Pro12Ala Polymorphism of the PPARγ2 Locus Modulates the Relationship Between Energy Intake and Body Weight in Type 2 Diabetic Patients

The study design was cross-sectional and observational. Participants were 343 unrelated type 2 diabetic patients (144 men and 199 women), aged 40–70 years, consecutively seen at the outpatient diabetes clinic of a health district of the province of Naples. The study was approved by the local ethics committee; informed consent was obtained from all participants. Patients with serum creatinine ≥2 mg/dl or cardiovascular events in the previous 6 months were excluded. All participants were regularly followed up at the clinic by their own doctors, according to current guidelines for good clinical practice. The study investigations were conducted by ad hoc trained observers unaware of the participant's genotype status. No intervention was implemented; the prescribed diet and medications were not modified by the study investigators. Weight, height, and waist circumference were measured with participants wearing light clothing and no shoes. BMI was calculated as body weight in kilograms divided by the square of height in meters. Waist circumference was measured at the level of the umbilicus: values ≥88 cm for women and ≥102 for men were used to define visceral adiposity. The waist-to-hip ratio was also calculated as an additional measure of fat distribution: hip circumference was measured at the widest part of the hip region, and visceral adiposity was defined as waist-to-hip ratio ≥0.85 for women and ≥0.90 for men. Fasting glucose, triglycerides, and total and HDL cholesterol were measured by standard laboratory methods on fresh plasma. Insulin was assayed by radioimmunoassay on frozen samples stored at −80°C for a maximum of 6 months. A1C was measured by high-performance liquid chromatography.

Dietary habits were investigated with the use of a 138-item semiquantitative food frequency questionnaire administered by trained dietitians and designed on the basis of previous validity and reliability studies (16,17). Briefly, participants were asked how often, on average, they had consumed a specified portion of a given food during the previous year. Daily nutrient intake was calculated by multiplying the nutrient content of the specified portion of a food item by the frequency of its daily consumption and then summing the results of all the items. Food values for energy and nutrients were taken from the tables of the European Institute of Oncology (18). Energy intake (kcal/day) and total saturated and polyunsaturated fat (g/day) were calculated; the polyunsaturated-to-saturated fatty acid ratio (P/S) and the glycemic load of the diet were also computed. Energy expenditure due to physical activity was evaluated by a standardized questionnaire (19). Participants were asked to fill in a questionnaire on habitual physical activity at work and during leisure time, which consisted of four increasing activity levels. For analytical purposes, participants with the lowest activity level were considered as sedentary, whereas those with an activity level higher than the lowest value were grouped together and defined as “physically active.” Medication use was assessed by interview.

Genomic DNA was isolated from whole blood using Biorobot EZ1 Qiagen. By PCR, all samples were genotyped for the Pro12Ala single nucleotide polymorphism. All the oligoprimers were tested by gradient PCR to optimize melting temperature. Genotyping was performed by an allele-specific amplification method using SYBR Green detection in a real-time ABI PRISM 7000 apparatus (PE Applied Biosystem).

Data are given as means and SDs or percentages. For non–normally distributed variables, log-transformed values were used in the analyses; the original values are given in the text and tables as geometric means and interquartile ranges. Group means were compared by unpaired Student's t test or ANOVA, as appropriate. Proportions were compared by contingency tables and χ² analysis. The separate and combined effect of the Pro12Ala polymorphism and diet on BMI was explored across quartiles of caloric intake using two-way ANOVA. Due to the different distribution of energy intake between men and women, sex-specific quartiles were computed. Multivariate analysis was conducted by linear regression, with BMI as the outcome variable; the independent variables included in the model were the Pro12Ala polymorphism, estimated daily energy intake, total fat, saturated fat, P/S, glycemic load, age, sex, hypoglycemic medications, A1C, and physical activity. The χ² goodness-of-fit test was used to assess deviation from Hardy-Weinberg equilibrium of the genotypic frequency by calculating expected frequencies of genotypes. A P value <0.05 (two tailed) was considered significant. All statistical analyses were conducted using SPSS for Windows version 12.0

The general characteristics of the study participants are shown in Table 1, together with the PPARγ2 genotype. As expected for type 2 diabetic patients, participants were middle aged and generally overweight. As for the PPARγ2 genotype, 301 subjects (88%) were Pro/Pro homozygotes, 41 (11.7%) were Pro/Ala heterozygotes, and only 1 was a homozygote for the Ala variant. Therefore, in subsequent analyses, those with the Ala/Ala or Pro/Ala genotype were considered as one group and were referred to as “Ala carriers,” whereas individuals with the Pro/Pro genotype were referred to as “non-Ala carriers.” The genotype distribution is in Hardy-Weinberg equilibrium.

