PPARG c.1347C>T polymorphism is associated with cancer susceptibility: from a case-control study to a meta-analysis

Recently, several studies suggested that PPARG c.1347C>T polymorphism was correlated with cancer risk. However, past results remained controversial. In this study, we performed a case-control study on the relationship of PPARG c.1347C>T polymorphism with risk of non-small cell lung cancer (NSCLC) and subsequently carried out a meta-analysis to further assess the association between PPARG c.1347C>T and overall cancer. In our case-control study, after adjusting by age, sex, body mass index (BMI), smoking and drinking, a tendency to increased NSCLC risk was noted (CT/TT vs. CC: adjusted OR, 1.21; 95% CI, 0.97–1.51; P = 0.097). In the meta-analysis, we found a significant association between PPARG c.1347C>T polymorphism and overall cancer risk (T vs. C: OR, 1.13; 95% CI, 1.03–1.23; P = 0.006; TT vs. CC: OR, 1.29; 95% CI, 1.07–1.56; P = 0.008, CT/TT vs. CC: OR, 1.11; 95% CI, 1.02–1.21; P = 0.014 and TT vs. CT/CC: OR, 1.26; 95% CI, 1.04–1.52; P = 0.016). In a subgroup analysis by ethnicity, evidence of significant association between PPARG c.1347C>T polymorphism and cancer risk was found among Asians and mixed populations. In a subgroup analysis by cancer type, PPARG c.1347C>T polymorphism was associated with risk of esophageal cancer and glioblastoma. In addition, in a subgroup analysis by origin of cancer cell, evidence of significant association between PPARG c.1347C>T polymorphism and cancer risk was also found among epithelial tumor. In conclusion, the findings indicate PPARG c.1347C>T polymorphism may increase the susceptibility of cancer.


INTRODUCTION
It is reported that about 14.1 million cancer patients and 8.2 million cancer-related deaths have occurred in 2012 worldwide [1]. In developing countries, the survival of cancer is poorer compared with the developed countries. The possible reason of this phenomenon is most likely due to limited access and lack of standard treatment. Cancer burden could be decreased through the application of tobacco control, healthier dietary intake, vaccine injection, early detection and treatment, and so on [2]. It is thought that cancer results from the interaction of individual's genetic components with environmental factors [3].

Meta-Analysis
Oncotarget 102278 www.impactjournals.com/oncotarget Peroxisome proliferator-activated receptor gamma (PPARG) involves three isoforms (e.g. PPARG1, PPARG2, and PPARG3). PPARG is an important nuclear receptor which acts as a transcriptional regulator and regulates energy metabolism [4]. In the pathological process of obesity, insulin insufficient/resistance and diabetes, PPARG may be activated, and then promotes the accumulation of fatty tissue [5]. PPARG agonists enhance insulin sensitivity [6]. PPARG may also possess anti-inflammatory roles [7,8]. Activation of PPARG could inhibit the production of many cytokines [e.g. tumor necrosis factor-alpha, interleukin (IL)-6, and IL-8] by antagonizing the activities of the signal transducer and activator of transcription, transcription factors activator protein 1, and nuclear factor-kappa-B, which inhibits the induction of inflammatory response [9]. A number of case-control studies demonstrated that obesity, insulin resistance/insufficient, metabolic syndrome and inflammation were correlative conditions in which PPARG could modify and regulate these actions, and influence the risk of cancer [10][11][12].
Recently, a number of studies focused on the association of PPARG polymorphisms with cancer risk [13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28]. PPARG NM_015869.4:c.34C>G (rs1801282 C>G) and NM_138712.3: c.1347C>T (rs3856806 C>T) polymorphisms are two common single nucleotide polymorphisms (SNPs). A meta-analysis indicated the PPARG c.34C>G polymorphism was associated with the risk of cancer in Asians [29]. However, the association of PPARG c.1347C>T polymorphism with cancer risk was not found. Several meta-analyses did not identify the association between this SNP and cancer risk [30,31]. Although more and more case-control studies focused on the relationship of the PPARG c.1347C>T polymorphism with cancer susceptibility, the obtained findings remained conflicting. In addition, the association between this polymorphism and lung cancer was not studied in Asians. Therefore, in this study, we designed a case-control study and assessed the relationship between PPARG c.1347C>T polymorphism and risk of nonsmall cell lung cancer (NSCLC) in Eastern Chinese Han population. Meta-analysis is a useful method of promoting the effective sample size by pooling of individual data from the enrolled studies, thus strengthening the power of the study for the assessment of genetic effects [32]. To address the association between PPARG c.1347C>T polymorphism and cancer risk more precisely, we carried out a comprehensive meta-analysis.

