TP53, MDM2, NQO1, and Susceptibility to Cervical Cancer

Infection with oncogenic human papillomavirus (HPV), particularly type 16– and/or type 18–related HPVs, is the major risk factor for invasive cervical cancer (ICC). However, disease progression is variable among infected women because host genetics in addition to environmental, hormonal, and possibly nutritional factors influences outcomes. For example, we previously showed that the HLA DQB1 and CD83 variants associated with ICC in a family-based association study (1-3). A single nucleotide polymorphism (SNP) in codon 72 of TP53 (P72R, rs1042522) has attracted wide attention over the past decade. The C to G base change in codon 72 replaces proline (Pro) with arginine (Arg). Storey et al. (4) found that women who are homozygous for TP53Arg are seven times more susceptible to HPV-associated squamous carcinoma of the cervix than are heterozygous women. Since then, many groups have reported an effect of the TP53 codon 72 polymorphism on cervical cancer. However, the results of those case-control studies have been inconsistent due to many factors, such as differences in the frequency of TP53 polymorphisms among ethnic groups, sample size, choice of control sample, and DNA sample source.

TP53 executes its tumor suppressor function by regulating DNA repair, cell cycle arrest, and apoptosis. Levels of its product, tumor protein 53 (p53), are tightly controlled, and disruption of the p53 pathway is a hallmark of most cancers. The E6 protein encoded by HPVs 16 and 18 complexes with the p53 protein, inducing its ubiquitin-dependent degradation. The resulting lack of p53 eliminates the p53-dependent control of the cell cycle (5).

Ubiquitin-dependent degradation of p53 in the proteasome is regulated by MDM2. An alternative pathway that does not require ubiquitin is regulated by NAD(P)H quinone oxidoreductase 1 (NQO1; refs. 6, 7). Therefore, stabilization of p53 requires either MDM2 or NQO1 activity.

MDM2 binds directly to and inhibits p53 by regulating its location, stability, and ability to activate transcription (8). A SNP in the MDM2 promoter (SNP309 T/G, rs2279744) increases affinity for Sp1 transcriptional activator, resulting in higher levels of MDM2 protein and subsequent attenuation of the p53 pathway (9). A meta-analysis of 25 published case-control studies found that MDM2 309GG associates with a significantly increased risk of all types of cancers [odds ratio (OR),1.17; 95% confidence interval,1.04-1.33; ref. 9].

Variants in the NQO1 gene also associate with susceptibility to various forms of cancer. A SNP in exon 6 (NQO1, 609 C/T, rs1800566) replaces proline with serine at amino acid position 187. The TT genotype gives rise to an inactive enzyme, and the CT genotype produces only mild activity compared with CC (10). These substitutions have been associated with cancer susceptibility in several solid tumors (11). Another SNP, in exon 4 (NQO1, 465 C/T, rs4986998), shifts arginine to tryptophan at amino acid position 139. This change disrupts the consensus sequence at the 5′-splice site of intron 4, increasing alternative splicing and decreasing the expression of NQO1 protein (12). The contribution of this polymorphism to cancer has received only limited attention, however.

We conducted a family-based study to investigate the association between these TP53, MDM2, and NQO1 SNPs and ICC and cervical intraepithelial neoplasia grade 3 (CIN3). We also examined the effect of the HPV type on the infected cervical tissue.

We genotyped a total of 577 family trios. Each trio consisted of a proband—a woman with ICC or CIN3 (CIN3 and/or adenocarcinoma in situ—and either her biological parents or one parent and usually one or more siblings. Blood or buccal cell samples were obtained from all participants. Cervical tissue either snap frozen in Tissue-Tek optimum cutting temperature compound (Sakura Finetek USA, Inc.) or formalin-fixed, paraffin-embedded blocks were used to type HPV. The characteristics of patients with ICC and CIN3 are shown in Table 1. The study was approved by the Washington University's Human Research Protection Office.

