Introduction

Nasopharyngeal carcinoma (NPC) arises from the epithelial cells that cover the upper part of the throat behind the nose and near the base of the skull. The disease is treatable at an early stage but the majority of NPC patients are diagnosed at a late stage because of the exhibition of nonspecific symptoms related to other head and neck illnesses.1, 2 General symptoms of NPC include trismus, otitis media, hearing loss, nasal regurgitation, cranial nerve palsies, nasal twang, bleeding and pain.3 The World Health Organization histopathological grading system classifies NPC into three types: keratinizing squamous cell carcinoma; non-keratinizing differentiated carcinoma; and undifferentiated carcinoma.4 The American joint committee on cancer established tumor, node and metastasis classification to determine the different stages of NPC.

Epidemiological studies suggest the association of food habits (alcohol, intake of salted fish containing nitrosamine, herbal tea and herbal medicine), lifestyle (occupational exposure to formaldehyde, chlorophenol, wood dust, tobacco users) and viral infection (Epstein–Barr virus and human papilloma virus) in the etiology of NPC.5, 6, 7, 8, 9 However, many individuals exposed to these parameters do not develop NPC, which indicates the involvement of genetic factors. To establish a link between genetic factors and NPC development, study of single-nucleotide polymorphism (SNP) in tumor suppressor genes has been the focus of many researchers.

p53 is a well-established tumor suppressor gene located on chromosome 17p13.1. It plays a critical role in response to genotoxic stress and tries to maintain genomic stability and control proper execution of the cell cycle and apoptotic pathways.10, 11, 12 Deregulated function of p53 may result in loss of this regulation, resulting in uncontrolled cell proliferation and cancer development.13, 14, 15 Polymorphisms in p53 or target genes impair the function of the p53 signaling pathway.16 The most studied polymorphism in p53 is located in exon 4 at codon72. It carries either the CGC sequence that encodes arginine or the CCC sequence that encodes proline due to G/C transversion.17, 18 As a result, two allelic forms (Arg and Pro) and three genotypes (ArgArg, ArgPro and ProPro) have evolved. These allelic variants and genotypes oscillate in their binding capacity to the transcriptional factors, induction of apoptosis and repression of transformation of human cells.7, 18, 19, 20 Arg variants induce apoptosis more efficiently than do the Pro variants, which may be due to their ability to localize into mitochondria and regulate the release of cytochrome C into cytosol.18 The released cytochrome C in turn activates caspase-3, one of the key executioners of apoptosis.21, 22 This difference between Arg and Pro variants may provide the plausible cause for Pro allele’s involvement in increased susceptibility to NPC. Earlier, several studies including our present study among the populations of north-eastern India have investigated the relation between p53 codon72 Arg>Pro polymorphism and NPC risk.23, 24, 25, 26, 27, 28, 29, 30, 31, 32 The purpose of a case–control study of north-eastern Indian populations was also to find out the incidence of different stages of NPC among them and to examine the clinical symptoms manifested by them. However, these findings were inconsistent and inconclusive. In view of the fact that a single study may have been underpowered in clarifying the association, we performed a meta-analysis to combine the findings of all earlier studies from public records and data from the present study according to PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses)33 guidelines to explore the overall association and derive a near-specific conclusion.