Association of PPARG c.1347C>T polymorphism with NSCLC
The risk factors, anthropometric data as well as demographics are listed in Table 1. Body mass index (BMI) of controls was significantly higher than it in NSCLC group (P < 0.001). This study was well-matched by age and gender. The SNP information of PPARG c.1347C>T is shown in Table 2. The genotyping success rate was 99.94% in 1,551 samples. Table 2
In a subgroup analysis by the ethnicity, evidence of significant association between PPARG c.1347C>T polymorphism and increased risk of cancer were also found among Asians, and mixed populations, but not Caucasians (Table 6). In a subgroup analysis by cancer type, c.1347C>T polymorphism was associated with the risk of esophageal cancer, and glioblastoma, but not biliary tract, breast, colorectal, melanoma, ovarian and other cancers (Table 6). In addition, in a subgroup analysis by the origin of cancer cell, evidence of significant association between PPARG c.1347C>T polymorphism and an increased risk of cancer were also found among epithelial tumor ( Table 6).
The quality score of the enrolled studies was determined by using Newcastle-Ottawa Quality    Assessment Scale [43].The results indicated that nineteen were high-quality and twelve were low-quality (Table 7). When we excluded the low-quality studies, the results were not substantially altered suggesting the reliability of our findings (   Figure 3). Using the oneway method (excluding an individual study in turn), sensitivity analysis was carried out to determine stability of our findings (Figure 4). The results indicated that our findings were stable and reliable.
Significant heterogeneities were found in this metaanalysis. Since the origin of cancer cell, quality score, ethnicity and cancer type could affect the results of meta-analysis, we carried out subgroup analyses by these factors and the findings were presented in Table 6. The results indicated that melanoma, non-epithelial tumor, Caucasians and quality score < 7.0 subgroups may contribute to the major heterogeneity. As shown in Table 6, significant heterogeneity was found in allele comparison, thus meta-regression was also performed to explore the source of heterogeneity.

DISCUSSION
The etiology of cancer was very complex. It is thought that many environmental and genetic factors may play important roles in the development of cancer. Multiple lines of evidence indicate a vital role for genetics in determining risk for cancer. PPARG is a member of the peroxisome proliferator-activated receptors (PPARs). PPARs interact with retinoid X receptors and then regulate the transcription process of many genes. PPARG has been implicated in the development of various diseases involving obesity, diabetes, inflammation, atherosclerosis and cancer [44][45][46][47]. PPARG is expressed in various cancer cells. There are accumulating evidences that obesity/ overweight, type 2 diabetes, inflammation, and malignancy are etiologically related [48,49]. Being at the crossroads of multiple diseases, PPARG may be a key component for understanding the pathophysiology of cancer. In this study, we explored the relationship of PPARG c.1347C>T polymorphism with NSCLC risk. Then, we conducted a comprehensive meta-analysis to further understand the potential role of this SNP for the susceptibility to overall cancer. In the case-control study, we found an association between PPARG c.1347C>T polymorphism and a tendency to increased risk of NSCLC. Along with a metaanalysis, we found that PPARG c.1347C>T polymorphism was associated with the increased risk of overall cancer. To the best of our knowledge, this study is the first casecontrol study focusing on the association between PPARG c.1347C>T polymorphism and NSCLC risk in Asians. And we first confirmed the relationship between this SNP and overall cancer risk.
With the increasing studies on genetic association, it is necessary to analyze the available data to obtain robust, replicable results. Considering the fact that a common SNP may make a small-to-moderate contribution to the risk of  [21,38,40]; however, others observed null association. Recently, a meta-analysis reported that this polymorphism was not associated with cancer risk [31]; however, this pooled-analysis only included four case-control studies. In this updated meta-analysis, overall findings among 21,404 subjects, evidence of significant association between this polymorphism and cancer risk were found, even in Asians, mixed populations, esophageal cancer, glioblastoma and epithelial tumor subgroups. In PPARG exon 6, a C to T substitution is a synonymous polymorphism which encodes histidine either with PPARG c.1347 C or T allele. The findings of previous epidemiological studies showed a relationship of this polymorphism with metabolic diseases such as type 2 diabetes and atherosclerosis [50][51][52][53]. It is proposed that the C to T substitution may modulate the expression of PPARG by altering mRNA processing or translation. A tendency of increased risk was observed for PPARG c.1347C>T polymorphism with NSCLC risk, and an increased risk was also found in the subsequent meta-analysis. These consistent findings demonstrated that PPARG c.1347C>TC>T polymorphism might influence the development of cancer. In the future, further evaluations with detailed environmental factors are warranted to confirm these results.
Additionally, some potential limitations should be further addressed when interpreting our findings. First, the design of our case-control study was hospital-based, and the selecting bias might have occurred. Second, in this meta-analysis, the included studies based on the published studies, unpublished articles might fail to be retrieved. Third, since the significant heterogeneities were found
Finally, we only focused on c.1347C>T polymorphism in PPARG gene, and did not consider other susceptibility genes or polymorphisms.
In conclusion, this case-control study in Eastern Chinese Han populations, along with a comprehensive meta-analysis, identify the association of PPARG c.1347C>T polymorphism with an increased risk of cancer, even in Asians, esophageal cancer, glioblastoma and epithelial tumor subgroups. Nevertheless, for some practical reasons, we hope that more case-control studies with the detailed environmental data to further explore the molecular mechanism of PPARG c.1347C>T polymorphism with development of cancer.

