Association of four GSTs gene polymorphisms with Parkinson disease : A meta-analysis

Parkinson disease (PD) is a neurological disorder with huge destruction to human body, which affects approximately 2% of the population aged 65 years or older. As antioxidants in the stress defence systems, glutathione S-transferases (GSTs) are dimeric cytosolic enzymes with an important role in the pathogenesis of PD. The aim of this study was to evaluate the association between the polymorphisms of GST genes and PD. Meta-analyses were conducted from 17 studies (38 stages) among 3419 cases and 5686 controls between four polymorphisms (GSTT1 deletion polymorphism; GSTM1 deletion polymorphism; GSTP1104: rs1695; GSTP1-114: rs1799811) and PD. There is no significant association between the four GST gene variants and PD. A further subgroup study by ethnicity observed a risky role of GSTM1 deletion polymorphism with PD in Europeans (p = 0.013, OR = 1.126, 95% CI = 1.025 1.236), and a protective role of GSTM1 deletion polymorphism with PD in Latin Americans (p = 0.032, OR = 0.750, 95% CI = 0.577 0.975). Our meta-analysis suggested that GSTM1 deletion polymorphism increased the risk of PD in Europeans, but reduced the risk of PD in Latin Americans. Future large-scale studies might be needed to confirm the ethnic difference of GSTM1 deletion polymorphism, and to check whether there was significant association of PD for other GST genetic polymorphisms.

PD is a complex disease that may be contributed by environmental and genetic factors.Environmental hypothesis was dominant for much of the 20th century [2].PD was shown to be associated with environmental factors such as encephalitis [12], oxidative stress [13], smoking and coffee [14], and environmental toxins [15].Although recent genetic studies have discovered a handful of genetic markers of PD [16][17][18], the aetiology of PD remains unknown.
Previous case-control studies showed inconsistent results between GST genes and PD (Table 1).Discrepancy among previous studies might be due to different ethnic background, or inefficient sample size [41], or the uncorrected physiological status among different studies [42].Meta-analysis is a wildly used method to augment statistical power and to draw a more convincing conclusion through the pooling of data from individual association study [43].The goal of our meta-analyses was to find out the causes of the inconsistent results among various case-control association studies, and to evaluate the contribution of GST variants to PD.

Data Collection
We performed a systematic literature searching in online databases (PubMed Wan Fang, Wei Pu and CNKI) without time and language restriction, using the following keywords "Parkinson GSTT1 association or Parkinson GSTT1 polymorphism", "Parkinson GSTM1 association or Parkinson GSTM1 polymorphism", "Parkinson GSTP1 association or Parkinson GSTP1 polymorphism" to identify available articles.The inclusion criteria of the literatures for the meta-analyses comprise the following items: (1) It was an original case-control study with an assessment of the association between GSTT1, GSTM1, GSTP1 and PD in humans; (2) It contains sufficient information to infer the odd ratios (ORs) and 95% confi-dential intervals (95% CIs); (3) Genotype distribution of each polymorphism in controls met Hardy-Weinberg equilibrium (HWE).All of the case-control studies about GSTs and PD were fully considered and carefully selected in August 2013.The following information was extracted or calculated from each study: Genetic locus, first author's name, year of publication, country, ethnicity, numbers of cases and controls, control source, HWE for controls, if the study had significant association with PD and power analysis for each study involved in the meta-analyses.

Statistical Analysis
Arlequin program was used to test HWE [44].Statistical heterogeneity across studies included in the meta-analysis was assessed by Cochran's Q statistic and I² test [45] to decide the type of analysis.For the studies with minimal to moderate heterogeneity (I² < 50%), the fixedeffect model would be used for the meta-analysis.For the studies with significant heterogeneity (I² > = 50%), the random-effect model would be used.Funnel plots were also drawn to observe the potential publication bias.Statistical analysis of meta-analysis was carried out in Review Manager 5 and STATA statistical software [46,47].Power and Sample Size Calculation program was applied to calculate the power of each study [48].
We observed no significant association between the four GST genes variants and PD (Table 2, Figure 2).A further subgroup study by ethnicity observed a risky role of GSTM1 deletion polymorphism with PD in Europeans (p = 0.013, OR = 1.126, 95% CI = 1.025 -1.236, Figure 2 and Table 2), and a protective role of GSTM1 deletion polymorphism with PD in Latin Americans (p = 0.032, OR = 0.750, 95% CI = 0.577 -0.975, Figure 2 and Table 2).Statistically significant heterogeneity was found for the meta-analysis of GSTP1-104 (I 2 = 79%, Table 2).No publication bias was observed for the meta-analyses of the four GST variants (Figure 3).

