HPV seropositivity synergizes with MDM2 variants to increase risk of oral squamous cell carcinoma

Abstract

The increasing incidence of oral squamous cell carcinoma (OSCC) in young adults has been associated with sexually transmitted infections of human papillomavirus (HPV), particularly HPV16. Given the roles of p53 in tumor suppression and of HPV E6 and MDM2 oncoproteins in p53 degradation, we evaluated HPV16 L1 seropositivity and MDM2 promoter variants to examine their possible associations with OSCC risk in a case-control study of 325 patients and 335 cancer-free matched controls. Compared with individuals having MDM2-rs2279744 GT or GG genotypes and HPV16 L1 seronegativity, the TT genotype and HPV16 L1 seronegativity were found to be associated with an odds ratio (OR) of 1.25 (95% confidence interval [CI],1.06–2.19) for OSCC risk, and GT/GG and HPV16 L1 seropositivity were associated with an OR of 2.81 (95% CI,1.67–4.74). For those with both the TT genotype and HPV16 L1 seropositivity, the associated OR was 5.57 (95% CI, 2.93–10.6). Similar results were observed for the MDM2-rs937283 polymorphism. Moreover, there was a borderline significant or significant interaction between the individual or combined MDM2 genotypes of the two polymorphisms and HPV16 L1 seropositivity (P_int = 0.060 for MDM2-rs2279744, P_int = 0.009 for MDM2-rs937283, and P_int = 0.005 for the combined MDM2 genotypes) on risk of OSCC. Notably, that effect modification was particularly pronounced in never smokers and never drinkers, and for oropharyngeal as opposed to oral cavity cancer. Taken together, our results indicate that the risk of OSCC associated with HPV16 L1 seropositivity is modified by MDM2 promoter polymorphisms.

Introduction

Oral squamous cell carcinoma (OSCC), which arises from several anatomic sites within the oral cavity and oropharynx and constitutes the majority of head and neck cancers, is common worldwide. The incidence rate of OSCC has risen dramatically in recent decades (1, 2). In the United States, it is estimated that approximately 35,700 new OSCC cases will be diagnosed and 7,600 deaths will occur from these cancers in 2009 (3). OSCC is characterized by local tumor aggressiveness requiring morbidity-inducing local-regional therapies. Even with such therapy, these local tumors have moderately high recurrence rates and common medical comorbidities and are associated with a high frequency of second primary tumors (4). The leading known risk factors for OSCCs are tobacco and alcohol use. However, despite declining smoking rates in the United States, the overall incidence of OSCC in young adults has been increasing in recent years, and this trend has been correlated with the increasing prevalence of infection with human papillomavirus (HPV) (2). Of the 120 known types of HPV, the high-risk oncogenic HPV16 is the most common frequent type, accounting for approximately 90% – 95% of HPV-positive OSCCs (5–8). Although HPV infection may be a major risk factor for OSCC (5, 9), only a small fraction of individuals actually develop OSCC associated with HPVs, thus implying that the patient's own genetic factors may modify the association between HPV infection and the risk of OSCC.

The p53 tumor suppressor has a highly conserved role as the `guardian of genome' (10) and can be activated by or interact with many other proteins in the network of signaling pathways. Upon cellular stress, such as DNA damage and oncogenic signals, the appropriate p53-mediated pathways are activated, and this ultimately leads to cell-cycle arrest, cellular DNA repair, senescence, or apoptosis, thus guarding normal cells against malignant transformation (11, 12). Therefore, p53 has a central role in this complex network of molecular interactions, and its degradation is implicated in the etiology of OSCC (11, 13).

Conversely, the human MDM2 gene promotes rapid degradation of p53 and inhibits growth arrest or p53-mediated apoptosis and cell-cycle control (11, 14–16). In humans, cellular expression levels of MDM2 seem to be critical for regulating p53; inactivation of p53 can be caused through amplification of MDM2 (17, 18). Overexpression of MDM2 in tumors is often associated with poor prognosis (19, 20).

The malignant transforming potential of oncogenic HPV is attributed to its oncoproteins, E6 and E7 (21). The HPV E6 oncoprotein can bind to tumor suppressor p53, promoting ubiquitination and rapid proteasome-mediated degradation (22, 23). Direct mutations can alter or inactivate p53, but interactions with other proteins, such as the HPV E6 oncoprotein of oncogenic viruses and MDM2 protein, can also cause aberrations in p53 regulation (14, 24). Therefore, both the HPV E6 oncoprotein and MDM2 have critical roles in regulating p53 in response to cellular stressors such as DNA damage and oncogenic signals.

