Polymorphic TP53BP1 and TP53 Gene Interactions Associated with Risk of Squamous Cell Carcinoma of the Head and Neck

Purpose: Tumor protein 53-binding protein 1 (TP53BP1) and TP53 interact during TP53-mediated transcriptional activation and during checkpoint activation in response to DNA damage. Because suboptimal repair of tobacco-induced DNA damage is associated with risk of squamous cell carcinoma of the head and neck (SCCHN), we hypothesized that potentially functional polymorphisms in TP53BP1 and TP53 may contribute jointly to SCCHN risk.

Experimental Design: In a case-control study, DNA samples from age- and sex-matched SCCHN patients (n = 818) and cancer-free controls (n = 821) were genotyped for the presence of three variants of TP53BP1 (T-885G, Glu³⁵³Asp, and Gln¹¹³⁶Lys) and three variants of TP53 (Arg⁷²Pro, PIN3, and MspI). Multivariate logistic regression was used to assess the adjusted odds ratios (OR) and 95% confidence intervals (95% CI).

Results: Although none of these six genetic variants alone was associated with SCCHN risk, the combined TP53BP1 genotypes were associated with a significant, dose response–dependent decrease in SCCHN risk among carriers of TP53Pro⁷²Pro, TP53PIN3del/del, and TP53Msp1AA genotypes (trend test: P = 0.024, 0.016, and 0.016, respectively). Furthermore, TP53BP1 variant haplotype GGC carriers who were also TP53 variant homozygotes had a significantly lower risk of SCCHN than did TP53BP1 haplotype TCA carriers (adjusted OR, 0.48; 95% CI, 0.25-0.94 for TP53Pro⁷²Pro; adjusted OR, 0.17; 95% CI, 0.04-0.69 for TP53PIN3del/de; and adjusted OR, 0.16; 95% CI, 0.04-0.65 for TP53Msp1AA). There was statistical evidence of interaction between TP53BP1 and TP53 diplotypes (P = 0.017).

Conclusion: Our data suggest that TP53BP1 variants may have protective effects on SCCHN risk but such effects were confined to TP53 variant allele/haplotype carriers.

Squamous cell carcinoma of the head and neck (SCCHN), which includes cancers of the oral cavity, pharynx, and larynx, is relatively common worldwide (1). Major risk factors for SCCHN are tobacco smoke and alcohol use. Tobacco carcinogens cause various kinds of DNA damage (2) that may lead to mutations in critical genes, such as oncogenes and the tumor suppressor gene TP53 (3, 4). However, the fact that some smokers and drinkers do not develop SCCHN suggests that a spectrum of genetic susceptibility exists in the general population (1). Our previous studies have shown that SCCHN risk is associated with suboptimal repair of tobacco-induced DNA damage (5) and low expression of DNA repair genes (6, 7). Although most of the major SCCHN susceptibility genes remain unknown, it is known that variations in genes involved in DNA repair (2, 8, 9) and cell cycle control (10–12) may contribute to tobacco-induced SCCHN.

The tumor suppressor gene TP53 encodes a key cellular component that helps maintain genomic stability by (a) arresting the cell cycle long enough to allow DNA repair, (b) inducing apoptosis, or (c) both (13, 14). Somatic mutations that inactivate the TP53 gene have been found in at least half of all human tumors (15, 16), suggesting that loss of TP53 function plays an important role in carcinogenesis. These mutations may either be acquired or occur naturally in the form of common genetic variants, such as the nonsynonymous single nucleotide polymorphism (SNP) at codon 72 (Arg⁷²Pro). The Arg⁷²Pro SNP has been extensively studied for its association with cancer risk, although the findings have ranged from conflicting (17) to conclusive (17–27).

To function properly, the TP53 protein must interact with many other proteins. One of the TP53-regulated genes is TP53-binding protein 1 (TP53BP1) that encodes a nuclear protein of 1,972 amino acids that contains numerous phosphatidylinositol-like kinase phosphorylation sites (S/TQ) and two NH₂-terminal BRCT motifs (28). TP53BP1 takes part in both DNA repair and cell cycle control and interacts specifically with the DNA-binding core domain of TP53 to enhance TP53-mediated transcriptional activation (29). It also helps mediate the DNA damage checkpoint by cooperating with damage sensors and signal transducers (30, 31).

