Diabetologia

, Volume 46, Issue 7, pp 977–983

Genetic epistasis of adiponectin and PPARγ2 genotypes in modulation of insulin sensitivity: a family-based association study

  • W.-S. Yang
  • C. A. Hsiung
  • L.-T. Ho
  • Y.-T. Chen
  • C.-T. He
  • J. D. Curb
  • J. Grove
  • T. Quertermous
  • Y.-D. I. Chen
  • S.-S. Kuo
  • L.-M. Chuang
  • the Sapphire Study Group
Article

DOI: 10.1007/s00125-003-1136-2

Cite this article as:
Yang, W., Hsiung, C.A., Ho, L. et al. Diabetologia (2003) 46: 977. doi:10.1007/s00125-003-1136-2

Abstract

Aims/hypothesis

Genetic interactions in modulating the phenotypes of a complex trait, such as insulin sensitivity, were usually taken for granted. However, this has not been commonly shown. Previous studies have suggested that both PPARγ2 and adiponectin genes could influence insulin sensitivity. Therefore it is likely that they could modulate insulin sensitivity through gene to gene interactions.

Methods

We genotyped 1793 subjects of Chinese and Japanese descendents from 601 hypertensive families recruited in Sapphire study for a T94G in the adiponectin gene exon 2 and the PPARγ2 Pro12Ala polymorphisms. Serum insulin concentrations and insulin resistance index (HOMAIR) were used as the markers of insulin sensitivity.

Results

We found that the T allele of adiponectin gene was associated with a higher Ins60 and higher area under curve of insulin (AUCi) in OGTT utilizing all subjects in a mixed model that corrected for family effects. Important interactions between adiponectin and PPARγ2 genotypes were found in fasting insulin concentrations (Ins0), insulin concentrations at 2-h (Ins120) in OGTT and insulin resistance index (HOMAIR). The main effects of the PPARγ2 genotypes were in the plasma glucose concentrations in OGTT. In contrast, the main effects of adiponectin genotypes were in every insulin variable, including Ins0, Ins60, Ins120, AUCi and HOMAIR. The subjects carrying the adiponectinG allele and the PPARγ2Ala12 allele seemed to be more insulin sensitive.

Conclusion/interpretation

These results showed that adiponectin is a genetic factor associated with insulin sensitivity. Interactions with PPARγ2 genotypes modified this association.

Keywords

AdiponectinPPARγ2polymorphismsgene interactioninsulin sensitivity

Abbreviations

PPAR

peroxisome proliferator-activated receptor

HOMA

homeotasis model assessment

T2DM

Type 2 diabetes mellitus

Type 2 diabetes mellitus (T2DM) has daunted many countries in the developed or developing world [1]. Insensitivity to insulin stimulation is generally accepted to be the fundamental pathophysiology of T2DM [2]. As a classical example of polygenic (also called multi-factorial or complex) diseases, the pathogenesis of T2DM is attributed to the collective actions of many genetic and environmental factors and the complex interactions among them [3]. The contribution of gene to gene interactions in shaping the phenotypes of T2DM, such as insulin sensitivity, is usually taken for granted. However, it is only occasionally shown [4, 5].

Among many potential thrifty genes, the Pro12Ala variants of the human PPARγ2 gene have been extensively surveyed in many genetic association studies by other investigators [6, 7, 8]. Previously we reported the association of the PPARγ2 Ala12 variant with better glucose tolerance and insulin sensitivity among subjects from Japanese and Chinese hypertension families in the Stanford Asia Pacific Program in Hypertension and Insulin Resistance (Sapphire) study [9]. Due to its family-based study design utilizing the comparisons between siblings, the genetic association study in Sapphire was expected to have better control in early childhood environmental factors and genetic background than a population-based genetic association study [9, 10, 11].

The other sensible genetic candidates for T2DM are adipose-derived secreted molecules regulated by PPARγ2, such as leptin, resistin and adiponectin. Among them, the recombinant adiponectin protein has been shown in mice to be capable of lowering plasma glucose and fatty acids, improving insulin sensitivity and reducing body weight [12, 13, 14]. It has also been reported that adiponectin genotypes in humans influence insulin sensitivity and the risk for T2DM [15, 16, 17, 18, 19]. Recent studies using knockout mice confirmed these biological functions of adiponectin [20, 21].

