PPARG (Pro12Ala) genetic variant and risk of T2DM: a systematic review and meta-analysis

Type 2 diabetes mellitus (T2DM) is a complex disease caused by the interaction between genetic and environmental factors. A growing number of evidence suggests that the peroxisome proliferator-activated receptor gamma (PPARG) gene plays a major role in T2DM development. Meta-analysis of genetic association studies is an efficient tool to gain a better understanding of multifactorial diseases and potentially to provide valuable insights into gene-disease interactions. The present study was focused on assessing the association between Pro12Ala variation in the PPARG and T2DM risk through a comprehensive meta-analysis. We searched PubMed, WoS, Embase, Scopus and ProQuest from 1990 to 2017. The fixed-effect or random-effect model was used to evaluate the pooled odds ratios (ORs) and 95% confidence intervals (CIs) depending on the heterogeneity among studies. The sources of heterogeneity and publication bias among the included studies were assessed using I2 statistics and Egger's tests. A total of 73 studies, involving 62,250 cases and 69,613 controls were included. The results showed that the minor allele (G) of the rs1801282 variant was associated with the decreased risk of T2DM under different genetic models. Moreover, the protective effect of minor allele was detected to be significantly more in some ethnicities including the European (18%), East Asian (20%), and South East Asian (18%). And the reduction of T2DM risk in Ala12 carriers was stronger in individuals from North Europe rather than Central and South Europe. Our findings indicated that the rs1801282 variant may contribute to decrease of T2DM susceptibility in different ancestries.

Type 2 diabetes mellitus (T2DM) is the most common form of diabetes and is described as a highly prevalent multifactorial disorder 1 . According to the recent statistics of the International Diabetes Federation (IDF), the global T2DM epidemic significantly grows at an alarming rate among populations and so it has become a common health problem worldwide 2 . Although T2DM usually affects older adults, it is also gradually seen in children, adolescents and younger adults due to increasing levels of obesity, low physical activity and poor diet 3 . T2DM is recognized as a major cause of morbidity and leads to premature coronary heart disease progression (CHD), stroke, peripheral vascular disease (PVD), renal failure, and amputation 4 . T2DM is characterized by hyperglycemia, impaired insulin secretion (IS) and insulin resistance (IR) that results from the interaction between numerous genes and environmental factors 5,6 . The genetic architecture of complex traits is now to be related to several numbers of causal variants. But, the most important common variants show small to modest effect sizes 7,8 .
Single nucleotide polymorphisms (SNPs), the most common type of genetic variations between individuals, are the key players in precision medicine approach. SNPs are responsible for more than 80 percent of the variation between individuals which makes them ideal for genotype and phenotype association studies. Genetic association studies as powerful approach have identified several SNPs that are significantly associated with T2DM susceptibility 9,10 . Search strategy. A comprehensive literature search was performed on several databases including Pub-Med, Scopus, Embase, Web of Science (Wos), and ProQuest to collect relevant literature published from January 1990 to October 2017. We found the different synonyms of the related terms for all subjects by using thesauri systems such as Medical Subject Headings (MeSH) and EMBAS subject headings (Emtree). The combination of the following terms was used to design the search strategy: ("Type 2 Diabetes Mellitus" OR "insulin independent diabetes mellitus" OR "Noninsulin-Dependent Diabetes Mellitus" OR …) and ("PPAR Gamma", "Peroxisome Proliferator Activated Receptor Gamma" OR "PPAR G" OR "PPAR-G" OR "PPAR Gamma" OR "PPAR-Gamma" OR …). The finalized search syntax in PubMed was adjusted in other databases for a comprehensive search (available in Supplementary file 1).
For the selection of possibly relevant studies, the titles and abstracts of the articles were independently screened according to eligibility criteria with two reviewers (MR and MH).
Two reviewers (MR and MH) also reviewed the full-text articles to determine whether the selected articles adapted to the eligibility criteria and could be included/excluded in the final investigation. Conflicting opinions Scientific RepoRtS | (2020) 10:12764 | https://doi.org/10.1038/s41598-020-69363-7 www.nature.com/scientificreports/ were resolved through further discussion to achieve a consensus. Moreover, the reference lists of all eligible studies were also checked to identify additional potentially relevant literature.

Eligibility criteria.
