A Case-Control Study on the Effect of Apoliprotein E Genotype on Head and Neck Cancer Risk Free

Apolipoprotein E (ApoE) is a small glycoprotein that plays a major role in the blood clearance of cholesterol-rich particles, known as remnant lipoproteins (1). Specifically, apoE mediates high-affinity binding of the remnant lipoproteins to the members of the low-density lipoprotein (LDL) receptor family and the cell surface heparin sulfate proteoglycans. ApoE has many other functions, however, including tissue repair, immune response and regulation, biliar acid synthesis and metabolism, as well as cell growth, cell differentiation, metastasis, and angiogenesis (2). Two single-nucleotide polymorphisms within the coding regions for amino acids 112 and 158 of the apoE gene (19q13.2) are known, resulting in three different alleles (ϵ2, ϵ3, and ϵ4) and six apoE genotypes (three homozygotes ϵ4/ϵ4, ϵ3/ϵ3, and ϵ2/ϵ2, and three heterozygotes ϵ4/ϵ3, ϵ3/ϵ2, and ϵ4/ϵ2), each showing different receptor-binding abilities (3, 4). A meta-analysis reported a nearly linear relationship between apoE genotypes and the levels of total and LDL serum cholesterol (LDL-C) when the six genotypes are ordered as follows: ϵ2/ϵ2, ϵ2/ϵ3, ϵ2/ϵ4, ϵ3/ϵ3, ϵ3/ϵ4, ϵ4/ϵ4 (5). In fact, the ϵ4 allele is associated with increased levels of total and LDL-C when compared with the common ϵ3 allele, whereas the ϵ2 allele is associated with decreased levels (5).

The effect of the apoE genotypes on Alzheimer's disease (AD) and coronary heart disease (CHD) is well established. As for the AD, the ϵ4 allele is among the few established risk factors because of its higher in vitro binding of amyloid β peptides compared with the ϵ3 allele (6). About CHD, the ϵ2 allele is associated with a 20% decreased risk, probably resulting from an advantageous lipid profile due to a high efficient binding of apoE ϵ2 isoform with small, phosholipid-enriched high-density lipoprotein (HDL) (7).

In the past two decades, observational studies exploring the potential causative effect of cholesterol on cancer reported an inverse relationship between plasma cholesterol levels and cancer risk (8-11), with individuals carrying low serum total cholesterol levels at increased risk for cancer. Among the possible explanations is confounding due to “reverse causation” in case-control studies, as low cholesterol levels could simply be the effect of cancer rather than the cause (12). Another possible reason reported for an increased incidence of cancer among individuals with low cholesterol levels is that this situation actually counteracts the lower cardiac mortality associated with lower serum cholesterol levels (13). Generally speaking, studying the association between sequence variants of genes (e.g., apoE) known to be related to an intermediate phenotype (e.g., serum cholesterol level) and a disease has the advantage of being less prone to the confounding effect of lifestyle factors or disease stage that might affect the association between the intermediate phenotype and the disease itself. This approach has been called “Mendelian randomization” (14, 15), and it has been recently adopted to evaluate the association between cholesterol levels and cancer risk by using the apoE genotype as a surrogate for cholesterol levels. Results, however, are largely conflicting (7, 16-18). Trompet et al. (19) recently reported the results of a large cohort study measuring the apoE genotype, plasma cholesterol levels, and overall cancer incidence and mortality during a 3-year follow-up period. Results show that individuals with the low plasma cholesterol levels experience a ∼2-fold increased risk of cancer incidence and mortality. Subjects with the apoE ϵ2 allele, however, had no increased risk of cancer. In addition, a recent meta-analysis (20) of observational and experimental studies focusing on the effect of cholesterol-lowering medications (e.g., statins) on cancer risk showed that these drugs do not increase the risk of cancer over short time.

