Genetic variants in p53 signaling pathway genes predict chemotherapy efficacy in colorectal cancer

Abstract Background The murine double minute‐2 gene (MDM2) was originally identified as predicting chemotherapy efficacy. However, little is known regarding the association between single nucleotide polymorphisms (SNPs) in the p53 signaling pathway and prognosis/chemotherapy sensitivity in colorectal cancer. Methods We analyzed the association between 111 SNPs in 22 p53 signaling pathway genes and both progression‐free survival (PFS) and disease control rate (DCR) using Cox regression and logistics regression analysis. The false discovery rate method was used for correction of multiple testing. Secondary structure was predicted by RNAfold. Expression qualitative trait locus analysis and mRNA expression differences were assessed using the GTEx and TCGA databases. Results We found that the rs747828 C allele of TP73 was significantly associated with reduced PFS (HR = 1.64, 95% CI = 1.27‐2.12, P = 2.00 × 10−4) in the additive model. In the stratified analysis, the rs747828 C allele was significantly associated with both reduced PFS (P = 1.40 × 10−3) and DCR (P = 1.82 × 10−2) in oxaliplatin‐based chemotherapy. The secondary structure of TP73 was altered in response to different rs747828 genotypes. Although the rs747828 C allele was not associated with messenger RNA (mRNA) TP73 expression, it was significantly associated with increased mRNA TP73‐AS1 expression levels in sigmoid tissues. TP73 mRNA was significantly overexpressed in tumor tissues compared to adjacent normal tissues (P = 2.36 × 10−19). Conclusion Our findings indicate that functional genetic variants of TP73 mediate the response to chemotherapy in colorectal cancer.


| INTRODUCTION
Colorectal cancer is the second leading cause of tumor-related mortality in men and the third in women in the United States. 1 First-line treatment, including fluorouracil/leucovorin combined with oxaliplatin or irinotecan, in advanced colorectal cancer has considerably improved survival. 2,3 Oxaliplatin and irinotecan are DNA-damaging agents and are usually combined with 5-fluorouracil as first-line chemotherapy. 4,5 Moreover, IRI/LV/5-FU and OXA/LV/5-FU regimens exhibit similarly substantial efficacy. 6 In spite of these improvements in treatment modalities, most patients eventually relapse due to the development of chemoresistance. Established predictive factors for chemotherapy efficacy include plasma mesothelin and single nucleotide polymorphisms (SNPs) in the ERCC1 gene. 7,8 Moreover, recent studies have shown that distinct patterns of gene expression are associated with both patient survival and response to chemotherapy. 9 Key genes in the p53 signaling pathway, including murine double minute-2 gene (MDM2), tumor protein p53 (TP53), tumor protein p63 (TP63), and tumor protein p73 (TP73), play important roles in cancer incidence, prognosis, and treatment response. [10][11][12][13] Moreover, MDM2 overexpression, which results in p53 dysfunction, increases resistance to chemotherapy. [14][15][16][17] Furthermore, in our previous study, we found that the MDM2 SNP309 polymorphism was a risk factor for colorectal cancers in Asians. 18 However, few studies have reported on the predictive role of p53 signaling pathway SNPs in response to colorectal cancer chemotherapy. 19 The magnitude of the association between chemotherapy sensitivity and p53 signaling pathway SNPs has not been thoroughly elucidated to date. To address this discrepancy, we performed a prospective study to investigate the role of p53 signaling pathway SNPs in colorectal cancer chemotherapy response.

| Study population
In the cohort, 344 patients with histologically diagnosed colorectal cancer were consecutively recruited from the First Affiliated Hospital of Nanjing Medical University and the Affiliated Nanjing First Hospital from September 2010 and were followed up by telephone interviews. 20 After removing 19 patients without receiving oxaliplatin-based or irinotecanbased chemotherapy, a total of 325 patients were retained for further analysis (Table S1). Among them, 188 colorectal cancer patients received oxaliplatin-based chemotherapy and 137 patients received irinotecan-based chemotherapy. Specifically, the clinical characteristics of 166 patients with oxaliplatin-based chemotherapy have been reported in the previous study. 21 Peripheral whole blood samples were collected in ethylenediaminetetraacetic acid (EDTA) tubes. Clinical data including age, sex, smoking status, drinking status, tumor site, Dukes stage, tumor grade, metastatic status, chemotherapy regimen, and response to chemotherapy were also collected. Progression-free survival (PFS) is defined as the time elapsed between chemotherapy initiation and objective disease progression, death, or last follow-up. Written informed consent was obtained from each patient, and this study was approved by the Institutional Ethics Review Board of the Nanjing Medical University.