Subjects
Genotyping analyses were carried out on genomic DNA of 521 NSCLC patients and 1,030 unrelated controls. All participants were come from Eastern Chinese Han population. The major included criterion of NSCLC patients were: (A) living in Eastern China area; (B) NSCLC was confirmed by pathological examination; (C) without autoimmune disease. The NSCLC patients comprised unrelated subjects who had been treated in Affiliated People's Hospital of Jiangsu University and Fujian Medical University Union Hospital. The blood samples were collected from January 2014 to December 2016. Index cases were first diagnosed with NSCLC. All patients gave a written informed consent.
The controls included healthy blood donors collected in the same hospitals, having the same ethnic background and similar lifestyle as the NSCLC patients. The controls were biologically unrelated to the NSCLC cases and were cancer-free. The age distribution of NSCLC cases and non-cancer controls was nearly identical (controls: 60.34 ± 9.11 years; cases: 59.76 ± 10.71years; P = 0.268). The sex distribution of NSCLC cases and controls was well-matched (P = 0.453). According to the guidelines of Chinese blood donation, each participant was examined by a questionnaire and wrote his/her informed consent. The controls were randomly collected during the years 2014-2016. The study was approved by the Ethics Committee of Jiangsu University (Zhenjiang, China) and Fujian Medical University (Fuzhou, China). Oncotarget 102287 www.impactjournals.com/oncotarget DNA extraction and genotyping EDTA anticoagulant vacutainer tube was used to collect blood sample. We used DNA Kit (Promega, Madison, USA) to extract the genomic DNA from the whole blood.

Statistical analysis
The continuous variables (e.g. age, and BMI) are presented as the mean ± SD. We used Student's t-test to examine the difference of continuous variables between NSCLC patients and non-cancer controls. In addition, we used χ 2 test to determine the difference of categorical variables (e.g. genotypes, smoking status, alcohol consumption, sex, age and BMI). HWE test in controls was undertaken using an internet-based χ 2 goodness-of-fit test (http://ihg.gsf.de/cgi-bin/hw/hwa1.pl). Genotype-specific ORs with their corresponding 95%CIs and P-values were calculated by SAS 9.4 software for windows (SAS Institute, Cary, NC). P-values were presented using twosided χ 2 -test.

Meta-analysis
To further determine the relationship between PPARG c.1347C>T variants and cancer susceptibility, we carried out a meta-analysis. All studies focusing on the association between this polymorphism and cancer risk were collected by searching of PubMed and Embase databases (the last search update on June 12, 2017). The search was performed with the terms of (Peroxisome proliferator activated receptor gamma or PPARG) and (NP_005028.4: p.His449His or His449His or H449H or C161T or C1431T or rs3856806 or c.1347C>T) and (polymorphism or variant) and (cancer or carcinoma). Additional studies were also supplemented by a hand search of the corresponding references in retrieved articles. In this study, the language of publication was restricted to English. In our analysis, eligible studies had to meet the inclusion criteria: (1) focusing on the association between PPARG c.1347C>T polymorphism and cancer risk; (2) designed as a case-control or cohort study; (3) data could be extracted from the publications (genotypes of cases and controls); (4) published in English language; (5) genotype distribution was consistent with HWE in controls. Two authors (H. Ding and H. Qiu) extracted the detailed information from the eligible publications independently. When they met the disagreement, the third reviewer (Y. Chen) was invited to discuss every item. Finally, a consensus was reached. The following characteristics were selected and collected: the first author, year, country, ethnicity, genotyping method, cancer type, sample size, the origin of cancer cell and genotype frequencies.
For each included study, we analyzed HWE in controls using goodness-of-fit test mentioned above and P < 0.05 was defined as violation of HWE. Crude ORs with their 95% CIs were used to examine the strength of relationship between PPARG c.1347C>T polymorphism and cancer susceptibility. The pooled ORs for this polymorphism were performed under four genetic models (e.g. TT+CT vs. CC, TT vs. CC+CT, TT vs. CC and T vs. C). Stratified analyses were extensively performed with respect to origin of cancer cell, ethnicity, cancer type and quality scores. The heterogeneity across the eligible studies was tested by using a χ 2 -based Q-test and I 2 test [54]. The pooled OR was calculated by a randomeffects model (the Der-Simonian and Laird method) if I 2 > 50% or P < 0.1, which indicated that heterogeneity was significant [55,56]. Otherwise, the pooled OR was assessed by a fixed-effects model (the Mantel-Haenszel method) [57]. Removing each study in turn, sensitivity analysis was carried out by one-way method to determine the stability of the results. Additionally, Begg's test and Egger's linear regression test were conducted to assess the potential publication bias [58] and P < 0.1 was regarded as a bias. Meta-regression was conducted to analyze the source of heterogeneity [59]. In the present meta-analysis, all statistical analyses were performed by using the STATA 12.0 software for windows (Stata Corporation, College Station, Texas). A P value (two-sided) less than 0.05 were considered significant. Newcastle-Ottawa Quality Assessment Scale was harnessed to determine the quality score of the enrolled studies. If scores ≥ 7 stars, the study was defined as high-quality [43,60].