TP53BP1 is polymorphic. Of over 178 SNPs reported to date, 70 are relatively common (e.g., minor allele frequency >0.05) but only 1 in promoter (i.e., T-885G) and 2 nonsynonymous (i.e., Glu³⁵³Asp and Gln¹¹³⁶Lys).⁴

A recent Chinese study of breast cancer (32) found variant genotypes of these three potentially functional TP53BP1 SNPs to be associated with increased breast cancer risk, particularly among TP53Pro⁷²Pro homozygotes. However, this study did not include two other TP53 SNPs known to be associated with cancer risk: TP53PIN3 (a 16-bp insertion/deletion variant in TP53 intron 3 associated with lung cancer risk; ref. 24) and TP53MspI (a 1798G>A SNP in TP53 intron 6 associated with colon cancer risk; ref. 27). We hypothesized that interactions between variants of TP53BP1 and TP53 may collectively contribute to SCCHN risk. To test this hypothesis, we conducted a case-control study in which we genotyped TP53BP1 variants T-885G, Glu³⁵³Asp, and Gln¹¹³⁶Lys SNPs and TP53 variants Arg⁷²Pro, PIN3, and MspI SNPs in SCCHN patients and cancer-free controls.

Study subjects. Our recruitment of study subjects has been described in detail elsewhere (11). In brief, we identified all patients whose newly diagnosed, untreated SCCHN was histologically confirmed at The University of Texas M. D. Anderson Cancer Center between May 1995 and March 2005. Excluded were any patients with a second primary SCCHN tumor, a primary tumor of the nasopharynx or sinonasal tract, a primary tumor outside the upper aerodigestive tract, cervical metastasis of unknown origin, or any histopathologic diagnosis other than SCCHN.

Cancer-free control subjects were recruited from among visitors at the clinics of our institution. Excluded as potential control subjects were any persons genetically related to any enrolled case subject or to any other control subject. Potential control subjects were asked to complete a short questionnaire to (a) determine their willingness to participate in research studies and (b) obtain demographic information for frequency matching to the cases by age (±5 years), sex, and ethnicity.

After giving informed consent, all eligible subjects who agreed to participate were interviewed to collect additional information about risk factors, such as tobacco smoking and alcohol use. Each study subject had 30 mL of blood drawn for later biomarker testing. Because relatively few minority subjects were recruited in this study, only non-Hispanic whites were included in the current analysis. The research protocol was approved by the Institutional Review Board of The University of Texas M. D. Anderson Cancer Center.

Genotyping. Genotyping analyses were done on genomic DNA obtained from the study subjects. Six SNPs were targeted for analysis: TP53BP1 T-885G, Glu³⁵³Asp, and Gln¹¹³⁶Lys SNPs and TP53 Arg⁷²Pro, PIN3, and MspI SNPs. The materials, methods, and PCR conditions for genotyping these six genetic variants have been described previously (24, 32). The laboratory personnel doing the genotyping analyses were blinded to the subjects' case-control status. Similar numbers of case and control DNA samples were assayed on each 96-well PCR plate. Approximately 10% of the DNA samples were reanalyzed, and the results of both sets of analyses were 100% concordant.

Statistical analysis. The case and control groups were compared in terms of selected demographic variables, smoking status, and alcohol use. Differences were evaluated using the χ² test. The associations between genotypes or diplotypes of the selected polymorphisms and SCCHN risk were estimated by computing the odds ratios (OR) and 95% confidence intervals (95% CI) from both univariate and multivariate unconditional logistic regression analyses. An analytic software program (PHASE 2.0; ref. 33) was used to infer haplotype frequencies based on observed genotypes. For each individual case or control, “diplotype” was defined as the most probable haplotype pair inferred using the PHASE 2.0 program. Potential gene-gene interactions were evaluated by logistic regression analysis and maximum likelihood testing as follows: the changes in deviance (−2 log likelihood) between the models were compared in terms of main effects with or without the interaction term. All statistical analyses were done using Statistical Analysis System software (v.9.1.3; SAS Institute).

A total of 835 cases and 854 controls of non-Hispanic whites frequency matched for age (±5 years) and sex was recruited. Because of the poor quality of their DNA samples, 17 of the case subjects and 33 of the control subjects did not have usable genotyping data, which were consequently excluded from the final analysis. Therefore, the final analysis included 818 case subjects and 821 control subjects. However, the cases were significantly more likely than the controls to be smokers (current smokers, 34.6% versus 16.0%; former smokers, 39.5% versus 37.3%) and drinkers (current drinkers, 50.9% versus 41.1%; former drinkers, 26.0% versus 19.1%; P < 0.001 for both smoking and alcohol use). Demographic and exposure variables were further adjusted for in multivariate logistic regression analyses. The most frequent SCCHN was cancer of the pharynx (46.8%, 383 of 818) followed by cancer of the oral cavity (30.1%, 246 of 818) and larynx (23.1%, 189 of 818).

