Abstract

The critical role of the human leukocyte antigen (HLA) system in presenting peptides to antigen-specific T cell receptors may explain why only some human papillomavirus (HPV)–infected women progress to cervical cancer. HLA class II DRB1 and DQB1 genes were examined in 315 women with invasive squamous cell cervical cancer (SCC) and 381 control subjects. Increased risks of SCC were associated with DRB1*1001, DRB1*1101, and DQB1*0301, and decreased risks were associated with DRB1*0301 and DRB1*13. Of squamous cell tumors, those containing HPV-16 were different from those not containing HPV-16 for 3 alleles: DRB1*0401, DRB1*07, and DQB1*06. Increased risks of SCC were associated with DRB1*0401-DQB1*0301 (odds ratio [OR], 1.7; 95% confidence interval [CI], 1.1–2.7) and DRB1*1101-DQB1*0301 (OR, 2.5; 95% CI, 1.4–4.5), and decreased risks were associated with DRB1*0301-DQB1*02 (OR, 0.7; 95% CI, 0.5–1.0) and DRB1*13-DQB1*06 (OR, 0.6; 95% CI, 0.4–0.9) haplotypes. These results add to the evidence that certain HLA class II alleles or allele combinations, or genes linked to them, make some women more susceptible to SCC

Cervical cancer is the second most common cancer among women, with 500,000 new cases and 200,000 deaths annually worldwide, and there is strong evidence that human papillomaviruses (HPVs) play a central role in cervical cancer etiology [1]. Many women become infected at sexual debut with genital HPV types that are potentially oncogenic, yet most infections regress spontaneously [2]. Persistence of HPV infections is associated with intraepithelial neoplasia, and a minority of those lesions progress to cancer [3]. Evidence of HPV DNA has been documented in 99% of cervical tumor tissues in a large, comprehensive study [4]. However, because HPV infection alone is not sufficient to cause cancer, a combination of environmental, viral, and host factors may work individually or together in the progression to invasive cervical cancer [5]

Impaired immunity is a host factor that has been associated with progression of HPV-related lesions, as reported in studies of various immunosuppressed patient populations. Renal transplant recipients, who have cell-mediated immune suppression, have an increased risk of HPV lesions [6, 7]. Generalized T cell deficiency also has been associated with increased anogenital neoplasia [8], as demonstrated by the 5-fold increase in the risk for cervical cancer among patients with human immunodeficiency virus infection or AIDS [9]. In addition, a rare genetic disease, epidermodysplasia verruciformis, is characterized by the combination of reduced CD4+ T cell counts and HPV-induced squamous cell skin lesions [10]

The critical role that the HLA system plays in presenting peptides to antigen-specific T cell receptors may explain why only some women with HPV infection progress to cervical cancer. Major histocompatibility gene products complexed with peptides derived from viral antigens, when on the surface of antigen-presenting cells, can induce T cell responses that clear viral infections. The HLA class II genes (DR, DQ, and DP) are expressed on antigen-presenting cells such as B cells and macrophages, where they present antigen fragments to CD4+ T cells. Although HLA gene products are not usually expressed on epithelial cells, HLA class II gene expression is increased in cervical cancer cells [11]. CD4+ T cells have been reported to have killer activity [12] and, thus, could potentially kill cervical cancer cells directly; however, CD4+ T cells more commonly provide helper functions that assist in the maturation of CD8+ T cells. The CD8+ T cells recognize peptides in conjunction with the more ubiquitously expressed HLA class I genes