Results and discussion

NPC is a public health problem in many countries; it has a complex etiology and ranks 24th among the most frequently diagnosed cancers.34 The incidence rate of this cancer is highest in south-east Asia and about 92% of new cases are being found in economically developing countries.34 In India, this rate is comparable to that of the United Kingdom with the younger age peak in the second decade.35, 36 Several susceptible genes have been implicated for NPC risk, such as tumor suppressor p53, TGFβ1, IL-12 p40 and DNA repair genes.28, 37, 38, 39 In contrast, FokI and Bsm I polymorphisms of vitamin D receptor gene, SNP of deleted in liver cancer-1 (−29A/T) showed no association with NPC.40, 41 However, polymorphisms in PIN-1, TNF-α and glutathione S-transferase genes are indirectly associated with NPC as they influence the p53 codon72 polymorphism.42, 43, 44 These studies suggest that genetic predisposition may play a role in NPC development. Hence, we conducted a study in the north-eastern Indian population among healthy controls and NPC patients to find out the prevalence of p53 codon72 Arg>Pro polymorphism (Table 1). In the control population, the wild-type homozygous ArgArg genotype (48.57%) was more prevalent than the mutant heterozygous ArgPro genotype (28.57%) and mutant homozygous ProPro genotype (22.86%). Also, the prevalence of the Arg allele (62.86%) was higher than that of the Pro allele (37.14%). In the NPC population, the ArgPro genotype (47.14%) was encountered more than the ArgArg genotype (20%) and the ProPro genotype (32.86%). However, the Pro allele (56.43%) was more prevalent than the Arg allele (43.57%). The allele and genotype frequencies of the p53 codon72 polymorphism observed in this population were comparable to those of previous studies conducted in other populations.23, 24, 25, 26, 27, 28, 29, 30, 31, 32 Significant association was observed in the distribution of p53 codon72 polymorphism in controls and NPC patients (Pro allele: P=0.001, OR=0.45, 95% CI=0.28–0.73; ArgPro genotype: P=0.03, OR=0.39, 95% CI=0.18–0.97; ProPro genotype: P=0.008, OR=0.28, 95% CI=0.11–0.69) (Table 2).

Table 1 Baseline and clinical characteristics of NPC cases and healthy controls
Full size table
Table 2 Genotyping and distribution of p53 codon72 Arg>Pro polymorphism in NPC cases and healthy controls of north-eastern Indian populations
Full size table

The p53 codon72 SNP has been studied by many groups across the world. In the Chinese population Birgander et al., in 1996, and Yung WC et al., in 1997 demonstrated no association between the mutant p53 codon72 and the risk for NPC.23 Subsequently, Tsai et al.,24 in 2002, reported that the p53 ProPro homozygote was a risk factor for NPC development. In the Thai population, Tiwawech et al.,25 in 2003, reported that the p53 gene polymorphism may be associated with NPC susceptibility, particularly the Pro/Pro genotype carriers in subjects older than 40 years. In Portugal, Sousa et al.26 in 2006 reported similar findings linking the susceptibility of the P53 codon72 polymorphism to NPC. Further, Hadhri-Guiga et al.,27 in 2007, found that individuals carrying the ProPro homozygote carried an elevated risk for NPC in Tunisia. Similarly, Xiao M et al.,29 in 2010, found that the p53 codon72 polymorphism carried an increased NPC risk independently or in combination with the murine double minute-2 (MDM2) polymorphism in a Chinese population sample, suggesting a gene–gene interaction in NPC pathogenesis. Furthermore, Li et al.,28 in 2013, reported that p53 codon72 and miR-34b/c rs4938723 polymorphisms may singly or collaboratively contribute to the risk for NPC. Two more studies reported this polymorphism as an independent prognostic marker for NPC, and hence one could speculate that this polymorphism means more risk for incidence and more risk for an aggressive disease.45, 46 Moreover, Zhang et al., in 2014, observed a weak effect of p53 polymorphisms on NPC risk. However, they found a significant risk with combination genotypes (i.e., p53 codon72 ArgPro+ProPro, MDM2 rs2279244 GT+GG, PTEN rs11202592 CC, AKT1 rs1130233 AA).30 Overall, variability in study results may be attributed to variation in study design, environmental factors, genetic backgrounds, racial heterogeneity, sample size, source of controls and enrollment criteria for NPC cases. A previous meta-analysis showed that the ProPro homozygote of p53 codon72 possesses an increased NPC risk.47 In another meta-analysis, Jiqiao Yang et al.48 analyzed publicly available data under five comparison models (allele contrast, homozygous, heterozygous, dominant and recessive) and showed the association of p53 codon72 Arg>Pro, MMP-1 (1G>2G), MMP-2 (−1306C>T), CYP2E1 (RsaI) and XRCC1 codon399 Arg>Gln polymorphism with increased risks for NPC.

