Pro Variant of TP 53 Arg 72 Pro Contributes to Gastric Cancer Risk in Asians : Evidence from a Meta-analysis

Gastric cancer (GC) was the sixth most common cancer worldwide (989,600 new cancer cases) and the second most frequent cause of cancer death worldwide (738,000 cancer deaths) in 2008 (Jemal et al., 2011). Over 70% of new cases and deaths occur in developing countries, and the highest incidence rate is in Eastern Asia (Jemal et al., 2011). Despite of the emergence of new cytotoxic drugs and targeted biologic agents, GC remains one of the most clinically challenging cancers among all gastrointestinal malignancies and need multidisciplinary management (Hartgrink et al., 2009; Jiang and Ajani, 2010). Thus, GC still is a serious fatal disease worldwide and has caused serious damage to human health (Hartgrink et al., 2009). As a complex and multi-factorial process, the gastric carcinogenesis is still not fully understood. Epidemiological studies have revealed that Helicobacter pylori, smoking, diets and environmental risk factors play important roles in the development of GC (Fukase et al., 2008; Hartgrink et al., 2009; Wroblewski et al., 2010). However, only a small proportion of individuals exposed to the known risk factors develop GC, while many cases develop GC among individuals without those risk factors, which suggest genetic factors also play an important role in GC etiology (Resende et al., 2010). The tumor suppressor p53 is a key player in stress responses and preserves genomic stability by responding to various insults including DNA damage, metabolic stress


Introduction
Gastric cancer (GC) was the sixth most common cancer worldwide (989,600 new cancer cases) and the second most frequent cause of cancer death worldwide (738,000 cancer deaths) in 2008 (Jemal et al., 2011).Over 70% of new cases and deaths occur in developing countries, and the highest incidence rate is in Eastern Asia (Jemal et al., 2011).
Despite of the emergence of new cytotoxic drugs and targeted biologic agents, GC remains one of the most clinically challenging cancers among all gastrointestinal malignancies and need multidisciplinary management (Hartgrink et al., 2009;Jiang and Ajani, 2010).Thus, GC still is a serious fatal disease worldwide and has caused serious damage to human health (Hartgrink et al., 2009).As a complex and multi-factorial process, the gastric carcinogenesis is still not fully understood.Epidemiological studies have revealed that Helicobacter pylori, smoking, diets and environmental risk factors play important roles in the development of GC (Fukase et al., 2008;Hartgrink et al., 2009;Wroblewski et al., 2010).However, only a small proportion of individuals exposed to the known risk factors develop GC, while many cases develop GC among individuals without those risk factors,

RESEARCH COMMUNICATION
Pro Variant of TP53 Arg72Pro Contributes to Gastric Cancer Risk in Asians: Evidence from a Meta-analysis Xiu-Li Su, Jian-Jun Jin* and oncogene activation (Vogelstein and Kinzler, 2004;Vousden and Lane, 2007).P53 can also interact with other cellular proteins, and these molecular interactions might contribute to the inhibitory role of p53 in carcinogenesis, while malfunction of the p53 pathway is an almost universal hallmark of human tumors (Chipuk et al., 2005;Riley et al., 2008).TP53 gene is a tumor suppressor gene encoding p53 and frequently mutates in various cancers, and those mutations can cause dramatic defects in p53 function and are often hallmarks of most human cancers (Chipuk et al., 2005;Riley et al., 2008).A polymorphism at codon 72 of the TP53 gene (Arg72Pro; rs1042522) has been most intensively investigated and reported to affect the functions of p53 network which is central to the development of cancer.TP53 Arg72Pro is a G-to-C polymorphism at the second position of codon 72 in exon 4 resulting in amino acid substitution from Arg to Pro.The current view is that the p53-Arg72 protein is more effective in inducing apoptosis and protecting cells from cancerization than the p53-Pro72 protein (Dumont et al., 2003;Whibley et al., 2009).A study published in 2004 showed that individuals with Pro/Pro genotype had a 2.98-fold higher risk of developing diffuse-type GC compared to individuals with Arg/Arg genotype (p=0.038)(Hiyama et al., 2002).Subsequently, nearly 20 studies have been published on this controversial issue, but it remains unclear whether there is an association between TP53 Arg72Pro polymorphism and GC risk (Hamajima et al., 2002;Hiyama et al., 2002;Gao et al., 2009).There were obvious inconclusive results among those studies, which were probably due to limited sample sizes, ethnic difference or publication bias.Meta-analysis is a statistical procedure for combining results from published studies to acquire a precise estimation of the major effect (Stroup et al., 2000).Two meta-analyses were published to assess the association between TP53 Arg72Pro and GC risk, but there was obvious between-study heterogeneity (Zhou et al., 2007;Gao et al., 2009).Besides, several large scale casecontrol studies have been published to further assess the association between TP53 Arg72Pro and GC risk in Asian population, which is the highest incidence rate of GC, but the association above in Eastern Asia is still uncertain (Kim et al., 2010;Shirai et al., 2010;Song et al., 2011).Thus, to assess the evidence regarding the association TP53 Arg72Pro and GC risk and to investigate potential sources of heterogeneity, we conducted a comprehensive meta-analysis of epidemiological studies investigating this association.