Genomic DNA was isolated as previously described (2). The TP53 codon 72 polymorphism (rs1042522) was determined using PCR-RFLP analysis as described by Ara et al. (13). PCR was done in a standard procedure with the following primers: forward 5′-TTGCCGTCCCAAGCAATGGATGA-3′ and reverse 5′-TCTGGGAAGGGACAGAAGATGAC-3′. The annealing temperature was 55°C. PCR products were run in a 2.0% agarose gel. After we obtained an amplified fragment of the expected size (199 bp), we digested 3 μL PCR products with 0.5 units of BstU1 at 60°C for 2 h. Then, 5 μL of each PCR product were electrophoresed in a 2.5% agarose gel. The Pro allele, which is not cleaved by BstU1, yields a single band with a fragment length of 199 bp. The Arg allele, which is cleaved, runs as two small fragments of 113 and 86 bp. Digesting the heterozygote samples yields three bands of 199, 113, and 86 bp, respectively.

MDM2 SNP309 (rs2279744) was also genotyped by PCR-RFLP, using the primers described by Bond et al. (14). The primer sets were forward 5′-CGGGAGTTCAGGGTAAAGGT-3′ and reverse 5′-AGCAAGTCGGTGCTTACCTG-3′. The annealing temperature was 63°C. The PCR product was digested with 0.2 units of MspA1 I at 37°C for 3 h. The GG genotype was cleaved into four fragments: 184, 85, 50, and 31 bp. The TT genotype was cleaved into three fragments: 234, 85, and 31 bp. The GT genotype produced all five fragments.

PCR-RFLP results were read by three investigators with no knowledge of the family structure. Inconsistent results were repeated (<1% of the samples).

SNPs rs1800566 and rs4986998 were genotyped by a Taqman genotyping assay, using the ABI Prism 7900 sequence detection system (Applied Biosystems) according to the manufacturer's instructions.

A gynecologic pathologist (PH) confirmed the histology of all the samples. HPV typing was done on DNA extracted from cervical cancers and CIN3 lesions as previously described (2, 15). Briefly, the DNA was amplified with primers to the conserved regions of HPV L1 (L1C1/L1C2M) and E6 (E6-L/E6-R). Aliquots of PCR were run on agarose gel and dHPLC and then sequenced. If tissue was infected with multiple HPV types (indicated by multiple peaks on dHPLC), the PCR fragments were first separated by the fragment collector and then sequenced. Families were grouped according to the HPV type detected in the probands' cervical neoplasia at diagnosis. HPV16-related types are HPV16, HPV31, and HPV52. HPV18-related types include HPV18 and HPV45. HPV16, HPV18-related types include single infection of above HPVs and multi-infection of above HPVs plus other types.

We used the family-based test of association implemented in the program TRANSMIT (16). This method is robust to population stratification, and our association study included trios of different ethnicities. The transmission/disequilibrium test was used to determine whether a particular focal allele was overtransmitted to the affected women and therefore should be associated with disease. Because heterozygosity is needed in the parents to discern which allele has been transmitted to an affected offspring, only doubly heterozygous matings are fully informative. However, heterozygosity in one parent provides partial information, and siblings and population allele frequencies can be used internally to impute any missing genotypes under a full maximum likelihood model of transmission. Thus, although the number of families analyzed remains constant, the number of informative families varies as a function of allele frequency and the consequent degree of heterozygosity. To aid in the interpretation of our results, we report the identities of overtransmitted alleles, the significance of the overtransmission test, and the number of informative families contributing to the result.

To identify possible heterogeneity of risk, we conducted follow-up tests by analyzing the subset of trios in which the proband had HPV16- and/or HPV18-related disease. Studies by us and others suggest this subgroup of high-risk HPV types have increased risk and poor prognosis (1, 17, 18).

Tumor tissues from 373 patients (64.6%) were available for HPV typing. Of those, 326 (87.4%) were positive for at least one HPV16- and/or HPV18-related type. The most common types detected among the cases (either alone or in coinfection with other types) were HPV16 (59.7%) and HPV18 (19.8%). Infection with multiple HPV types was detected in 20.1% of the cases.

Table 2 shows the overall analysis of codon 72 of TP53 in all the cases. There is no significant link between the G allele and cervical cases (ICC and CIN3, P =0.213). However, the G allele of that codon was overtransmitted in Caucasian cases (P = 0.043), and that association was even stronger (P = 0.026) in cases with HPV16- and/or HPV18- related infection. The G allele frequency was 71% for the Caucasian cases and 65% for the African-American cases.

