (4.5 to 5.7) | 5.0 (4.4 to 5.5) | 0.2 | 0.2 | 4.9 (4.6 to 5.2) | 5.0 (4.6 to 5.0) | 0.6 | 0.6 |
| fS-insulin (μU/mL) | 16.0 (10.4 to 22.2) | 16.0 (10.9 to 22.0) | 0.7 | 0.7 | 6.4 (4.7 to 8.8) | 7.0 (5.2 to 8.6) | 0.3 | 0.4 |
| HOMA-IR | 3.52 (2.32 to 5.46) | 3.38 (2.29 to 5.11) | 0.5 | 0.5 | 1.30 (1.02 to 2.03) | 1.47 (1.08 to 1.97) | 0.3 | 0.3 |
| fP-HDL (mM) | 1.1 (0.9 to 1.3) | 1.2 (1.0 to 1.4) | 0.008 | 0.006 | 1.3 (1.1 to 1.7) | 1.3 (1.1 to 1.5) | 0.8 | 0.9 |
| fS-TG (mM) | 1.95 (1.30 to 2.68) | 1.60 (1.10 to 2.40) | 0.03 | 0.02 | 1.15 (0.70 to 1.88) | 1.10 (0.90 to 1.70) | 0.7 | 0.8 |
- Data are median (interquartile range). p values refer to Mann-Whitney statistics except for the gender comparison where they represent Fisher's test statistics. Both unadjusted (p) and adjusted (padj) p values are presented. NA, not applicable.
In the follow-up replication study, the C-allele failed to show association to obesity (p = 0.094), but genotype frequencies differed between obese and nonobese subjects both separately (p = 0.043) and when grouping carriers of the T-allele (C/C vs. C/T and T/T, p = 0.013) but not when grouping C/C and C/T vs. T/T (p = 0.6) (Table 2). As judged by logistic regression models, homozygosity for the C-allele was significantly associated with obesity (p = 0.01); OR of 2.01 (1.15 to 3.52). Adding age (p = 0.05) and gender (p = 0.03) as independent variables, homozygosity for the C-allele (p = 0.02) contributed to explaining the affection status (obese or nonobese) with an OR of 1.95 (1.11 to 3.44). Obese C-allele carriers had decreased high-density lipoprotein (HDL)-cholesterol (padj = 0.006) and increased TG levels (padj = 0.02) compared with subjects homozygous for the T-allele (Table 3). No other differences in parameters related to insulin resistance were observed. In contrast to the initial association study, there were no significant differences in the replication study between genotype carriers concerning parameters of body composition.
We have previously presented data from a family-based study (genotype discordant sibling pair analysis) implicating FOXC2 in human insulin resistance and hypertriglyceridemia (2). Here, we used a case-control association study design and could show that FOXC2 C-512T was associated with obesity with an OR ranging between 1.74 and 2.01 depending on the population and model. The association was weak but consistent, found both in the present investigation and in Pima Indians (3). Data from our initial association study suggested a dominant effect for the C-allele, whereas data from the replication study suggested a recessive effect of this allele. These differences may be explained by population differences but are most likely a reflection of the lack of power that is inherent to the design (6). In view of this, the most probable explanation is a multiplicative model. When pooling the two study populations to explain the affection status in a logistic regression model including age, gender, and study center in the present investigation, FOXC2 C-512T genotype was a significant independent predictor of the phenotype (p = 0.01; data not shown).
A theoretical link between FOXC2 and obesity can be easily envisioned (1,2,7). The mechanism for developing obesity with respect to FOXC2 most likely involves decreased mRNA transcription rather than protein function per se (2,8). Transgenic mice overexpressing human FOXC2 in white adipose tissue are partially resistant to diet-induced obesity and insulin resistance, and FOXC2 mRNA levels are decreased in adipose tissue from insulin-resistant subjects (1,9). We have previously shown that only subjects homozygous for −512T have higher expression of FOXC2 mRNA in visceral than in subcutaneous fat and, less pronounced, a loss of insulin regulation in skeletal muscle in subjects carrying the C-allele (2). However, whether this polymorphism is of functional importance is not yet known.
Obesity, particularly abdominal obesity, usually precedes insulin resistance and type 2 diabetes, and genetic factors have been considered to explain 60% of the variance in abdominal fat (10,11,12,13). Therefore, in light of the association between FOXC2 and obesity, it is, perhaps, surprising that three independent studies have failed to identify an association between FOXC2 and type 2 diabetes (3,4,5). This indicates that the biological significance of the weak association between FOXC2 and obesity is of minor importance. Alternative explanations include methodological issues such as the case-control approach or sample size, or that the exact phenotype associated with FOXC2 has not yet been identified. It is tempting to speculate that the dysmetabolic syndrome may be one such phenotype because it includes both obesity and insulin resistance, and many, but certainly not all, subjects with obesity and/or type 2 diabetes suffer from it (14,15). There are indications that single genes may promote several of the components of the dysmetabolic syndrome (16). Although speculative at this point, the association between the FOXC2 −512C and increased TG and decreased HDL-cholesterol levels underscores this hypothesis. This difference was found exclusively in the replication study, but this is likely to be due to sample size and lack of data in the original association study. Increased TG levels in relation to FOXC2 C-512T have been reported previously (2,3), but the HDL finding is novel. These findings were seen in the obese subjects but are not likely to be secondary to insulin resistance because there were no differences in HOMA-IR between the genotype carriers. The FOXC2 C-512T polymorphism was recently implicated in familial combined hyperlipidemia (FCHL), a trait characterized by increased levels of TGs and decreased levels of HDL-cholesterol, where the common allele was associated with increased TG levels and a haplotype combination consisting of the common alleles of the C-512T polymorphism and another downstream polymorphism (hCV2649496) was more often transmitted to affected subjects in families with FCHL (17). Whether FOXC2 is associated specifically with FCHL or dyslipidemia in general merits further investigation.
