Gene variants of TCF7L2 influence weight loss and body composition during lifestyle intervention in a population at risk for type 2 diabetes.

OBJECTIVE: The impact of the diabetes risk gene transcription factor 7-like 2 (TCF7L2) on body weight is unclear. As TCF7L2 is expressed in adipose tissue and involved in Wnt-dependent regulation of adipogenesis, we studied the impact of TCF7L2 variants on body composition and weight loss during lifestyle intervention. RESEARCH DESIGN AND METHODS: We genotyped 309 German subjects at increased risk for type 2 diabetes for single nucleotide polymorphisms (SNPs) rs7903146, rs12255372, rs11196205, and rs7895340 in TCF7L2 and performed oral glucose tolerance tests before and after a 9-month lifestyle intervention. Fat distribution was quantified using whole-body magnetic resonance imaging/spectroscopy in a subgroup of 210 subjects.
RESULTS: After adjustment for confounding variables, we observed a negative impact of the type 2 diabetes allele of SNP rs7903146 on change in BMI (P = 0.0034) and on changes in nonvisceral (P = 0.0032) and visceral fat (P = 0.0165) during lifestyle intervention. An association of rs7903146 with lifestyle intervention-induced changes in insulin secretion, glucose concentrations, liver fat, or insulin sensitivity were not detected (all P > 0.2). Essentially the same results were obtained with SNP rs1255372. In contrast, we found no effects of SNPs rs11196205 and rs7895340 on change in BMI (all P > or = 0.5).
CONCLUSIONS: Our data reveal that diabetes-associated alleles of TCF7L2 are associated with less weight loss in response to lifestyle intervention. Thus, diabetes-associated TCF7L2 gene variation predicts the success of lifestyle intervention in terms of weight loss and determines individual susceptibility toward environmental factors.

Variation in the transcription factor 7-like two (TCF7L2) gene strongly determines the risk of type 2 diabetes (1) mainly via what is believed to be its effects on β-cell function, and the insulin secretion deficit of carriers of the diabetes risk alleles (2) may be explained at least in part by incretin resistance (3,4).

The association of TCF7L2 with weight-related traits and body fat in humans is unclear. Conflicting data are reported on the issue of whether there is an influence of TCF7L2 variants on insulin resistance (5–7) or liver fat (8), and a modulation of BMI (9,10) and obesity was postulated (11). Helgason et al. (12) defined two haplotypes, HapBtype 2 diabetes (rs7903146 efficiently tags HapB in all HapMap groups), which is mainly associated with type 2 diabetes, and HapA (rs7924080 efficiently tags HapA in all HapMap groups), which shows a suggestive association with BMI and altered concentrations of hunger-satiety hormones.

A role of TCF7L2 in adipocyte differentiation is hypothesized for the following reasons: 1) TCF7L2 is expressed in subcutaneous and visceral adipose tissue (11,13,14), and a decrease in TCF7L2 expression in adipose tissue could be demonstrated during caloric restriction (11); 2) TCF7L2 is part of the Wnt signaling cascade, and Wnt signaling inhibits adipogenesis (15–17); and 3) a polymorphism in the Wnt receptor LRP6 is associated with obesity, the metabolic syndrome, and bone loss (18).

In this study, we investigated whether TCF7L2 variants, first, are associated with body fat composition and ectopic lipid storage in our cross-sectional setting or, second, interfere with changes in body composition or weight loss during lifestyle intervention. To this end, we used a whole-body magnetic resonance imaging (MRI) approach enabling precise quantification of body fat stores and magnetic resonance spectroscopy (MRS) to measure ectopic fat deposition in the liver. In our analyses, we focused on single nucleotide polymorphisms (SNPs) rs7903146 and rs1255372, which have yielded the best association with type 2 diabetes across multiple Caucasian populations and tag haplotype HapBtype 2 diabetes (1–3). In addition, we also investigated SNPs rs1196205 and rs7895340, which were available in our cohort, are in near-complete linkage with SNP rs7924080 (both r2 ≥ 0.89, HapMap data), and therefore efficiently tag haplotype HapA (12).

RESEARCH DESIGN AND METHODS

We studied 309 nondiabetic German Caucasian subjects who participated in the ongoing Tuebingen Lifestyle Intervention Program (TULIP) (19). The subjects were at increased risk for type 2 diabetes (20). Most of these subjects (∼70%) had a family history of diabetes. About 13% of the study subjects were related to each other. Oral glucose tolerance tests (OGTTs) revealed that 60% of the nondiabetic participants had normal glucose tolerance, 13% had impaired fasting glycemia, 14% had impaired glucose tolerance, and 13% had both impaired fasting glycemia and impaired glucose tolerance.

OGTT.

