TP53 Arg72Pro Polymorphism and Colorectal Cancer Risk: A Systematic Review and Meta-Analysis

Colorectal cancer is the third most common type of cancer in the United States and is responsible for approximately 50,000 deaths per year (1). Family-based studies have suggested that the disease has a significant genetic component, with a large twin study conducted in Scandinavian countries suggesting that as many as 35% of colorectal cancers may be due to inherited susceptibility (2). However, the recognized Mendelian predisposition syndromes, such as hereditary nonpolyposis colorectal cancer and adenomatus polyposis coli, account for less than 5% of the overall incidence of colorectal cancer (3). Therefore, common, low-penetrance polymorphisms may confer a substantial part of the genetic risk, but given that the estimated effect of each polymorphism is expected to be small, large studies are necessary to reduce the size-related uncertainty of effects and provide robust evidence of association.

The TP53 gene, located at 17p13, is a prototypical tumor suppressor gene encoding a 53-kDa protein (p53) with important functions in cell cycle control, apoptosis, and maintenance of DNA integrity (4-6). The importance of p53 in cell cycle regulation (via gene transcription) and DNA integrity is such that it has been called the “guardian of the genome” (7). The function of p53 is to reduce the incidence of cancers by mediating apoptosis in cells that have activated oncogenic pathways. Similarly, DNA damage or genotoxic stress may cause the induction of p53, leading to growth arrest or apoptosis (8). Germ-line mutations in TP53 are known to cause a number of recognized human cancers including Li-Fraumeni syndrome (9). When TP53 itself is not genetically inactivated, other mechanisms, such as loss of heterozygosity by deletion of the 17p locus or gene methylation, may contribute to reduced p53 activity (10-15). The polymorphic nature of the TP53 gene and its central role in cell cycle regulation have highlighted it as a good potential candidate susceptibility gene for colorectal cancer. Although several TP53 polymorphisms have been investigated as risk factors for cancer, by far the most extensively investigated is a nonsynonymous polymorphism in a proline-rich domain located in exon 4, where a cytosine (C; variant allele) for guanine (G) substitution results in the substitution of proline (Pro) for arginine (Arg) at codon 72 of the p53 protein (Arg72Pro, refSNP no. rs1042522; ref. 16). Various lines of evidence indicate that these two alleles differ in their capacities to induce target gene transcription, their interaction with p73, their targeting of the proteasome, and their susceptibility to degradation by human papillomavirus E6 protein (17-20). They are also recognized as modulating apoptosis at differing rates (21).

Several epidemiologic studies have addressed the influence of this polymorphism on cancer risk for most common cancer types, including colorectal cancer; however, small sample sizes and deficiencies in study design have contributed to conflicting results (22-26). To offer a comprehensive evaluation of the potential association of this polymorphism with colorectal cancer risk, we conducted a systematic review and meta-analysis of candidate genetic association studies.

We sought to identify genetic association studies published before July 31, 2009, investigating the association between the rs1042522 polymorphism located within the TP53 gene and colorectal cancer, using computer-based searches (last search: July 31, 2009) of MEDLINE (PubMed), the Human Genome Epidemiology Network (HuGE Net) Literature Finder, and the NIH Genetic Association Database (27), using keywords related to the TP53 gene and colorectal cancer (the full search strategy is available from the authors on request). Additionally, we searched two TP53-specific databases that collect information related to TP53 polymorphisms: the IARC TP53 database (28) and the p53 website (29). We also hand-searched the reference lists for all retrieved studies and relevant review articles, as well as journals known to publish studies relevant to the topic.

Studies using an analytic design (case-control, nested case-control, or cohort) and employing validated genotyping methods to examine the frequency of rs1042522 among colorectal cancer patients and controls were eligible for inclusion. Family-based studies were not considered eligible owing to different design considerations. Studies that included patients known to have hereditary colorectal cancer syndromes, such as nonpolyposis colorectal cancer or familial adenomatous polyposis, were excluded. If studies reported on mixed populations of syndromic and sporadic colorectal cancer, we only used the genotype information for patients with sporadic disease (when available). We only considered studies published in English.

The following information was abstracted from each study: first author, journal, year of publication, study design, matching, ethnicity of participants, definition and numbers of cases and controls, DNA extraction and genotyping methods, source of genetic material for genotyping cases, frequency of genotypes, anatomic location of the tumor (colon versus rectum), and the number of cases and controls for each TP53 genotype. Data extraction was done independently by two reviewers (I.J.D. and V.V.) and discrepancies were resolved by consensus including a third reviewer (S.M.).