All selected studies had to fulfill the following inclusion criteria: (1) Studies published in peer-reviewed journals.
(2) All case-control studies just conducted on the human that assessed any association between PPARG rs1801282 (Pro12Ala) polymorphism and risk of T2DM. (3) The data about the allele or genotype frequencies should be sufficient to calculate the odds ratios (ORs) with the corresponding 95% confidence intervals (CIs) of the polymorphism in both the case and control groups. (4) The control group included people without T2DM. (5) Studies that just published in the English language. (6) Studies that their full text was available. (7) Short communication and brief genetic report with sufficient data.
Following studies were excluded: (1) Family based association studies, including case reports and case series.
(3) In vitro, ex vivo or animal studies. (4) Studies lack sufficient data about allele frequency or data that could not be calculated. (5) Duplicate publications and redundant studies of duplicated data; for duplicate reports that published by the same authors using the same case series, only the most recent and the one with the largest sample size one was included.
Data extraction. Two researchers (LH and KH) extracted the following items were selected from all eligible studies using a redesigned form according to the aforementioned inclusion and exclusion criteria: (1) First author's names and year of publication (2) Country of setting or ethnicity of participants (3) The number of cases and controls (4) Mean age and body mass index (BMI) of participants (5) Genotyping method (6) The genotype and allele frequencies of PPARG gene variant in cases and controls (7) Hardy-Weinberg equilibrium (HWE) was calculated based on genotype frequencies of certain PPARG rs1801282 (Pro12Ala) gene polymorphisms in the control group.
The agreement between two researchers (LH and KH) was achieved by a discussion with a third expert person (MH) in the research team.
If there was a lack of genotype information, the reviewers contacted the corresponding author to get the required data.
Quality assessment. The Newcastle-Ottawa Scale (NOS) was used to assess the quality of included studies 33 .
The following items were selected for the inclusion of the study, including the selection of cases and controls (4 items, 1 point each), comparability between cases and controls (1 item, up to 2 points) and exposure in cases and controls (3 items, 1 point each).
The NOS has a score range of zero to nine, and studies with a rating of 7-9 were presumed to be of high quality, 4-6 as moderate quality, 4 or less was classified as low-quality studies. Quality assessment was also conducted by two investigators independently.
First, any disagreement regarding the quality assessment was resolved by checking and discussion between the two reviewers. If the two authors could not achieve a consensus, then a third reviewer would resolve the cases of conflict.
Statistical analysis. Within the study, the results were combined using RStudio (3.51). Genetic association studies do not follow a specific model and therefore multiple genetic models need to be investigated 34 . The odds ratio (OR) and 95% confidence interval (CI) were used to assess the association between rs1801282 polymorphism in PPARG gene and the risk of T2DM in seven genetic models as follow: allele model (G vs. C), homozygote model (GG vs. CC), heterozygote model (CG vs. CC), additive model (GG vs. CG), dominant model (GG + CG vs. CC), recessive model (GG vs. CC + CG), and co-dominant model (CG vs. CC + GG) which a ' 'C'' denotes a major allele; ' 'G'' denotes a minor allele.
The Cochrane Q-test index was used for detecting the existence of heterogeneity between the results of the primary studies 35 and I-square index (I 2 ) determined the degree of the heterogeneity in meta-analysis based on I2 value of 25%, 50%, and 75% were nominally regarded as low, moderate, and high estimates, respectively 36 .
We applied the random effect model (REM) for an inverse variance which was used to calculate the combining of primary results (the pooled OR estimate) if the heterogeneity was significant (P-value of Q-test < 0.05 or I 2 > 50%). Otherwise, the fixed effects model (FEM) was occupied to assess the primary effect of the genotype 37 .
The agreement on the genotype frequency with HWE in the controls was calculated using the Pearson's χ2 test for each study. Sensitivity analysis. Sensitivity analysis was accomplished by removing those studies that did not meet with HWE. Studies with a very poor quality score (equal to two or three) were also excluded from the meta-analysis to getting possible stronger results. Moreover, the leave-one-out method in the sensitivity analysis was conducted through consecutive excluding only one study in each time to assess the cause of heterogeneity and to determine whether any individual study influences the stability of final results (pooled OR) in meta-analysis.
