The Gly482Ser Missense Mutation of the Peroxisome Proliferator-Activated Receptor γ Coactivator-1α (PGC-1α) Gene Associates with Reduced Insulin Sensitivity in Normal and Glucose-Intolerant Obese Subjects

Among the putative candidate genes for insulin resistance, the peroxisome proliferator-activated receptor γ coactivator-1α (PGC-1α) is a transcriptional coactivator of PPARγ and α, regulating a wide range of processes involved in energy production and utilization, such as thermogenesis, liver gluconeogenesis, glucose uptake in muscle. In population studies a Gly482Ser substitution in PGC-1α has been reported to be associated with increased risk of type diabetes 2 and insulin resistance. In the present study we have analysed the association between the Gly482Ser missense mutation of the PGC-1α gene and insulin sensitivity and glucose tolerance in a population of obese non-diabetic subjects. The Gly482Ser SNPs was detected by PCR-RFLP in a cohort of 358 Caucasian obese subjects (223 with normal glucose tolerance (NGT) and 125 with impaired glucose tolerance (IGT). We observed a significant association (p < 0.007) between carriers of the Gly482Ser variant of the PGC-1α gene and insulin resistance measured by HOMAIR. Multivariate analysis confirmed that the Gly482Ser SNP was a significant (p < 0.02) determinant of decreased insulin sensitivity, independently from other well-known modulators of insulin action. In conclusion, we have found significant association between the Gly482Ser variant of the PGC-1α gene and reduced insulin sensitivity in obese subjects. This association resulted independent from all other known modulators of insulin resistance, and suggests a primary role for the PGC-1α gene on the genetic susceptibility to insulin resistance in obesity.


Introduction
Insulin resistance precedes and predicts the development of common disorders such as type 2 diabetes, hypertension and atherosclerosis. Familial transmis-sion and variations in ethnic distribution suggest that insulin resistance is genetically determined [17], and it is likely that impaired insulin action results from several inherited mutations in a variety of genes, each with different effects.
Molecular studies have shown that PGC-1α interacts with several nuclear receptors involved in glucose metabolism. For example it has been shown that the hepatocyte nuclear factor-4α (HNF4α) is absolutely required for PGC-1α to induce the expression of the key gluconeogenic genes PEPCK and G6Pase. Also the interaction with the forkhead transcription factor FOXO1 appears to be important for insulin-regulated hepatic gluconeogenesis [12].
Furthermore, in support of the key role played by PGC-1α it has been shown that its expression is greatly increased in mouse models of diabetes, such as the ob/ob mouse, contributing to the insulin resistance of these animals [5]. Finally, adenovirus-mediated delivery of recombinant PGC-1α in primary hepatocytes determines a dramatic rise in the expression of key gluconeogenic enzymes in the liver and the subsequent production of glucose [5].
Previous studies have reported linkage between the PGC-1α genomic region on chromosome 4p15.1 and fasting serum insulin concentration in Pima Indians [11], and this region has also been linked to increased BMI in Mexican Americans [1].
In population studies a Gly482Ser missense mutation in PGC-1α has been reported to be associated with increased risk of type diabetes 2 in a large Danish population [2]. Association has also been reported between the Gly482Ser substitution and insulin resistance in Japanese subjects [4]. Finally, association of this SNP with altered lipid oxidation and reduced early insulin release has been reported in Pima Indians [10].
Recently, however, the Gly482Ser variant was found not associated with diabetes-related traits in nondiabetic German and Dutch populations [16]. Different ethnicities, bias in the selection of the study groups or in the clinical parameters under study are all possible explanations for these contrasting results.
To help to clarify if the PGC-1α gene plays a primary role in insulin resistance, we have investigated in a population of obese, non-diabetic subjects the possible association of the Gly482Ser polymorphism with insulin sensitivity and glucose tolerance.

Subjects
A total number of 358 Caucasian obese subjects (223 with normal glucose tolerance (NGT) and 125 with impaired glucose tolerance (IGT), were recruited from the Day-Hospital of the Division Endocrinology of the Department of Clinical Sciences, University of Rome "La Sapienza". All were unrelated individuals with a BMI > 28 Kg/m 2 , and all subjects underwent a 75 g oral glucose tolerance test (OGTT).
In order to avoid the confounding effect of multiple environmental, pharmacological and genetic factors on insulin sensitivity, one of the exclusion criteria was the presence of type 2 diabetes. Diagnosis of type 2 diabetes was based on history of hypoglycaemic treatment and/or confirmed fasting blood glucose > 126 mg/dl (7.0 mmol/l) [15].