The role of apoE genotypes on head and neck cancer (HNC) etiology has not been specifically studied so far, as Trompet et al. (19) included only 14 HNC cases. Results of a cohort study, however, showed a significant inverse trend for the association between serum cholesterol level and oral/pharyngeal/esophageal cancers combined (21), and a case-control study focusing on plasma lipid profile patterns and HNC confirmed this inverse relationship (22). With the question of whether hypocholesterolemia is a predisposing factor for HNC or a preclinical stage of HNC still under debate, our hospital-based case-control study aims to overcome this issue by directly looking at the relationship between apoE genotypes and HNC as well as their interaction with potential effect modifiers.

Study design has been described elsewhere (23-25). Study participants were recruited among patients admitted to the teaching hospital “Agostino Gemelli” of the Università Cattolica del Sacro Cuore (Rome, Italy) from May 2002 to September 2009, and eligibility was restricted to Caucasian individuals born in Italy. Cases were selected from patients with untreated HNC admitted to the Department of Otorhinolaryngology with histologically confirmed HNC according to the International Classification of Disease (9^th revision, codes 140-149 and 161). Tumors were staged according to the International Union Against Cancer tumor-node-metastasis classification; 47.2% were staged I to II and 53.4% were classified T1 to T2 (tumor stage was missing for 25 patients, whereas tumor grade was missing for 16 HNC cases). Cases presented the following distribution according to tumor site: 62.8% laryngeal, 15.9% oral cavity, and 21.3% pharyngeal. Controls were selected from among cancer-free patients admitted to the same hospital during the same time period with a broad range of diagnoses. The study sample size comprised 417 cases and 436 controls, with a participation rate of 98% among cases and 93% among controls. Written informed consent was obtained from all study subjects, after which each subject provided a venous blood sample that was collected into EDTA-coated tubes. This study was done according to the Declaration of Helsinki and approved by the ethics committee of the Università Cattolica del Sacro Cuore.

DNA was extracted from the peripheral blood lymphocytes, and genotyping of apoE was done using RFLP. To carry out the genotyping of apoE, 20 ng of genomic DNA were RFLP amplified using oligonucleotide primers apoE-F (5′-TCCAAGGAGCTGCAGGCGGCGCA-3′) and apoE-R (5′-GCCCCGGCCTGGTACACTGCCA-3′). Initially, the RFLP reactions were denatured for 3 minutes at 95°C, followed by 35 two-step cycles consisting of 10 seconds at 95°C and then 10 seconds at 66°C. After the cycles were completed, a final extension of 5 minutes at 95°C was done. A 10-μL aliquot of each RFLP product was digested with 5 units of AflIII, and a separate 10-mL aliquot was digested with 5 units of HaeII. Both apoE ϵ2 and apoE ϵ3 alleles were cut with AflIII to yield products of 50 and 168 bp, whereas the apoE ϵ4 allele remains uncut at 218 bp. Using HaeII, both apoE ϵ3 and apoE ϵ4 alleles yield products of 23 and 195 bp, whereas the apoE ϵ2 allele remains uncut at 218 bp. The six possible genotypes were assigned by analyzing the patterns produced by the restriction digest. Quality controls were used at each RFLP by using sequenced DNA samples for all the possible six apoE genotypes sent by Seripa et al. (26).

Cases and controls were interviewed by trained medical doctors using a structured questionnaire to collect information on demographic data, cigarette smoking, drinking history, dietary habits, physical activity, and family history of cancer focusing on HNC family history specifically. Questions about lifestyle habits focused on the time period ending 1 year before diagnosis for cases and the year before the interview date for controls. Pack-years were calculated as years smoked multiplied by the current number (or previous number, for those who had quit) of cigarettes smoked per day divided by 20. Fruit and vegetable intake was categorized according to portions per week, given the lowest consumption (0-13 portions per week) as the reference category. Physical activity was classified as “high” if the individual has physical activity at least two times per week. Family history of cancer referred to parents, siblings, and offspring. The response rate for interview completion was 94.8% among cases and 93.8% among controls, with the exception of data relating to family history of cancer in any site (unknown in 24% of cases and 11% of controls).