Previous epidemiologic studies demonstrated that HPVs (serologic or tumor DNA status) are strongly associated with the risk of OSCC (7, 25–28), but no studies have investigated the association between MDM2 functional promoter polymorphisms and the risk of cancers associated with HPVs, including OSCC. To date, only one case-control study has examined the association between MDM2-rs2279744 polymorphism alone and the risk of head and neck cancers; however, that study included fewer than 100 OSCC patients and found no association (29). Furthermore, as mentioned above, HPVs may share common pathways with MDM2 during carcinogenesis in OSCC, but no published study has yet explored the joint effects of HPV16 L1 seropositivity and MDM2 promoter variants on the risk of OSCC.

We hypothesized that genetic variation in MDM2 modifies the association between HPV16 L1 seropositivity and the risk of OSCC. To test this hypothesis, we evaluated the interactions between HPV16 serologic status and MDM2 promoter polymorphisms on the risk of OSCC.

Materials and Methods

Patients and control samples

All patients with histologically confirmed OSCC were consecutively recruited through the Head and Neck Surgery Clinic at The University of Texas MD Anderson Cancer Center between May 1996 and May 2002. Of patients initially contacted for participation, approximately 95% of eligible incident cases agreed to participate. Excluded from participation were patients with second primary tumors; primary tumors of the sinonasal tract, nasopharynx, hypopharynx, and larynx; primary tumors outside the upper aerodigestive tract; cervical metastases of unknown origin; and histopathologic diagnoses other than squamous cell carcinoma. In addition, patients who had received recent blood transfusions (in the last 6 months) or who were receiving immunosuppressive therapy were excluded. As a result, this study included 325 non-Hispanic white patients with primary squamous cell carcinoma of the oral cavity (n = 137; 42.2%) and oropharynx (n = 188; 57.8%).

A pool of cancer-free subjects was recruited from the Kelsey-Seybold Foundation, a multispecialty physician practice with multiple clinics throughout the Houston metropolitan area, and from healthy visitors who accompanied cancer patients to outpatient clinics at MD Anderson Cancer Center but who were genetically unrelated to these patients. In this pool of cancer-free controls, each individual was first surveyed by means of a short questionnaire to determine his or her willingness to participate in research studies and then interviewed. Each eligible subject provided demographic and epidemiologic information, such as age, sex, ethnicity, smoking status, and alcohol consumption status. The overall proportion of responders was approximately 78%. Exclusion criteria for the control groups included receiving immunosuppressive therapy, having had previous cancer, and having received recent (in the last 6 months) blood transfusions.

Those subjects who had smoked more than 100 cigarettes in their lifetime were defined as `ever smokers' and the rest as `never smokers,' as is traditionally done in epidemiologic studies. Individuals who drank alcoholic beverages at least once a week for more than 1 year were defined as `ever drinkers' and the rest as `never drinkers'. After informed written consent was given, each individual provided 30 mL of blood collected in heparinized tubes. The research protocol was approved by both the MD Anderson Cancer Center and Kelsey-Seybold Institutional Review Boards.

In this study, 335 cancer-free control individuals were selected from the pool of potential controls by frequency matching. These controls were frequency-matched to the patients by age (±5 years), gender, ethnicity, and smoking and drinking status using SAS software (Version 9.1; SAS Institute, Cary, NC). These variables were further adjusted in later multivariable logistic regression analyses, because they were factors on which the controls were frequency matched to the cases.

HPV16 serologic testing

We used HPV16 L1 virus-like particles generated from recombinant baculovirus-infected insect cells to test for antibody against HPV16 L1 capsid protein in the plasma of study participants by using a standard enzyme-linked immunosorbent assay, as described previously (30, 31). Control sera known to be positive and negative were also tested in parallel with the study samples in duplicate on each plate. The cutoff level for HPV16 L1 seropositivity was determined from a standard pooled serum known to be at the cutoff point for HPV16 L1 seropositivity in a previous study (30). Samples within 15% of the cutoff point were tested twice more; those positive in all three tests were considered positive. To eliminate potential binding interference by heparin, we treated the plasma samples with 43 U/mL heparinase I (Sigma, St. Louis, MO) before testing (32). We tested heparinized plasma, as well as serum, obtained from three individuals and did not detect discernible differences between the reactions of the serum samples and those of the heparinized plasma samples treated with heparinase. We also retested a randomly chosen 10% of the samples and obtained 100% concordance on the repeat assays.

MDM2 polymorphism genotyping

We extracted genomic DNA from the buffy-coat fraction of the blood samples using a DNA Blood Mini Kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. To genotype MDM2 promoter polymorphisms, we identified four single nucleotide polymorphisms (SNPs; MDM2-rs2279744, MDM2-rs937282, MDM2-rs937283, and MDM2-rs2870820) that have minor allele frequencies (MAF) > 10% in a published SNP database (33). In our previous pilot study, we found that SNPs MDM2-rs937282 and MDM2-rs2870820 were rare (MAF < 5%) in our study population. Therefore, we only included the common SNPs (MAF > 10%) MDM2-rs2279744 and MDM2-rs937283 in this study.

