Major Histocompatibility Complex Class II Polymorphisms and Risk of Cervical Cancer and Human Papillomavirus Infection in Brazilian Women¹

Cervical cancer is one of the leading causes of death in women in developing countries. Incidence and mortality rates for invasive cervical cancer in Brazil are 27.6 and 7.7 cases per 100,000 women,respectively, and some of the highest rates are found in the northern and northeastern regions of the country (1). Epidemiological studies have indicated that infection with high-risk HPV³types is the main risk factor for the development of malignant lesions in the uterine cervix. HPV DNA has been found in over 90% of cervical cancer specimens, in which HPV-16 is the most prevalent HPV type detected (2). Infection with oncogenic HPV types is frequent among sexually active women, with prevalence rates ranging from 15–40% although most of these infections are transient (3). Development of malignant cervical lesions occurs in a small proportion of infected women that harbor persistent infections with oncogenic HPV types (4, 5).

Experimental and clinical evidence demonstrate that the immunological and genetic backgrounds of the host play an important role in the outcome of HPV-associated diseases. Certain polymorphisms for MHC genes have been associated with genetic susceptibility to cervical cancer (6, 7). Many of the attributed polymorphisms are located in HLA genes found in the MHC region that encodes for cell surface class I (HLA-A, -B, and -C) and class II(HLA-DR, -DQ, and -DP) molecules. HLA class I molecules, which are expressed in most nucleated cells, present antigenic peptides to CD8⁺ T cells. Expression of HLA class II molecules is restricted to professional antigen-presenting cells. Thymus-derived CD4⁺ lymphocytes (T helper cells) recognize the antigenic peptides bound to HLA class II molecules through their antigen-specific TCRs (8).

Previous studies conducted in different populations have obtained disparate results regarding MHC polymorphisms and susceptibility to cervical cancer, although most studies were restricted to HLA class II alleles. Elevated risks for cervical cancer or malignant precursor lesions have been detected for HLA-DQB1*03 (DQ3)allele carriers among different ethnic groups (6, 9, 10, 11, 12). When focused on specific combinations of alleles or haplotypes, such as DRB1*15-DQB1*0602 and DRB1*04-DQB1*0302, studies have reported both positive (7, 13, 14, 15, 16) and negative (17) risk associations. Lower frequencies of women with DR13 alleles and haplotypes have been found among women with cervical cancer as compared with healthy women (7, 13, 17, 18). Furthermore, these associations seem to depend on HPV infection status, predominantly that for HPV-16 (7, 13, 15).

To investigate the association pattern between HLA class II genes,cervical cancer, and HPV infection in Brazil, we analyzed polymorphisms of HLA-DQA1, -DQB1, and -DRB1 genes with their promoters (QAP and QBP for HLA-DQA1 and -DQB1, respectively) in a case-control study of women from the city of João Pessoa. This region of Brazil has one of the world’s highest incidence rates of cervical cancer and has an admixed population with a distinct ethnic composition to which all major ethnic groups contributed (19).

The diversity found in this population allowed us to investigate disease susceptibility for a wide range of MHC alleles and haplotypes and the effects of linkage disequilibrium patterns in the MHC region on the risk of cervical cancer (20).

We randomly selected 161 of 344 cases and 257 of 683 controls from an epidemiological study of cervical cancer and HPV infection conducted in northeastern Brazil (21). The study was approved by the local ethics committee, and all subjects consented to participate. Controls were women with normal or nondysplastic Pap smears who were randomly selected from a city-wide cervical cancer screening program carried out at the Napoleão Laureano Hospital (João Pessoa,Brazil). Cases included women with cervical cancer admitted at the same hospital for surgery or women who had histopathological confirmation of invasive SCC between 1986 and 1990. Exclusion criteria were any prior treatment for cervical disease or hysterectomy. Epidemiological data from all subjects were obtained during a standardized interview carried out by a trained nurse using a structured questionnaire. Information was obtained regarding sociodemographic variables, sexual behavior,smoking and alcohol consumption, diet, personal hygiene habits, access to health care, and reproductive history. Subjects’ ethnicity was identified by the categories (white, mulatto, and black) individually designated by the interviewer.