Adjusted odds ratios (ORs) and relative 95% confidence interval (CI) estimates were calculated using a logistic regression model to measure the association between HNC and putative risk factors. Possible risk factors were considered to be confounders if the addition of that variable to the model changed the OR by 10% or more, and once a confounder of any estimated main effect was identified, it was kept in all models. Based on these criteria, we controlled for age, sex, alcohol intake, cigarette smoking (pack-years), fruit and vegetable intake, physical activity, and family history of cancer.

A χ² test of Hardy-Weinberg equilibrium for the three apoE alleles was done among controls. A gene-environment interaction analysis was done by using those carrying the homozygous wild-type genotype (ϵ3/ϵ3 related to the apoE3 isoform) as the reference group. In this analysis, the genotype was categorized as follows: presence of at least one apoE ϵ2 allele or presence of at least one apoE ϵ4 allele (the genotype ϵ2/ϵ4 was not included in either category), providing the other two apoE isoforms (apoE2 and apoE4).

In this analysis, age was categorized as ≤45 and >45 years old, alcohol consumption as drinkers/nondrinkers, smoking status as ever/never cigarette smokers, and physical activity as ≤1 and ≥2 times per week. As for fruit and vegetable intake, on the basis of the weekly combined number of fruit and vegetable portions, the following categories have been used: ≤13 portions per week, 14 to 23 portions per week, and ≥24 portions per week. The cutoff of 45 years old was chosen because it is the median age of early-onset HNC (27). Finally, to test for interaction between two exposure variables, the likelihood ratio test was used, with the individuals homozygous for wild-type genotype (ϵ3/ϵ3) and not exposed to the variables of interest used as the reference group.

General characteristics of our population made of 417 HNC cases and 436 controls are presented in Table 1. The percentage of males is higher among cases than controls (81.5% versus 58.3%). An increased risk of HNC is associated with high alcohol consumption (OR, 10.03; 95% CI, 4.91-20.51) and ever cigarette smoking (OR, 2.76; 95% CI, 1.61-4.71 for 1-24 pack-years of cigarette, and OR, 8.33; 95% CI, 4.97-13.96 for ≥25 pack-years). Consistent with previous findings (28, 29), high fruit and vegetable intake is associated with decreased HNC risk, with an OR of 0.22 (95% CI, 0.12-0.42) for those consuming at least two portions of fruit and vegetables per day (Table 1). Additionally, family history for HNC is associated with an increased HNC risk, with an OR of 4.69 (95% CI, 1.34-16.38), whereas no significant association was found in relation to a positive family history for other cancer types (OR, 1.47; 95% CI, 0.691-2.38). The apoE genotype frequencies in the control group were found to be in Hardy-Weinberg equilibrium (P > 0.05). As shown in Table 2, there are no significant differences in the distribution of the six apoE genotypes between the compared groups. Results were consistent even when cases were stratified according to tumor site (data not shown). From the multivariate analysis, however, results show a borderline significant 40% decreased risk of HNC (OR, 0.58; 95% CI, 0.31-1.05) among individuals carrying at least one apoE ϵ2 allele compared with those homozygous for the wild-type (ϵ3/ϵ3; Table 2). Results of the gene-environment interaction analysis are presented in Table 3. A significant interaction between apoE ϵ2 allele and age was observed (P for interaction = 0.01), most probably due to the opposite effect on HNC risk observed in the highest level category compared with the two others. Additionally, a 60% risk reduction of HNC risk (OR, 0.43; 95% CI, 0.21-0.90) was noted among females carrying the ϵ2 allele compared with those ϵ3 homozygotes. Lastly, a statistically significant interaction was found between alcohol intake and the ϵ4 allele (P for interaction = 0.044), with 74% significant reduced risk for HNC among never drinkers carrying the ϵ4 allele, and a 2-fold increased risk (OR, 2.06; 95% CI, 0.95-4.48) among ever drinkers with the ϵ4 allele, with respect to those nondrinkers with the ϵ3/ϵ3 genotype.