The methods for genotyping SNPs MDM2-rs2279744 and MDM2-rs937283 have been previously described elsewhere (34). We performed the genotyping and evaluated the results without knowing the samples' patient or control status. A random 10% of the samples were retested, and the results of retesting were 100% concordant.

Statistical analysis

Differences between the patients and controls in the distributions of HPV16 status and MDM2 genotypes were examined using the χ² test. We estimated the association of HPV16 status and MDM2 genotypes with the risk of OSCC by computing the odds ratios (ORs) and their 95% confidence intervals (CIs) using both univariate and multivariable logistic regression analyses. We also evaluated the joint effects of HPV16 serology and MDM2 genotypes on the risk of OSCC: the joint effects were further stratified by smoking and drinking status and tumor site. We assessed the trends in the risk of OSCC associated with the number of variant alleles of both polymorphisms. Logistic regression analysis was also used to assess potential interaction effects by evaluating departures from the model of multiplicative interaction between selected variables. A more-than-multiplicative interaction was suggested when OR₁₁ > OR₁₀ × OR₀₁, for which OR₁₁ = OR when both factors were present, OR₁₀ = OR when only factor 1 was present, and OR₀₁ = OR when only factor 2 was present. We assessed the interaction by reporting the P values from the Wald test for testing the coefficients (β_{MDM2 polymorphism, HPV16 L1 seropositivity}) were different from 0, where the interaction term consisted of the product of the two variables: MDM2 polymorphism and HPV16 L1 seropositivity. To account for the multiple tests of interaction, we used Benjamini and Hochberg's method to calculate the False Discovery Rate adjusted P-values (35). All tests were 2-sided, and a P < 0.05 was considered the cutoff for statistical significance. All of the statistical analyses were performed with Statistical Analysis System software (Version 9.1; SAS Institute, Cary, NC).

Results

Demographics and risk factors for study subjects

The distribution of demographic characteristics and known OSCC risk factors are summarized in Table 1. We found that HPV16 L1 seropositivity was significantly more common in patients than in controls (P < 0.001) and that HPV16 L1 seropositivity was associated with ORs of 3.17 for OSCC (95% CI, 2.11–4.77), of 0.75 for oral cavity cancers (95% CI, 0.38–1.48) and 5.37 for oropharyngeal cancers (95% CI, 3.70–8.89), respectively, after adjusting for age, sex, smoking status, and drinking status.

Table 1.

Frequency distribution of demographic and risk factors in OSCC patients and controls recruited from The University of Texas MD Anderson Cancer Center and general hospitals in Houston between 1996 and 2002

	Patients n = 325)		Controls* (n = 335)
Characteristics	Number	Percentage	Number	Percentage
Age (years)
< 40	31	9.5	27	8.1
41 – 55	126	38.8	105	31.3
56 – 70	119	36.6	154	46.0
> 70	49	15.1	49	14.6
Sex
Male	241	74.1	269	80.3
Female	84	25.9	66	19.7
Ethnicity
Non-Hispanic white	325	100	335	100
Tobacco smoking
Ever	227	69.8	239	71.3
Never	98	30.2	96	28.7
Alcohol drinking
Ever	250	76.9	240	71.6
Never	75	23.1	95	28.4

^*The controls were selected so as to be frequency matched to the cases on the factors shown in the table.

Association of MDM2 variants with the risk of OSCC

The distributions of MDM2 genotypes among the controls were in agreement with the Hardy-Weinberg equilibrium (P = 0.835 for MDM2-rs2279744 and P = 0.413 for MDM2-rs937283; Table 2). Compared with the MDM2-rs2279744 TT genotype, a significantly reduced risk of OSCC was associated with the GT genotype (OR, 0.64; 95% CI, 0.44–0.90), GG genotype (OR, 0.60; 95% CI, 0.37–0.95), and the combined GT/GG genotypes (OR, 0.62; 95% CI, 0.45–0.87). However, compared with the MDM2-rs937283 AA genotype, there was a significantly increased risk of OSCC associated with the AG genotype (OR, 2.20; 95% CI, 1.51–3.22) and combined AG/GG genotypes (OR, 2.05; 95% CI, 1.42–2.95). For each polymorphism, the risk of OSCC may be reduced or increased with increasing numbers of variant alleles (P_trend = 0.010 for MDM2-rs2279744 and P_trend = 0.014 for MDM2-rs937283). However, the actual estimates of association did not reflect such significant trends.

Table 2.

