Association of p21 SNPs and risk of cervical cancer among Chinese women

We ascertained 393 patients with cervical cancer between October, 2008 and September, 2011 at the Shengjing Hospital of China Medical University. Diagnosis of cervical cancer was confirmed by routine histopathological examination. A total of 434 control samples were randomly selected from the physical examination center from the same hospital and during the same period. The selection criteria for the controls included no individual history of cancer, no history of cervical intraepithelial neoplasia, normal cervical cytology and HPV examination, or normal pathological examination of the cervix within the past two years. HPV examination was performed as described previously [21]. At recruitment, written informed consent was obtained from all study subjects.

Age, smoking status, HPV status, and pathological stage of all 393 cervical cancer cases and 434 controls are presented in Table 1. The two groups seemed to be adequately matched in terms of age and smoking. Among the patients, 89.3% were HPV positive. The patients’ disease severity was assessed using the FIGO (1988) cervical cancer staging system (Table 1).

Genomic DNA was isolated from peripheral blood using the Tiangen Blood Genome Kit. Detailed primer and reaction information for the determination of the p21 SNP genotypes is given in Additional file 1: Table S1. Briefly, genotyping of rs762623 and rs2395655 was determined by DNA sequencing (Additional file 2: Figure S1). Genotyping of rs1801270, rs3176352, and rs1059234 was determined using a PCR-RFLP method. Each PCR reaction consisted of 0.1 mg DNA, 1 U Taq polymerase, 10 mM Tris–HCl (pH 8.3), 50 mM KCl, 1.0 mM MgCl₂, 20 mM of each dNTP, and 0.2 μM of each primer. PCR products were digested with the indicated restriction enzymes and visualized on agarose gels (Additional file 3: Figure S2, Additional file 4: Figure S3, Additional file 5: Figure S4).

Genotype and allele frequencies were used to assess conformity to Hardy–Weinberg equilibrium. Differences in age and smoking status between patients and controls were evaluated using the Chi-square test. The statistical effect of this study was calculated by Gpower 14 software (Faul & Erdfelder, Bonn, Germany). Post-hoc analysis (affect scale set to 0.2 and α set to 0.05) determined the statistical power of this study to be 0.86. Linkage disequilibrium (LD) between five loci of p21 was estimated by the likelihood-ratio test. A Bayesian statistical method was used to estimate the haplotypes for each subject based on their known genotypes. These analyses were implemented in Haploview software (version 4.2; Massachusetts Institute of Technology, Cambridge, MA). Haplotype construction and haplotype frequency analysis were performed by PHASE (version 2.1) software [22]. All other statistical analyses were performed using SPSS 13.0 software (SPSS Inc., Chicago, USA). After Bonferroni correction, differences were deemed significant at p < 0.01.

The frequencies of each allele of the five p21 SNPs (rs1801270, rs762623, rs2395655, rs3176352, and rs1059234) for patients and controls are given in Table 2. After performing Bonferroni correction to eliminate bias of multiple comparison, we found that the frequency of the rs1801270 A allele in cases (0.421) was statistically significantly lower than that found in controls (0.494; p = 0.003). Frequency of the rs3176352 C allele in cases (0.319) was also statistically significantly lower than that in controls (0.417; p < 0.001). The frequency of the rs1059234 alleles between cases and controls was close to statistical significance (p = 0.055). The frequencies of the other two polymorphisms were not statistically significant between the two groups (Table 2).

The genotype distributions among both patients and controls (Table 3) were all in accordance with Hardy–Weinberg equilibrium. We used an unconditional logistic regression model adjusted for age and smoking to find any association between genotypes and the risk of cervical cancer (Table 3). We found that the rs1801270 AA genotype was associated with a decreased risk for the development of cervical cancer (OR = 0.583, 95% CI: 0.399 - 0.853, P = 0.005). This observation is consistent with the increased frequency of the rs1801270 A allele among controls. However, we did not observe a similar protective effect of the rs3176352 CC genotype even though the rs3176352 C allele frequency was lower in controls. No other genotypes of p21 SNPs were found to be associated with the risk of cervical cancer (Table 3).

Linkage disequilibrium (LD) analyses gave D’ values and r² values (Additional file 6: Table S2), which suggested that three of the p21 SNP loci (rs1801270, rs3176352, and rs1059234) were in LD (Figure 1). The most common haplotypes were AGT, CCC, CGC, and CGT, accounting for 94.1% of the haplotypes observed in all individuals studied (Table 4). The frequency of AGT haplotype in cases (37.88%) was slightly lower than that in controls (44.42%), while the frequency of the CGC and CGT haplotypes were slightly higher in cases (20.92% and 7.07%, respectively) than in controls (11.63% and 3.91%, respectively; Table 4).

To determine any effect of a particular haplotype on the risk of cervical cancer, we looked at the frequency distribution of the four common p21 haplotypes among cases and controls. To further confirm the association between the p21 haplotype and the risk of cervical cancer, we selected the haplotypes of which the sum frequency in case and control is more than 5% and analyzed the association between the numbers of variants in each haplotype (ht1-AGT, ht2-CCC, ht3-CGC and ht4-CGT) and the risk of cervical cancer (Table 3). We found that homozygosity or heterozygostiy for the AGT haplotype provides a protective effect (Table 5). The OR value of the homozygous AGT haplotype is 0.643 (95% CI: 0.434 - 0.954, P = 0.028), while the OR value of the heterozygous AGT haplotype is 0.636 (95% CI: 0.467 - 0.865, P = 0.004). Homozygosity of the CGC haplotype significantly increased the risk of cervical cancer (OR=3.047; 95% CI: 1.788-5.192, P=0.000), heterozygostiy for the CGC haplotype had no effect on risk of cervical cancer (P = 0.321). Heterozygosity for the CGT haplotype was more common among patients, suggesting an increased risk of cervical cancer among CGT heterozygotes (OR = 1.846; 95% CI: 1.156 - 2.945, P = 0.010). Homozygosity for the CGT haplotype was not statistically significantly different among groups, possibly because of the small sample size (P=0.338). The frequency of the CCC haplotype was not statistically significantly different among the two groups and was therefore not associated with risk of cervical cancer in this population.