The genotype frequency distributions of the six selected polymorphisms in the controls were consistent with the Hardy-Weinberg equilibrium: TP53BP1 T-885G (P = 0.311), Glu³⁵³Asp (P = 0.271), and Gln¹¹³⁶Lys (P = 0.650) and TP53 Arg⁷²Pro (P = 0.411), PIN3 (P = 0.916), and MspI (P = 0.670). In the single-locus analysis (Table 1), none of the six variants was associated with SCCHN risk.

Further linkage disequilibrium analysis revealed high but incomplete linkage disequilibrium among the three loci in TP53BP1 (r² = 0.684, D′ = 0.851 for T-885G and Glu³⁵³Asp; r² = 0.681, D′ = 0.833 for T-885G and Gln¹¹³⁶Lys; and r² = 0.615, D′ = 0.815 for Glu³⁵³Asp and Gln¹¹³⁶Lys) and the three loci in TP53 (r² = 0.295, D′ = 0.860 for Arg⁷²Pro and PIN3; r² = 0.295, D′ = 0.860 for Arg⁷²Pro and MspI; and r² = 0.721, D′ = 0.849 for PIN3 and MspI). We estimated eight inferred haplotypes for both TP53BP1 and TP53. Of these haplotypes, two common TP53BP1 haplotypes [i.e., TCA (66.9%) and GGC (22.2%)] and three common TP53 haplotypes [Arg-ins-G (71.7%), Pro-ins-G (14.2%), and Pro-del-A (10.6%)] accounted for 89.1% and 95.8%, respectively, of the 1,642 chromosomes in control subject DNA; however, only one rare TP53BP1 haplotype (i.e., TGA) was associated with a significantly reduced risk of SCCHN (adjusted OR, 0.50; 95% CI, 0.28-0.87; Table 2).

Because altered risk was associated with the number of TP53BP1 variant alleles only in TP53 variant carriers, we also stratified and analyzed the TP53BP1 haplotypes by TP53 genotype. Compared with the most common TP53BP1 haplotype TCA, the TP53BP1 haplotype GGC containing all variant alleles was consistently associated with a significantly reduced SCCHN risk only in TP53 variant homozygotes (adjusted OR, 0.48; 95% CI, 0.25-0.94 for TP53Pro⁷²Pro; adjusted OR, 0.17; 95% CI, 0.04-0.69 for TP53PIN3del/del; and adjusted OR, 0.16; 95% CI, 0.04-0.65 for TP53Msp1AA; Table 3).

Finally, we analyzed diplotypes using TP53BP1 GGC and TP53 Pro-del-A as the variant haplotypes. Because too few carriers of the TP53BP1 diplotype GGC/GGC also carried the TP53 Pro-del-A/Pro-del-A diplotype, this group was combined with the heterozygotes of the TP53BP1 haplotype GGC. Diplotypes of one or two copies of the TP53BP1 haplotype GGC harboring all three TP53BP1 variant alleles were associated with a significantly reduced SCCHN risk, if the carrier had two copies of the TP53 variant haplotype Pro-del-A (adjusted OR, 0.08; 95% CI, 0.01-0.50) but with a nonsignificantly increased risk, if the carrier had zero copies of the TP53 variant haplotype Pro-del-A (adjusted OR, 1.10; 95% CI, 0.87-1.38]); this interaction was statistically significant (P = 0.017; Table 4).

In this hospital-based case-control study, we investigated the association between six potentially functional SNPs of the TP53BP1 and TP53 genes and the risk of SCCHN in a U.S. non-Hispanic white population. Despite observing no detectable main effect of any single genetic variant on SCCHN risk, we did note that the main effect of haplotypes/diplotypes of the TP53BP1 SNPs on SCCHN risk seemed to be modified by the TP53 Arg⁷²Pro SNP, particularly in TP53 variant homozygotes and haplotypes/diplotypes. The finding of an interaction, but in a different direction, was also reported in Chinese breast cancer patients (32), although any conclusions drawn from these two studies are obviously limited by differences in ethnic diversity and cancer sites. Moreover, the present finding, as they are based on a limited number of possible interactions, needs to be validated in larger studies that allow for robust subgroup analyses.