Association of MDM2 genotypes and their combinations with OSCC risk in patients and controls recruited from The University of Texas MD Anderson Cancer Center and general hospitals in Houston between 1996 and 2002

Variables	Patients (325)		Controls (335)		P*	Adjusted OR (95% CI)†
	Number	Percentage	Number	Percentage
MDM2-rs2279744‡					0.010
TT (Ref.)	146	44.9	112	33.4		1.00
GT	132	40.6	165	49.3		0.64 (0.45–0.90)
GG	47	14.5	58	17.3		0.60 (0.37–0.95)
Trend test						P = 0.010
GT+GG	179	55.1	223	66.6		0.62 (0.45–0.87)
MDM2-rs937283§					0.005
AA (Ref.)	69	21.2	114	34.0		1.00
AG	209	64.3	169	50.5		2.20 (1.51–3.22)
GG	47	14.5	52	15.5		1.55 (0.93–2.60)
Trend test						P = 0.014
AG+GG	256	78.8	221	66.0		2.05 (1.42–2.95)
Combined MDM2 risk genotypesα					0.0002
Low risk group (Ref.)	57	17.6	95	28.4		1.00
Medium risk group	134	41.2	146	43.6		1.57 (1.03–2.39)
High risk group	134	41.2	94	28.0		2.48 (1.60–3.85)

^*Genotype distributions for MDM2-rs2279744, MDM2-rs937283, and combined risk genotypes between patients and controls.

^†ORs were adjusted for age, sex, smoking, drinking, and HPV16 serology.

^‡The observed genotype frequencies among controls were in agreement with Hardy-Weinberg equilibrium (P = 0.835).

^§The observed genotype frequencies among controls were in agreement with Hardy-Weinberg equilibrium (P = 0.413).

^αLow-risk group: MDM2-rs2279744 G carriers and MDM2-rs937283 AA genotype; Medium-risk group: MDM2-rs2279744 G carriers and MDM2-rs937283 G carriers or MDM2-rs2279744 TT and MDM2-rs937283 AA genotypes; and High-risk group: MDM2-rs2279744 TT and MDM2-rs937283 G carriers.

Because the two polymorphisms are not closely linked to each other (D' = 0.777, R² = 0.255 and P < 0.01) and the distribution of haplotypes between the patients and controls did not differ significantly (data not shown), we performed a combined analysis of both polymorphisms. We categorized the patients and controls into three groups based on the combination of the two MDM2 polymorphism genotypes, as follows: 1) the low-risk group (MDM2-rs2279744 G carriers and MDM2-rs937283 AA genotype); 2) the medium-risk group (MDM2-rs2279744 G carriers and MDM2-rs937283 G carriers or MDM2-rs2279744 TT and MDM2-rs937283 AA genotypes); and 3) the high-risk group (MDM2-rs2279744 TT and MDM2-rs937283 G carriers), as shown in Table 2.

The frequencies of the combined genotypes in low-, medium-, and high-risk groups were statistically significantly different between the patients and controls (P = 0.0002). Compared with the combined genotypes in the low-risk group, the combined genotypes in the medium- and high-risk groups were associated with a significantly increased risk of OSCC (OR, 1.57; 95% CI, 1.03–2.39 for medium-risk group and OR, 2.48; 95% CI, 1.60–3.85 for high-risk group) in the multivariable logistic regression analysis.

Interaction between HPV16 L1 seropositivity and MDM2 variants for the risk of OSCC

Table 3 showed that MDM2 variants borderline significantly or significantly modified the association between HPV16 serology and the risk of OSCC. Compared with individuals with MDM2-rs2279744 GT or GG genotypes and HPV16 L1 seronegativity, the risk of OSCC increased among those with the TT genotype and HPV16 L1 seronegativity (OR, 1.25; 95% CI, 1.06–2.19), GT/GG genotypes and HPV16 L1 seropositivity (OR, 2.81; 95% CI, 1.67–4.74), and TT genotypes and HPV16 L1 seropositivity (OR, 5.57; 95% CI, 2.93–10.6), respectively. Similarly, compared with individuals with MDM2-rs937283 AA genotypes and HPV16 L1 seronegativity, the risk also increased among those with AG or GG genotypes and HPV16 L1 seronegativity (OR, 1.85; 95% CI, 1.23–2.78), AA genotypes and HPV16 L1 seropositivity (OR, 2.29; 95% CI, 1.09–4.80), and AG or GG genotypes and HPV16 L1 seropositivity (OR, 6.88; 95% CI, 3.87–12.3), respectively.

Table 3.

Joint effect of HPV16 L1 seropositivity and MDM2 genotypes on risk of OSCC in the patients and controls recruited from The University of Texas MD Anderson Cancer Center and general hospitals in Houston between 1996 and 2002