| Chemotherapy regimen
All patients included in this study received oxaliplatin-based or irinotecan-based chemotherapy. Oxaliplatin-based regimen consisted of a combination of oxaliplatin and short-term infusional FU (FOLFOX) or capecitabine (XELOX). The irinotecan-based regimen was a combination of irinotecan and short-term infusional FU (FOLFIRI) or capecitabine (XELIRI). Therapy was continued until disease progression, unacceptable toxicity, or patient refusal.

| Clinical evaluation
Bidimensionally measurable lesions were evaluated on CT scans before treatment and after a minimum of two cycles of chemotherapy. The primary endpoint was the tumor response to chemotherapy, evaluated according to the Response Evaluation Criteria in Solid Tumors (RECIST 1.1) as follows: (a) complete response (CR): disappearance of all target lesions; (b) partial response (PR): at least 30% decrease in the sum of the diameters of target lesions; (c) progressive disease (PD): at least 20% increase in the sum of the diameters of the target lesions; and (d) stable disease (SD): does not qualify for either PR or PD. All results were confirmed at 4 weeks. A disease control rate (DCR) was defined as the proportion of CR, PR, and SD.

| SNP genotyping
Genomic DNA was extracted from EDTA-treated blood of all subjects using the Qiagen Blood Kit (Qiagen). Genotyping was performed using Illumina Human Qmni ZhongHua Bead Chips in all samples that met the sequencing requirements. A uniform quality control protocol was used to filter samples and SNPs.

| Expression analysis
Expression qualitative trait locus (eQTL) analysis was performed to assess for correlations between genotypes of selected SNPs and expression levels of nearby genes using the Genotype-Tissue Expression (GTEx) project dataset, including 203 sigmoid tissues and 246 transverse tissues. Differential gene expression of RNA-sequencing data in colorectal cancer was analyzed from The Cancer Genome Atlas (TCGA) database (http://cance rgeno me.nih.gov/). Colorectal cancer tissues and normal adjacent tissues in TCGA database were used to calculate logarithmic fold change in expression levels of selected genes. All individuals included in the TCGA database were of European descent.

| Statistical analysis
We extracted SNPs using the Han Chinese from Beijing (CHB) and Japanese from Tokyo (JPT) data from the 1000 Genomes Project and HaploView 4.2 software. Unconditional univariate and multivariate Cox regression analyses were used to calculate hazard ratios (HRs) and their 95% confidence intervals (CIs) for evaluating the association between PFS and genetic variants in colorectal cancer. To calculate the crude and adjusted odds ratios (ORs) and their 95% CIs for evaluating the correlation between DCR and genetic variants, we used an unconditional univariate and multivariate logistic regression model. The false discovery rate (FDR) method was applied for significance testing to restrict the probability of false-positive findings in light of the large number of SNPs tested. The sequence kernel association test (SKAT) was performed to conduct gene-based analysis. We used a two-sided Student's t test to compare significant differences in gene expression between colorectal cancer tumor tissues and adjacent normal tissues. The relationship between BMI and the corresponding gene expression was examined using a linear regression model. PLINK 1.07 was used for primary statistical analysis, and other statistical analyses were performed using SAS (version 11.0; SAS Institute, Inc Cary, NC) and R software (version 3.2.3). Kaplan-Meier curves were used to estimate the effects of identified genotypes on the cumulative probability of PFS and OS. Linkage disequilibrium (LD) between SNPs in the genes of interest was explored using LD mapping in HaploView 4.2. P-values < 0.05 were considered to be statistically significant.

