Peroxisome Proliferator-Activated Receptor α Gene Variation Influences Age of Onset and Progression of Type 2 Diabetes

European type 2 diabetic subjects were selected from two studies, UDACS (16) and EDSC. UDACS and EDSC subjects were recruited consecutively from the diabetes clinics at University College London Hospital and Ealing Hospital, respectively. Patients completed a questionnaire with details of age, ethnicity, smoking habit, fasting status, duration of diabetes, family history of diabetes, history of heart attack or stroke, current medication, and other clinical details. Blood was collected for plasma and DNA analysis. BMI, blood pressure, glucose, HbA_1c, cholesterol, HDL, LDL, triglyceride, urea, creatinine, albumin-to-creatinine ratio, potassium, and proteinuria were measured. All patients had type 2 diabetes according to World Health Organization criteria (17). No subjects requiring renal dialysis were recruited. Ethical approval was obtained from University College London/University College London Hospital and Ealing Hospital ethics committees and from data protection at University College London. Middle-aged U.K. men (50–61 years, mean age 56.1 ± 3.5 years) participating in the Second Northwick Park Heart Study were used as a population-based sample for comparison of allele and haplotype frequencies (15).

The intron 1 A/C variant (refsnp 135539) was genotyped using forward primer CCAGGGGGAGGAAAGAGTGAA and reverse primer GCCACAACTAAGCAGGCAGTG to generate a product of 210 bp, which, when digested with HinfI, gave products of 148 and 62 bp in the presence of the C-allele. The L162V (refsnp 1800206) and intron 7 G/C (refsnp 4253778) variants were genotyped as previously described (15).

Statistical analysis was performed by univariate ANOVA and linear regression using SPSS (version 12; SPSS, Chicago, IL). Factors and covariates, together with all two-way interactions, were entered into the statistical model and removed in a stepwise manner until a parsimonious model containing only factors or interactions that were statistically significant was obtained. Mean age at diagnosis, adjusted for family history of diabetes, sex, and smoking, was generated by calculating residuals and adding to the overall mean value. Haplotype frequencies and effects were determined using THESIAS (18). Smoking status and BMI at diagnosis were not available; therefore, measures at recruitment were used. BMI, triglycerides, and duration of diabetes before insulin treatment were log₁₀ transformed to normalize distribution. The three V162 allele homozygotes were combined with V162 allele carriers in all analyses. Data are shown as means ± SE, and a P value of 0.05 was considered statistically significant.

The clinical characteristics of type 2 diabetic subjects are shown in Table 1. The intron 1 A→C variant is situated ∼12.5 kb 3′ of the site of transcriptional initiation of the PPARα gene, 55 kb 5′ of L162V, and 71 kb 5′ of the intron 7 G→C variant. Rare allele frequencies were intron 1 A→C 0.404 (95% CI 0.381–0.427), L162V 0.064 (0.053–0.076), and intron 7 G→C 0.187 (0.169–0.205). The intron 1 A→C variant was selected due to its rare allele frequency and close proximity to the 5′ end of the PPARα gene, whereas the L162V and intron 7 G→C variants have previously shown positive results in association studies (12–15). No other PPARα variants were examined. Genotype distributions were in Hardy-Weinberg equilibrium and were not different between the two study centers. Allele frequencies were not significantly different between type 2 diabetic subjects and those previously reported in healthy U.K. middle-aged men (15). All three variants were in linkage disequilibrium (intron 1-L162V D′ = 0.57, P = 0.0001; intron 1-intron 7 D′ = 0.46, P = 2.6 × 10⁻⁸; L162V-intron 7 D′ = 0.58, P = 2.7 × 10⁻²³).

Individually, PPARα gene variants did not influence age at diagnosis when examined by ANOVA (Table 2), but when examined in combination, both the intron 1 A→C (P < 0.001) and intron 7 G→C (P = 0.025) variants significantly influenced age at diagnosis and showed a significant interaction (intron 1 × intron 7 interaction, P < 0.001), acting synergistically to lower age at diagnosis (Table 3). In multivariate linear regression, sex (P = 0.001), currently smoking (P = 0.026), family history of diabetes (P < 0.001), intron 1 genotype (P = 0.001), intron 7 genotype (P = 0.045), and the intron 1 × intron 7 interaction (P = 0.003) were independent predictors of age at diagnosis. The effects of PPARα genotype were the same by study center, which did not independently influence age at diagnosis. Haplotype analysis revealed that PPARα haplotype significantly influenced age at diagnosis (P = 0.037) (Table 4). Haplotypes 6 and 8, comprising the intron 1 C- and intron 7 C-alleles with the L162V L162 allele (haplotype 6, C-L-C) or the V162 allele (haplotype 8, C-V-C), were associated with an age at diagnosis of 5.86 ± 2.57 (P = 0.02) and 9.98 ± 4.65 (P = 0.03) years, respectively, earlier than the common haplotype.

Given the dramatic haplotype effect on age at diagnosis, the role of PPARα in early-onset type 2 diabetes was examined. PPARα haplotype frequency was calculated in subjects stratified by age of diagnosis of ≤45 or >45 years, based on the previous use of this cutoff age in a linkage scan for age at diagnosis (19). Overall, PPARα haplotype frequency was highly significantly different between young-onset and later-onset subjects (P = 1.4 × 10⁻⁷), with haplotype 6 increasing risk of early-onset type 2 diabetes (odds ratio 3.75, 95% CI 1.65–8.56, P = 0.002). Haplotype frequencies were not different between the entire type 2 diabetes sample and middle-aged U.K. men without overt type 2 diabetes (P = 0.50) but were significantly different between middle-aged U.K. men and early-onset subjects (P = 0.024) (Table 4).

