TP73 G4C14-A4T14 polymorphism and cancer susceptibility: evidence from 36 case–control studies

G4C14-A4T14 polymorphism of TP73 gene has been reported with a potential association in cancer risks through affected cell homeostasis; however the results were not consistent. We performed a comprehensive meta-analysis to explore the associations between G4C14-A4T14 polymorphism and cancer susceptibility. Extensive retrieve was performed in PubMed, EMBASE, Google Scholar, Web of Science, Wanfang database and CNKI database up to May 20, 2018. Odds ratios (ORs) and 95% confidence intervals (CIs) were conducted to evaluate the overall strength of the associations in five genetic models, as well as in subgroup analyses. Q-test, false-positive report probability analysis and trial sequential analysis, Egger’s test and Begg’s funnel plot were applied to evaluate the robustness of the results. In silico analysis was managed to demonstrate the relationship of TP73 expression correlated with cancer tissues. Finally, 36 case–control studies with a total of 9493 cancer cases and 13,157 healthy controls were enrolled into the meta-analysis. The pooled results present a significantly higher risk of G4C14-A4T14 polymorphism in all the five genetic models, as well as in the subgroups of Caucasian, cervical cancer, colorectal cancer, H-B subgroup and comfort to Hardy–Weinberg equilibrium subgroup. In silico analysis revealed that the expression of TP73 in cervical cancer tissue is higher than it in corresponding normal tissue, as well as in cervical cancer. All in all, TP73 G4C14-A4T14 polymorphism causes an upgrade cancer risk, especially in Caucasian population. G4C14-A4T14 polymorphism might be a potential biomarker for judging the tumorigenesis of cervical cancer and colorectal cancer.


Introduction
Cancer is a pivotal public health and leads to the second cause of death problem around the world. In 2018, there are almost 4700 new cancer diagnoses per day, as well as about 1700 cancer-related deaths in United States [1]. Breast cancer, lung cancer and colorectal cancer are the most three frequently cancer of female in United States, while prostate cancer occupied the first diagnosis cancer in male [1]. Attributed to the increasing population growth and aging, cancer has also been the leading cause of death around China. In 2015, there are about 12,000 newly diagnosed invasive cancer cases on average per day, while over 7500 cancer death [2]. In the past decades, biological scientists have reported that environmental factors, genetic mutations and the multiple interactions between them mainly affect the process of tumorigenesis, and the new research results are also on the road, such as epigenetic control [3][4][5].
Tumor protein P73 (TP73), also known as P53-like transcription factor, is a pivotal member of TP53 family, which affects cell proliferation, apoptosis and cell-cycle regulation [6][7][8]. Compared with frequently mutant TP53 gene, TP73 is rarely mutated [9]. p73 protein, the encoded product of TP73, is homologous with p53, 63% of p73 has the same amino acid sequence with p53, so it plays a critical role in normal cell homeostasis, while it can partially compensate the loss of p53 protein function [10,11].

Data extraction
Two independent investigators separately extracted the relative data with any disagreement resolved by rechecking and discussion. For every eligible study, the following data were extracted: the name of the first author, the data of publication, ethnicity, sample size, genotyping methods, and genotype frequency of the cases and controls. In the subgroup analysis by race, the Caucasian population typically lived in Europe or America, and the Asian population typically lived in Asia.

Statistical methods
All the statistical calculation was conducted with STATA 12.0 software (Stata, College Station, Texas) in the present study. ORs with corresponding 95% CIs were performed to measure the strength of the relationship between G4C14-A4T14 polymorphism and cancer susceptibility. Five common genetic models applied for assessing gene-disease associations are allele contrast model (GC vs. AT), homozygote comparison model (GC/GC vs. AT/AT), heterozygote comparison model (GC/AT vs. AT/AT), dominant comparison model (GC/GC+GC/AT vs. AT/AT) and recessive comparison model (GC/GC vs. GC/AT+AT/AT) (AT/AT, homozygotes for the common allele; GC/AT, heterozygotes; GC/GC, homozygotes for the rare allele). Stratified analyses were also calculated by ethnicity, cancer type and the source of control. In addition, we applied the chi-squared (χ 2 )-based Q-test to calculate between-study heterogeneity [18]. P<0.1 was indicated as a substantial level of heterogeneity, and a random-effects model (the Der-Simonian and Laird method) was selected to pool the data [19]; or else, the fixed-effects model (the Mantel-Haenszel method) was adopted. Moreover, we also conducted the Begg's funnel plots and Egger's test to evaluate the publication bias [20,21]. Hardy-Weinberg equilibrium (HWE) of controls was calculated by the χ 2 test to compare the expected and actual genotype frequencies among the controls in each study. All the statistical tests in this meta-analysis were two-tailed, and P-values ≤ 0.05 were considered statistically significant.

False-positive report probability analysis and trial sequential analysis
We also use the false-positive report probability (FPRP) method to evaluate the results. 0.2 was set as an FPRP threshold and assigned a prior probability of 0.1 to detect the odds ratio (OR) of 0.67/1.50 (protective/risk effects). The significant result with the FPRP values less than 0.2 was considered as a worthy finding [22,23]. Trial sequential analysis (TSA) was conducted with the guideline of a former publication. We set a significance of 5% for type I error, as well as a 30% significance of type II error, to calculate the required sample size, and built the TSA monitoring boundaries.