In our meta-analysis, all eligible reports that fulfilled the inclusion criteria were identified from publication search and the data from the north-eastern Indian population were also included for evaluation.23, 24, 25, 26, 27, 28, 29, 30, 31, 32 Thus, a total of 10 case–control studies counting 1842 NPC patients and 2330 controls comprising populations from India, China, Tunisia, Portugal and Thailand were included in the final meta-analysis (Figure 1). The characteristics of all studies considered for the p53 codon72 Arg>Pro polymorphism were given. Minor allele frequency, Hardy–Weinberg equilibrium and a post hoc power of each study were calculated to detect the probability of association between p53 codon72 Arg>Pro polymorphisms and NPC at the 0.05 level of significance, assuming small effect size (w=0.15). In the north-eastern Indian population, the minor allele frequency of the p53 codon72 Arg>Pro was 0.37 for controls and 0.56 for NPC. The power of this case–control study was too weak (23%) to detect any mild effect of the polymorphisms on disease susceptibility.

Figure 1
figure 1

Flow chart showing the overall process of the study identification and selection. All case–control studies related to p53 codon72 and NPC were searched for in Pub Med, Science direct and manually from printed editions in different journals published up to December 2015. The search items included the combination of the following key words: p53, p53 codon72, p53 codon72 ArgPro, p53 codon72 Arg>Pro, p53 Arg72Pro or rs1042522; and nasopharyngeal cancer, nasopharyngeal carcinoma or NPC; and mutation, polymorphism, single nucleotide polymorphisms or SNPs. The inclusion criteria were case–control studies in peer-reviewed journals and articles containing useful allele and genotype frequency. The exclusion criteria were case reports without control, overlapping data with previous publications, and review articles.

Full size image

The distribution of genotype frequency among controls in all these studies did not deviate from Hardy–Weinberg equilibrium since P>0.05, except the study in the north-eastern Indian population (Table 3).

Table 3 Main characteristics of studies included in the meta-analysis
Full size table

It is worth noting that the small size of samples from the north-eastern population may be due to the low incidence of NPC.

Significant associations between p53 codon72 Arg>Pro polymorphism and NPC risk were observed in the combined analysis of overall studies (Pro vs Arg: OR=1.28, 95% CI=1.17–1.40, POR <0.001; ProPro vs ArgArg: OR=1.70, 95% CI=1.41–2.04, POR<0.001; ArgPro vs ArgArg: OR=1.24, 95% CI=1.07–1.43, POR=0.004; ProPro+ArgPro vs ArgArg: OR=1.35, 95% CI=1.17–1.55, POR<0.001; ProPro vs ArgArg+ProPro: OR=1.54, 95% CI=1.18–2.01, POR=0.002) (Figure 2). Stratified analysis was performed according to ethnicity (Asian, Caucasian), source of sample (hospital-based and population-based studies) and in Chinese studies. The pooled ORs (Table 4) and forest plots (figures not shown) indicated that the p53 codon72 polymorphism among Asians and population-based studies was associated with the development of NPC in all five comparison models (Pro vs Arg, ProPro vs ArgArg, ArgPro vs ArgArg, ProPro+ArgPro vs ArgArg and ProPro vs ArgArg+ProPro). In Caucasian and hospital-based studies a similar risk was noted in three comparison models (Pro vs Arg, ProPro vs ArgArg and ProPro vs ArgArg+ProPro). In the overall Chinese studies NPC risk was found for all comparison models except for the recessive model. Sensitivity analyses were carried out to assess the stability of the results in the overall and stratified analysis by sequential omission of individual study each time. It was observed that the influence of individual data sets on the significance of pooled ORs was not markedly influenced by any single study (data not shown). Funnel plot and Egger’s test were conducted in five comparison models to assess the publication bias in the overall combined meta-analyses. The shape of funnel plots did not reveal any evidence of asymmetry (Figure 3). Stratified analysis in Asian, population-based and Chinese studies also showed similar trends in the shape of the funnel plots (figures not shown). Furthermore, Egger’s test in overall, Asian, population-based and Chinese studies did not show evidence of publication bias in any of the comparison models as P-values were larger than 0.05 (Table 4). However, publication bias (Funnel plot and Egger’s test) was not possible in Caucasian and hospital-based studies because the numbers of studies were less than three. Heterogeneity within and among different studies were tested with Q-value, P-value of heterogeneity (PH) and I2 statistics (Table 4). The random-effects model was used for meta-analysis if the Q-statistic was significant (PH<0.05), which indicates heterogeneity across studies. The fixed-effect model was employed when PH⩾0.05. In the overall population, the fixed-effect model was employed for meta-analysis of the p53 codon72 Arg>Pro polymorphism in four comparison models (Pro vs Arg, ProPro vs ArgArg, ArgPro vs ArgArg and ProPro+ArgPro vs ArgArg). However, the ProPro vs ArgArg+ArgPro comparison model showed heterogeneity among studies in the overall population and the random-effect model was used.