This hospital-based case-control study including 417 cases and 436 controls evaluated for the first time the effect of three isoforms of the apoE gene, and their interactions with selected demographic and lifestyle factors, on HNC risk. Results show a 40% borderline significant protective effect of the ϵ2 allele, which is related to low serum cholesterol level when compared with the wild-type ϵ3 allele. This effect was particularly strong among females. Additionally, results show that alcohol drinkers carrying at least one ϵ4 allele have a 2-fold increased risk of HNC compared with nondrinkers with the ϵ3/ϵ3 genotype.

ApoE is a small glycoprotein playing a major role in remnant lipoprotein clearance (1), and it is coded by the highly polymorphic apoE gene. The allele distribution among Caucasians for the most common ϵ3 allele (Cys112;Arg158) ranges from 70% to 85%, whereas the frequency of the ϵ4 (Arg112;Arg158) and ϵ2 (Cys112;Cys158) alleles ranges from 10% to 20% and 5% to 10%, respectively (3). A different affinity for the apoE-binding receptors of cell surfaces and its effect on intestinal absorption of cholesterol can be detected based on the apoE genotypes, resulting in typically lower serum LDL-C levels among those carrying the ϵ2 allele and increased LDL-C levels for those carrying the ϵ4 allele (30). The role of apoE isoforms in CHD and AD has been widely investigated, with results showing that the ϵ2 allele provides a 20% decreased risk of CHD (5) compared with the homozygous ϵ3 genotype, and that the ϵ4 allele increases the risk of AD by 3 times in heterozygotes and by 15 times in homozygotes (31, 32). These evidences are in line with those from observational studies relating cholesterol levels with CHD and AD, as cohort studies showed an increased risk of both diseases among those with high serum cholesterol levels (33, 34).

The effect of apoE genotypes has been investigated in relation to breast, colorectal, biliary tract, prostate cancer and hematologic malignancies (7, 16-18, 35), with conflicting results. Trompet et al. (19) recently addressed the relationship between cholesterol and cancer by studying both the effect of plasma cholesterol levels and apoE isoforms on overall cancer risk during a 3-year follow-up study in a large elderly cohort (19). Results show an inverse relationship between cancer incidence or mortality and cholesterol levels, whereas no effect was shown for the apoE alleles. Even with some limitations on the selected cohort, the authors conclude by suggesting that there is a substantial lack of causal effect of cholesterol on cancer risk, and state that the apparent contradiction between results relating plasma cholesterol levels and cancer risk might be indicative of a preclinical cancer stage involving the increased uptake of cholesterol from the blood for the cell growth and proliferation, lowering cholesterolemia before the clinical cancer diagnosis (19).