After an overnight fast, subjects ingested a solution containing 75-g glucose at 08:00 h. Plasma glucose, insulin, and C-peptide concentrations were determined at 0, 30, 60, 90, and 120 min thereafter.

Body composition.

Total body fat was measured by bioelectrical impedance (BIA-101; RJL Systems, Detroit, MI). In a subgroup of 210 subjects, an exact quantification of fat distribution was performed by MRI and MRS. Magnetic resonance examinations were performed on a 1.5-T whole-body imager (Magnetom Sonata; Siemens Medical Solutions, Erlangen, Germany). A whole-body imaging protocol was applied for recording a set of 90–120 parallel transverse slices. Slice thickness was 10 mm for the entire body. T1-weighted contrast was applied, allowing semiautomatic quantitative assessment of fatty tissue and other tissue types in each cross-section. MRI-derived visceral, nonvisceral, and total fat mass are given in kilograms. Hepatic fat content was determined by localized stimulated echo acquisition mode (STEAM) 1H-magnetic resonance spectroscopy (repetition time [TR] = 4 s, echo time [TE] = 10 ms; 32 scans). The lipid content was quantitatively assessed by analyzing the signal integral using the liver water signal integral as an internal reference. MRS-derived hepatic fat is given in percent of water signal (21).

Lifestyle intervention.

TULIP consists of exercise and dietary intervention for 2 years, which was adopted from the intervention used in the Finnish Diabetes Prevention Study (DPS) (10) and was previously reported (19). Since this is an ongoing project and most data are available for the baseline and 9-month visit, we analyzed these data in the present study. The participants aimed at a weight loss of at least 5%, a reduction of caloric intake from fat to <30% and an increase of fiber intake to at least 15 g/1,000 kcal. Individuals were asked to perform at least 3 h of moderate exercise per week, which was monitored using a heart rate monitor (Polar, Büttelborn, Germany). All individuals had up to 10 sessions with a dietitian during the 9 months. Informed written consent was obtained from all participants, and the local ethics committee approved the protocol.

Genotyping.

DNA from whole blood was isolated using a commercial DNA isolation kit (NucleoSpin; Macherey & Nagel, Düren, Germany). Genotyping was performed using the TaqMan assay (Applied Biosystems, Forster City, CA). The TaqMan genotyping reaction was amplified on a GeneAmp PCR system 7000, and fluorescence was detected on an ABI PRISM 7000 sequence detector (Applied Biosystems). The genotyping success rates were 99.9% for SNP rs7903146, 99.6% for SNP rs12255372, 99.1% for rs11196205, and 98.9% for rs7895340. The genotypes were verified in 50 randomly selected subjects by bidirectional sequencing, and both methods yielded 100% identical results.

Analytical procedures.

Plasma insulin and C-peptide levels were determined with a commercial chemiluminescence system (ADVIA Centaur; Siemens Medical Solutions, Fernwald, Germany). Blood glucose was measured using a bedside glucose analyzer and the glucose oxidase method. (Yellow Springs Instruments, Yellow Springs, OH).

Calculations.

The area under the curve (AUC) of glucose and C-peptide during the OGTT was calculated according to the trapezoid method. Insulin sensitivity during the OGTT was estimated according to the formula: 10,000/(fasting glucose × fasting insulin × mean glucose [OGTT] × mean insulin [OGTT])1/2.

Statistical analyses.

Unless otherwise stated, data are given as means ± SD. Data that were not normally distributed were logarithmically transformed prior to statistical analysis. Hardy-Weinberg equilibrium was tested using the χ2 test. Differences in parameters between genotypes were tested using multivariate linear regression analysis. Differences with a P value <0.05 were considered statistically significant. The JMP 7.0 (SAS Institute, Cary, NC) statistical software package was used.

Power calculation.

The study was sufficiently powered (1 −β = 0.93) to detect effect sizes as small as SD/5 in the overall cohort for change in BMI.

RESULTS

The observed minor allele frequency for SNP rs7903146 was 34%, for rs12255372 it was 29%, for rs11196205 it was 48%, and for rs7895340 it was 46%. All SNPs were in Hardy-Weinberg equilibrium (P ≥ 0.4).

All data were adjusted for sex and age. Glucose concentrations, insulin secretion, and insulin sensitivity derived from the OGTT were additionally adjusted for BMI. The effect of the lifestyle intervention on glucose metabolism, body composition, and weight loss was reported previously (19). Subject characteristics and effect of lifestyle intervention are presented in Table 1.