Publication bias. We also appraised the fundamental sources of potential publication bias in Egger's linear regression test and visual inspection of the asymmetry of the Begg funnel plot 39 . If there was publication bias, the Duval and Tweedie trim-and-fill technique was accomplished to explore the impact of the publication bias on the results 40 .

Results
Characteristics of the included study. During the first stage of our comprehensive search, 6,622 studies were identified through electronic databases and hand searches. As illustrated in Fig. 1, 3,938 articles remained after excluding the duplicates. After reviewing the titles, abstracts of the primary studies, 3,746 papers were identified to be irrelevant. 192 potentially relevant articles were retrieved for further evaluation. Among those remaining studies, 120 studies were excluded for different reasons (56 studies had not sufficient or relevant data about T2DM including studies that evaluated the association of Pro12Ala with type 1 diabetes, metabolite traits, or diabetes complications, also assessed the link of other SNPs and genes with T2DM, and GWAS studies ; 15 studies were not English studies or not available full text; 34 studies were the exclusion of study design such as a clinical trial, cohort, case-series (had no control group), family-based studies, review and meta-analysis, letter to editors/research letters, meeting abstract, commentary, report, news, pilot study; and 15 studies were duplicate) that details are provided in Supplementary Table S1 online. www.nature.com/scientificreports/ Two articles due to the insufficient data were removed from further step (meta-analysis). Finally, a total of 73 studies with 62,250 T2DM patients and 69,613 controls met our inclusion criteria for overall meta-analysis after reading the full texts. It should be noted that these 73 studies had been reported by 66 articles.
Besides, other 13 studies data from four articles were lack genotypes or alleles frequency while only OR was reported for these. So we did not exclude from meta-analysis and analyzed using separately command from Stata 41 that is available in Supplementary file 2 online.
The characteristics and genotype frequencies of included studies are listed in Table 1 and Supplementary file 3 online.
The NOS score of eligible articles ranged from two to eight stars. 11 of included studies were evaluated to be high quality, 33 were low quality, and 28 studies were considerate as moderate quality.
There were 21 studies of Europeans, 17 studies of East Asians, 10 South Asians, 13 Greater Middle Eastern, five other, and two Hispanic or Latin American. Other ancestry groups such as South East Asian, Asian unspecified, other admixed ancestry, Native American, and African American or Afro-Caribbean included only one study.
Different genotyping methods consist of TaqMan, tetra-primer amplification refractory mutation system (TETRA-ARMS), restriction fragment length polymorphism (RFLP-PCR), mass spectrometry, direct sequencing, real-time PCR, and etc. which was listed in Table 1.
The genotype frequency of the control group met to HWE in the included studies except for five case-control studies and one study that not reported the p-value of HWE.
The results of meta-analysis. Combining the results of the primary studies showed a significant association between the Pro12Ala polymorphism and T2DM risk under REM in seven genetic models including Allelic (OR: 0.82, 95% CI: 0.76-0.89, P < 0.01) with high between-study heterogeneity (I 2 = 71%), homozygous, heterozygous, additive, dominant, recessive, and co-dominant genetic models. Further details on the genetic coding of the variant are provided in Table 2, and the forest plots are shown in Supplementary Fig. S1 online.
The OR analysis results designed for the allele (5 studies), additive (4 studies), and dominant (4 studies) genetic models that were analyzed according to just OR (totally 13 studies) were consistent with the main results of the meta-analysis (Supplementary file 2 online).  (Table 3).
There was not a significant association detected under all genetic models in the Greater Middle Eastern population.
Other ancestry categories including South East Asian, Asian unspecified, African American or Afro-Caribbean, Hispanic or Latin American, Native American, and Other admixed ancestry were not reliable to report due to the low number of publications.
Same to the results of the study by Ludovico et al. 27 , it was observed a significant between-study heterogeneity among Europeans, whereas it was not in other ancestry categories. So, to further subgroup analysis, data from Europe was stratified to "North European" (Scottish, British, and Finnish), "Central European" (Poles, Germans, French, Czechs), "South European" (Italians and Spaniards), and "not available subgroup data" (such as Caucasian) to confirm the previously reported findings. As presents in Fig. 2, the reduction of T2DM risk in G allele carriers from Northern and Southern Europe was almost equal (23% and 21%, respectively) but did not influence in Central Europe. However, it should be noted that the negative result may have been due to the low sample size in Central Europe studies.