An ecto- and endocervical cell specimen was collected from each control by using a cytobrush, and tumor biopsies were obtained from all cases. Pap smears and tumor biopsies were prepared and submitted for pathological confirmation. The remaining cells or tissues were stored at 4°C in sterile buffered saline and shipped in dry ice to the Ludwig Institute for Cancer Research (São Paulo, Brazil) for DNA extraction and HPV testing (21). An aliquot of the purified DNA was shipped to the Human Molecular Genetics Laboratory (Curitiba, Brazil) for HLA typing.

HPV detection and typing involved PCR-based amplification of a 450-bp fragment of the viral L1 gene with primers MY09 and MY11 (22). A 268-bp fragment of the β-globin gene was also amplified to ensure DNA quality. A negative control tube containing all PCR reagents except template DNA was included in all PCR reactions. As positive controls, we used plasmid DNA with the target HPV sequence inserted and DNA from HeLa cells that harbor HPV integrated into the host genome. The PCR products were dot blotted onto nylon membranes and hybridized with specific ³²P-labeled probes that discriminate between HPV-6, -11, -16, -18, -26, -31, -33, -35, -39,-40, -42, -45, -51–59, -66, and -68 (22).

The protocols, primers, and probes for PCR and hybridization reactions used in this study followed the 12^thInternational Histocompatibility Workshop and Conference recommendations (23). This high-resolution approach allows for discrimination of most alleles of the genes analyzed.

For this study, 257 controls and 161 cases were HLA typed. QAP and QBP typing was done in 320 samples (92 cases and 228 controls). We amplified the polymorphic second exon of the HLA-DQA1, -DQB1, and -DRB1 genes and the upstream regulatory regions of the HLA-DQA1 and -DQB1genes with PCR. HLA alleles and promoter variants were discriminated by hybridization with sequencespecific oligonucleotide probes. PCR reactions were done using MJ Research PTC-200 or Amersham Pharmacia TC-341 thermocyclers. One-tenth (5 μl) of the amplified products was subjected to electrophoresis on 1.5% agarose gels. Gels were then stained with 0.1 μg/ml ethidium bromide solution and visualized under UV light.

The PCR products (1.5 μl from each sample) were spotted on positively charged nylon membranes (Hybond N⁺; Amersham,Buckinghamshire, United Kingdom). Membranes were soaked for 5 min in 0.4 N NaOH for denaturation and soaked for 10 min in 3× SSPE [20×SSPE (pH 7.4) = 3 m NaCl, 200 mmNaH₂PO₄·H₂O,and 20 mm EDTA] for neutralization. UV radiation was used to cross-link the DNA. The membranes were prehybridized for 30 min in 10 ml of TMACl solution [3 m TMACl, 50 mm Tris(pH 8.0), 0.1% SDS, and 2 mm EDTA (pH 8.0)] at 54°C. The appropriate 5′-biotinylated oligonucleotide probe (40 pmol) was added to the prehybridization solution and incubated for 1–16 h. After hybridization, membranes were washed once in 2× SSPE and 0.1% SDS for 10 min at room temperature and washed once in TMACl solution for 20–30 min at 59°C. The membranes were incubated for 30 min in 10 ml of conjugated horseradish peroxidase-streptavidin (0.4 μg/ml) solution. Hybridized probes were detected by color reaction with tetramethylbenzidine after a 3-min incubation in a solution containing 5 ml of citrate buffer (pH 5.0), 5 μl of 30% (w/w) hydrogen peroxide, and 150 μl of tetramethylbenzidine (2 mg/ml; dissolved in water). The reaction was interrupted with distilled water.

By amplifying only the second exon of HLA genes, we could not discriminate between some alleles. For this reason, the DQA1*0101/DQA1*0104, DQB1*0201/DQB1*0202, and DQB1*06051/DQB1*0609 alleles were grouped as DQA1*0101/04, DQB1*0201/02, and DQB1*06051/09,respectively. The QBP 6.2 and 6.3 promoter variants were distinguished using amplification with specific primers (23).