TABLE 1

Anthropometric and metabolic parameters at baseline and changes during lifestyle intervention

	Mean ± SD	Range	Δ = Tfollow up − Tbaseline	Mean ± SD	P
Age (years)	46 ± 11	18–69
BMI (kg/m2)	30.1 ± 5.6	19.4–51.0	ΔBMI	−0.9 ± 1.5	<0.0001
Total body fat (%)	33.0 ± 8.6	9.8–58.3	ΔTotal body fat	−1.1 ± 4.5	<0.0001
Fasting glucose (mmol/l)	5.3 ± 0.5	4.1–6.5	ΔFasting glucose	−0.1 ± 0.5	0.0011
Glucose120-min (mmol/l)	7.0 ± 1.6	3.6–11.1	ΔGlucose120-min	−0.3 ± 1.6	<0.0001
AUCC-Peptide/AUCGlucose (×10−9)	1,217 ± 660	60–4,385	ΔAUCC-Peptide/AUCGlucose	−0.4 ± 74	0.9
Insulin sensitivity, OGTT (AU)	12.8 ± 6.9	2.6–32.5	ΔInsulin sensitivity	1.8 ± 6.2	<0.0001
Nonvisceral fat (kg)*	23.2 ± 10	3.4–60.5	ΔNonvisceral fat	−2.0 ± 3.4	<0.0001
Visceral fat (kg)*	3.0 ± 1.8	0.3–10.1	ΔVisceral fat	−0.4 ± 0.6	<0.0001
Liver fat content (% of water signal)*	6.6 ± 8.3	0.2–44.7	ΔLiver fat content	−2.1 ± 5.9	<0.0001

n (female/male) = 194/115. Changes during lifestyle intervention were statistically analyzed by MANOVA.

*MRI/MRS subcohort.

In the cross-sectional setting, we found no association of the type 2 diabetes risk SNP rs7903146 with insulin sensitivity, BMI, bioimpedance-derived total body fat, visceral fat mass, nonvisceral fat mass, or liver fat (all P ≥ 0.2) (Table 2). However, we observed a trend for lower insulin secretion as measured by AUCC-Peptide/AUCGlucose (P ≤ 0.07) and higher 120-min glucose (P ≤ 0.1) in type 2 diabetes risk allele carriers (Table 2).

TABLE 2

Associations of TCF7L2 SNP rs7903146 with anthropometric and metabolic parameters in the cross-sectional setting and with changes in these parameters during lifestyle intervention by genotype

	CC	XT	P1	P2		CC	XT	P1	P2
n (female/male)	148 (95/53)	161 (99/62)	—
Age (years)	46 ± 10	47 ± 12	—
BMI (kg/m2)	30.3 ± 5.5	30.0 ± 5.8	0.5	0.7	ΔBMI	−1.2 ± 1.6	−0.7 ± 1.5	0.0034	0.0108
Total body fat (%)	33.7 ± 8.5	32.5 ± 8.7	0.3	0.3	ΔTotal body fat	−1.7 ± 4.5	−0.5 ± 4.3	0.0155	0.07
Fasting glucose (mmol/l)	5.3 ± 0.5	5.3 ± 0.5	0.95	0.9	ΔFasting glucose	−0.1 ± 0.4	0.0 ± 0.4	0.2	0.1
Glucose120-min (mmol/l)	6.8 ± 1.5	7.2 ± 1.7	0.05	0.1	ΔGlucose120-min	−0.2 ± 1.6	−0.4 ± 1.5	0.4	0.8
AUCC-Peptide/AUCGlucose (×10−9)	317 ± 96	295 ± 88	0.06	0.07	ΔAUCC-Peptide/AUCGlucose	−2.6 ± 78	1.5 ± 70	0.3	0.8
Insulin sensitivity, OGTT (AU)	12.5 ± 7.0	13.0 ± 6.8	0.4	0.6	ΔInsulin sensitivity	2.2 ± 6.0	1.5 ± 6.4	0.6	0.5
Nonvisceral fat (kg)*	23.5 ± 9.5	22.8 ± 10.3	0.5	0.8	ΔNonvisceral fat	−2.7 ± 3.6	−1.3 ± 2.9	0.0022	0.06
Visceral fat (kg)*	3.0 ± 1.8	3.0 ± 1.9	0.2	0.3	ΔVisceral fat	−0.5 ± 0.6	−0.3 ± 0.6	0.0165	0.06
Liver fat content (% of water signal)*	7.6 ± 9.3	5.7 ± 7.0	0.3	0.2	ΔLiver fat content	−2.7 ± 5.8	−1.5 ± 5.9	0.09	0.2

Data are means ± SD. BMI and total body fat were adjusted for sex and age. Glucose concentration, insulin secretion, and insulin sensitivity were additionally adjusted for BMI. Δ = Tfollow up − Tbaseline. The change in the parameters was additionally adjusted for baseline values. The change in glucose concentration, insulin secretion, and insulin sensitivity was adjusted for the change in BMI. Data were statistically analyzed by multivariate linear regression analysis in the dominant inheritance model, P1, and the additive inheritance model, P2. Data in bold represent statistical significance.