We have shown that treating T2DM patients with rosiglitazone, a PPARγ2 agonist raised their plasma adiponectin concentrations by more than two-fold [22]. The other investigators have also shown that the administration of glitazones increased the expression of adiponectin in humans as well as in animals and in cultured adipocytes [23, 24, 25, 26]. These results together suggest a possibility of genetic interactions between PPARγ2 and adiponectin, and its attribution to insulin sensitivity.

In this study, we investigated the association of adiponectinT94G polymorphism with metabolic phenotypes and the effects of its interaction with the PPARγ2 Pro12Ala polymorphism on these phenotypes among the large cohort from the Sapphire [9, 10]. We found that the T allele of adiponectin was associated with post-glucose load hyperinsulinaemia. The main effects of the adiponectin and PPARγ2 genotypes and the genetic interactions between them on the metabolic variables were revealed in further analyses. The subjects having both the G allele of adiponectin and the Ala12 allele of PPARγ2 seemed to be more insulin sensitive.

Subjects and methods

Subjects and phenotypic characterization

The characteristics of study population, the inclusion and exclusion criteria of Sapphire study were detailed in several previous publications [9, 10]. Notably, diabetic subjects diagnosed based on the WHO criteria were excluded [27]. This study incorporated both the concordant (both hypertension) and the discordant sibling-pairs (one hypertension, the other hypotension) in design. A total of 2525 subjects of Japanese or Chinese descendents were recruited from 6 centres at San Francisco, Hawaii and Taiwan. In this report, 1793 subjects including parents and siblings from 601 families were genotyped. Written informed consent was obtained from all participants. The study was approved by the ethics board of each participating institute. Characterization of the subjects by fasting plasma glucose, insulin, triglyceride, total cholesterol, lipoprotein profile, and anthropometric measurements, including height, weight, waist and hip circumference were described in detail previously [9, 10]. A 75-g OGTT was carried out and plasma glucose and insulin concentration at 1-h and 2-h post-glucose load were measured [9, 10]. The fasting plasma sample was obtained right before the OGTT was conducted.

Extraction of genomic DNA and genotyping

Total genomic DNA was purified from peripheral blood leukocytes using a DNA extraction kit of Puregene (Minneapolis, Minn., USA), following the manufacturer's protocol. The primers used for PCR amplification for the exon 2 of human adiponectin gene were: 5′-TAG AAG TAG ACT CTG CTG AGA TG-3′ and 5′-CTC CCT GTG TCT AGG CCT TAG-3′. The PCR reactions were carried out in a total volume of 15 µl containing 20 ng of genomic DNA with an initial denaturation at 94°C for 5 min, followed by denaturing at 94°C for 30 s, annealing at 68 to 60°C for 1 min with the annealing temperature stepping down 2°C for every five cycles, and polymerization at 72°C for 40 s; then by a final extension at 72°C for 10 min. Of the amplified DNA 4 µl was digested with the BspHI enzyme (New England BioLabs, Beverly, Mass., USA), and then electrophoresed on a 2% agarose gel. The resulting fragment of PCR was 422 base pairs (bp). As the T94G substitution abolished a BspHI restriction site, two fragments of 265 and 157 bp after digestion indicated the presence of T allele. This polymorphism is equivalent to the SNP45 described elsewhere [16]. The genotyping of PPARγ2 Pro12Ala was published previously [9].

Statistical analyses

Association analysis for adiponectin genotypes—All variables were in SI units except that in HOMAIR. The data were given in means and S.E. In order to compare each outcome variable between the G variant-carrying individuals (including the genotypes T/G and G/G) and their siblings with the genotype T/T, only the siblings with discordant genotypes in T94G polymorphism of the adiponectin gene were included in the analysis. Paired analysis was used to test whether there was difference in variables of interest between the siblings discordant for genotypes after adjustment for age, sex, BMI, ethnicity and area of enrollment by using analysis of covariance with these variables as covariates. The analyses were done using the SAS version 8.0 PROC GLM. We also carried out analysis of covariance using a mixed model to assess whether the adiponectin genotypes affect the outcome variables controlling for covariates age, sex, BMI, ethnicity and area of enrollment as fixed effects, and controlling for clustering among families as a random effect. The analyses were performed using the SAS 8.0 PROC MIXED. All statistical tests were two-tailed. A p value of less than 0.05 was considered statistically significant.