| Characteristics of the study population
As shown in Table S1, 205 patients were male and 120 were female, with a mean age of 58.62 years being observed.

| Selection of genes and SNPs from the p53 signaling pathway
As shown in Figure 1, 87 key p53 signaling pathway genes were selected from KEGG and BioCarta. To specifically investigate the association between SNPs in p53 signaling pathway genes and prognosis of patients with colorectal cancer, we identified 33 genes located on autosomal chromosomes reported by previous studies in PubMed (Table S2). Two hundred and eighty-six SNPs were located in these 33 candidate gene regions, including 2 kb upstream. After functional annotation, 111 putative functional SNPs in 22 genes were retained in the study ( Figure S1).

| In silico analysis and genebased analysis
To investigate the function of selected SNPs, we performed in silico analysis using RegulomeDB, SNPinfo Web Server, and HaploReg. SNPs correlated with colorectal cancer prognosis were predicted to have active biological functions due to the integration of three online functional annotation tools (Table S5). We found that rs747828 in TP73 possessed enhancer histone marks, altered motifs and DNAse, and its RegPotential and RegulomeDB scores were 0.102 and 5, respectively.
Furthermore, we conducted gene-based analysis using SKAT to confirm the most significant associations between genes and colorectal cancer prognosis (Table S6). However, no significant differences were observed in association between TP73 and prognosis in colorectal cancer patients (P = 0.318).
It is worth noting that the chemotherapy regimen was an effect modifier between TP73 rs747828 and colorectal cancer prognosis. We analyzed the relationship of TP73 rs747828 with PFS and DCR stratified by treatments (Table S7). In the oxaliplatin-based chemotherapy subgroup, results of the additive and dominant models indicated that the rs747828 C allele is associated with reduced PFS and decreased DCR. The effect of the TC genotype was significantly associated with reduced PFS and decreased DCR in the codominant model (adjusted HR = 1.86, 95% CI = 1.27-2.71, P = 1.30 × 10 −3 ; adjusted HR = 2.80, 95% CI = 1.21-6.44, P = 1.59 × 10 −2 , respectively). Due to the limited sample size, results could not be calculated in the codominant (CC genotype) or recessive models.
In the irinotecan-based chemotherapy subgroup, results of the additive model indicated that the rs747828 C allele was associated with reduced PFS and decreased DCR (adjusted HR = 1.46, 95% CI = 1.02-2.10, P = 1.03 × 10 −2 ; adjusted HR = 2.49, 95% CI = 1.12-5.55, P = 2.55 × 10 −2 , respectively). However, the rs747828 C allele was neither associated with reduced PFS nor decreased DCR in the dominant model after adjusting for age, sex, smoking, and drinking status. We also examined Kaplan-Meier curves of PFS for rs747828 in patients suffering from colorectal cancer in the dominant model (Figure 2A). Patients with the TC/CC rs747828 genotypes exhibited reduced PFS compared to patients with the TT genotype.

| Prediction of rs747828 TP73 folding structures and eQTL analysis
We conducted in silico analysis using RNAfold to predict the TP73 secondary structure of rs747828. Results showed that the secondary structure was dramatically altered in rs747828 T/C alleles ( Figure S2), with the minimum free energy decreasing from −13.90 kcal/mol to −16.00 kcal/mol. We further conducted eQTL analysis to evaluate the effect of rs747828 in TP73. No significant association was observed between rs747828 and TP73 in the GTEx or TCGA databases ( Figure S3A-B). Moreover, no significant difference was found in 246 colon-transverse samples (NES = 0.020, P = 0.786), but rs7474828 was significantly correlated with the expression of TP73-AS1 in 203 colon-sigmoid samples (NES = 0.369, P = 1.35 × 10 −3 ) ( Figure S3C).