Nevertheless, no significant association was found among age and year of publication (see Supplementary  Fig. S3, Fig. S4 online).
Sensitivity analysis. Although, the combined results remained stable after removing single studies in the allele, homozygote, heterozygote, dominant, recessive, and co-dominant models, but the pooled OR was go away from significantly after omitting the study by Motavallian et al. (2013) 18  Furthermore, we excluded those HWE-violating studies for sensitivity analyses. However, the pooled ORs in overall was not statistically altered, indicating that the results were stable (see Supplementary Fig. S6  www.nature.com/scientificreports/ The evidence of sensitivity analysis suggested that removing poor-quality studies could not influence the combined results (see Supplementary Fig. S7 online). No evidence of asymmetry was observed among the primary studies by Begg funnel plots in any comparisons (see Supplementary Fig. S8 online). The Egger's regression test indicated that there was no evidence of potential statistical publication bias in either of genetic models except in allele model and the results were constant and sturdy ( Table 2).

Discussion
Diabetes is one of the major driver of morbidity and mortality worldwide and in spite of introducing approximately 100 identified susceptibility loci with robust interaction signals with T2DM but most of them show little value in clinical practice 43 .
It seems that the PPAR-γ plays an important role in the pathological process of diabetes. The functional role of PPAR-γ has been well described, and its variations in association with TDM and obesity have been extensively investigated in different ethnicities 44  www.nature.com/scientificreports/ Pro12Ala (rs1801282) is considered to be the most analyzed common variants in the PPARG gene which decreases the receptor binding affinity to the responsive elements and consequently inducing a reduction in transcriptional activity both with and without PPAR-γ agonists using effect on receptor structure which eventually leads to insulin sensitivity and abnormalities of adipose tissue formation 47 . The more common (C) and rare (G) alleles of rs1801282 encode the 'Pro' and ' Ala' amino acids, respectively. According to the previous GWAS, the Pro allele of this variant was reported to increase the risk of T2DM. But, the Ala allele has a protective effect on T2DM development 12,46,48 .
The result of the present systematic review and meta-analysis consists of 62,250 cases and 69,613 controls from 73 studies in order to achieve substantial evidence of any association between PPAR-γ rs1801282 and T2DM risk. The findings of this meta-analysis showed that the G allele of Pro12Ala polymorphism could cause approximately an 18% reduction in the probability of developing T2DM. The reduction of the T2DM risk was also detected vary across different ethnicities; European (18%), East Asian (20%), and South Asian (18%) while no association was founded in the Greater Middle Eastern population.
As is obvious, the differences in the reduction of T2DM risk between those of European, South Asian, and East Asian ancestries are not really very different in the present study whereas in a study by Ludovico et al. 27 the reduced risk in the Asian population more than European (35% vs. 14%) was reported. Consistent with the previous report, the reduced T2DM risk was stronger in North European populations in stratified Europe from Northern to Southern gradient 27 .
The contradictory results from the different ethnic populations appear interesting that it can be partially attributed to the small sample size and the fact that different genetic backgrounds and various environmental factors might be lead to conflicting results from the same polymorphism among primary studies with different ethnicities 49  It should be noticed that the results of our study could be closer to reality due to the number of cases, controls, and studies of different ancestries. Furthermore, a study by Huguenin et al. showed a significant effect of the Ala allele on reduction of T2DM risk in Caucasians 28 . Also, a meta-analysis by Gouda in 2010 observed that the PPAR-γ Pro12Ala variant is positively associated with a reduction of T2DM risk 50 .
The analysis of subjects harboring polymorphisms within PPAR-γ has made an important contribution in providing convincing genetic evidence of a role in glucose homeostasis, lipid metabolism, and determination of fat mass. Such studies also provided data for the underlying mechanisms of insulin sensitivity, PPAR-γ action and, T2DM risk. However, neither environmental triggers nor genetics alone can explain T2D pathogenesis because of its multifactorial nature. Hence, more functional studies and large population-based validation surveys are needed to perform. To the best of our knowledge, this is one of the most comprehensive meta-analysis of the association of PPAR-γ rs1801282 (Pro12Ala) polymorphism and T2DM risk.