Research Methods and Procedures
Subjects
Two separate populations were selected and genotyped for the FOXC2 C-512T polymorphism. First, 127 nondiabetic Scandinavian white obese subjects (BMI > 30 kg/m2, age 25 to 50 years old) attending the obesity outpatient clinic at the Department of Endocrinology, University Hospital MAS, in the south of Sweden were randomly and individually matched for gender and age (±5 years) with nondiabetic normal-weight subjects (BMI < 25 kg/m2, age 25 to 50 years old) from the Botnia study (10). Second, a replication study was performed in 223 obese (BMI > 30 kg/m2, age 25 to 50 years old) and 231 nonobese (BMI <30 kg/m2, age 25 to 50 years old) subjects participating in an ongoing study on the genetic and metabolic background of obesity at Huddinge Hospital in the east of Sweden (18). All patients gave their written informed consent, and the study was approved by the local University ethics committees.
Phenotypic Characterization
Phenotypic characteristics of the case-control and replication association studies are shown in Table 1. All laboratory specimens were obtained after a 12-hour overnight fast. Measurements of glucose, insulin and lipid concentrations, and bioimpedance analysis to approximate the total body fat mass were performed as previously described (10,18). The HOMA-IR index [fasting serum (fS)-insulin × fasting plasma (fP)-glucose/22.5] was used to estimate the degree of insulin resistance (19). Body weight, height, waist (measured with a soft tape midway between the lowest rib and the iliac crest), and hip circumference (at the widest part of the gluteal region) were measured with subjects in light clothing without shoes and BMI and waist-to-hip ratio (WHR) accordingly calculated.
Genotyping of FOXC2 C-512T
DNA was extracted from blood using a conventional method (20). The FOXC2 C-512T region was amplified with primers 5′-GTC TTA GAG CCG ACG GAT TCC TG-3′ and 5′-TGG GGA CCA AGG TGG ACC CTC G-3′. Polymerase chain reaction was carried out in a 20-μL volume containing 10 μM of each dNTP, 1.5 mM MgCl2, 5% dimethyl sulfoxide, 0.5 U of Taq polymerase (Amersham Pharmacia Biotech, Piscataway, NJ), 10 pmol of each primer, and 25 ng of genomic DNA. The cycling conditions were 94 °C for 5 minutes, 35 cycles of 94 °C for 30 seconds, 63 °C for 30 seconds, and 72 °C for 30 seconds, followed by a final extension at 72 °C for 10 minutes. Polymerase chain reaction products were digested with BbsI (New England Biolabs, Beverly, MA) in 37 °C overnight. The digested products were separated on a 3% agarose gel.
Statistical Analyses
Because individual matching for gender and age was performed in the original case-control association study, McNemar statistics were used to compare allele and genotype frequencies. Allele and genotype frequencies in the replication study were compared by Fisher's exact test and χ2 statistics. A logistic regression analysis including age, gender, and genotype as independent variables was performed to explain affection status in the replication study. Continuous variables were compared by Mann-Whitney statistics between different genotype carriers after grouping those with the rare allele (i.e., comparing C/C and C/T vs. T/T). Because only two major and highly interrelated phenotypic traits were investigated, body composition and insulin resistance, two-tailed p values are reported without correction for multiple testing. A p value below 0.05 was considered statistically significant. Data were adjusted for age and gender (BMI, body fat percentage, and WHR) or age, gender, and BMI (fS-insulin, fP-glucose, HOMA-IR, fS-HDL-cholesterol, and fP-TGs) using the general linear model, and adjusted p values (padj) are presented along with unadjusted values. Data are presented as median with interquartile range (25th to 75th percentile) in parenthesis. All statistical operations were performed using the Number Cruncher Statistical Software (NCSS, Kaysville, UT).
Acknowledgement
This investigation was funded by the Crafoord Foundation, by the Malmö University Hospital Foundation, by the Albert Påhlsson Foundation, by the Swedish Research Council, by the Diabetes Association in Malmö, by the Juvenile Diabetes-Wallenberg Foundation, by the Lundberg Foundation, by EC-GIFT, by the Novo Nordisk Foundation, by Region Skåne, by ALF, by the Wiberg Foundation, by the Swedish Society of Medicine, by the Magnus Bergvall Foundation, by the Fredrik and Ingrid Thurings Foundation, and by the Borgströms Foundation. We thank the study subjects for their participation.