Search strategy and selection criteria
We searched PubMed, Embase and CBM database using the following search strategy: ('gastric cancer' or 'stomach cancer' or 'gastric carcinoma' or 'stomach carcinoma') and ('p53' or 'TP53' or 'Arg72Pro' or 'codon 72') and ('polymorphism' or 'polymorphisms' or 'mutation') for papers published from January 1980 to July 2011.Additionally, we searched the internet for unpublished data.We also reviewed reference lists in those published articles.There was no language limitation.The inclusion criteria were: (1) case-control studies which evaluated the association between TP53 Arg72Pro polymorphism and GC risk; (2) used an unrelated case-control design; (3) had available genotype frequency for estimating an odds ratio (OR) with 95% confidence interval (CI); (4) genotype distribution of the control population was in Hardy-Weinberg equilibrium (HWE).Overlapping study or studies comparing different laboratory methods were all excluded.

Data extraction and study design
The following information was extracted from included studies: first author, year of publication, number of cases and controls, characteristics of cases and controls, genotypes frequency and allele frequency of cases and controls.Difference was settled by reaching an agreement between all investigators.To test the population stratification in the controls, a chi-square test using a web-based program (http://ihg2.helmholtz-muenchen. de/cgi-bin/hw/hwa1.pl)was applied to determine if TP53 Arg72Pro genotype distribution in the control population reported conformed to HWE (P<0.05 was considered significant).Subgroup analyses were mainly done by study size and published language.Study size was divided into those with at least 300 GC cases and those with fewer than 300 GC cases.

Statistical analysis
The strength of the association between TP53 Arg72Pro polymorphism and GC risk was measured by odds ratio (OR) corresponding to 95% confidence interval (CI).We assessed the association between Pro variant of Arg72Pro polymorphism and GC risk on allele genetic comparison model (Pro vs. Arg), the homozygote comparison model (Pro/Pro vs. Arg/Arg), the recessive genetic comparison model (Pro/Pro vs. Arg/Pro+Arg/Arg) and the dominant genetic comparison model (Pro/Pro + Arg/Pro vs. Arg/ Arg).
In our study, two models of meta-analysis for dichotomous outcomes were conducted: the randomeffects model and the fixed-effects model.The randomeffects model was conducted using the DerSimonian and Laird's method, which assumed that studies were taken from populations with varying effect sizes, calculating the study weights both from in-study and betweenstudy variances, considering the extent of variation, or heterogeneity (DerSimonian and Laird, 1986).The fixedeffects model was conducted using the Mantel-Haenszel's method, which assumed that studies were sampled from populations with the same effect size, making an adjustment to the study weights according to the in-study variance (Mantel and Haenszel, 1959).To assess the between-study heterogeneity more precisely, both the chi-square based Q statistic test (Cochran's Q statistic) to test for heterogeneity and the I2 statistic to quantify the proportion of the total variation due to heterogeneity were calculated (Cochran, 1954;Higgins and Thompson, 2002;Higgins et al., 2003).Heterogeneity was considered significant for P Cochran's Q statistic < 0.05, and the randomeffects model was used to pool the results; on the contrary, the fixed-effects model was used to pool the results when P value of Cochran's Q statistic was more than 0.05.The significance of the pooled OR was determined by the Z test and a p value of <0.05 was considered significant.Besides, Galbraith plot was also used to spot the outliers which were the sources of heterogeneity (Galbraith, 1988).To validate the credibility of outcomes in this meta-analysis, sensitivity analysis was performed by sequential omission of individual studies.Publication bias was investigated by funnel plot, and an asymmetric plot suggested possible publication bias in this meta-analysis.Publication bias was investigated by funnel plot, in which the standard error of logor of each study was plotted against its logor (Stuck et al., 1998).An asymmetric plot suggested possible publication bias.Furthermore, Funnel-plot asymmetry was assessed by the method of Egger's linear regression test (Egger et al., 1997).
Statistical analysis was performed with the software programs RevMan (version 5.1) and STATA (version 12).All the P values were two-sided.

