Peroxisome Proliferator-Activated Receptor-γ Coactivator-1 Gene Locus: Associations With Obesity Indices in Middle-Aged Women

Peroxisome proliferator–activated receptor-γ coactivator-1 (PPARGC1) is a coactivator of PPAR-γ and other nuclear hormone receptors and plays an essential role in energy homeostasis. Studies in rodents and cell culture models showed that Pgc1 stimulates mitochondrial biogenesis and activates genes of the oxidative phosphorylation pathway and thermogenesis (1–4). Human PPARGC1 exhibits 94% amino acid identity with the mouse ortholog and was mapped to chromosome 4p15.1 (5). This chromosomal region has been associated with basal insulin levels in Pima Indians (6) and abdominal subcutaneous fat in the Quebec Family Study (7). Because of these associations and PPARGC1’s role in energy metabolism in animal models, we tested the hypothesis that markers at the PPARGC1 gene locus are associated with obesity indices.

By sequencing eight cDNA alleles transcribed from muscle RNA, we identified four polymorphisms in PPARGC1 transcripts. Two polymorphisms, a G-to-A transition at position +1,564 in exon 8 (that predicted a glycine to serine amino acid substitution at position 482, and therefore referred to as G482S) and an A-to-G substitution at position +2,962 in the 3′-untranslated region (Genbank accession no. NM-013261) were selected because of their heterozygosity indices of 0.449 (95% CI 0.436–0.462) and 0.499 (0.498–0.501), respectively, to genotype 591 unrelated men and 467 unrelated women of the Salzburg Atherosclerosis Prevention Program in Subjects at High Individual Risk (SAPHIR) (8). The two other polymorphisms, a silent G-to-A transition at +1,302 in exon 8 and an A-to-T transversion at +3,010 in exon 13 were not considered because heterozygosity indices were <0.10. Frequencies of the wild-type, heterozygous, and variant genotypes at the G482S site were 0.433 (n = 457), 0.453 (n = 481), and 0.114 (n = 120). The respective frequencies at the +2,962 A/G site were 0.264 (n = 280), 0.488 (n = 516), and 0.247 (n = 262). The wild-type and variant allele frequencies for the G482S polymorphism were 0.659 or 0.341. Respective frequencies for the +2,962 polymorphism were 0.509 and 0.492. Genotype frequencies at both polymorphic sites were in agreement with Hardy-Weinberg expectations (χ² for G482S or +2,962 A/G were 0.117 or 0.624, respectively). The two polymorphisms showed a highly significant standardized pairwise linkage disequilibrium (D/Dmax = +0.971, P < 0.001). Estimated haplotype frequencies were 0.503, 0.157, 0.006, and 0.335 for the double–wild-type allele (0,0), the allele with the variant nucleotide at +2,962 (0,1), the allele with the variant amino acid at G482S (1,0), and the double-variant allele (1,1), respectively. Sex-specific analyses revealed no significant differences in genotype, allele, or estimated haplotype frequencies.

Genotypes associated with each of the two polymorphic sites showed associations with obesity indices and metabolic parameters in women but not in men (online appendix, Tables 1 and 2). Interestingly, women heterozygous at either site displayed lower average values for BMI, waist circumference, and body fat and higher values for HDL cholesterol than women homozygous for the wild-type allele at the respective sites. Several considerations argued against either of the two polymorphisms being a major functional site. The amino acid exchange at G482S is conservative and does not create or eliminate protein motifs known to be functional. Furthermore, the variant amino acid is identical to the wild-type amino acid at this position in mice and rats. In addition, transient transfections of PAZ-6 cells showed that the magnitude of PPAR-γ–mediated transactivation of the uncoupled protein-1 promoter was identical for PPARGC1 constructs harboring the wild-type or variant amino acid at G482S (H.O., H.E., V.L., A.D. Strosberg, F.K., and W.P., unpublished observations). Moreover, the possibility that the +2,962 polymorphism in the 3′-untranslated region was a major determinant of mRNA expression appeared unlikely based on preliminary allele-specific expression studies in muscle and adipose tissues of obese subjects. The ratio of the wild-type to variant transcripts (with respect to the +2,962 site) showed wide variations in subjects carrying one variant or two wild-type amino acids at the G482S site (online appendix, Table 3).