		Patients (325)		Controls (335)		Adjusted OR (95% CI)*	Pint
		Number	Percentage	Number	Percentage
HPV16 status	MDM2-rs2279744
−	GT/GG(Ref.)	128	39.4	195	58.2	1.00	0.060
−	TT	97	29.8	98	29.2	1.25 (1.06–2.19)
+	GT/GG	51	15.7	28	8.4	2.81 (1.67–4.74)
+	TT	49	15.1	14	4.2	5.57 (2.93–10.6)
	MDM2-rs937283
−	AA (Ref.)	49	15.1	97	29.0	1.00	0.009b
−	AG/GG	176	54.1	196	58.5	1.85 (1.23–2.78)
+	AA	20	6.2	17	5.1	2.29 (1.09–4.80)
+	AG/GG	80	24.6	25	7.4	6.88 (3.87–12.3)
−	Low risk group (Ref.)a	42	12.9	82	24.5	1.00	0.005b
−	Medium risk group	93	28.6	128	38.2	1.49 (0.94–2.37)
−	High risk group	90	27.7	83	24.8	2.20 (1.36–3.57)
+	Low risk group	15	4.6	13	3.9	2.27 (0.98–5.26)
+	Medium risk group	41	12.6	19	5.6	4.39 (2.25–8.57)
+	High risk group	44	13.6	10	3.0	9.55 (4.32–21.1)

^*ORs were adjusted for age, sex, smoking, and alcohol drinking.

P_int: P values for interaction between the individual or combined MDM2 genotypes of the two polymorphisms and HPV16 serology.

^aLow-risk group: MDM2-rs2279744 G carriers and MDM2-rs937283 AA genotype; Medium-risk group: MDM2-rs2279744 G carriers and MDM2-rs937283 G carriers or MDM2-rs2279744 G TT and MDM2-rs937283 AA genotypes; and High-risk group: MDM2-rs2279744 TT and MDM2-rs937283 G carriers.

^bSignificant at the 5% level after adjusting for multiple comparisons.

To investigate the modifying effects of the combined genotypes of the two polymorphisms on the risk of OSCC associated with HPV16, we used the individuals in the low-risk group with HPV16 L1 seronegativity as the comparison group and found that the risk of OSCC increased among individuals in the medium-risk group with HPV16 L1 seronegativity (OR, 1.49; 95% CI, 0.94–2.37), the high-risk group with HPV16 L1 seronegativity (OR, 2.20; 95% CI, 1.36–3.57), the low-risk group with HPV16 L1 seropositivity (OR, 2.27; 95% CI,0.98–5.26), the medium-risk group with HPV16 L1 seropositivity (OR, 4.39; 95% CI, 2.25–8.57), and the high-risk group with HPV16 L1 seropositivity (OR, 9.55; 95% CI, 4.32–21.1), respectively. The modification effect may suggest a more-than-multiplicative interaction. Moreover, the testing for interaction remained significant for individual (MDM2-rs937283) or combined MDM2 risk genotypes after taking into account multiple comparisons.

Stratified analysis of interaction of HPV16 L1 seropositivity and MDM2 variants on OSCC risk by smoking/drinking status

We stratified the analyses of effect modification between HPV16 serology and MDM2 variants by smoking and drinking status (Table 4). Overall, we did not observe a similarly above-significant interaction between HPV16 L1 seropositivity and MDM2 variants on the risk of OSCC in each of these subgroups (including never smokers, ever smokers, never drinkers, and ever drinkers, respectively) except for MDM2-rs937283 polymorphism in ever smokers and ever drinkers. Although the interaction between MDM2-rs937283 polymorphism and HPV16 L1 seropositivity on the risk of OSCC was significant in ever smokers and ever drinkers, the interaction did not remain significant after adjusting for multiple comparisons. We did find that for each polymorphism, the OSCC risk for patients with HPV16 L1 seropositivity was much stronger in the never smokers than in the ever smokers and in the never drinkers than in the ever drinkers (Table 4). In addition, results were similar when we stratified the joint effect of HPV16 L1 seropositivity and the combined risk genotypes of both polymorphisms on the risk of OSCC by smoking and drinking status (data not shown).

Table 4.

Joint effect of HPV16 L1 seropositivity and MDM2 genotypes on risk of OSCC stratified by smoking and drinking status in the patients and controls recruited from The University of Texas MD Anderson Cancer Center and general hospitals in Houston between 1996 and 2002

HPV16 status	MDM2-rs2279744	Never smokers		Ever smokers		Adjusted OR (95% CI)*
		Patients (98)	Controls (96)	Patients (227)	Controls (239)	Never smokers	Ever smokers
−	GT/GG(Ref.)	29	54	99	141	1.00	1.00
−	TT	30	34	67	64	1.79 (0.89–3.62)	1.53 (0.99–2.36)
+	GT/GG	21	6	30	22	8.27 (2.84–24.0)	1.99 (1.07–3.70)
+	TT	18	2	31	12	25.2 (5.21–121.7)	3.89 (1.88–8.06)
Pint.						0.153	0.119

		Never drinkers		Ever drinkers		Adj. OR (95% CI)
		Cases (75)	Controls (95)	Cases (250)	Controls (240)	Never drinkers	Ever drinkers
−	GT/GG(Ref.)	28	53	100	142	1.00	1.00
−	TT	22	33	75	65	1.37 (0.64–2.90)	1.35 (1.08–2.53)
+	GT/GG	11	6	40	22	4.01 (1.25–12.8)	2.60 (1.44–4.70)
+	TT	14	3	35	11	13.0 (3.20–52.8)	4.54 (2.18–9.42)
Pint.						0.180	0.172