Ala carriers and noncarriers were comparable with respect to age, diabetes duration, A1C, blood pressure, glucose, insulin, homeostasis model assessment of insulin resistance index, and total and HDL cholesterol (Table 1). BMI was significantly higher in carriers than noncarriers (P < 0.02), whereas no significant differences were observed for waist circumference and waist-to-hip ratio between the two groups (Table 1). The proportion of patients with central body fat distribution defined according to either waist circumference (≥102 cm in men and ≥88 cm in women) or waist-to-hip ratio (≥0.90 in men and ≥0.85 in women) was not significantly different in Ala carriers or non-Ala carriers (64 vs. 67% and 80 vs. 81%, respectively). Fasting plasma triglycerides were significantly higher in carriers than noncarriers. This difference was largely driven by BMI and was no longer evident after correction for BMI.

Differences in BMI were not explained by differences in dietary habits. Estimated energy intake or the macronutrient composition of the diet (i.e., intake of total fat, saturated fat, polyunsaturated fat, P/S, and carbohydrates) was not significantly different between the two groups. Medications for diabetes are known to affect body weight; however, we did not observe any significant difference in the proportion of patients using insulin, insulin secretagogues, or insulin sensitizers (namely metformin, as the thiazolidenidiones were not marketed in Italy at the time the study was conducted) between the two genotype groups (Table 2). Study participants were generally sedentary; the proportion of physically active participants was low in both groups, with no significant differences between Ala carriers and non-Ala carriers (13.3 vs. 21.4%, respectively, P < 0.24).

To explore the separate and combined effect of the Pro12Ala polymorphism and diet on BMI, participants were stratified according to sex-specific quartiles of energy intake and genotype. BMI increased progressively with increasing energy intake in both genotype groups with a significant linear trend (P < 0.03 for the effect of energy intake; P < 0.016 for the effect of genotype, with no significant interaction). Figure 1 clearly shows that in the first quartile of energy intake BMI was similar in carriers and noncarriers (30.0 vs. 30.1 kg/m², P = 0.1), whereas in the highest quartile of caloric intake the Ala carriers had a significantly greater BMI than Pro/Pro homozygotes (36.0 vs. 32.1 kg/m², P < 0.016). Interestingly, average daily energy intake and diet composition (i.e., total fat, saturated fat, P/S, and carbohydrates) were comparable within each quartile for carriers or noncarriers of the Ala allele (Table 3). Relative to the noncarriers, Ala carriers had a significantly lower energy intake per kilogram body weight (Table 3), thus suggesting that the Ala allele is associated with a higher food efficiency (i.e., for the same body weight, a lower energy intake is required to maintain a stable body weight). Possible confounders such as glycemic control, physical activity, and proportion of patients on insulin, sulfonylureas, or metformin were comparable between Ala carriers and non-Ala carriers within quartiles.

Multivariate regression analysis (Table 4) was performed to explore the independent effect on BMI of energy intake, diet composition, and genotype (presence/absence of the Ala allele); since age and sex are associated with both BMI and energy intake, these two variables were included in the model. Among the variables included in the model, only energy intake and presence of the Pro12Ala polymorphism were significantly and independently associated with BMI. This finding did not change when type of hypoglycemic medications, A1C, and physical activity were also included in the model.

This study shows that type 2 diabetic patients carrying the Pro12Ala polymorphism of PPARγ2 have a significantly higher BMI than noncarriers despite a similar energy intake. As a matter of fact, BMI increases progressively with increasing energy intake in both groups; however, Ala carriers had a significantly lower energy intake per kilogram body weight, thus suggesting that the Ala allele is associated with a higher food efficiency. The confounding effect of hypoglycemic medications, glycemic control, and physical activity was ruled out, thus conferring consistency to the finding.