To our knowledge, the effect of apoE genotypes on the risk of HNC has not been previously investigated. To overcome the issues related to reverse causation or confounding by lifestyle factors, we decided to clarify the role of cholesterol levels on HNC by using a Mendelian randomization approach, in other words by studying directly the effect of the apoE genotypes on a large series of Italian HNC cases and controls. Our results show a borderline 40% decreased risk of HNC associated with the ϵ2 allele. Because considerable research suggests that ϵ2 carriers have a lower serum cholesterolemia than non-ϵ2 carriers, this finding seems not to be in line with the previously reported inverse relationships between serum cholesterol levels and HNC risk from the two published observational studies (21, 22, 36). Our results are also partly discordant with those from the aforementioned large cohort study by Trompet et al. (19), which found no association between apoE genotypes and overall cancer risk. By avoiding an overinterpretation of our results, however, we wish to keep in mind that apoE has many other functions beside its well-known role in lipid metabolism, which are potentially involved in cancer risk. ApoE has also been shown to be involved in tissue repair, inflammatory and immune response, cell growth, and angiogenesis (2), and also shows antioxidant properties (37). Of importance, apoE protein has certain antioxidative properties, with decreasing antioxidant activity in the order ϵ2 > ϵ3 > ϵ4 alleles. Even if the molecular mechanisms responsible for the antioxidant properties of apoE have not yet been fully explained, several studies have examined the mechanisms through which apoE genotypes could affect the oxidative status-dependent mediators or biomarkers of oxidative stress (38-40). ApoE ϵ2 carrier smoking individuals, who are exposed to nicotine, an important source of oxidative stress, have an ∼30% higher total antioxidant status compared with apoE ϵ3 carriers, measured as the capacity to inhibit the peroxidase-mediated formation of the 2,2-azino-bis-3-ethylbenzothiazoline-6-sulfonic acid radical, whereas apoE ϵ4 subjects show a 30% increased oxidized LDL (38). Oxidative stress as a consequence of an imbalance between the formation and inactivation of reactive oxygen species might be involved in the pathogenesis of HNC. A recently published study (41) shows a wide magnitude of oxidative stress in HNC cases compared with healthy individuals, as shown by elevated levels of lipid peroxidation products and depletion of enzymatic and nonenzymatic antioxidants. In view of all these findings, the results of our study, which reports a borderline protective effect of the apoE ϵ2 allele on HNC, might be explained by the better antioxidant properties of ϵ2 allele compared with the ϵ3 or ϵ4 alleles, and this evidence can be especially true for HNC, whose pathogenesis is strongly affected by smoking-related oxidative stress. If our model holds true, we would expect an interaction between apoE ϵ2 allele and smoking status; however, the limited power of our interaction analysis may have obscured it.

According to our results, a 60% reduction of HNC risk was noted among females carrying the ϵ2 allele compared with ϵ3 homozygotes, probably due to the combined presence of the ϵ2 allele and the well-known lower rate of oxidative damage exhibited by females (42). Last, our analysis shows a statistically significant 74% reduced risk for HNC among never drinkers carrying the ϵ4 allele and a 2-fold increased risk for ever drinkers carrying the ϵ4 allele, if compared with ϵ3/ϵ3 nondrinkers. Acknowledging the recent finding of a direct effect of alcohol drinking on telomere shortening, a marker of chromosome instability associated with cancer risk (43), this result could be attributable to the combined presence of the ϵ4 allele linked to a higher oxidative stress and the direct effect of alcohol intake on DNA mutations.

Some limitations of the study should be considered to correctly interpret our results. First, on the basis of the prevalence of the apoE alleles in our control population, this study has a priori 80% power to detect an OR of 0.67 for the effect of the apoE ϵ2 allele (at 5% significance level). Therefore, due to power limitations, the 42% protective effect associated with the ϵ2 allele should be confirmed by increasing our sample size. The sample size of the study also limits the possibility to explore gene-environment interactions. Second, as in all case-control studies, there is the possibility of information bias leading to misclassification of the exposure, given that information on lifestyle habits was self-reported and referred to at least 1 year before the onset of the disease. Last, data about human papillomavirus infection and serum cholesterol levels were not available in our study population. Despite these limitations, our study provides for the first time evidence of a possible protective effect of the ϵ2 allele against HNC, and may suggest a different role of apoE in carcinogenesis. To overcome these limitations and clarify the possible effect of apoE genotype in HNC etiology, a coordinated research effort within the International Head and Neck Cancer Epidemiology consortium (44) including several case-control studies is desirable.

No potential conflicts of interest were disclosed.

We thank Benedetto Simone for the linguistic revision of the final manuscript.

Grant Support: Università Cattolica del Sacro Cuore (D1 projects 2008).

Author contributions: E. De Feo, G. Ricciardi, and S. Boccia conceived the study and implemented the final draft of the manuscript; E. De Feo, J. Rowell, and N. Nicolotti did the statistical analysis and wrote the paper; G. Cadoni, A. Giorgio, D. Arzani, and G. Paludetti recruited HNC cases and controls; D. Arzani and R. Amore processed blood samples and genotyped for apoE.

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.