Figure 2
figure 2

Forest plots for association between p53 codon72 Arg>Pro polymorphism and NPC risk. The squares and horizontal lines correspond to the study-specific OR and 95% CI, respectively. The area of the squares reflects the study-specific weight and the diamond represents the pooled OR and 95% CI.

Full size image
Table 4 Summary of overall and stratified meta-analysis results
Full size table
Figure 3
figure 3

Funnel plots of Egger’s test to detect publication bias. Each point represents a separate study. The OR was plotted on a logarithmic scale against the precision of each study.

Full size image

In the stratified analysis, the fixed-effect model was employed in all comparison models of Asian studies except the Pro vs Arg comparison, in which the random-effect model was used. In Caucasian studies the fixed-effect model was employed in all comparison models. In population-based and Chinese studies the fixed-effect model was employed in all comparison models except in the ProPro vs ArgArg+ArgPro comparison model, which reflects the combined results of the overall study. In hospital-based studies the fixed-effect model was employed in three comparison models (Pro vs Arg, ProPro vs ArgArg and ProPro vs ArgArg+ProPro) and the random-effect model in two comparison models (ArgPro vs ArgArg and ProPro+ArgPro vs ArgArg). The overall pooled results indicate that the p53 codon72 polymorphism is a significant risk factor in the pathogenesis of NPC. Stratified analyses in Asian, Caucasian, hospital-based, population-based and Chinese case–control studies corroborate this association. This meta-analysis supports the findings in north-eastern Indian populations. To our knowledge, the current study is the first to analyze the p53 codon72 polymorphism and association with NPC in the Indian population.

In conclusion, our case–control study in North Indian populations and meta-analysis results as evidenced from five genetic models suggest that the p53 codon72 Arg>Pro polymorphism could be employed as a risk factor for NPC. However, some limitations exist in the current meta-analysis. Association of p53 codon72 polymorphism with susceptibility to the histological and clinical grade of NPC patients has not been investigated because of lack of available data on the subject.

The p53 Arg form is more susceptible to degradation than the Pro form by human papilloma virus E6 protein.49 Notably, Epstein–Barr virus infection modulates the effect of the p53 family22 and is a well-nown risk factor for NPC. Nevertheless, whether the p53 Arg or the Pro form is also susceptible to degradation by viruses or by other infectious agents needs to be investigated. Further, as there are a large number of SNPs for p53, the SNP studied in the present analysis was limited only to the functionally important one. In future, screening of all p53 and related polymorphisms in larger samples based on ethnicity in view of confounding factors such as age, sex, cigarette smoke, tobacco use, alcohol intake, dietary habit, stages of NPC and socioeconomic status is required to validate the findings.