Tables 3 to 5 show the overall analysis of MDM2 SNP309 and NQO1 SNPs. No significant association was found between MDM2 SNP309 (rs2279744) and the study subjects, although overtransmission of the T allele was marginally significant when all cases were analyzed (ICC and CIN3, P = 0.053). This contrasts with a previous meta-analysis that found the G allele associated with a slight increased risk in various tumor types (9). Significant overtransmission of the C allele (P = 0.026) of NQO1 SNP609 (rs1800566) was seen in all cervical cancer cases (Table 4). However, there was no association between NQO1 SNP465 (rs4986998) and cervical cancer (Table 5).

Using a family-based association study, we confirmed that the TP53 codon 72 G (arginine) is significantly overtransmitted in Caucasian ICC and CIN3 subjects, especially in cases infected with HPV16- and/or HPV18-related HPVs. The stronger association with HPV16- and/or HPV18-related HPVs enhances the plausibility of our findings. The NQO1 SNP609 C allele was also significantly overtransmitted in all cases. We observed no significant association between NQO1 SNP465 or MDM2 SNP309 and cervical cancer.

The association between the TP53 codon 72 polymorphism and cervical cancer has been investigated in many studies and in various populations since Storey et al. (4) first identified the Arg/Arg genotype as a risk marker for cervical neoplasia. However, research on this polymorphism has produced conflicting results. Two recent meta-analyses confirmed the association of homozygous Arg with ICC (but not with preinvasive lesions), the overall OR being 1.1 to 1.2 (19, 20). One reason for this association may be that the Arg variant of the p53 protein is more vulnerable than the Pro form to being bound and degraded by the E6 oncoprotein of HPVs 16 and 18 (4, 5). A third meta-analysis failed to confirm the association in most European countries, except in Italy and the United Kingdom, in which it was significant for ICC only (21).

Several factors could be contributing to the discrepancies, such as study design, ethnicity of the subjects, DNA sample source, HPV infection status, sample size, and laboratory methods. To date, all of the TP53 polymorphism studies have had case-control designs. Control populations have included blood donors, age-matched populations, laboratory personnel, HPV-positive individuals only, and HPV-negative subjects only. Traditional case-control designs may cause spurious associations because of unrecognized population substructure. Our family-based approach is robust to population stratification.

The meta-analysis by Koushik et al. identified the deviation from the Hardy-Weinberg equilibrium as the principal source of divergent results (19). Studies with observed departures from the equilibrium suggest a possible issue with the control group, such as bias in control selection, ethnic admixture in the population, or misclassified genotypes due to poor laboratory techniques (such as errors in genotype calls or using tumor DNA for genotyping, which could misclassify individuals as homozygous due to loss of chromosome segment in tumors).

In addition, allele frequency varies across ethnic populations with different genetic backgrounds. For example, the frequency of the arginine variant of TP53 codon 72 changes with latitude, ranging from 31% in South Africa to 50% in India and 76% in Finland (22). Hence, the power to detect the TP53 codon 72 Arg variant with cervical cancer is related to the frequency of that variant in a population: there is generally more power to detect the effect of the variant when it is common rather than rare (22-24). A pooled analysis of 49 studies showed a significant association for arginine and cervical cancer limited to Caucasians in which the frequency of arginine in the population was higher than the other ethnic groups (25). This held true in our study, which is unbiased to population stratification. The Arginine is more frequent in the Caucasian than African-American population and significantly overtransmitted in the Caucasian cervical cancer patients. (Table 2). However, the sample size for African-American families was smaller and no association may be due to statistical power or other biological factors in linkage with p53 in the Caucasian population.

Most studies to date have analyzed relatively small samples and have provided little or no information on tumor histology or HPV type/variants. Furthermore, only a few studies have evaluated the HPV status of cervical lesions. Lack of data on HPV status weakens a study's power, given that the association of the TP53 polymorphism with the risk of cervical cancer is HPV-correlated. In our study, the tumor histology was confirmed and HPV was typed. The association was stronger in subjects affected with HPV16- and/or HPV18-related HPVs. However, family trios are unable to provide estimates of population parameters, such as the genotype relative risk.