*MRI/MRS subcohort.

During lifestyle intervention, the risk allele carriers displayed lower reduction in BMI (P ≤ 0.0108) and total body fat (Pdom = 0.0155; Padd = 0.07) (Table 2). SNP rs7903146 explained 2.6% of the variation in the intervention-induced change of BMI. The variant was not associated with changes in fasting or 120-min glucose, insulin sensitivity, liver fat, or insulin secretion (all P ≥ 0.09). Because of the known effect of the variant on insulin secretion, we additionally adjusted the effect on change in BMI for insulin secretion and 120-min glucose. This however did not alter the results.

In the MRI/MRS cohort, we observed in the risk allele carriers a lower reduction in nonvisceral (Pdom = 0.0022; Padd = 0.06) and visceral fat mass (Pdom = 0.0165; Padd = 0.06) (Table 2). As adjustment for sex and age does not completely account for sex- and age-specific interaction effects of rs7903146 on body composition and weight loss, we performed separate analyses for male and female and old and young (separated by the median) study participants (supplemental Tables 1 and 2, available in an online appendix at http://diabetes.diabetesjournals.org/cgi/content/full/db09-1050/DC1). After stratification for sex, most of the associations observed in the overall cohort were also seen in the subgroups, and no clear sex effects could be detected. After stratification for age, however, the associations of the overall cohort were only reflected in the young subgroup.

To minimize the chance of statistical type 1 errors, we additionally analyzed SNP rs12255372, which is in incomplete linkage disequilibrium with rs7903146 (r2 = 0.77, HapMap data). Essentially the same results were obtained with rs12255372 (supplemental Table 3).

In contrast, SNPs rs11196205 and rs7895340 (tagging haplotype HapA) were not associated with body composition cross-sectionally or with weight loss during lifestyle intervention (all P > 0.5) (supplemental Tables 4 and 5).

DISCUSSION

In the present study, we found in a population at risk for type 2 diabetes that risk variants for type 2 diabetes of the TCF7L2 gene (rs7903146, rs12255372) influence lifestyle intervention-induced changes in BMI, total fat, and nonvisceral and visceral fat, while changes in insulin secretion, insulin sensitivity, or blood glucose are unaffected. Like our cross-sectional setting, most cross-sectional studies found no reliable association of variants in TCF7L2 gene with BMI (1,3,22). Nevertheless, Helgason et al. (12) observed a moderately leaner phenotype in carriers of the haplotype HapBtype 2 diabetes (0.44% lower BMI per copy). The finding of a leaner phenotype is confirmed in the cross-sectional analyses of the Diabetes Prevention Program (DPP) and the DPS (9,10). The authors of the DPP concluded that their finding is unexpected and possibly artificial (9). Subjects with lower insulin secretion, like the TCF7L2 risk allele carriers, had a higher risk to develop diabetes even though they were leaner. Therefore, there may be a bias in the study population when enrolling only subjects at risk for type 2 diabetes (9). Although the HapA haplotype is reported to associate with BMI and altered concentrations of hunger-satiety hormones (12), the HapA-tagging SNPs tested in our cohort were not associated with weight loss in the longitudinal setting.

In the DPS as well as in the DPP, it was shown that the diabetogenic effect of the SNPs rs7903146 and rs12255372 in TCF7L2 is mitigated by lifestyle intervention (9,10). Changes in BMI and fat depots though were not reported in the DPS and the DPP; therefore, we cannot draw conclusions as to how changes in body weight influence the outcome in these studies. However, it may be speculated that smaller reductions in BMI and body fat due to the risk allele may not necessarily influence type 2 diabetes incidence, at least within the 3- to 4-year observation period.

We currently have no explanation for the conflicting findings with regard to the phenotype of TCF7L2 SNP carriers in cross-sectional settings (9,10,12) versus our longitudinal setting. In the present study, the intervention was applied to all individuals so that each individual served as his/her own historic control subject. This approach may explain the different findings of studies that are fully randomized.

The effects of TCF7L2 on body fat found in this study need replication in other larger cohorts and longer-term lifestyle intervention studies to verify their clinical relevance. These studies should include a control group without weight loss and subjects not at high risk for diabetes. In addition, studies at the molecular level should also be undertaken to identify the possible mechanisms of how TCF7L2 influences body fat change.

In conclusion, we show that diabetes-associated variation in TCF7L2 influences changes in BMI and body fat and thus predicts the success of lifestyle intervention in terms of weight loss. Therefore, TCF7L2 represents a gene that influences individual susceptibility toward environmental factors. The mechanisms of how the variants interact with environmental factors to influence BMI are not understood and need further investigation.