When required for analysis, the high-resolution typing results were translated into groups of alleles in a way that resembles the detection power of HLA serology (e.g., DRB1*1501,DRB1*1502, and DRB1*1503 alleles were grouped as DRB1*15). HLA haplotypes were inferred from known linkage disequilibrium patterns among HLA-DRB1, -DQA1, and -DQB1 alleles (24). Initially, we analyzed only DRB1-DQB1 haplotypes. Complete DRB1-DQA1-DQB1haplotypes were investigated when there was evidence of association for a DQA1 allele.

The proportions of cases and controls bearing a certain HLA allele,allele group, or haplotype were ascertained. Comparisons of exposure profiles between cases and controls were done byχ ² and Student’s t tests for independent samples. The magnitude of association between the HLA markers and the occurrence of cervical cancer or HPV infection was measured by OR and respective 95% CI. Crude and adjusted risk associations were estimated by unconditional logistic regression using the statistical program SPSS version 9.0 (SPSS, Chicago, IL). Initially, crude ORs and 95% CIs were calculated. Confounding by other exposure variables was ascertained using a change-in-estimate criterion for the crude observations. For the first level of adjustment, age and ethnic group were considered. ORs were also derived after additionally controlling for the following empirical confounders: religion; marital status; level of education; income; and consumption of alcoholic beverages other than beer or sugar cane distillates. These variables appeared to be associated with classification for ethnic group in our population and were included as confounders to minimize the potential bias in the point estimates for the association between HLA polymorphisms and cancer.

Sociodemographic information was collected for all subjects,whereas HLA typing was done for 418 (161 cases and 257 controls) of 1027 samples. As shown in Table 1, the distributions of most sociodemographic variables between non-HLA-typed and HLA-typed samples were similar. Considering the variable ethnic group, we observed that the proportion of cervical cancer specimens from black patients who were HLA typed was lower than that among non-HLA-typed samples (3.1% and 9.3%, respectively). However, the distribution of ethnic groups between HLA-typed cases and controls was statistically comparable (P > 0.05,χ ² test), with a slightly higher proportion of white women among controls.

In the HLA-typed group, cases were significantly older than controls. The mean age was 52.1 years for cases and 39.1 years for controls(P < 0.001, Student’s t test), and age ranged from 16–84 years for controls and from 27–84 years for cases. The majority of cases and controls had low income levels, although a higher educational level was attained by controls (P <0.001, χ² test). Whereas controls were mainly from urban zones (71.2%), most cases came from rural areas (51.6%).

To verify whether specific HLA-DQA1, -DQB1, or -DRB1 alleles were associated with risk of SCC, we calculated crude and adjusted ORs and 95% CIs (Tables 2 and 3). For the HLA-DQA1 gene, the proportion of samples carrying the DQA1*0101/04 allele was lower in cases than in controls. The ORs also remained similar and marginally significant even after adjustment for all covariates. The QAP 1.3 promoter variant was also found in a smaller proportion of cases than controls. The OR,adjusted for age and ethnic group, was 0.48 (95% CI, 0.24–0.98), but the association lost precision when adjusted for all variables (OR,0.54; 95% CI, 0.25–1.16). QAP 1.3 was linked with the DQA1*0101/04 allele in DR10, DR12, and DR14 haplotypes and with DQA1*0103 only in DR13 haplotypes. In DR1 haplotypes including the DQA1*0101/04 allele, the promoter variants were QAP 1.1, QAP 1.4, or QAP 1.5. We did not find significant deviations in OR from unity for the other DQA1 alleles or QAP variants.

The ORs and 95% CIs for HLA-DQB1 alleles are also shown in Table 2. The DQB1*05 group, including DQB1*0501, DQB1*0502, and DQB1*0503 alleles, was significantly associated with a decreased risk for SCC. Each individual DQB1*05 allele, as well as the QBP 5.12 variant found only in the promoter region of DQB1*05 alleles, showed ORs below the null value, although deviations were not statistically significant. Analysis of risk associations for other DQB1 alleles or QBP promoter variants did not show any significant relationships. ORs could not be estimated for DQB1*0601 because this allele was present only among cases,and ORs could not be estimated for DQB1*06051/09, which was detected only among controls.