For our main analysis, we compared allele frequencies (the proline-encoding allele C versus the arginine-encoding allele G) between cases and controls. We also evaluated a recessive (CC versus CG+GG) and a dominant model (CC+CG versus GG) for the C allele. All associations were presented as odds ratios (OR) with their corresponding 95% confidence interval (95% CI). Between-study heterogeneity was tested using the χ²-based Q-statistic and was considered statistically significant at P < 0.1 (30). Between-study inconsistency was quantified using the I² statistic (31). A pooled OR was estimated based on the individual study ORs using random-effects (DerSimonian and Laird) models (32). Cumulative meta-analysis was carried out to evaluate the trend of the random-effects OR over time (33).

Cancer subtype (colon versus rectum), participant ethnicity (individuals of White ancestry versus East Asian), Hardy-Weinberg equilibrium (HWE) in the control group, matching of cases and controls, genotyping quality control (repeat genotyping of a random selection of samples), blinded genotyping (genotyping by individuals blinded to the case/control status of each individual versus lack of blinding or no mention of blinding), and source of DNA for cases (use of tumor tissue obtained during surgery versus blood/normal tissue) were prespecified as characteristics for assessment of heterogeneity by subgroup analysis. When a study explicitly stated that pathologic examination was used to select healthy tissue obtained by surgery, we considered the study along with studies that used blood samples, given the high sensitivity and specificity of pathologic diagnosis for discriminating healthy and cancer tissues. We also performed sensitivity analysis by excluding such studies.

The differential magnitude of effect in large versus small studies was assessed using the Harbord modification of the Egger test (34, 35). A test for interaction was used to compare the results of the first study with the pooled estimate of all subsequent studies and to compare pooled effect estimates between studies (36). The distribution of the genotypes in the control group was tested for HWE using an exact test (37). Studies with controls not in HWE were subjected to a sensitivity analysis in which the effect of excluding specific studies was examined. Analyses were performed using Stata (version 11/SE, Stata Corp.) and statistical significance was defined as a two-sided P value <0.05 for all tests except those for heterogeneity.

Our initial search identified 5,035 studies, of which 62 were considered potentially eligible for inclusion in this review and were retrieved in full text. Of those, 37 were excluded (9 did not include colorectal cancer patients, 8 did not include control groups, 8 did not assess the polymorphism of interest, 5 were not published in English, 4 included patients with hereditary colorectal cancer syndromes, 2 were preclinical studies, and 1 was an editorial) and 25 were considered eligible for the meta-analysis (references to excluded studies are available on request). Of those, one study did not provide extractable data (38) and one used an unconventional genotyping method (DNA pooling; ref. 39) and was included only in sensitivity analyses. For the main analysis, 23 studies were considered eligible, of which one provided data only for the recessive model (40); the first was published in 1991 and the last in 2009 (40-62). Detailed study characteristics are presented in Table 1 and Supplementary Table S1. In total, 23 studies investigated 6,514 colorectal cancer cases and 9,334 controls for the Arg72Pro polymorphism (mean number of cases, 283; median, 115; min, 53; max, 2,558). Sixteen studies had healthy individuals as controls, and seven studies matched cases and controls (all for age, and five of those for gender). Nine studies used some form of genotyping quality control and only three reported that genotyping was blinded to the case-control status of participants. In one study, the distribution of the genotypes in the control group was not in HWE (Fisher's exact test, P < 0.05). Five studies used tumor tissue obtained during surgery for determining the case genotype.

The overall analysis investigating the association between C allele and risk of colorectal cancer relative to the G allele (C versus G) revealed significant between-study heterogeneity (P_Q < 0.001; I² = 63%) and the random-effects OR was nonsignificant (OR, 0.99; 95% CI, 0.89–1.09; P = 0.80; Fig. 1). Moreover, the recessive model for the C allele (CC versus CG+GG) showed moderate heterogeneity (P = 0.02, I² = 44%) and no evidence of an association (OR, 1.01; 95% CI, 0.84–1.21; Supplementary Fig. S1A). Similarly, the dominant model for the C allele (CC+CG versus GG) showed significant heterogeneity (P < 0.001, I² = 65%) and the random-effects OR was nonsignificant (OR, 1.00; 95% CI, 0.88–1.15; Supplementary Fig. S1B).

In cumulative meta-analysis of allele-frequency contrasts, the pooled OR has remained centered on 1 over time, indicating that rs1042522 is an unlikely risk variant for colorectal cancer (Supplementary Fig. S2).

There was no evidence of a differential magnitude of effects in large versus small studies (Harbord/Egger test P > 0.5 for all genetic contrasts). In addition, there was no significant difference between the OR of the first study versus the pooled random-effects OR of all subsequent studies under any genetic model, and between-study heterogeneity remained significant after excluding the first study from all analyses (P_Q < 0.001 for the allele-frequency and dominant models, and P_Q = 0.02 for the recessive genetic model).