The large number of polymorphisms of the HLA alleles leads to variations in the antigen-recognition site on the cell surface, which may confer susceptibility or resistance to HPV infection and neoplastic progression. Malignant transformation, as well as regression of cottontail rabbit papillomavirus–induced lesions, was clearly shown to be associated with class II DR and DQ gene expression [13]. The first reported associations between HLA class II genes and cervical cancer [14] were with DQw3, a serologic designation that has subsequently been subdivided into 3 other serologic designations: DQ7, DQ8, and DQ9. The specific alleles that were later determined to contribute to these serologic specificities include DQB1*0301 and *0304 for DQ7, DQB1*0302 and *0305 for DQ8, and DQB1*0303 for DQ9. A second report on the same samples was updated by use of DNA-based HLA typing methods and assigned the increased risk to DQB1*0301 and *0303 [15]. A variety of studies of different ethnic populations have confirmed this association with cervical cancer [16–20 ], yet other groups did not find statistically significant associations for these alleles [21–24 ]. Other allele groups that have been reported in >1 study to be associated with an increased risk include DRB1*11 [14, 19], DRB1*15 [20, 24, 25], and DQB1*06 [18, 25] and the related haplotypes DRB1*0401-DQB1*0301 [20, 26], DRB1*1101-DQB1*0301 [23, 26], and DRB1*1501-DQB1*0602 [20, 24, 27]. In addition, DRB1*13 and the DRB1*13-DQB1*06 haplotype have been reported to confer a lower risk of cervical cancer [22, 27–29 ]. Nevertheless, for each of these associations, there are contradictory studies that have not found the same relationship

Issues such as small sample sizes, inappropriate control groups, different laboratory methods, and chance findings hamper comparisons of results across studies. The likelihood of chance findings is increased by multiple comparisons between the extensive number of polymorphisms and disease. Comparing results from older studies that used serologic typing or low-resolution DNA typing with newer studies that use high-resolution methods also complicates direct comparisons

There is compelling evidence that immunologic control of HPV-associated lesions plays an important role in the pathway to cervical cancer [30], and it is likely that HLA plays an important role in regression of HPV-related lesions [31]. Prophylactic and therapeutic HPV vaccine trials are currently underway [32], so it is important to understand the mechanism by which some women may have a deficient immune response to HPV. Therefore, we conducted a large, population-based case-control study to examine the associations between HLA class II polymorphisms and invasive squamous cell cervical cancer (SCC) in the Seattle area. High-resolution HLA typing was used to identify specific HLA class II alleles. The HPV type within the cervical tumors was determined by polymerase chain reaction (PCR), and HPV serologic testing was done for all participants

Methods

Study populationThis investigation was conducted as part of an ongoing population-based epidemiologic study of HPV and invasive cervical cancer, as described elsewhere [33]. Eligible case subjects included all 18–74-year-old women residing in 3 western Washington counties, with invasive SCC diagnosed between January 1986 and June 1998. These patients were identified from the files of the Cancer Surveillance System of Western Washington, a population-based cancer registry that is part of the Surveillance, Epidemiology, and End Results program of the National Cancer Institute [34]. Eligible diagnoses were coded on the registry as International Classification of Diseases for Oncology morphology codes 801–808 [35]

Population-based control subjects were selected by use of random-digit telephone dialing from among residents of the same 3 counties in which case subjects were identified. Before 1996, control subjects were selected by use of simple unrestricted random-digit dialing; beginning in 1996, control subjects were selected by use of the Waksberg-Mitofsky modification of random-digit dialing, with a clustering factor (denoted “k” by Waksberg) of 2 residences per sampling unit [36]. One-step recruitment was used [37], with a stratified sampling design that recruited control subjects evenly throughout the ascertainment period into age (5-year) and county strata, to approximate the age and county distribution of the case subjects. Control subjects were matched to the case subjects on the year of diagnosis and then assigned a randomly chosen month

About 66% of the eligible case subjects and 68% of the eligible control subjects participated in the parent study [33]. Only white women, the largest racial group in the Seattle area, were included in the present study, to limit the heterogeneity in HLA polymorphism reported in different ethnic populations. In total, 315 case subjects and 381 control subjects were studied