	MDM2-rs937283	Never smokers		Ever smokers		Adj. OR (95% CI)
		Cases (98)	Controls (96)	Cases (227)	Controls (239)	Never smokers	Ever smokers
−	AA (Ref.)	10	38	39	59	1.00	1.00
−	AG/GG	49	50	127	146	3.57 (1.56–8.19)	1.39 (0.87–2.24)
+	AA	8	2	12	15	19.5 (3.38–112.9)	1.25 (0.52–2.99)
+	AG/GG	31	6	49	19	23.9 (7.51–76.4)	4.28 (2.16–8.47)
Pint.						0.833	0.014a

		Never drinkers		Ever drinkers		Adj. OR (95% CI)
		Cases (75)	Controls (95)	Cases (250)	Controls (240)	Never drinkers	Ever drinkers
−	AA (Ref.)	9	26	40	71	1.00	1.00
−	AG/GG	41	60	135	136	2.39 (0.96–5.97)	1.76 (1.11–2.78)
+	AA	3	3	17	14	4.81 (0.67–34.4)	2.08 (0.92–4.70)
+	AG/GG	22	6	58	19	15.1 (4.26–53.5)	5.48 (2.84–10.6)
Pint.						0.176	0.031a

^*ORs were adjusted for age, sex, smoking, and alcohol drinking.

P_int.: P values for interaction between the individual MDM2 genotypes and HPV16 serology.

^anot significant at the 5% level after adjusting for multiple comparisons.

Because the difference in the tumor's HPV status between patients with oropharyngeal and oral cavity cancers may result from different etiologies at the two different sites, we further evaluated the modifying effect of individual or combined MDM2 risk genotypes for both polymorphisms on the association between HPV16 L1 seropositivity and the risk of OSCC stratified by tumor site (Table 5). We found that the modifying effects of MDM2 variants on the risk associated with HPV16 L1 seropositivity were more pronounced for oropharyngeal as opposed to oral cavity cancer.

Table 5.

Joint effect of HPV16 L1 seropositivity and MDM2 genotypes on risk stratified by tumor site in patients and controls recruited from The University of Texas MD Anderson Cancer Center and general hospitals in Houston between 1996 and 2002

HPV16 status	MDM2 genotypes	Oropharynx		Oral cavity
		Pa/Co (188/335)	Adj. OR (95% CI)*	Pa/Co (137/335)	Adj. OR (95% CI)*
	MDM2-rs2279744
−	GT/GG(Ref.)	50/195	1.00	78/195	1.00
−	TT	51/98	2.06 (1.29–3.29)	46/98	1.23 (0.78–1.94)
+	GT/GG	42/28	5.37 (3.00–9.60)	9/28	0.82 (0.36–1.85)
+	TT	45/14	12.6 (6.37–25.1)	4/14	0.79 (0.25–2.53)
	MDM2-rs937283
−	AA (Ref.)	25/97	1.00	24/97	1.00
−	AG/GG	76/196	1.62 (0.96–2.73)	100/196	2.18 (1.29–3.69)
+	AA	16/17	3.30 (1.44–7.54)	4/17	0.94 (0.28–3.12)
+	AG/GG	71/25	11.5 (6.03–22.1)	9/25	1.68 (0.67–4.20)
	MDM2 combined genotypes
−	Low risk group (Ref.)a	21/82	1.00	21/82	1.00
−	Medium risk group	33/128	1.11 (0.59–2.07)	60/128	1.98 (1.10–3.57)
−	High risk group	47/83	2.37 (1.29–4.36)	43/83	2.20 (1.18–4.12)
+	Low risk group	11/13	2.99 (1.15–7.77)	4/13	1.26 (0.36–4.41)
+	Medium risk group	36/19	7.30 (3.45–15.4)	5/19	1.05 (0.34–3.25)
+	High risk group	40/10	17.2 (7.26–40.7)	4/10	2.05 (0.57–7.42)

Pa/Co: Patients/Controls

^*ORs were adjusted for age, sex, smoking, and alcohol drinking.

^aLow-risk group: MDM2-rs2279744 G carriers and MDM2-rs937283 AA genotype; Medium-risk group: MDM2-rs2279744 G carriers and MDM2-rs937283 G carriers or MDM2-rs2279744 G TT and MDM2-rs937283 AA genotypes; and High-risk group: MDM2-rs2279744 TT and MDM2-rs937283 G carriers.