We therefore analyzed associations with two loci haplotype combinations. Because three of the possible haplotype combinations, i.e., 1,0/0,0, 1,0/1,0, and 1,1/1,0 were each observed in less than five men or five women, these haplotype data are not shown. Haplotype combinations are unambiguous except for the double heterozygous genotype. Based on estimated haplotype frequencies, >98% of the double heterozygous individuals were predicted to carry a double–wild-type (0,0) and a double-variant allele (1,1), whereas <2% were predicted to carry alleles with one variant nucleotide each (0,1/1,0). In men, no significant association of haplotype combinations with obesity indices was observed (Table 1). In women, however, significant associations with age-adjusted values for BMI, waist and hip circumference, and body fat, with or without adjustment for lean body mass, were observed (Table 2). Abdominal visceral and subcutaneous fat showed borderline significant differences. Age-adjusted plasma levels of glucose, triglycerides, and HDL cholesterol also differed, but these differences were not maintained after adjustment for BMI. Adjustment for postmenopausal hormone use had no substantial effect on results. Possible confounding by menopause was excluded, because BMI and other obesity indices differed significantly among haplotype combinations in women >52 years of age and not using hormone replacement therapy (all P < 0.03).

In women, the haplotype combination of the double-variant and double–wild-type alleles was associated with significantly lower BMI, waist and hip circumferences, fat mass, triglyceride levels, and fasting blood glucose in comparison to the haplotype combination of two wild-type alleles, whereas HDL cholesterol was increased (Table 2, all P < 0.05). However, subjects homozygous for the double-variant or double–wild-type alleles displayed similar obesity indices, as exemplified for BMI in Table 3. To determine associations with obesity in this population, we compared genotype and haplotype frequencies between obese (BMI >30 kg/m²) and nonobese (BMI <30 kg/m²) women. Frequencies of G482S genotypes and haplotype combinations restricted to double–wild-type and variant alleles significantly differed between obese and nonobese women. Estimated risks for heterozygous subjects were lower than for wild-type subjects (online appendix, Table 4).

The apparent association of heterozygosity at two polymorphic sites with reduced obesity indices in women are likely to reflect linkage disequilibria with putative causative site(s). Our allele-specific expression studies showing tissue-specific differences in +2,962 A/G transcript ratios within and among G482S wild-type or heterozygous subjects imply functional polymorphisms in regulatory gene regions that are under tissue-specific control and may therefore confer tissue-specific advantages. Interestingly, several reports have shown that heterozygous PPAR-γ–deficient mice are protected from high-fat diet–induced or aging-induced adipocyte hypertrophy, obesity, and insulin resistance (9–11).

While this manuscript was in review, Ek et al. (12) reported the association of the G482S polymorphism with type 2 diabetes. These authors observed no association with BMI in their male/female type 2 diabetes and control populations. Because sex-specific data were not reported in this study, these results are difficult to compare with our findings. The association of the PPARGC1 gene locus with obesity indices in women suggests, but does not prove, that this locus may contribute to the genetic susceptibility for the multifactorial disorder of obesity. Our results are consistent with concepts emerging from animal studies and recent reports from the Quebec family study (7,13–15). We have no clear explanation why the associations were only observed in women. There is, however, precedence for sex-specific associations of genetic markers with obesity indices (16–18). In addition, PPARGC1 has been shown to interact with the estrogen receptor and enhances its transcriptional activity (3,19). Moreover, animal experiments have demonstrated sex-specific effects of other nuclear hormone receptors whose transcriptional activity is regulated by PPARGC1. In PPAR-α null mice, inhibition of cellular fatty acid flux caused hepatic and cardiac lipid accumulation, hypoglycemia, and death in all male mice but in only 25% of female mice, and treatment with β-estradiol rescued the metabolic phenotype in male PPAR-α null mice (20). Clearly, identification of the functional sites at the PPARGC1 gene locus will be required to gain insight into the associations at the molecular level. If the associations reported are confirmed in other populations, PPARGC1 may be a target for obesity treatment.