Meta-analysis results
Table 2 listed the main results of this meta-analysis.When all 17 studies were pooled into meta-analysis, significant heterogeneity existed in three genetic comparison models, including Pro vs. Arg, the recessive genetic comparison model, and the homozygote comparison model; thus, the random-effects model was used to pool the results in these three comparison models except the dominant genetic comparison model (P H =0.11>0.05,I 2 =31%<50%).The combined results based on all studies showed the variant genotypes of TP53 Arg72Pro were associated with an elevated GC risk in three genetic comparison models (OR Pro vs. Arg =1.13, 95%CI 1.03-1.25,P OR =0.01; OR Homozygote comparison model =1.33, 95%CI 1.07-1.64,P OR =0.009) OR Dominant genetic model =1.13, 95%CI 1.05-1.22,P OR =0.002) (Figure 2).Subgroup analyses by sample also suggested the variant genotypes of TP53 Arg72Pro were associated with an elevated GC risk in all four genetic comparison models.
The between-study heterogeneity was obvious in most comparisons before adjustment for heterogeneity.Galbraith plots spotted Cao YY's study and Li XZ's study as the outliers and the major source of the heterogeneity (Figure 3) (Li et al., 2004, Cao et al., 2007).Interestingly,   the heterogeneity remarkably decreased in those three genetic comparison models after excluding Cao YY's study and Li XZ's study (Liet al., 2004, Cao et al., 2007), which further indicated that Cao YY's study and Li XZ's study were the major source of the heterogeneity (Table 2).Besides, a more obvious association was observed after the heterogeneity was decreased by excluding Cao YY's study and Li XZ's study in those three genetic comparison models above (Table 2).Sensitivity analysis was performed by sequential omission of individual studies, and the significance of all pooled ORs was not influenced excessively by omitting any single study (data were not shown).Besides, sensitivity analysis was also performed by adding Zhou Y's study in which the genotype distribution of the control population was significantly deviated from HWE (Zhou et al., 2010), and the significance of all pooled ORs was also not influenced.

Publication bias
Funnel plot and Egger's test were performed to access the publication bias of this meta-analysis.The shape of the funnel plots for most genetic contrast models seemed symmetrical, and all the P values of Egger's tests were more than 0.05, providing statistical evidence of funnel plot symmetry.The results above suggested that publication bias was not evident in our meta-analyses.As shown in Figure 4, in the meta-analysis investigating the association between TP53 Arg72Pro and GC risk, the funnel plot's shape for the allele genetic comparison model (Pro vs. Arg) seemed symmetrical, suggesting no presence of publication bias; the P value of the Egger's test for the allele genetic comparison model was 0.236, providing statistical evidence for funnel plot's symmetry.