Overall, we did not find evidence of effect heterogeneity between studies that reported the use of quality control for genotyping or those where genotyping was performed blinded to the case/control status of participants, compared with those that did not. Although few studies stratified cases into colon and rectal cancer subgroups, the effect of rs1042522 was null in both. Studies using RFLP genotyping methods suggested that the Pro allele was associated with an increased risk of colorectal cancer (OR, 1.12; 95% CI, 0.96–1.30), compared with studies using alternative genotyping methods (OR, 0.91; 95% CI, 0.82–1.01). The difference was statistically significant (P_interaction = 0.03). Furthermore, the ORs from studies using tumor tissue for case genotyping showed a protective effect for the Pro allele (OR, 0.75; 95% CI, 0.60–0.94). In contrast, studies using blood/normal tissue did not detect an association (OR, 1.04; 95% CI, 0.94–1.15); this difference was statistically significant (P_interaction = 0.03). In addition, studies of East Asian populations seemed to produce more exaggerated effect sizes compared with studies in Caucasian populations (P_interaction = 0.03). Finally, inclusion of the study that used DNA pooling also did not affect the results under any genetic model (data not shown; ref. 39). Table 2 summarizes the results of subgroup analysis for the allele-frequency comparison. Supplementary Table S2 summarizes the results of subgroup analyses using different genetic models.

Colorectal cancer is estimated to have a significant heritable component, which is not completely accounted for by the high-penetrance mutations responsible for the known Mendelian colorectal cancer predisposition syndromes (3). The TP53 rs1042522 polymorphism, one of the most widely investigated polymorphisms in genetic epidemiology, has been considered a good candidate genetic risk factor for many cancers (23). This meta-analysis, based on 6,514 cases and 9,334 controls, shows that this polymorphism is an unlikely risk factor for colorectal cancer. Our results are supportive of the findings of the largest study of this polymorphism in colorectal cancer conducted to date by a collaborative effort of British investigators (61), which did not identify an association between rs1042522 and colorectal cancer. Most importantly, several potential factors that may lead to bias seem to be active in this field. For example, only three studies specifically mentioned that genotyping was blinded to case/control status, and only nine studies implemented some form of genotyping quality control.

Subgroup analysis showed that the ORs from studies of East Asian individuals were exaggerated compared with those investigating Caucasian individuals (P_interaction = 0.03). Most likely, this is a spurious finding and does not represent a true biological difference (63, 64). Furthermore, use of RFLP methods for genotyping produced significantly different results compared with other genotyping methods (P_interaction = 0.03). This discrepancy between genotyping methods highlights the need for implementing rigorous quality control procedures in future studies, but it is unclear whether the interaction of genotyping method and genetic effect is due to bias in calling uncertain results, or if use of RFLP methods is a surrogate for study quality in general (48). In addition, subgroup analysis revealed that the ORs from studies using cancer tissue for genotyping the cases were significantly (P = 0.03) different from the pooled point estimate of studies using blood/normal tissue. In general, use of tumor-derived DNA for determining the constitutional genotype is discouraged because multiple deletional somatic events that occur in tumor cells early in the carcinogenetic process may skew the overall results. Loss of heterozygosity often occurs in colorectal cancer, in many cases involving large genomic regions. There is also evidence that allelic loss at the TP53 locus is nonrandom, and that tumor cells preferentially retain the Arg allele, in different cancer types (19, 57, 65, 66). Regarding colorectal cancer in particular, studies in heterozygous individuals have shown that there is a preferential retention of the Arg allele that may cause genotype misclassification (57). This misclassification would tend to bias the results of genetic association studies toward a detrimental effect of the Arg allele (i.e., a spurious protective effect for the Pro allele). Interestingly, similar results were reached in a recent pooled analysis of individual patient data from 49 studies in cervical cancer (24). Overall, laboratory artifacts, rather than true biological effects, seem to drive the observed associations of Arg72Pro with colorectal cancer.

It should be noted that the pooled effect estimate based on studies using appropriate DNA sources is in itself not accurate enough to exclude the possibility of a small effect of the rs1042522 polymorphism on colorectal cancer risk, indicating that further research may be necessary to provide conclusive evidence for this variant. Another important limitation of the existing literature is the lack of information about potential gene-gene or gene-environment interactions. Given that the role of several environmental factors in the pathogenesis of colorectal cancer is established, further research should be performed in this direction.

In conclusion, this systematic review and meta-analysis of genetic association studies shows that TP53 Arg72Pro is unlikely to be a major risk factor for colorectal cancer. Several sources of bias, including the use of inappropriate genotyping material and the lack of quality control, need to be addressed in the design of future studies.

No potential conflicts of interest were disclosed.

Grant Support: I.J. Dahabreh was supported by a research fellowship provided by the “Maria P. Lemos” Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.