Data collectionInformation elicited during the in-person interview included demographic characteristics, such as education and income, as well as reproductive, sexual, contraceptive, hormonal, Pap testing, and smoking histories. Control subjects were assigned a reference date corresponding to the diagnosis date of case subjects, and interview information pertained to the time period before diagnosis (case subjects) or reference date (control subjects). After the interview, all subjects were asked to provide a venous blood sample, from which serum and lymphocytes were isolated and stored at −80°C. Case subjects were asked to sign a consent form that would allow us to retrieve tumor blocks for HPV DNA testing

HLA typingDNA was extracted from buffy coat specimens for DNA-based typing of HLA-DRB1 and -DQB1 genes. The high-resolution HLA typing method used PCR amplification of genomic DNA, followed by sequence-specific oligonucleotide probe hybridization with nonradioactive detection [38]. Allele-level DRB1 typing was accomplished primarily by amplification with sequence-specific primers designed to amplify separately the polymorphic DRB1 allele families (DRB1*01, *02, *04, *03/11/13/14, and *08/12), followed by hybridization to expanded sequence-specific oligonucleotide probe panels [39] designed to identify all known alleles for samples tested between 1997 and 1999. After that time, ∼100 samples were typed for their DRB1 alleles by DNA sequencing methods, with use of fluorescence-labeled dideoxynucleotide terminator chemistry; analysis was done on an ABI 377-XL96 DNA sequencer with use of ABI Sequence Navigator and HLA Matchmaker software (all from Applied Biosystems). DQB1 typing was done with generic, locus-specific DQB1 primers and probe panels that allow for identification of the known DQB1 alleles

HLA class II DRB1 and DQB1 loci have been shown to exhibit strong positive linkage disequilibrium between particular alleles at each locus [40]. Among the unrelated persons included in this study, DRB1/DQB1 haplotype linkage was inferred on the basis of the well-known common linkages for whites, as published in International Histocompatibility Workshop Proceedings [41, 42] and determined by HLA family study typing in the Clinical Immunogenetics Laboratory, Seattle Cancer Care Alliance (A.G.S., unpublished data). The inferred haplotype refers only to the DRB1-DQB1 linkage and does not account for the possibility that alleles at other loci could differ in persons assigned to the same haplotype

HPV DNA typingTissue blocks were obtained for 80% (251/315) of the case subjects. The principal reason tissue samples were unavailable for laboratory testing was that the pathology laboratory had routinely discarded them. The study pathologist examined histologic sections from all available tissue blocks and selected the most representative for HPV DNA testing. HPV nucleic acids were detected with L1 consensus primers and 16E6 and 18E6 type-specific primers. The identity of the PCR products was confirmed by Southern hybridization, and the L1 consensus products were typed by 2 methods. Samples tested before 1997 (n=95) were typed by sequential hybridizations with probes for HPV-6, -11, -16, -18, -31, -33, and -35 [43]. Samples tested after 1997 were typed by restriction-fragment analysis (n=156), in which HPV type was assigned by comparison of restriction patterns of samples with HPV recombinant plasmid patterns. HPV DNA PCR results were available for 242 (96.4%) of the 251 case subjects included in this analysis; the remaining 9 women had tumor samples from which a fragment of the β-globin gene could not be amplified, indicating poor sample quality for HPV testing. No HPV DNA testing was done for the 381 control subjects

HPV serologic testsSerum samples from case subjects and control subjects were tested for antibodies to HPV-16 and HPV-18 by use of an antigen-capture antibody ELISA, as described elsewhere [33]. In brief, HPV capsids were produced by use of HPV-16–L1 recombinant vaccinia virus and purified on cesium chloride gradients. Capture antibodies were provided by Neil Christensen (Milton S. Hershey Medical Center, Hershey, PA). Human serum samples were tested in triplicate with and without capsids by researchers blinded to case-control status and other characteristics of the participants. Bound human IgG was detected by use of a goat anti–human alkaline phosphatase–conjugated antibody (Roche Diagnostics) and developed with alkaline phosphatase substrate 104 (Sigma). Positive controls (a pool of HPV-16–reactive serum samples) were included on each plate. For each serum sample, a value was calculated as follows: average of the natural log of the optical density values from the 3 wells containing antigen minus the average of the natural log of the optical density values from the 3 wells with no antigen. The cutoff points were determined by receiver operator characteristic analysis on the basis of the reactivity of serum samples from women with a low probability of HPV exposure, as described elsewhere [33]. Ten subjects, 1 control subject and 9 case subjects, did not have serum samples available for HPV antibody testing