Limitations. Despite our promising findings, multiple limitations should also be addressed. Firstly, T2DM is a complex disorder and we only discussed individual genetic variant without having to consider the interaction with other genetic variants or environmental variables (lifestyle, smoking, etc.). Table 2. The meta-analysis results of association between Pro12Ala variant and T2DM risk. OR odds ratio, CI confidence interval, REM random effect model, FEM fixed-effects model, I 2 I-squared metric of the heterogeneity, P H P value of heterogeneity, Q test, P Egger P value of Egger linear regression test. I 2 value of 25%, 50%, and 75% were nominally regarded as low, moderate, and high estimates, respectively.

Genetic model
No of studies www.nature.com/scientificreports/ www.nature.com/scientificreports/ Secondly, owing to the restriction of the accessibility of original research information, the study did not consider other appropriate variables such as gender, age, and genotype frequency data as the genotype frequency data was not available in some articles (11 from 73) and only the allele model was evaluated in order to assess the association among the overall population. Therefore, a more precise association with sufficient data should be explored. These results should be interpreted with caution until further sequencing approaches verification and greater meta-analysis is required.
Thirdly, significant publication bias was observed in some T2DM comparisons including the allele model. This may be due to the fact that the ethnicity of the populations in the early studies is mostly European or Asian, or that there is a greater number of low or medium quality studies rather than the high-quality ones. And also, significant heterogeneity was detected in the primary study results, indicating that the inconsistent results of the included studies could not be fully explained by differences in ethnic background, BMI, age, and other unmeasured variables of participants that may also partially attribute heterogeneity to the inter-study.
Fourthly, there are some gaps about particular ancestry groups including; Aboriginal Australian, African unspecified, Asian unspecified, Central Asian, Oceanian, and Sub-Saharan African that should be addressed.
Finally, obesity is also a significant intermediate factor in the rise of T2DM and having BMI information would be important and useful in the association analysis. But the definitions of obesity were not the same or accessible in our included studies. Therefore, in our subgroup analysis with BMI, the mean BMI of populations was used which does not indicate the exact BMI individual level of the study. So, this may be causing the contradictory result of this stratification.
Moreover, despite these limitations, our comprehensive research can still make a valid conclusion.

Conclusion
Genomic association studies help in disease predispositions by using genomic variants which have been discovered by GWASs 10 . The introduced genetic variants can be used to detect high-risk individuals for certain diseases. Thereby personalized medicine goals for improving patient outcomes will be achieved through such studies. A genetic variant that is associated with disease in one ethnic group but not in another may indicate ethnic differences in risk disease predisposition. So the result of genetic association studies represents only the tip of the iceberg and meta-analysis study shows great benefit for the personalized medicine approach 10 . The PPARG Pro12Ala variant in current meta-analysis indicated enough evidence for the presence of a significant association of individuals carrying the PPARG Ala12 variant with a reduced risk of T2DM. Additionally, the results of analysis under diverse ancestries confirm the importance of SNPs association studies in different ethnicities. But this effect is not very different among European compared to other ancestries. And among Europeans, existence stronger in North European, and barely significant in South European, and not being in South European. The genetic architecture of diabetes, including polygenicity and most risk variants, has been discussed in previous studies with important implications for precision medicine 51 . But several obstacles complicate the translation www.nature.com/scientificreports/ of novel loci and variants into the clinical decision practice, overcoming these will lead to the development of new drugs to treat T2DM. It appears that the lack of efficacy in anti-diabetic drugs is returned to the preclinical models in clinical trials. Human genomic advancement provides a better condition for proper assessment of drug development efficacy in pharmaceutical R&D through combining a targeted pathway or genetic alteration to a desirable phenotype (T2DM).
Decision-making through precision medicine needs therapeutic approaches which are obtained by the genetic association study of the common variants/loci.
The identification of the right drugs that are most effective and safe for each patient and reducing the global economic impact will be possible when the genetic information of the diabetic patients will provide a valuable resource to predict T2DM progression. Genetic studies are one of the most important approach in order to predict and prevent T2DM in the near future. It is hoped, therefore, that the decades ahead will elucidate the extent to which the inherited variation and its interaction with the environmental factors help clinicians' diagnostics.

Data availability
Data sharing is not applicable to this type of article.