Biological measurements
Plasma glucose was determined by the glucose oxidase method (Autoanalyzer, Beckman Coulter, USA; coefficient of variation (CV), 1.9 ± 0.2%). Plasma insulin concentration was measured on frozen samples using a radio immunoassay (Biodata Insulin Kit, Milan, Italy) with an interassay CV of 7.5%. Cholesterol and triglyceride concentrations in total plasma were measured with a Technicon RA-1000 Autoanalyzer. High-density lipoprotein (HDL) cholesterol was determined in the whole plasma after precipitation of apoB-containing lipoproteins with phosphotungstic acid/MgCl 2 .
The OGTT was performed after a 10-h overnight fast: subjects ingested a solution containing 75 g dextrose, and venous blood samples were obtained after 0 and 120 min to measure plasma glucose and plasma insulin. Insulin-resistance was estimated with the homeostasis model assessment (HOMA IR = fasting glucose (mmol/l) × fasting insulin (µU/ml)/22.5 [8]. The HOMA %B index was used to estimate beta-cell function as described [18]. This HOMA %B index has been reported to have a good correlation with insulin release estimated from the hyperglycaemic clamp and from the acute insulin response from the intravenous glucose tolerance test (IVGTT) [18].

Genotyping
Genomic DNA was extracted from peripheral blood lymphocytes and the 500 bp fragment of the PGC-1α gene containing the Gly482Ser SNP was amplified by modification of the PCR method described by Hara et al. [4]. PCR conditions were as follows: 34 cycles of denaturation at 94 • C for 30 s, annealing at 54 • C for 30 s, and extension at 72 • C for 45 s. The sequences of the new primers to detect the Gly482Ser polymorphism were 5'-tttggaggcaagcaagcag-3'and 5'tattagggttttgccaagg-3'. By use of Webcutter 2.0 (www. firstmarket.com/cutter/cut2.html), it was found that the Gly482Ser variant of the PGC1 gene creates an Msp1 restriction site, resulting in the presence of two bands of 380 bp and 120 bp. Thus, restriction enzyme digestion of the 500 bp PCR fragment of the intron 4 was carried out at 37 • C for 2 h in 22 µl reactions containing 12 µl PCR product, 2.2 µl buffer 10x and 6 units of Msp1 restriction enzyme. Fragments were analysed on 3.5% high-resolution agarose gel stained with ethidium bromide.

Statistical analysis
Categorical variables were compared by chi-square or Fisher's exact test. Differences between continuous variables were evaluated by two-tailed Student's t-test. Logarithmic transformation was used to normalize distributions of HOMA IR , plasma insulin, HOMA %B index, HDL cholesterol, and triglycerides. Genotype distributions and allele frequencies between the study groups were compared by 2 X 2 and 2 X 3 contingency tables and chi-square analysis.
To evaluate the possible influence of the Gly482Ser SNP on insulin resistance (measured by HOMA IR ) and on β-cell function (HOMA %B), multiple linear regression analysis was employed to adjust for possible confounding factors known to have an effect on insulin resistance and β-cell function such as age, gender, BMI, and lipids. p values < 0.05 or less were taken as statistically significant. All statistical analyses were performed with SPSS statistical package.

Results
The study group comprised 223 NGT and 125 IGT obese subjects. The clinical characteristics of the two groups are described in Table 1. IGT obese subjects, as expected, had significantly higher BMI, systolic and diastolic blood pressure, fasting and 120' blood glucose, fasting and 120' plasma insulin, HOMA IR , and HDL cholesterol, but did not differ in gender, total cholesterol, triglycerides and HOMA %B.
Genotype distributions and allele frequencies of the Gly482Ser SNP of the PGC-1α gene were not different between NGT and IGT subjects (Table 2), similarly to what reported by Stumvoll and co-workers [16].
Clinical and metabolic parameters were then compared between genotypes of the Gly482Ser SNP of the PGC-1α gene. A significant difference in parameters of insulin-resistance was observed when carriers of the variant (both Gly/Ser heterozygous and Ser/Ser homozygous) where compared to carriers of the Gly/Gly genotype (Table 3). In particular, heterozygous and homozygous carriers of the Gly482Ser SNP had higher HOMA IR (p < 0.01), and higher fasting plasma insulin (p < 0.01). Also a significant difference in HOMA %B (p < 0.03) was observed between carriers and non-carriers of the Gly482Ser variant,with non-carriers showing a significant lower β-cell function. No difference was observed between heterozygous Gly/Ser and homozygous Ser/Ser genotypes when compared for any of the clinical and metabolic parameters analyzed.
The same analysis was performed in the NGT and IGT groups independently. In NGT subjects a significant association between carriers of the Ser allele and HOMA IR (p < 0.01), fasting insulin (p < 0.005) and HOMA %B (p < 0.01) was observed (data not shown). In IGT subjects, however, this association did not reach significance, although a similar trend was found, with IGT carriers of the Ser allele being more insulin resistant.
To confirm the independent contribution of the PGC-1α Gly482Ser variant to insulin sensitivity we performed multiple linear regression analysis controlling for the confounding effect on HOMA IR of well-known modulators of insulin resistance (age, gender, BMI, HDL-cholesterol. and triglycerides). A second regression model was used to evaluate if the glucose tolerance status could influence the analysis. This in order to verify if IGT status, which is probably more strongly related to insulin-resistance, could play a role as confounding factor.
Both regression models confirmed that the Gly482Ser variant was independently associated with higher HOMA IR (p=0.024), together with BMI (p < 0.0001) and gender (p < 0.002) (data not shown). When IGT status was added to the analysis, all associations remained significant, including the glucose tolerance status (p < 0.0001). This second model better  Multiple regression analysis was also employed to evaluate the proportion of variance of HOMA %B explained by the gene variant independently from the other variables (including sex, age, BMI, HDL, triglycerides). The analysis showed that the association of the gene variant with the HOMA %B index was just above marginal significance (p = 0.052) (data not shown).