Very few studies have assessed the impact of genetic polymorphisms and diet on weight, and none of these were performed in diabetic individuals. PPARγ2 is one of the most promising candidate genes of common obesity, although so far results of association studies have been somewhat inconsistent. Cross-sectional studies have shown no difference or a lower or modestly greater BMI in Ala carriers compared with noncarriers; the few available prospective studies suggest that the Pro12Ala polymorphism is associated with higher insulin sensitivity and may confer increased susceptibility to weight gain over time, particularly in obese individuals (8). However, no information on habitual diet was collected in these studies. Results of intervention studies in nondiabetic patients indicate that the Pro12Ala polymorphism may modulate physiological responses to dietary fat intake in humans (9–15). In the Quebec Family Study, the Ala carriers had higher BMI, waist circumference, and fat mass than noncarriers but were more resistant to weight gain and metabolic deterioration when exposed to a high fat intake (10). At least three other studies have confirmed that the weight response to the amount and type of dietary fat differs according to the PPARγ2 genotype (9,12,15). In our study, the Pro12Ala polymorphism did not seem to modulate the impact of the fat content of the diet on BMI. It is relevant to note in this regard that the present study was conducted in a Mediterranean region where on average the habitual dietary fat intake is lower than in Northern European and American countries. Furthermore, the study was conducted in diabetic patients who are usually prescribed, as part of their treatment, a diet reduced in both total and saturated fat. All study participants were regularly attending a diabetes clinic, and, although most patients were not fully compliant with the prescribed diet, the average intake of total fat and saturated fat was substantially lower in this sample than in previous studies (i.e., average total fat intake was 60 g in our study, 90 g in the Canadian study [10], and 72 g in the Finnish study [11]). Likewise, P/S was higher in our study than in others. It is possible that the modifying effect of the Pro12Ala variation on the relationship between dietary fat intake and BMI may not be evident for a low total fat intake, predominantly of the unsaturated type. Alternative explanations for not confirming previous findings include the difficulty in accurately assessing fat intake based on food frequency records.

Our finding of a differential susceptibility to fat accumulation by genotype is in keeping with a study in which women carriers of the Ala variant were shown to be more susceptible to weight regain when resuming spontaneous energy intake after a weight loss treatment (20). These results are compatible with the hypothesis that Ala carriers have a higher food efficiency (i.e., for the same body weight, they need a lower energy intake to keep their weight stable).

As to mechanisms responsible for the effects of the Ala variant on individual weight regulation, we can only make speculations. The cellular and molecular mechanisms by which PPARγ affects adipogenesis are not entirely clear; it has been suggested that the Pro12Ala polymorphism is associated with greater insulin sensitivity, and this could be linked to a greater increase in body weight (21–24). This hypothesis would also be in line with results of functional studies showing that one likely mechanism through which PPARγ reduces insulin resistance is via upregulation of several adipocyte genes and metabolic pathways that favor adipocyte uptake of circulating fatty acids and promote a net flux of fatty acids from the blood and other tissues into adipocytes. A more efficient suppression of lipolysis by insulin in the Ala carriers could also contribute to a greater weight gain in this group by further tilting toward lipogenesis the minutely regulated balance between lipogenesis and lipolysis (24,25). Consistent with this hypothesis, Tschitter et al. (26) reported that Ala carriers had a 50% reduction in circulating levels of free fatty acid during a euglycemic-hyperinsulinemic clamp, although no association between the Pro12Ala allele and fasting free fatty acid levels was found in a large population studied by our group (27).

One limitation to this study is the small sample size; we calculated that the study has an 80% power of detecting a difference of 1 unit BMI between carriers and noncarriers with a P < 0.05. Furthermore, focusing on diabetic subjects may limit the generalizability of our findings, although the genotype distribution in this population is in Hardy-Weinberg equilibrium and similar to that observed in other Caucasian populations (28). Limitations are also built in the assessment of habitual diet: nutrient intake was retrospectively assessed using a questionnaire validated against a 7-day food record. The major pitfalls of this method are recall bias and errors in the estimate of portion size by visual inspection of pictures; however, we do not see why these limitations should apply differently to the two genotype groups, thus introducing a systematic error in the estimate of dietary habits. In any case, expected relations, such as that between energy intake and body weight, are confirmed in this study, thus conferring consistency to the findings.

Overall results of this study suggest a differential susceptibility to fat accumulation, and hence weight gain, for Ala carriers and non-Ala carriers when exposed to habitual excess energy intake. Based on this and other findings, the hypothesis can be formulated that Ala carriers are more prone to weight gain when exposed to an obesogenic environment but may benefit more from energy restriction or increased energy expenditure. The role of a combined gene-environment effect in the etiology of complex traits such as obesity and insulin resistance needs to be further explored, as it may provide a basis for identifying at-risk individuals at a young age and enable the selection of more responders to preventive measures based on lifestyle modifications.

This work was supported in part by funds from the Italian Ministry of University, Research and Technology (MURST prot. 2004-062128-001).

We thank Rosanna Scala for the linguistic revision of the manuscript.