Discussion

In this study, we found that the two MDM2 polymorphisms, individually or in combination, significantly contributed to the risk of OSCC independent of HPV16 L1 seropositivity, and that the interaction between the individual or combined MDM2 genotypes of the two polymorphisms and HPV16 L1 seropositivity on risk of OSCC was borderline significant (MDM2-rs2279744) or significant (MDM2-rs937283 and the combined MDM2 genotypes). We also found that the joint effect of HPV16 L1 seropositivity and MDM2 variants on the risk of OSCC was higher in never smokers than in ever smokers and higher in never drinkers than in ever drinkers, but the similarly significant modifying effect of MDM2 polymorphisms on the risk of OSCC associated with HPV16 was not observed in each of these subgroups (including ever smokers, never smokers, ever drinkers, and never drinkers, respectively) after adjusting for multiple comparisons. Furthermore, we found that such an effect modification was particularly pronounced for oropharyngeal as opposed to oral cavity cancer. To the best of our knowledge, this is the first association study of gene-virus interaction and OSCC risk.

The strongest evidence of an association between HPV16 and risk of oral cancer has been from molecular epidemiologic (HPV16 serologic or tumor DNA status) studies showing a range of ORs from 3 to 60 (7, 25–28, 36, 37). A nested case-control study of 292 patients and 1568 controls within a prospective Scandinavian cohort of almost 900,000 subjects from whom blood was collected prior to any cancer diagnosis found that HPV16 L1 seropositivity was significantly associated with a 14.4-fold increased risk of oropharyngeal cancer (26). An American case-control study identified HPV as a risk factor for laryngeal cancer (OR, 3.0), similar to a recent Mexican case-control study that demonstrated an association (OR, 3.4) between HPV and risk of oral cancer after adjustment for other confounding factors, including age and smoking and drinking status (38, 39). The estimated association between HPV16 L1 seropositivity and oral cancer (OR, 3.17) in the current study was consistent with the above reported results. In addition, other studies also have demonstrated that HPV, particularly that of HPV16, was associated with an increased risk of oral cancer or oropharyngeal cancer independent of exposure to alcohol and tobacco (25, 36).

p53 is maintained at a low level during normal cell growth and is upregulated in response to cellular stress, such as DNA damage. The MDM2 protein is an important regulator of p53 and together their expression is dependent on a negative feedback loop (40). Expression of MDM2 is induced via p53 binding, leading to MDM2-facilitated repression of p53 expression. MDM2 is thought to exert its effect on p53 through the inhibition of p53-mediated transcriptional activity and direct mediation of p53 degradation by the proteasome (41). p53 is mutated in up to 50% of all cancers; but in tumors expressing wild-type p53, carcinogenesis occurs through a different mechanism (40, 41). One possibility for this mechanism is the overexpression of MDM2, which has been found in 5% – 10% of tumors (40, 41). It is plausible that these two MDM2 polymorphisms might affect MDM2 expression which, in turn, could affect p53 expression levels, thereby influencing cancer susceptibility.

Epidemiologic studies have evaluated the association between the MDM2-rs2279744 polymorphism and the risk of different types of cancer in recent years. This study and our previous study along with other recent studies have shown that the variant G allele was associated with a reduced risk of some types of cancer (34, 42). Moreover, one Asian study showed that the G allele had an inverse association with oral cancer risk and was associated with later onset of the disease (43). However, some published studies found no association between this polymorphism alone and risks of cancers in different populations (29, 42, 44), including a recent meta-analysis that showed no association between this polymorphism and the risk of breast cancers (including 5737 patients and 6703 controls) and colorectal cancers (including 1602 patients and 866 controls) (45).

In contrast, other studies have found that the G allele of the MDM2-rs2279744 polymorphism was associated with an increased risk of several types of cancer (46, 47). The interpretation of these studies is complex, and the possibility that the MDM2-rs2279744 polymorphism confers dissimilar risks for different tissue types as well as for the possible confounding effects of ethnic differences among the study groups should be taken into account. In fact, when stratified by ethnicity, the G allele appears to increase cancer risk in Asians but not in Europeans or Africans (47).

The finding of a synergistic effect of MDM2 polymorphisms on the risk of OSCC associated with HPV16 is consistent with the idea that HPVs and MDM2 may act synergistically through common pathways in p53 degradation in the development of OSCC. Moreover, such a synergistic interaction remained statistically significant for MDM2-rs937283 or combined MDM2 risk genotypes even after adjusting for multiple comparisons in the test for interaction (or heterogeneity test).

In vivo studies have provided evidence that p53 degradation by MDM2 and HPV E6 occurs through different mechanisms (48). When the normal MDM2-p53 ubiquitination pathway is inactive (e.g., following DNA damage), HPV E6 targets p53 for degradation. On the other hand, the p53 pathway could be inactivated by amplification or overexpression of MDM2 through either directly binding and masking p53's transactivation domain or acting as an E3 ubiquitin ligase for p53 degradation (40,41). Inappropriate excesses of MDM2 could lead to exaggerated silencing of p53, abrogating its protective tumor suppressor effects. Indeed, at least 5% – 10% of human tumors possess inappropriate MDM2 overexpression due to either gene amplification or to transcriptional and post-transcriptional mechanisms (40, 41).