In silico analysis of TP73 expression
In order to further explore the relationship between TP73 expression and cancer, we used a newly developed interactive web server, GEPIA (http://gepia.cancer-pku.cn/), to see the difference between tumor tissue and normal tissue. GEPIA provided the mRNA sequencing expression data of tumors and normal samples from the TCGA and the GTEx projects [24].

Study characteristics
As shown in Figure 1, we found 1740 potentially relevant studies from PubMed, EMBASE, Google Scholar, Web of Science, Wanfang database and CNKI database. After reviewing titles and abstracts, we excluded 1537 publications not investigating the association between TP73 G4C14-A4T14 polymorphism and cancer risk. And then, full texts of remaining articles were evaluated. In the end, 36 case-control studies with a total of 9493 cancer cases and 13,157 healthy controls were enrolled into the meta-analysis [14][15][16][17]. The characteristics of these studies were showed in Table 1. Among these publications, there are 6 concerned about cervical cancer [17,34,40,[50][51][52], 5 about lung cancer [16,29,33,35,38], 4 about colorectal cancer [28,37,46,47], 4 about esophageal cancer [14,27,28,39], 4 about gastric cancer [28,39,43,48], 3 about breast cancer [15,26,30], 3 about squamous cell carcinoma of the head and neck [32,41,45], as well as other 7 publications focus on Endometrial cancer [36], lymphoma [31], melanoma [42], nasopharyngeal carcinoma [53], neuroblastoma [25], ovarian cancer [44] and prostate cancer [49], respectively. As to the ethnicity, 14 studies were performed in Caucasians, while the other 22 studies were managed in Asian population. The characteristics of each case-control study, genotype frequencies and HWE examination results were presented in Table 1. Four case-control studies were not comforted to HWE [16,32,37,45], and we further conducted a sensitive analysis to validate the influence of the three studies on the integrated data. In order to evaluate the quality of each enrolled studies, we applied Newcastle-Ottawa Scale (NOS) [45] and fill the result in Table S1, the result of PRISMA2009 checklist was also listed to present our meta-analysis work (Table S2). Table 2 listed the main results of current meta-analysis work of polymorphisms in G4C14-A4T14 and risk of cancer. The pooled results of the 36 included studies had shown that G4C14-A4T14 polymorphism conferred a significantly higher overall risk to cancer susceptibility in all the five genetic models, allelic contrast model (GC vs. AT: OR =   Table  2 and Figure 2).

Stratification analysis by cancer type
After overall pooled analysis, we also conducted stratification analysis by cancer type, in order to obtain more precise result about the G4C14-A4T14 polymorphism and cancer susceptibility. As shown in Table 2 and Figure 3, the subgroup analysis of six enrolled colorectal cancer related studies have shown that G4C14-A4T14 polymorphism was related to an increased cancer risk in allelic contrast model (GC vs. AT: OR = 1.204, 95% CI = 1.044-1.389,    Figure 4). We also performed subgroup analysis of breast cancer, esophageal cancer, gastric cancer, lung cancer and squamous  cell carcinoma of the head and neck, no significant association was found between G4C14-A4T14 polymorphism and these carcinomas in all five genetic models ( Table 2 and Figures S1-S4).  Table 2 and Figure S5).   (Table 2 and Figure S6).

Stratification analysis by HWE status
In order to exclude the influence of allele frequency changing, we calculated whether the control group conform to HWE, and conducted the stratification meta-analysis in subgroups of HWE status. As shown in Table 2 (Table 2 and Figure  S7).

Sensitivity analysis and publication bias
Sensitivity analysis was performed to assess the influence of each individual study on the pooled OR by sequential removal of individual studies, the results showed that the study material alteration did not influence the corresponding pooled ORs for the overall meta-analysis ( Figure 5 and Table S3). In addition, Begg's funnel plot and Egger's test were presented to assess the potential publication bias, no evidence of publication bias was revealed by evaluating the shape of Begg's funnel plot and by Egger's regression test ( Figures S8, S9 and Table S4).

Result of FPRP and TSA
The FPRP values for significant findings at different prior probability levels are shown in Table 3. In the result of overall group in five genetic models, all the statistical power is about 1, and the FPRP values are all less than 0.2, under the prior probability of 0.1. On the subgroup of cervical cancer and colorectal cancer, the FPRP values are also less than 0.2. The result of TSA is shown in Figure 6, the required sample size is 21,728 samples, and the cumulative z-curve crossed the trial sequential monitoring boundary before reaching the required sample size, which means that our conclusions are robust with these sufficient evidence.