Discussion
The present study provides evidence of association between the Gly482Ser SNP of the PGC-1α gene and reduced insulin sensitivity in obese subjects. This association was present in the whole population of obese subjects, and was most evident within NGT carriers of the Ser allele. IGT subjects showed a similar trend, with carriers of the Ser allele being more insulin re-sistant, but this difference did not reach significance. Possible explanation for this result could be that IGT subjects may already be the result of insulin-resistance, and therefore making it more difficult to discriminate between subjects within this group. Also, we should point out that the IGT group was smaller, possibly reducing the power of this stratified analysis. Genotype distributions did not differ between IGT and NGT groups, similarly to what reported by Stumvoll and coworkers [16]. It could be speculated that the Gly482Ser variant influences insulin sensitivity in a similar way in obese NGT and IGT subjects, predisposing obese subjects to become glucose intolerant when other genetic factors come into play.
The association between the Gly482Ser variant of the PGC-1α gene and reduced insulin sensitivity was also independent from all other known determinants of insulin resistance. Our results are consistent with previous studies reporting association with insulin resistance and related traits in various populations [2,4,10]. At variance with these results is the only negative study published so far [16], where no association was found in German and Dutch populations. As mentioned before, differences in ethnicity, in case selection, or in the clinical parameters under study are all possible explanations for this discrepancy It should be pointed out that our population included only obese subjects, and obesity may be a necessary cofactor for the Gly482Ser SNP of the PGC-1α gene to exert its effects on insulin sensitivity. In favour of this hypothesis is the study by Esterbauer and co-workers [3], where a significant association between the Gly482Ser variant and obesity -related indices in women has been reported.
Muller and co-workers have also shown association of the Gly482Ser variant with decreased early insulin release [10], and explained this observation with the hypothesis that this variant, by increasing free fatty acids and consequently inducing lipotoxicity, determines a reduction in beta-cell insulin secretion. We were not able to replicate this finding in the multiple regression analysis, where the association of the PGC-1α gene variant resulted not independent from other determinants of β-cell function. Nevertheless, our HOMA %B index may be less sensitive than the index used by Muller et al. [10], and further studies are needed to find conclusive evidences. In our study, however, noncarriers of the Gly482Ser variant showed a lower β-cell function than carriers, which were probably hyperinsulinemic to compensate the higher insulin-resistance.
The mechanism by which the PGC-1α gene Gly482Ser missense mutation may affect insulin sensitivity is not clearly understood. To our knowledge no functional studies have been performed with this vari-ant. It could be speculated that defects in PGC-1α gene expression or activity may negatively influence any of the metabolic steps in which this gene is involved, such as hepatic gluconeogenesis, glucose transport in the muscle and thermogenesis [13]. Defects in any of these processes could then contribute to determine insulinresistance.
One limitation of our study, but also a possible point of strength, is that the Gly482Ser variant was investigated only in obese subjects. Obesity per se is a strong risk factor for insulin resistance. The evidence of a significant association in the presence of obesity may indicate that this variant has an effect that is additive to obesity, where the clinical abnormalities are already the result of a few susceptibility genes interacting with environmental factors.
In conclusion, we have found significant association between the Gly482Ser missense mutation of the PGC-1α gene and reduced insulin sensitivity in obese subjects. This association resulted independent from all other known modulators of insulin resistance, and suggests a primary role for the PGC-1α gene on the genetic susceptibility to insulin resistance in obesity.

Aknowledgements
Financial support to this work was provided to M.G. Baroni. by the Faculty of Medicine of the University of Rome "La Sapienza" (grant n. C26F022871), by the Italian Ministry of the University and Scientific Research (COFIN 2003), and by the Ministry of Health (grant RF2003).