Analyses on adiponectin gene and PPARγ2 gene interactions

In order to assess the interactions between the adiponectin and PPARγ2 genotypes, we carried out an analysis of covariance using a mixed model to evaluate the contribution of the main effects of adiponectinT49G and PPARγ2 Pro12Ala respectively, and their interactions for each variable of interest in the same model controlling for covariates age, sex, BMI ethnicity and area of enrollment as fixed effects and entering family as a random effect. The analyses were done using the SAS 8.0 PROC MIXED.

The differences of several variables between the subjects with different adiponectin genotypes within fixed PPARγ2 genotypes after adjusting for age, sex, ethnicity, BMI, area of enrollment and family effects were compared. The values/bars reported are the least square means for each subgroup. The p values were obtained using the SAS 8.0 PROC MIXED. The SAS statistical program is from the SAS Institute (Cary, N.C., USA).

Results

The genotype and allele frequencies of adiponectinT94G polymorphism in Japanese and Chinese subjects were comparable (Table 1). The genotype and allele frequencies in the hypertensive probands were similar and did not deviate from Hardy-Weinberg equilibrium (Table 1). The genotype and allele frequencies, and the basic characteristics of these subjects categorized by the PPARγ2 genotypes were reported previously [9]. Using a sibling-based comparison for phenotypic variables, we found that the subjects with the T/T genotype (n=282) had higher mean plasma insulin concentration at 1 h (Ins60) during OGTT than that of the sibs with the discordant genotypes (T/G and G/G, n=312) after adjustment for age, sex, BMI, ethnicity and the area of enrollment (562±27 vs. 483±18 pmol/l, p=0.0067). The difference in the AUCi in OGTT was of borderline statistical significance (828±37 vs. 743±27 pmol.h/l, p=0.0578). The differences between discordant adiponectin genotypes in the other variables did not reach statistical significance (data not shown).
Table 1.

The proportions (%) and the numbers of subjects with specific adiponectinT94G genotypes and alleles among Chinese or Japanese subjects and hypertensive probands

  

Genotypes

Alleles

   

TT

TG

GG

T

G

Subjects

Chinese

50.15%

39.77%

10.08%

70.04%

29.96%

667

529

134

1863

797

Japanese

50.28%

41.25%

8.47%

70.90%

29.10%

273

224

46

770

316

Probands

All

47.83%

41.52%

10.65%

68.59%

31.41%

265

230

59

760

348

Chinese

47.90%

41.48%

10.62%

68.64%

31.36%

194

168

43

556

254

Japanese

47.65%

41.61%

10.74%

68.46%

31.54%

71

62

16

204

94

To confirm the above findings, we also used a mixed model to correct for familial effects. This method enabled us to include more subjects for analyses. The basic characteristics of the subjects included in this analysis were shown in Table 2. After adjusting for age, sex, BMI, ethnicity and area of enrollment, only the means in Ins60 and AUCi in OGTT were different between subjects with the T/T genotype and those otherwise (Table 3). Post-glucose load hyperinsulinaemia among subjects with the T/T genotype in order to normalize plasma glucose concentrations shown in these two analyses suggests that they could be more insulin resistant. We also carried out the analyses by categorizing the subjects into three adiponectin genotypes (n=844 of T/T, n=701 of T/G and n=168 of G/G). The results were quite similar, only the mean Ins60 (525±18 vs. 455±20 vs. 461±35 pmol/l, p=0.0035) and AUCi (785±26 vs. 702±28 vs. 719±50 pmol.h/l, p=0.0204) were significantly different among the three genotypes. The differences were primarily between the T/T and the other two genotypes, suggesting a dominant effect of the G allele in association with insulin sensitivity. Thus this allowed us to pool G/T and G/G genotypes together in all the analyses. No significant difference among the three genotypes was observed in the other variables (data not shown).
Table 2.

Characteristics of the subjects discordant for the adiponectin genotypes are compared in subjects with the T/T vs. those with the T/G or G/G

Variables

T/T (n=844)

T/G or G/G (n=869)

Means±S.E.