The proportions of samples carrying DRB1 alleles are shown in Table 3. We found increased risks of SCC for the HLA-DRB1*15 allelic group. In this study, the DRB1*15 group included DRB1*1501, DRB1*1502, and DRB1*1503 alleles. For all of these alleles, we obtained ORs above unity. The magnitude of associations with DRB1*15alleles remained similar with both partial and full adjustment. DRB1*0408 and DRB1*1103 alleles showed substantial susceptibility for cancer, but estimates were relatively imprecise. The allele DRB1*1302 was found in a smaller proportion of cases, producing a protective risk association for SCC. A significant negative association between DRB1*0101 and SCC was also observed after complete adjustment.

The DRB1*1501 and DRB1*1503 alleles displayed strong linkage disequilibrium with DQB1*0602, whereas DRB1*1502 was accompanied primarily by DQB1*0601or DQB1*0502 alleles. Therefore, to separate out the effects of linkage disequilibrium between DRB1 and DQB1genes, we investigated associations for the most frequent DRB1-DQB1 haplotypes found in our population (Table 4). Adjustment was done only for age and ethnic group because differences in point estimates after full adjustment were small. As expected from our initial results for the HLA-DRB1*15 group and from the level of linkage disequilibrium associated with the DQB1alleles, a significant relationship was found between DRB1*15-DQB1*0602 haplotypes and increased risk for SCC. Because linkage disequilibrium between DRB1*15 and DQB1*0602 was not absolute, a separate analysis of these alleles was possible. The magnitude of association for DRB1*15 more than doubled, increasing from 2.04 (95% CI,1.15–3.61) to 4.56 (95% CI, 0.50–41.4) in the absence of DQB1*0602. In contrast, the adjusted OR for DQB1*0602 was reduced substantially to 0.46 (95% CI,0.13–1.66) when DRB1*15 was absent (DQB1*0602was also found in DRB1*11 haplotypes). Similar tendencies were observed for DQB1*0602 in the absence of alleles DRB1*1501 and DRB1*1503, with adjusted ORs decreasing to 1.49 (95% CI, 0.77–2.90) and 1.05 (95% CI,0.54–2.07), respectively (data not shown). No significant associations were found for the other DRB1-DQB1 haplotypes, even when DQA1 alleles were also considered.

To verify if the associations detected were dependent on HPV type, we estimated ORs for HLA-DQA1, -DQB1, and -DRB1alleles considering only HPV-16-positive cases (Table 5). This was the most frequent HPV type in our sample (69.6% of cases). The association pattern for HPV-16-positive cases did not diverge from that seen in all subjects. Positive associations were found with the DRB1*15 group and the DRB1*1503 allele. We observed a slight increase in relative frequency of DRB1*03alleles that remained marginally significant after adjustment for additional variables. No significant negative associations were found in this analysis.

We analyzed the distribution of HLA-DRB1, -DQA1, and -DQB1 alleles and haplotypes in controls with respect to HPV status (Table 6). Twenty-six of the 257 HLA-typed controls were excluded because the samples were inadequate for HPV testing. A higher proportion of HPV-positive samples carried DRB1*15 alleles compared with HPV-negative samples. At the allele-specific level, significant positive associations were found for DRB1*1503, DQB1*0602, and DRB1*0405 alleles, although a trend in risk was also observed for DRB1*0411. No significant associations were detected regarding haplotypes.

Of the 231 controls included in the analysis, 21 harbored high-risk HPV DNA. We further examined the risks associated between HLA alleles and HPV types after grouping the latter on the basis of their oncogenic potential (Table 6). The OR for DRB1*15 alleles decreased in magnitude, mainly for DRB1*1503 but increased for DRB1*1501. Risk associations of HPV infection for women carrying the DRB1*0411 allele were also higher for oncogenic types. Due to the small number of HPV-positive controls, an individual analysis of different HPV types was not possible.