Discussion
The TP53 gene is one of the most widely investigated genes because of its role as a tumor suppressor gene, which plays a key role in the development and progression of cancers.TP53 gene is frequently mutated in various cancers, and it has been suggested that genetic polymorphisms in TP53 gene could affect the functions of p53 and TP53 gene variants have drawn increasing attention in the etiology of several cancers.Although there were many genetic association studies investigating the effect of TP53 genetic variants on GC cancer risk, the impact of TP53 Arg72Pro polymorphism on GC risk remained to be elucidated because the published data are controversial.Two meta-analyses were published to assess the association between TP53 Arg72Pro and GC risk, but there was obvious between-study heterogeneity.Besides, several large scale case-control studies have been published to further assess the association between TP53 Arg72Pro and GC risk in Asian population, which is the highest incidence rate of GC, but the association above in Eastern Asia is still uncertain (Zhou et al., 2007;Gao et al., 2009).Thus, to assess the evidence regarding the association TP53 Arg72Pro and GC risk and to investigate potential sources of heterogeneity, we conducted a comprehensive meta-analysis of epidemiological studies investigating this association.We performed a pooled analysis of all eligible studies.The present meta-analysis included 17 case-control studies including a total of 5,990 GC cases and 6,812 controls, providing the most comprehensive assessment of the association between TP53 Arg72Pro polymorphism and GC risk up to now.The outcomes of this meta-analysis indicate that variant genotypes of TP53 Arg72Pro were associated with an elevated GC risk in three genetic comparison models (OR Pro vs. Arg =1.13, 95%CI 1.03-1.25,P OR =0.01; OR Homozygote comparison model =1.33, 95%CI 1.07-1.64,P OR =0.009; OR Dominant genetic model =1.13, 95%CI 1.05-1.22,P OR =0.002).Besides, a more obvious association was observed after the heterogeneity was decreased (All P values were less than 0.001).This association was further identified by both subgroup analyses and sensitivity analysis.Subgroup analyses were performed based on smoking status, study size, published language and study type, and the results of subgroup analyses further validated this statistically significant association (Table 2).
Heterogeneity is a potential problem when interpreting the results of all meta-analyses, and finding of the source of heterogeneity is one of the most important goals of meta-analysis (Ioannidis et al., 2007).The present metaanalysis showed that significant heterogeneity existed in most analyses before we excluded Cao YY's study and Li XZ's study (Li et al., 2004, Cao et al., 2007), and Galbraith plot also spotted Cao YY's study and Li XZ's study (Li et al., 2004, Cao et al., 2007)as the outliers and further identified it as the major source of the heterogeneity (Figure 4).Therefore, Cao YY's study and Li XZ's study (Li et al., 2004, Cao et al., 2007) were the major source of the heterogeneity in this meta-analysis and might have some differences from other studies in the selection of controls, prevalence of lifestyle factors, or other aspects.Besides, heterogeneity between studies in the pooled analysis of total studies was removed after we excluded Cao YY's study and Li XZ's study (Table 2).
Somatic mutation of TP53 that results in the absence or dysfunction of p53 is one of the most common mechanisms by which the p53 pathway is damaged during carcinogenesis.The two alleles of TP53 Arg72Pro differ not only biologically but also functionally in their ability to bind components of the transcriptional machinery, to activate transcription, to induce apoptosis and to repress the transformation of primary cells.The Arg allele induces apoptosis more effectively than the Pro variant, suggesting that the TP53 Pro variant might be a weaker tumor suppressor than its Arg counterpart.Thus, there is obvious biological evidence for the different effects on cancer development between the two different variants and the TP53 Pro variant might be a weaker tumor suppressor compared with Arg allele.In 2008, Ioannidis JP et al suggested an interim guideline to develop guidance criteria for assessing cumulative epidemiologic evidence in genetic associations, such as the amount of biological evidence, epidemiological credibility and clinical publichealth impact (Ioannidis et al., 2008).As is argued above, there is obvious biological evidence that the TP53 Pro variant might be a weaker tumor suppressor compared with Arg allele.In addition, our pooled analysis adds strong epidemiological evidence for the association between TP53 Arg72Pro polymorphism and GC risk in Asians.Finally, there are studies suggesting p53 codon 72 Pro/Pro is associated with poor relapse-free survival and overall survival, and there is also convincing evidence of clinical relevance between TP53 Arg72Pro polymorphism and GC (Huang et al., 2009;Kim et al., 2009).Thus, biological evidence, epidemiological evidence, and clinical evidence all confirm the association between Arg72Pro polymorphism and GC risk.
However, some possible limitations in our metaanalysis should be acknowledged.Firstly, the eligibility criteria for inclusion of controls were different from each other.The controls in some studies were selected from non-cancer patients who underwent gastroscopy, while the controls in other several studies were just selected from asymptomatic individuals (Table 1).Additionally, misclassification bias was possible.For example, most studies could not exclude latent cancer cases in the controls (Table 1).Finally, gene-gene and gene-environmental interactions were not fully addressed in this meta-analysis for the lack of sufficient data.As we know, aside from genetic factor, smoking is a major risk factor for GC; however we didn't perform subgroup analyses in smokers or nonsmokers owing to the limited reported information on such associations in the included studies.
Despite of those limitations, this meta-analysis suggests the Pro variant of TP53 Arg72Pro contributes to gastric cancer risk in Asian population.

Figure
Figure 1.Flowchart of Selection of Studies for Inclusion in Meta-analysis

Figure 2 .
Figure 2. Forest Plot of Pooled OR with 95% CI for TP53 Arg72Pro Polymorphism with Gastric Cancer Risk (A, Pro vs. Arg, Random effects model; B, Homozygote comparison model, Random effects model; C, Recessive genetic comparison model, Random effects model; D, Dominant genetic model, Fixed effects model ) (The squares and horizontal lines corresponded to the study-specific OR and 95% CI.The area of the squares reflected the study-specific weight (inverse of the variance).The diamond represented the pooled OR and 95% CI)

Table 1 . Characteristics of Studies Included in this Meta-analysis
Risk in Asians: Evidence from a Meta-analysis males, 46 females; median age, 58.6 years) with histologically confirmed gastric cancer from hospitals of Hexi area 125 age-and gender-matched healthy controls (85 males, 40 females; median age, 59.1 years) confirmed by gastroscopy Song HR, 2011 Korea PCR-TaqMan 0.48 2213 patients (1510 males, 703 females; median age, 60.2 years) with histologically confirmed gastric cancer 1700 healthy controls (576 males, 118 females; median age, 52.2 years) recruited from the Thyroid Disease Prevalence Study of Korea

Table 2 . Summary of Odds Ratios (OR) with Confidence Interval (CI) in the Meta-analysis
Adjustment for Heterogeneity was performed by excluding Cao YY's study and Li XZ's study as the outlier and the possible major source of heterogeneity # P H , the P value of heterogeneity test; *