Means±S.E.

p=**

Gender (male/female)

372/472

389/480

NS

Ethnicity (Chinese/Japanese)

619/225

631/238

NS

Area (Taiwan/San Francisco/Hawaii)

513/96/235

509/92/268

NS

Age (years)

50.78±0.38

50.88±0.38

NS

BMI (kg/m2)

25.65±0.16

25.96±0.16

NS

Waist (cm)

85.71±0.47

86.64±0.47

NS

Waist to hip ratio

0.88±0.00

0.89±0.00

NS

** NS for p>0.1

Table 3.

Comparisons of metabolic variables between subjects with discordant adiponectin genotypes after adjustment for age, sex, BMI, ethnicity and area of enrollment using a mixed model corrected for family effects

Variables*

T/T (n=844)

T/G or G/G (n=869)

Means±S.E.

Means±S.E.

p=***

Glu0

5.35±0.05

5.31±0.05

NS

Glu60

9.52±0.14

9.35±0.14

NS

Glu120

7.81±0.14

7.71±0.14

NS

AUCg

16.11±0.21

15.86±0.21

NS

Ins0

54.5±1.5

53.4±1.5

NS

Ins60

525.0±18.4

455.9±18.4

0.0008

Ins120

475.3±20.0

447.1±20.0

NS

AUCi

784.7±26.3

705.2±26.3

0.0056

HOMAIR**

1.84±0.06

1.84±0.06

NS

*Abbreviations: Glu, plasma glucose at the specific time point in OGTT; Ins, plasma insulin at the specific time point in OGTT; AUCg, area under curve of plasma glucose in OGTT, AUCi, area under curve of plasma insulin in OGTT

** HOMA: homeostasis model assessment calculated by [Insulin/22.5e-In(Glucose)]

*** p values greater than or equal to 0.1 are marked NS (not significant), p values less than 0.05 are in bold

Next we investigated the effects of genetic interactions between the adiponectinT94G and the PPARγ2 Pro12Ala polymorphisms. Previously, we found in the same study population that the PPARγ2 Ala12 variant was associated with better glucose tolerance at 0 (Glu0) and 1 h (Glu60) in OGTT and with lower insulin resistant index by HOMA (HOMAIR) [9]. Using two-way ANOVA analysis, we found that there were significant interactions between the adiponectin and PPARγ2 genotypes in insulin concentrations at 0 (Ins0) and 2 h (Ins120) in OGTT and HOMAIR, with adjustment for age, sex, BMI, ethnicity and area of enrollment, whereas their interactions in Glu0 and AUCi were of borderline significance (Table 4). Interestingly, the main effects of adiponectin genotypes were all significant on every parameter (Ins0, Ins60, Ins120 and AUCi) of insulin in OGTT and HOMAIR. The main effect of the adiponectin genotypes on Glu0 was of borderline significance (Table 4). In contrast, the main effects of PPARγ2 genotypes were only significant on Glu60 and area under curve of glucose (AUCg) in OGTT (Table 4).
Table 4.

Significance levels of the adiponectin or PPARγ2 genotypes on the main effects and their interactions after adjustment for age, sex, ethnicity, BMI, and area of enrollment using a mixed model corrected for family effects

Variables*

Adiponectin T94G

PPARγ2 Pro12Ala

Interactions

Main effects

Main effects

F(d.f.s)=**

p=

F(d.f.s)=

p=

F(d.f.s)=

p=

Glu0

3.61(1,811)

0.0671

0.36(1,811)

0.5491

2.96(1,811)

0.0859

Glu60

0.47(1,687)

0.4951

5.76(1,687)

0.0167

0.01(1,687)

0.9271

Glu120

0.37(1,715)

0.5459

2.47(1,715)

0.1167

0.09(1,715)

0.7640

AUCg

0.69(1,678)

0.4063

5.35(1,678)

0.0210

0.08(1,678)

0.7751

Ins0

4.72(1,808)

0.0301

0.41(1,808)

0.5243

4.81(1,808)

0.0285

Ins60

8.64(1,696)

0.0034

0.58(1,691)

0.4468

1.81(1,696)

0.1795

Ins120

5.54(1,722)

0.0189

0.68(1,722)

0.4083

4.09(1,722)

0.0435

AUCi

9.07(1,682)

0.0027

0.73(1,682)

0.3937

3.32(1,682)

0.0691

HOMAIR

4.20(1,807)

0.0408

0.64(1,807)