We analyzed the proportions of HLA class II DQA1, DQB1, and DRB1 polymorphisms in 161 patients with SCC and 257 healthy women from a high-incidence area in northeastern Brazil. We found a significant association between HLA-DRB1*15 group alleles and an increased risk for SCC, as well as for the presence of HPV DNA in controls. The strongest associations involved DRB1*1503, although all DRB1*15 alleles tended to be more frequent among cases. Associations with allele DRB1*1503 have not been reported previously. This allele is prevalent in African populations but not in Caucasian populations, where DRB1*1501 is more frequent (24). Northeastern Brazilian populations are characterized by extensive racial admixture, with the majority of subjects being of Caucasian or African origin. Previous reports have described an increased risk for SCC associated with allele DRB1*1501 or haplotype DRB1*1501-DQB1*0602(7, 13). On the other hand, in a large study of women from the United States (17), haplotype DRB1*1501-DQB1*0602 was associated with a decreased risk for high-grade squamous intraepithelial lesions. These studies were conducted in Hispanic and Caucasian populations and did not reveal an association with DRB1*1503.

Positive associations between DQB1*0602 and SCC or squamous intraepithelial lesions were also found by studies investigating only the DQB1 gene (14, 16). In all populations studied, the DQB1*0602 allele was in strong linkage disequilibrium with DRB1*1501 and DRB1*1503(7, 13, 24), making the individual analysis of the effect of each allele very difficult. In our sample, some individuals carried either the DRB1*15 or DQB1*0602 allele but not both. The analysis of such informative individuals revealed that the adjusted risk attributable to DRB1*15 alleles was increased 2-fold in haplotypes that did not carry DQB1*0602. The DQB1*0602-associated risk decreased roughly 5 times in the absence of DRB1*15 alleles, although the values were not statistically significant due to the small number of subjects with such haplotypes. Nevertheless, these results indicate that DQB1*0602 could not have made an important contribution to the DRB1*15-associated risk for SCC in our population. The presence of the DQB1*0602 allele was also associated with increased HPV positivity among controls; thus, the role of this allele in HPV infection susceptibility cannot be discounted. However, most of the DQB1*0602 HPV-positive controls were also DR15 positive; therefore, the independent effects of each allele were not estimable.

Decreased risk for SCC associated with the presence of DR13alleles has been reported in several studies (7, 13, 18). We found moderate protection for SCC associated with the DRB1*13 allelic group, of which the main effect was mediated by DRB1*1302. The strongest negative associations in our study were between DRB1*0101, DQA1*0101/04, and DQB1*05 alleles and SCC. These DQA1 and DQB1 alleles present positive linkage disequilibrium with DR1, DR10, DR12, and DR14 group alleles. A decreased susceptibility for cervical dysplasia associated with the haplotype DRB1*0101-DQB1*0501 has been described in British women (25).

An increased risk for SCC associated with DQ3 alleles has also been reported for different populations (6, 11, 12). However, we did not find any association with the DQB1*03group or individual DQB1*03 alleles or haplotypes. Although associations detected in different populations frequently involve the same HLA groups, there is no consensus about which HLA alleles or haplotypes contribute to the risks of SCC in any particular population. HLA allele distributions can be population specific; therefore, to reduce colinearity and control for social risk factors that may have influenced development and treatment of SCC, we used a broader definition of ethnicity reflecting socioeconomic status. It is possible, however, that some residual confounding by ethnicity may have remained due to misclassification. Different patterns of association have arisen even when analyses were carried out in samples from the same population (9, 13, 14, 26). Contradictory results could be an indication that such populations have intrinsic features characterizing their risk and that interaction of environmental and host factors might be different. However, we cannot refute the hypothesis that such controversial results originate from biases in case-control selection, lack of appropriate adjustments, the effect of HPV infections, histopathological misclassifications in outcome, or chance associations. Other genes closely linked to HLA genes may be involved in the genetic susceptibility to papillomavirus infections and SCC development and could be responsible in part for the discrepancies among different studies. Although some misclassification may exist in the definition of haplotypes following reported linkage disequilibrium between DRB1, DQA1, and DQB1alleles (23), such instances would be rare because coding was based on three loci, and the degree of linkage disequilibrium between these loci is strong. The patterns of haplotype distribution did not vary from those observed in other Brazilian populations (24).