The DNA source is yet another factor attributed to the inconsistency of the association. The sources used in the TP53 polymorphism assays varied from WBC, cervical cells, tumor biopsies, paraffin-embedded tissue, or remains unreported. Using tumor cells may lead to spurious results because cancer DNA often loses one allele at the TP53 locus (26). In addition, poor-quality DNA, such as that derived from formalin-fixed tissue, can inhibit PCR amplification. One review showed the overall OR that Arg/Arg women would develop cervical cancer varied by DNA source. The OR seemed to increase when the DNA came from formalin-fixed tissues (20). In the present study, we used genomic DNA from blood or buccal cells for cases and their parents.

Another possible reason for discrepancies is that the significant SNPs identified in this study may predispose to rapid-onset cervical cancer because the probands in this study are predominantly young women. If age of onset is a factor, studies of women with older age may dilute that effect.

Finally, one study showed that interlaboratory differences may impair or destroy the ability to detect an association between p53 arginine and cervical cancer. The proportions of the Arg/Arg, Pro/Pro, and Arg/Pro genotypes varied substantially among laboratories, with κ coefficients ranging from 0.49 to 0.63 (27). The OR for the Arg/Arg genotype in cervical cancer increased from 1.5 to 8.0 when discordant genotypes were excluded. By using stringent allele typing and restricting the comparison with HPV-positive controls, these investigators increased the OR to 21.5 (95% confidence interval, 3.4-137.8; ref. 27). To reduce errors in our study, the genotypes were read by three different investigators with no knowledge of family structure. Inconsistent results were <1% and all were repeated.

The combination of high levels of MDM2 with the G-allele of SNP309 can severely weaken the p53 tumor suppressor pathway, resulting in a higher mutation rate, less efficient DNA repair, and reduced apoptosis. All of these deficits lead to faster and more frequent tumor formation (28). This SNP has been evaluated in several tumor types in diverse populations. A meta-analysis of 25 population case-control studies representing >2,900 individuals found that the homozygous 309GG variant associated with a significantly increased risk of all types of cancers (OR, 1.17; ref. 9). A subsequent case-control study from Brazil, which found no association of MDM2 with cervical cancer, was very small [72 cervical cancer cases (65.5% CIN3) and 100 healthy controls; ref. 29]. Our study may also have lacked the power to identify an association at such a small OR.

A meta-analysis suggested that the NQO1 SNP609 TT genotype, with null enzyme activity, may affect individual susceptibility to lung, bladder, and colorectal cancer (11). In a Japanese study, TT genotype was a risk factor for cervical squamous cell carcinoma but not for adenocarcinoma or adenosquamous carcinoma (30). In addition, a study of head and neck squamous cell carcinomas found no association with NQO1 SNP609 or SNP465 (31). Our study showed that the C allele of SNP609 was significantly overtransmitted in ICC and CIN3, the opposite allele from these previous reported studies. This conflicting result may be attributable to differences in sample size and population characteristics. Niwa et al. (30) studied a relatively small sample (total 131 cervical cancers), and the frequency of the T allele is known to differ among ethnic groups, being higher in Japanese people and lower in Caucasians (0.398 versus 0.217). Additional studies are needed to clarify this association and determine the reason for the variability.

Our study is the first to use a family-based sample to evaluate the risk of variants in TP53, MDM2, and NQO1 genes in women with CIN3 and ICC. However, the number of informative family trios might have been too small to detect relatively small ORs, as identified in the meta-analyses for MDM2 and TP53. For example, we may not have had sufficient power in our subgroup analyses.

In conclusion, this family-based association study indicated a significant association between TP53 codon 72 and cervical cancer in a Caucasian population, and that association was even stronger in subjects infected with HPV16- and/or HPV18-related types. These results provide evidence at the molecular level for the heritability of cervical cancer risk. The study suggests that further work on interactions among genetic variants in the p53 pathway, HPV, and other environmental stresses are warranted.

No potential conflicts of interest were disclosed.

Grant Support: National Cancer Institute grant R01 CA095713.

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.