| Gene expression analysis in colorectal tumors and adjacent normal tissues
We used TCGA database to analyze differential expression of TP73 between adjacent normal and tumor tissues. In Figure 3, we observed that TP73 expression was significantly increased in TCGA database (P = 2.36 × 10 −19 ). Among these tissues, TP73 expression was significantly different (P = 1.31 × 10 −14 ) in paired tumor tissues and adjacent normal tissues. Furthermore, we analyzed differential expression of TP73 in colorectal cancer tissues based on age, sex, site, and KRAS mutation status ( Figure S4). We also examined differential expression of TP73 in colorectal cancer tissues based on major cancer stages, metastasis, and BMI ( Figure S5). Each stage and metastasis were significantly different (all P < 1.00 × 10 −4 ). Moreover, TP73 expression was decreased in BMI over 27 (P = 3.25 × 10 −2 ), and TP73 expression decreased with increasing BMI (P = 4.00 × 10 −3 , r 2 = 0.027) by linear regression analysis. Finally, we observed that no significant association existed between TP73 expression levels and overall survival in patients suffering from colorectal cancer ( Figure 2B).

| DISCUSSION
Predictive markers for chemotherapy resistance are highly useful to prospectively identify patients who will benefit from the treatment. Previous studies have suggested that genetic variants are associated with prognosis for patients suffering from metastatic colorectal cancer who are treated with oxaliplatin-based chemotherapy. 22 Several reports indicate that SNPs related to chemotherapy resistance exist in p53 signaling pathway genes. [23][24][25] Most studies have shown that SNPs affect cancer prognosis by altering messenger RNA (mRNA) expression or by combining with microRNA. 26 However, the mechanisms whereby SNPs reportedly affect mRNA are inconsistent. Several studies failed to find mutations in the TP73 gene, suggesting that its mutation plays little role in tumor progression. 27 In this study, we evaluated the relationship between SNPs in p53 signaling pathway genes and colorectal cancer prognosis in a Chinese population. We observed that the TP73 rs747828 C allele might predict reduced PFS and decreased DCR. In addition, rs747828 altered TP73-AS1 expression and affects the secondary structure of TP73. Collectively, our study provides evidence of the relationship between TP73 genetic variants and prognosis in colorectal cancer.
Although rs747828 does not affect mRNA TP73 expression, it was significantly associated with mRNA TP73-AS1 expression levels in sigmoid tissues. In particular, TP73-AS1 is the antisense of the coding gene TP73, which encodes a protein sharing notable similarities to TP53 in structure and function. 28 Previous studies reported that global genomic analysis indicates the transformation of the antisense RNA can affect expression of the sense gene, and SNPs may functionally regulate mRNA expression. 29 Given that TP73-AS1 mantles substantial portions of TP73, TP73-AS1 may function through posttranscriptional regulation of TP73. 30 Therefore, rs747828 may affect TP73 by altering TP73-AS1 expression. Of note, the mutation frequency of rs747828 was only 0.01 in both American and European samples, while it was 0.16 in Asian samples. However, research employing the TCGA database helps to elucidate the state of American and European populations. Furthermore, functional biological experiments are necessary to validate our findings in the future.
Several studies have reported that alcohol consumption is a negative prognostic factor in colorectal cancer, but the results have been inconsistent. [31][32][33] In the present study, rs747828 was robustly associated with PFS and DCR in the drinking subgroup. These differences might be caused by population heterogeneity to some extent. Furthermore, a previous study reported that irinotecan and oxaliplatin regimens have similar chemotherapy efficacy. 6 However, oxaliplatin-based chemotherapy resulted in superior DCR and overall survival compared with irinotecan-based chemotherapy in a metaanalysis. 34 Obviously, these studies did not take different genotypes into account in their analysis. In our study, rs747828 was significantly associated with reduced PFS and decreased DCR in oxaliplatin regimens but not in irinotecan regimens by stratified analysis. Hence, we provided evidence that genotype influences treatment effect, and different chemotherapy drugs possess diverse therapeutic targets and mechanisms.
In conclusion, our study provides a new view of the development of biomarkers for predicting chemotherapeutic efficacy in colorectal cancer. Genetic variants in TP73 may predict chemotherapy sensitivity of colorectal cancer patients treated with oxaliplatin-based chemotherapy. Moreover, increased TP73 mRNA expression was observed in colorectal tumor tissues compared to corresponding normal tissues.