We investigated whether PPARα variants influence the progression of type 2 diabetes, as estimated by the duration of diabetes before commencing insulin monotherapy. In UDACS, carriers and homozygotes of the rare C-allele progressed to insulin treatment significantly earlier than A allele homozygotes (AA 9.4 ± 1.5 and AC + CC 5.3 ± 1.1 years, P = 0.002). The L162V and intron 7 G→C variants did not influence duration before insulin therapy. PPARα genotype did not influence whether subjects were on an oral hypoglycemic agent, insulin, or both an oral hypoglycemic agent and insulin therapy (not shown). Measures of progression were not available in EDSC.

Treatment with PPARα activators beneficially alters plasma lipid profile, and we previously demonstrated that variation in the PPARα gene influences plasma lipid levels in dyslipidemic type 2 diabetic subjects (12). We therefore examined whether PPARα gene variation influences plasma lipid levels in type 2 diabetic subjects, but observed no significant association with plasma triglycerides or LDL, HDL, or total cholesterol concentrations (not shown).

This study demonstrates that variation in the PPARα gene influences the onset and progression of type 2 diabetes, in concurrence with the recent observation that bezafibrate reduces the incidence and delays the onset of type 2 diabetes (11). Allele and haplotype frequencies were not different between healthy middle-aged U.K. men and the type 2 diabetes sample, but haplotype frequency was different between subjects with age at diagnosis ≤45 years and both those with later age at diagnosis and with healthy middle-aged U.K. men without overt type 2 diabetes, indicating that PPARα predisposes to early-onset type 2 diabetes. PPARα gene variants influenced age at diagnosis in combination and in haplotype analysis. Carriers of the intron 1 and 7 rare C-alleles had a dramatically younger age at diagnosis, an effect confirmed in haplotype analysis, in which haplotypes 6 and 8, comprising the intron 1 C- and intron 7 C-alleles with the L162 or V162 alleles, respectively, were associated with an age at diagnosis 5.8 and 10.0 years earlier than the common haplotype. Genome scans have identified several loci linked to age at diagnosis (19,20), one of which maps in a subset of families with age at diagnosis <45 years to chromosome 22q11 (19), 28 Mb centromeric to the PPARα gene.

The PPARα intron 1 genotype was also associated with a more rapid progression to insulin monotherapy, suggesting that PPARα influences the rate of development of insulin resistance and/or progression to β-cell failure. PPARα is expressed in liver, skeletal muscle, and pancreas (5) and could therefore affect the disease process at multiple levels. In rodents, PPARα activators lowered fasting plasma triglycerides, and free fatty acids, as well as lowered visceral adiposity and muscle and hepatic steatosis, resulting in improved hepatic and muscle insulin sensitivity and improved pancreatic β-cell function. Genetically determined reduced PPARα levels may influence any of the above-mentioned factors. We did not observe association between PPARα variation and plasma triglyceride levels (fasting triglyceride levels were not available for all subjects), thus decreasing the power to observe an effect. Studies that directly assess the impact of PPARα gene variants on levels of intramuscular and intrahepatic triglycerides, whole-body glucose uptake, and β-cell function will elucidate the mechanisms underlying the observed effects.

The PPARα genotype influences the risk of coronary artery disease (15), and it is conceivable that it could increase the likelihood of subjects presenting with heart disease and subsequently being diagnosed with type 2 diabetes, thus confounding the current study. The presence of cardiovascular disease was in fact associated with later age at diagnosis (absence 53.4 ± 0.5 years, presence 56.5 ± 0.7 years, P = 0.001) and did not influence the effect of PPARα genotype on age at diagnosis, indicating that these data are not confounded by cardiovascular disease. The differences between UDACS and EDSC in BMI, measures of glycemic control, blood pressure, and plasma lipids did not confound the effect of PPARα haplotype on age at diagnosis.

Mechanistically, several lines of evidence lead us to speculate that the intron 1 C- and intron 7 C-alleles mark a haplotype with reduced PPARα gene expression. First, in previous studies, the intron 7 C-allele shows opposing effects to the V162 allele (14,15), which encodes a more active PPARα in vitro (12). Furthermore, the intron 7 C-allele is associated with reduced response to fenofibrate in the DAIS (Diabetes Atherosclerosis Intervention Study) (21). Finally, PPARα activators delay onset of type 2 diabetes in humans (11) and rats (10), indicating that increased PPARα activity protects against the onset of type 2 diabetes.

PPARα interacts with several genes that influence type 2 diabetes. Hepatic PPARα gene expression is regulated by hepatocyte nuclear factor 4α (22), a closely related member of the nuclear hormone receptor superfamily that is associated with risk of type 2 diabetes (23). Additionally, genetic variation in PPARγ coactivator-1, which also coactivates PPARα, influences the insulin secretory response in Pima Indians (24). The retinoid X receptor α, with which PPARα forms heterodimers, also influences insulin resistance (25). It will be interesting to examine interaction between PPARα and hepatocyte nuclear factor 4α, PPARγ coactivator-1, and retinoid X receptor α gene variants in subsequent studies.

In summary, we demonstrate that PPARα gene variation influences the onset and progression of type 2 diabetes in human subjects. The mechanisms underlying these effects remain unclear and warrant future investigation.

D.M.F. and S.E.H. are supported by grants from the British Heart Foundation (FS 2001/075 to D.M.F. and RG2000015 to S.E.H.). J.W.S. is supported by a clinical training fellowship from Diabetes U.K. (BDA: RD01/0001357). Recruitment of patients in EDSC was supported by The Coronary Thrombosis Trust.

The authors thank the medical staff and patients who contributed to the UDACS and EDSC.