Additionally, the regulation of HPV16 gene expression by the transcriptional regulator, HPV16's E2 protein, is critical for the viral life cycle. HPV16's E2 protein requires an active proteasome for its optimal transcriptional activator function. Mechanically, MDM2 interacts with HPV16's E2 to dramatically activate the HPV16 promoter, demonstrating that HPV16's E2 can actively recruit MDM2 to the HPV promoter for cooperative activity in HPV16's E2-activated gene expression (49). It is possible that these two MDM2 promoter polymorphisms may affect MDM2 expression levels and subsequently modify HPV16 expression.

In this study, with further stratification by smoking and drinking for each polymorphism, we found that the modifying effect of MDM2 variants on the risk of OSCC associated with HPV16 L1 seropositivity was higher in never smokers than in ever smokers and higher in never drinkers than in ever drinkers, and the size of such effect modification differs between the subgroups by smoking or drinking status.

Therefore, MDM2 risk genotypes of the two polymorphisms may play a role in the development of OSCC associated with HPV16 among never smokers and never drinkers in the general population. However, the modifying effect of MDM2 polymorphisms on the risk of OSCC associated with HPV16 was not statistically significant in each of these subgroups (including ever smokers, never smokers, ever drinkers, and never drinkers, respectively). This lack of significance could be either because there was no such interaction effect in these subgroups or because the small sample sizes in each substratum limited the statistical power to detect a significant interaction effect. Therefore, the significance and degree of such interaction in each subgroup needs to be further investigated in future studies with larger sample sizes. From these findings, we suggest that when assessing the modifying effects of MDM2 variants on the risk associated with HPV L1 seropositivity, smoking or drinking status might also need to be taken into account.

Additionally, we found that MDM2 variants synergize with HPV16 L1 seropositivity to tend toward the development of oropharyngeal cancer rather than oral cavity cancer. These results are in line with the notion that OSCC associated with HPVs is more commonly found in the oropharynx than in the oral cavity. This finding is in agreement with those of previous studies, in which a majority of oropharyngeal cancers were caused by HPV infection, whereas most nonoropharyngeal cancers were found to be caused by smoking and drinking (8, 31).

The finding that the modifying effects of MDM2 polymorphisms on the association between HPV16 L1 seropositivity and the risk of OSCC was higher in never smokers and never drinkers than ever smokers and ever drinkers may further support a role for MDM2 polymorphisms in the development of OSCC associated with HPVs (e.g., oropharyngeal cancer) rather than OSCC associated with non-HPVs (e.g., oral cavity cancer). However, these findings need to be further tested in studies with larger sample sizes.

Thus far, only one small case-control study in a Finnish population has examined the association between the MDM2-rs2279744 polymorphism and the risk of head and neck cancers, and no significant association was found (29). The study had small sample sizes and did not examine whether HPV and the MDM2-rs2279744 polymorphism jointly increased the risk of head and neck cancers. Our current study is the first one to include the effect of HPV16 L1 seropositivity along with smoking and drinking status on the risk of OSCC among those with different MDM2 polymorphisms.

Although our study had a relatively large sample size and minimizes potential confounding factors, there were several limitations. A possible selection bias cannot be ruled out because this was a hospital-based case-control study and the controls may not represent the same population from which the patients arose. In addition, stratified analyses included a limited number of individuals in each subgroup, so our results could be chance findings and should be confirmed in larger studies. Moreover, because our study included only non-Hispanic white participants, it is uncertain whether these results are generalizable to other ethnic populations. Finally, HPV16 L1 seropositivity actually indicated whether the patients and controls had had prior HPV exposure or were not, in fact, HPV16-related, and thus might not reflect the tumor's actual HPV16 status, leading to some misclassifications (false negatives for cases). However, an early study confirmed a reasonable concordance between HPV16 L1 seropositivity (52% for HPV16 L1 antibodies and 65.4% for HPV16 E6 or E7 antibodies) and HPV16 DNA positivity in tumor tissues (7). More importantly, the use of serologic status allowed for the inclusion of a cancer-free control group.

In conclusion, MDM2 risk genotypes and HPV16 L1 seropositivity may interact synergistically to increase susceptibility to OSCC, particularly in never smokers, never drinkers, and for oropharyngeal as opposed to oral cavity cancer. However, for more rigorous analyses of gene–environment or gene–gene interactions, and for more precise estimates of OSCC risk, future functional and molecular epidemiologic studies with larger sample sizes in different populations are warranted. In particular, since the p53 ubiquitination pathway involves many genes that may potentially interact with HPV oncoproteins, future studies should investigate the effects of HPV status on multiple genes in this pathway. Additionally, future studies are needed to elucidate the mechanisms behind the interactions between HPV16 L1 seropositivity and MDM2 variants in the development of OSCC.

Abbreviations

CI

confidence interval

HPV

human papillomavirus

OR

odds ratio

PCR

polymerase chain reaction

OSCC

oral squamous cell carcinoma