That the effects of TP53BP1 on cancer risk might depend on the status of TP53 is biologically plausible. Studies of the association between TP53 polymorphisms and cancer risk have produced inconsistent results (34). It has been variously suggested that the wild-type TP53 Arg⁷² allele enhances tumor development (via increased inactivation of TP53) in cells expressing the mutated TP53 protein but inhibits tumor development (via increased apoptotic ability) in cells expressing the wild-type TP53 protein (34), that the TP53PIN3 variant causes alternative splicing and RNA instabilization (27), and that the TP53MspI variant influences apoptosis and cell survival (35). Together, these suggest that the joint effect of these three TP53 polymorphisms on cancer risk may be more significant than the individual effect of any one of them alone. Indeed, cells bearing the TP53 haplotype that contains all three variant TP53 alleles (Pro-del-A, the so-called “mutant” MMM haplotype) are associated not only with lung cancer risk but also with a decreased capacity for apoptosis and DNA repair (24). Therefore, the association of this Pro-del-A haplotype with decreased SCCHN risk and the interaction between the TP53BP1 variant diplotypes and the TP53 diplotypes that carry it suggest that the TP53BP1 variant alleles may evolutionarily compensate for the adverse effects of the TP53 variant alleles/haplotype. However, these findings need to be validated in larger studies. Additionally, we found no significant effect of the TP53 Pro-del-A haplotype by itself but a profound effect of the homozygous TP53 Pro-del-A genotype on the effect of the TP53BP1 diplotype on cancer risk.

It is known that the p53 codon 72 variants have markedly different apoptotic potential by differentially activating p53-responsive promoters. For example, the Arg⁷² allele has a higher apoptotic potential than the Pro⁷² allele because Arg⁷² has an enhanced association with MDM2 and CRM1 and a greater ability to localize to the mitochondria (36); earlier study also suggested that Arg⁷² allele enhanced the ability of mutant p53 to bind p73 and neutralized p73-induced apoptosis (37). Therefore, it is likely that TP53 and TP53BP1 variant allele or haplotype may enhance their protein interaction in the absence of TP53 mutations, a hypothesis consistent with a protective effect observed in this study. Larger studies are needed to test this hypothesis.

To date, only two case-control studies have investigated the role of TP53BP1 variants in cancer susceptibility. One was a relatively large German study of 353 breast cancer patients and 960 control subjects that found no overall association between four TP53BP1 SNPs (i.e., D353E, G412S, K1136Q, and 1347_1352delTATCCC) and breast cancer risk (38). The other was the Chinese study of 404 breast cancer cases and 472 cancer-free controls that found no significant main effect of any TP53BP1 genotype (of T-885G, Glu³⁵³Asp, and Gln¹¹³⁶Lys) or haplotype (except for GGC) on risk but an increased risk associated with the combined genotypes in TP53 variant Pro⁷²Pro homozygotes only (32). However, we noticed that the allele/haplotype frequency distribution of both TP53BP1 and TP53 variant genotypes in the Chinese study differed dramatically from that in ours and that our study had a much smaller number of TP53 variant homozygotes and diplotypes containing two copies of the TP53 mutant haplotype, thus limiting the statistical power of our study. One possible reason for the difference in frequency distribution between studies is ethnicity. The distribution of TP53BP1 Glu³⁵³Asp genotype frequencies in our U.S. non-Hispanic white population (51.6% for CC, 39.3% for CG, and 9.0% for GG) was more comparable with that in the German case-control study (47.6% for CC, 42.5% for CG, and 9.9% for GG in controls; P = 0.233; ref. 36) than to that in the Chinese study (30.9% for CC, 50.9% for CG, and 18.2% for GG; P < 0.0001; ref. 32). The frequency distribution of TP53 Arg⁷²Pro genotypes among the control subjects in our current study was also significantly different from that in the Chinese study (P < 0.0001; ref. 32). Another possible reason for the difference between the Chinese findings and ours is the type of cancer studied. SCCHN and breast cancer likely involve fundamentally different pathways of DNA repair. Indeed, the nucleotide excision repair pathway seems to be more relevant in the etiology of SCCHN (6, 7), whereas DNA double-strand break repair, as suggested by the roles of BRCA1 and BRCA2 (39), seems to be more relevant in the etiology of breast cancer (22).

In summary, our data from a relatively large, although racially homogenous, case-control population suggest that none of the TP53BP1 and TP53 SNPs we studied affects individually the risk of SCCHN. However, the data also suggest that gene-gene interaction between TP53BP1 and TP53 may alter the risk of SCCHN, a notion that warrants further evaluation in larger studies. In any case, our findings warrant functional elucidation of the six SNPs we have studied here to better understand the mechanisms underlying carcinogenesis in SCCHN.

We thank Margaret Lung, Kathryn Patterson, and Leanel Fairly for their assistance in recruiting the subjects; Zhensheng Liu, Xiaodong Zhai, and Jiachun Lu for their technical support; Yawei Qiao, Jianzhong He, and Kejing Xu for their laboratory assistance; Monica Domingue for manuscript preparation; and Jude Richard, ELS, for scientific editing.