0.4255

5.90(1,807)

0.0154

** F(d.f.s): F-value (degrees of freedom for the numerator and denominator of the relevant F statistics)

To clearly illustrate the effects of interactions, these subjects were grouped by their genotypes. It was apparent that the subjects carrying both the adiponectinG and PPARγ2 Ala12 alleles were more insulin sensitive. They tended to have lower fasting plasma glucose and insulin, plasma insulin concentrations in OGTT, and HOMAIR (Table 5, Fig. 1). Subjects with the adiponectinG allele even when coupled with PPARγ2 Pro/Pro still had lower insulin concentrations at 1 h and AUCi in OGTT than subjects with the T/T genotype (Table 5 and Fig. 1A).
Table 5.

Comparisons between the subjects with different combinations of adiponectin and PPARγ2 genotypes. The p values in the 1st column from the right indicate the comparisons between the different adiponectin genotypes within a group with the same PPARγ2 genotypes. The p values at the bottom indicate the comparisons between the different PPARγ2 genotypes within a group with the same adiponectin genotypes

Variables

PPARγ2 genotypes

Adiponectin genotypes

p=

T/T

G/-

n

Pro/Pro

779

804

Ala/-

63

62

Glu0

Pro/Pro

5.34±0.05

5.33±0.05

NS

Ala/-

5.46±0.15

5.08±0.16

0.0649

  p=

NS

NS

Glu60

Pro/Pro

9.57±0.14

9.41±0.14

NS

Ala/-

8.90±0.38

8.69±0.40

NS

  p=

0.0833

0.0721

Glu120

Pro/Pro

7.85±0.14

7.76±0.14

NS

Ala/-

7.46±0.39

7.21±0.42

NS

  p=

NS

NS

AUCg

Pro/Pro

16.18±0.22

15.95±0.21

NS

Ala/-

15.27±0.58

14.82±0.61

NS

  p=

NS

0.0618

Ins0

Pro/Pro

54.2±1.6

54.2±1.5

NS

Ala/-

59.1±4.4

45.1±4.7

0.0234

  p=

NS

0.0554

Ins60

Pro/Pro

519.3±19.0

459.1±18.9

0.0050

Ala/-

599.8±52.9

439.5±55.3

0.0256

  p=

NS

NS

Ins120

Pro/Pro

467.0±20.5

453.2±20.5

NS

Ala/-

479.5±54.8

410.8±58.6

0.0225

  p=

0.0430

NS

AUCi

Pro/Pro

774.5±27.1

711.8±26.8

0.0357

Ala/-

916.4±74.3

665.8±76.8

0.0118

  p=

0.0605

NS

HOMAIR

Pro/Pro

1.83±0.07

1.88±0.07

NS

Ala/-

2.04±0.19

1.44±0.20

0.0201

  p=

NS

0.0294

Fig. 1.

AComparisons of AUCi between the subjects with different combinations of the adiponectin and PPARγ2 genotypes, the bars show means ± S.E., *p=0.0357, **p=0.0118. B Comparisons of AUCg between the subjects with different combinations of the adiponectin and PPARγ2 genotypes, the bars showed means ± S.E. C Comparisons of HOMAIR between the subjects with different combinations of the adiponectin and PPARγ2 genotypes, the bars showed means ± S.E., ***p=0.0201

Discussion

In this report, we showed that the adiponectinT94G genotypes indeed were associated with the variation in insulin sensitivity, indicated indirectly by the serum insulin concentrations and insulin resistance index by HOMA (HOMAIR), in a cohort of Chinese and Japanese hypertension families recruited in the Sapphire study. The subjects with the T/T genotype had higher post-glucose load insulin concentrations but similar glucose concentrations than those with the other genotypes, indicating insulin resistance associated with T/T genotypes. It should be noted that diabetic subjects were excluded in the recruitment of Sapphire study [10]. Compared with four recent population-based genetic association studies [15, 16, 17, 18], the advantage of the current study is that this is a family-based association study utilizing siblings as control subjects. As they share similar genetic and environmental backgrounds with the population control subjects, the observed phenotypic differences between them are more likely the results of genetic differences in interest [11].