HLA class I genes have been less investigated than HLA class II genes with regard to susceptibility to HPV infection and SCC. Nevertheless,some studies have demonstrated associations with particular HLA class I alleles (10, 17, 27). Similarly, several polymorphic non-HLA genes located in the MHC class III region that also participate in immune responses have been related to the occurrence of several diseases (20, 28) and are emerging as candidate genes for SCC and HPV infection susceptibility. As more risk associations are identified for combinations of MHC genes, we may be able to explain some of the discrepancies observed regarding HLA-related susceptibility among individuals.

HPV infection outcomes result from a combination of features intrinsic to some HPV types (particularly those of the high-risk types) and interactions between HPV and the host. It is likely that only women persistently infected with oncogenic HPV types are prone to develop malignant cervical lesions (3, 4). Furthermore,persistent HPV infections associated with a high viral load are considered to be risk factors for developing cervical lesions (4). We found that among controls HPV infection was also positively associated with DRB1*15 alleles. Our analysis,however, is based on HPV measurements taken at a single point in time,and therefore the interactive relationship between HLA polymorphisms,HPV infection persistence, and development of malignant lesions cannot be confirmed. The fact that DRB1*15 is associated with increased risks for both SCC and HPV positivity in this population is an indication that HLA polymorphism may be important in directing the course of HPV infections. Genetic susceptibility may influence persistence in HPV infections and high viral loads, augmenting the risk for development of malignant lesions of the cervix. To examine this hypothesis we are currently looking at the effect of HLA class II polymorphisms on the incidence of cervical lesions in a large follow-up study (29).

Appropriate interactions among MHC molecules, antigenic peptides, and TCRs are crucial for T-cell activation. MHC molecules encoded for by different genes and alleles have distinguishable properties in antigen binding, and the difference in combinations of individual MHC genes and TCR variability is important in determining the level of immune response against pathogens and tumor cells. Modifications in viral epitopes may affect proper antigenic recognition by the immune system and impair viral infection clearance. Variants of HPV-16 have been identified in a large proportion of HLA-B7-positive cancer patients who carry a point mutation in the E6 protein that alters an HLA-B7 binding epitope (30). Furthermore, infection rates for different HPV variants also differ between populations (31), and their interaction with specific HLA alleles may explain some of the differences in cancer risk observed among these populations. For example, HPV-16 and HPV-18 E2 variants, which cosegregate with E6 gene variants at the 350 nucleotide position, were detected more frequently in DRB1*0401-DQB1*0301 and DRB1*1101-DQB1*0301carriers (32). However, other authors have failed to demonstrate a correlation between HLA class I and class II genes and HPV variants based on the variability of E6 and E7 sequences (27).

There is strong evidence that cellular immune response is essential in HPV infection clearance. A higher prevalence of HPV in malignant anogenital lesions has been observed in patients infected with HIV (33). Furthermore, higher levels of cytokines that stimulate the cellular immune response, such as IFN-γ, have been correlated with anogenital wart regression (34) and other less severe HPV-associated diseases (35). In mice, there is experimental evidence that the MHC class II-peptide complex plays a role in T-cell fate through activation and cytokine production (36). However, in humans, there is still no strong evidence supporting this hypothesis. Nevertheless, different combinations of HLA class II molecules and antigenic peptides may influence cytokine production during the early stages of the immune response against an HPV infection.

In conclusion, our results suggest that HLA polymorphisms are involved in genetic susceptibility to SCC and HPV infection in Brazilian populations. In particular, we found an increased risk for SCC and HPV positivity associated with DRB1*15 alleles, even after adjustment for potential confounding. On the other hand, individuals that carry DQB1*05, DRB1*0101, or DRB1*1302alleles have a decreased risk for SCC. Understanding these associations in different populations is necessary to describe the role that HLA molecules play in immune responses against HPV infections and subsequent SCC pathogenesis. Such information may help to determine better approaches for prevention and treatment of SCC through vaccination and immunotherapy.

We are indebted to the following clinicians who helped with subject accrual: Drs. J. Simoes; M. Farias; C. Seixas; L. Araujo; I. Arruda; P. Honorato; and F. Honorato (Napoleão Laureano Hospital,João Pessoa, Brazil). We are also grateful to A. Silva for conducting the interviews.