Discussion
TP73 gene is located at chromosome 1p36 and comprises 15 exons [54]. TP73 could be transcribed from two individual promoters, one is in the upstream of exon 1, it could produce p53-like proteins containing transactivation domain (TAp73) and another TA lacking protein ( TAp73). The second promoter is situated in intron 3, it could turn out the N-terminal truncated isoform ( Np73) [55]. What's more, both TAp73 and Np73 undergo the alternative splicing and initiation of translation, and lead to several splicing isoforms [56,57]. While sharing the similar sequence with p53, TAp73 could activate the expression of downstream genes through specifically binding domain of p53 response element, regulating cell apoptosis or cell-cycle arrest [58,59]. On the meanwhile, Np73 could present a potent anti-oncogenic function through inhibiting the key role of TAp63, TAp73 or p53 [60]. Several publications had reported that the TP73 expression plays critical role in tumorigenesis, combined with different isoforms or several mutations [61][62][63][64].
In the past decades, almost 146 unique variations had been reported (shown in the Biomuta database [65]), while numerous studies had probed into the relationship of G4C14-A4T14 polymorphism and cancer genomics. G4A (rs2273953) and C14T (rs1801173) polymorphisms are located at position 4 (G to A) and 14 (C to T) of exon 2 5'-untranslated region, which might influence the initiating AUG codon through constructing a stem-loop [54]. Zheng et al. [40] and Niwa et al. [34] reported that G4C14-A4T14 polymorphism was not associated with the cancer susceptibility of cervical cancer in Uighur and Japanese, respectively. However, Craveiro et al. [51] revealed that G4C14-A4T14 polymorphism leads to an increasing risk of cervical cancer, as well as the newest study conducted by Feng et al. [17].As colorectal cancer, Hamajima et al. [28] presented that no significant differences in the genotype frequencies were observed among the enrolled cases and controls in his study. On the contrast, Lee et al. [47] reported that GC/AT and AT/AT genotypes were significantly associated with colorectal cancer risk in Korean population. Arfaoui et al. [66] also uncovered that no remarkable differences of genotype frequencies in cancers and controls, but they found that AT/AT genotype might cause the poor prognosis of colorectal cancer. Several researches also managed in lung cancer. Hu et al. [35] indicated that both AT/AT and GC/AT variants were associated with a remarkable decreased risk for lung cancer, distinguishingly, Li et al. [64] suggested that the AT/AT and GC/AT genotypes were related with a statistically significantly increased risk for lung cancer. Choi et al. [38] did not agree with each of them, they revealed that TP73 G4C14-A4T14 polymorphism does not affect the susceptibility to lung cancer in Korean population.
Among these publications concerned about G4C14-A4T14 polymorphism and cancer risk, the result is not consistent. Liang et al. [67] conducted a meta-analysis about G4C14-A4T14 polymorphism and cervical cancer, they only enrolled 5 studies, as well as Liu et al. [68], they only enrolled 5 studies about lung cancer. Yu et al. [69] had performed a meta-analysis with only 23 eligible studies; however, they draw a decreased risk of G4C14-A4T14 polymorphism, this mistake may cause by the fewer samples. Therefore, our team carried out the present comprehensive meta-analysis aiming at shedding light on the multiple lines of evidence. Finally, 36 case-control studies comprise 9493 cases and 13,157 controls were enrolled and analyzed. All in all, our recent updated meta-analysis draws a comprehensive, precise and convincible result, which is that G4C14-A4T14 polymorphism of TP73 is strongly associated with the increasing cancer risk, especially for Caucasian, cervical cancer and colorectal cancer. Therefore, in the future, G4C14-A4T14 polymorphism might be a useful diagnostic marker for cervical cancer and colorectal cancer, especially in Caucasian population. On the other hand, for researchers, other polymorphisms of TP73 should be focused on to assess whether they change cancer risks.
The current result about G4C14-A4T14 polymorphism and cancer risk should be cautiously interpreted, because there are some limitations. First, an insufficient capacity that slight effects on cancer susceptibility occurred when a stratified analysis was conducted by the cancer type, ethnicity and source of control. Second, several potential confounding factors were ignored, such as age, gender, smoking, drinking and etc., so we are unable to perform a further assessment of potential gene-environment interactions. Third, we only enrolled publications written in English or Chinese, missing publications from other languages may cause potential bias. On the meanwhile, the advantages of this research should not be buried. First, a comprehensive search was conducted to identify more qualified studies, so this analysis is persuasive and substantive. Second, the quality of each registered research was evaluated by NOS scale, low-quality studies were eliminated to raise the credibility of results. Third, stratification analysis was performed by ethnicity, source of controls, tumor type or ethnicity, in order to decrease the impact of heterogeneity and obtain the real conclusion.
In conclusion, our meta-analysis had successfully elaborated that TP73 G4C14-A4T14 polymorphism causes an upgrade cancer risk, especially in Caucasian population. G4C14-A4T14 polymorphism might be a potential biomarker for judging the tumorigenesis of cervical cancer and colorectal cancer.

Author Contribution
M.J.L. and W.S. accessed information from literature for this article. M.J.L., Z.M. and F.S. designed and performed the meta-analysis. M.J.L., Z.L. and L.C.Z. contributed toward writing, discussing and editing the manuscript. All authors discussed the results and contributed to the final manuscript.

Funding
This manuscript was supported by National Natural Science Foundation of China [81630019 (to L.C.Z)].