DISCUSSION
Our meta-analyses showed that GSTM1 deletion polymorphism was a risky and protective factor in PD for Europeans and Latin Americans, respectively (Table 2 and Figure 2).GSTM1 encodes a key GST that plays an important role against oxidative stress [49], and thus defenses PD.GSTM1 deletion polymorphism showed a wide association with diseases such as coronary heart disease [50], thyroid carcinoma [51], myelodysplastic syndrome [52], oral cancer [53] and PD [22,32,34].Homozygous deletion genotypes of GSTM1 and GSTT1 were relatively common in Caucasian populations [54], and disease prevalence by WHO showed that Europe was the most severe region of PD in the world.We performed a further subgroup study by ethnicity that showed a significant association between GSTM1 deletion polymorphism and PD in Europeans (p = 0.013, OR = 1.126, 95% CI = 1.025 -1.236, I² = 0%, Table 2).On the contrary, there was a significant protective role of GSTM1 deletion polymorphism for PD in Latin American population (p = 0.032, OR = 0.750, 95% CI = 0.577 -0.975,Table 2).Since the latter result was produced by only one study [34] with relative low power (power = 0.458), future study was needed to confirm this ethnic difference of GSTM1 deletion polymorphism in the risk of PD.
Although we didn't observe a significant association between GSTP1-104 and PD in the combined samples, subgroup meta-analyses by ethnicity indicated a trend of significance in Europeans (p = 0.104, Table 2) and Asians (p = 0.096, Table 2).GSTP1-104-A allele frequency showed a significant ethnic difference between Asians (Hapmap-JPT: 0.907) and Europeans (Hapmap-CEU: 0.593, Fst = 0.16).As is shown in Table 2, the power is 0.476 in Europeans and 0.223 in Asians, indicating that larger sample size is needed to validate the association of GSTP1-104 and PD in the future.
To be noted, a former meta-analysis with 4 studies in GSTT1 and GSTM1, and 1 study in GSTP1 in 2000 [55] showed a significant association between GSTT1 and PD, and no significant results were found in GSTM1 and GSTP1.Compared with the former meta-analysis, our study included 8, 12 and 8 more studies in GSTT1, GSTM1 and GSTP1, respectively.We also performed an ethnic subgroup study that previous study did not perform.We performed HWE test for the controls in GSTP1 and excluded 3 case-control studies that failed to meet HWE in the control samples [26,29,31].With an enhanced power, more comprehensive analysis and stricter selection criteria, our meta-analyses produced a more reliable conclusion than the previous meta-analysis studies.
Several limitations in our study needed to be carefully considered.Firstly, the number of associations in non-Caucasian populations was limited.This suggested that   non-significant results in Asians needed to be taken with caution.Future studies with larger samples size are required to establish the association of GST gene polymorphisms with PD.Secondly, PD is a complex disease that different physiological status of PD may exist in the cases.All the existing case-control studies didn't perform a stratified analysis by the PD stage.This may partially explain the discrepancies in the current case-control studies.Thirdly, genetic heterogeneity may exist for the polymorphisms of GST genes.Our meta-analyses only focused on four polymorphisms that might not fully represent the overall contribution of GST gene polymorphisms.Other GST gene polymorphisms need to be analyzed for their contribution to PD in the future.Fourthly, all the p values are not corrected by the number of tests.The positive findings in the present study need to be validated in the future study in order to prevent the false positives caused by multiple testing in our analyses.
In conclusion, our meta-analysis observed a significant association of GSTM1 deletion polymorphism with PD in Europeans.Further large-scale studies are required to evaluate the polymorphisms of GSTs that might contribute to the risk of PD.

Figure 1 .
Figure 1.Flowchart of selection process in the meta-analyses.

Figure 2 .
Figure 2. Forest plot of GSTM1 deletion polymorphism with PD.

Figure 3 .
Figure 3. Funnel plots of four SNPs with PD.A) Funnel plot of GSTT1 deletion polymorphism with PD; B) Funnel plot of GSTM1 deletion polymorphism with PD; C) Funnel plot of GSTP1-104 polymorphism with PD; D) Funnel plot of GSTM1-114 polymorphism with PD.

Table 1 .
Characteristics of the case-control studies in the current meta-analyses.
* a, Queensland, Australia, America; Mix * b, Italy, Malta, Romania, Scotland, Sweden; GHC, Group Health Cooperative; HWE, Hardy-Weinberg equilibrium; L.A, Latin Americans; NA, Not applicable; Result * , The association between GST genes and PD; NS, No significant; S, Significant.0.05).We were unable to test HWE for GSTT1 and GSTM1 deletions because there was no genotype information in the original reports.At last, we included 17 studies (38 stages) among 3419 cases and 5686 controls in the current meta-analyses of four GST gene variants (Table
* : Amount of stages; NA, Not applicable.