There has been a great interest in tackling the genetic make-up of human complex traits or diseases, such as diabetes, obesity and hypertension. Although a huge amount of genetic data has been accumulated in the past, not very many have addressed the issue of genetic interactions [4, 5]. To our knowledge, our study is one of the earliest to investigate the genetic interactions in connection with insulin resistance. Significant interactions between the genotypes of adiponectin and PPARγ2 in relation to fasting insulin and post-glucose load insulin concentration at 2 h and HOMAIR were noted, providing some genetic evidence in humans that these two genes could participate either in the same pathway or in two interdependent pathways related to the biological processes of insulin sensitivity. Consistent with this, previous studies in cell culture, in animals and in humans have shown that adiponectin expression either at mRNA or at protein levels could be up-regulated by PPARγ2 activation [21, 22, 23, 24, 25]. This study has shown that a genetic association study is capable of detecting genetic epistasis.

The molecular mechanisms behind the genetic interactions between adiponectin and PPARγ2 genotypes are not clear. The T94G polymorphism of adiponectin is a silent mutation for Gly15 (GGT to GGG). We speculated that it might be in linkage dis-equilibrium with the other genetic alterations, probably a regulatory mutation. In contrast, the PPARγ2 Ala12 variant was previously shown to have reduced transactivating ability [8]. Although only a half-site of PPARγ response element (PPRE) was recognized in the proximal promoter of adiponectin by sequence analysis, synthetic PPARγ2 agonists were shown to enhance transcription of luciferase driven by the proximal 2 kb of adiponectin promoter [23]. How PPARγ2 variants might activate the adiponectin gene in different promoter contexts is an important question to address in the future.

In conclusion, we found that the T94G polymorphism of the adiponectin gene was associated with insulin sensitivity, represented by serum insulin levels and HOMAIR, in the subjects from a large hypertensive family cohort. More importantly, the effect of adiponectin was modified by the PPARγ2 Pro12Ala genotypes, indicating a genetic interaction in association with insulin sensitivity. These observations suggest that both the adiponectin and PPARγ2 genes and the interactions between them could play a role in the etiopathogenesis of insulin resistance.

Acknowledgements

The authors would like to thank Ms. C.-L. Chao and K.-C. Lee for their technical assistance. We thank patients and their families for participating in this study. We also thank S. Mockrin and S. Old of the National Heart, Lung and Blood Institute, and the other members of the Sapphire project for their help. This study was supported by grants (NSC 85-2331-B002-350Y, NSC 86-2314-B002-345Y, NSC 87-2312-B002-021Y, and NSC 91-3112-B-002-019) from the National Science Council, a grant (BS-090-pp-01) from National Health Research Institutes, a grant from the Department of Education (89-B-FA01-1-4), Taiwan, R.O.C. and a grant (UO1 HL54527-0151) from the National Heart Lung and Blood Institute (USA). We also apologize to the authors whose work on PPARγ2 could not be cited here because of the limitations on the number of references.

Copyright information

© Springer-Verlag 2003

Authors and Affiliations

  • W.-S. Yang
    • 1
    • 2
  • C. A. Hsiung
    • 3
  • L.-T. Ho
    • 4
  • Y.-T. Chen
    • 5
  • C.-T. He
    • 6
  • J. D. Curb
    • 7
  • J. Grove
    • 7
  • T. Quertermous
    • 8
  • Y.-D. I. Chen
    • 9
  • S.-S. Kuo
    • 1
  • L.-M. Chuang
    • 1
    • 2
  • the Sapphire Study Group
  1. 1.Department of Internal MedicineNational Taiwan University HospitalTaipeiTaiwan
  2. 2.Graduate Institute of Clinical MedicineNational Taiwan UniversityTaipeiTaiwan
  3. 3.Division of Biostatistics and BioinformaticsNational Health Research InstitutesTaipeiTaiwan
  4. 4.Department of Medical Research and Education, Taipei Veterans General Hospital; and Faculty of Medicine, School of MedicineNational Yang-Ming UniversityTaipeiTaiwan
  5. 5.Department of Endocrinology and MetabolismTaichung Veterans General HospitalTaichungTaiwan
  6. 6.Department of Endocrinology and MetabolismTri-Service General HospitalTaipeiTaiwan
  7. 7.University of HawaiiHonoluluUSA
  8. 8.Stanford University School of MedicineStanfordUSA
  9. 9.Cedars-Sinai Medical CenterLos AngelesUSA