Data analysisUnadjusted analyses are presented to compare associations in this population with results from other studies, which usually do not account for multiple comparisons. There are a large number of polymorphisms at each locus; thus, the risk of chance findings due to multiple comparisons is inherent in the data. As proposed by Farewell and Dahlberg [44], likelihood ratio tests based on logistic regression models were used to address multiple comparisons. There were 4 such tests, 2 of the allele-specific variables (37 DRB1 alleles in 1 model and the 16 DQB1 alleles in the second model) and 2 of the allele family group variables (13 DRB1 family groups in the third model and 7 DQB1 family groups in the fourth model). The tests measure the extent to which allelic variation at a given locus is associated with cervical cancer. The P values of the 4 tests were .032 for DRB1-specific alleles, .037 for DQB1-specific alleles, .002 for the DRB1 allele family group, and .0335 for the DQB1 allele family group, indicating a basic case-control difference at the locus level

Most analyses used binary logistic regression to estimate the relative risk of SCC, as estimated by the odds ratio (OR). In the haplotype association analysis, polytomous regression was used to compare simultaneously the HPV-16–positive and HPV-16–negative case groups, to determine whether the case groups were statistically different with respect to the haplotypes of interest. The first case group contained the 162 HPV-16–positive tumors (i.e., any HPV-16 DNA detected, even when other HPV types were present), and the second case group had 80 tumors that did not contain HPV-16 DNA (i.e., HPV types other than HPV-16 detected and HPV DNA negative). In this setting, the log likelihood for the unconstrained model (containing the 2 case groups and the control group) and a model in which the exposure estimates were constrained to be similar were compared by a χ2 distribution, where P<.05 was compatible with a difference between the case groups. In a subanalysis to investigate the risk of SCC among subjects who had evidence of prior exposure to HPV-16, the analysis of haplotypes was restricted to subjects positive for HPV-16 antibody

The DR-DQ–inferred haplotype combinations we investigated were chosen on the basis of significant findings in the previous literature and on the frequency of combinations in our population. The following haplotypes were determined to be of interest on the basis of previous studies: DRB1*0401-DQB1*0301, DRB1*0401-DQB1*0302, DRB1*07-DQB1*02, DRB1*07-DQB1*0303, DRB1*1101-DQB1*0301, DRB1*1301-DQB1*0603, and DRB1*1501-DQB1*0602.The second group comprised haplotypes that were present among ⩾ 5% of the study subjects and were not hypothesized to be important a priori: DRB1*0101-DQB1*0501, DRB1*03-DQB1*02, DRB1*13-DQB1*06, and DRB1*1401-DQB1*0503

We also examined, by 2 methods, whether certain allele combinations other than the inferred haplotypes increased or decreased the risk of SCC. First, to determine whether there was evidence that homozygosity of individual alleles increased or decreased the risk of SCC in a dose-dependent manner, carriership of 1 or 2 copies of an allele was compared with carriership of 0 copies. Second, in an attempt to determine the relative importance of individual alleles versus combinations of alleles, a separate categorical variable was constructed for each of the statistically significant haplotype combinations. For example, for the DRB1*0301-DQB1*02 haplotype, the categories were DRB1*0301-DQB1*X (where X is any DQB1 allele other than DQB1*02), DRB1*X-DQB1*02 (where X is any DRB1 allele other than DRB1*0301), DRB1*07-DQB1*02 (a haplotype that accounts for most of the subjects who have DQB1*02 but not DRB1*0301-DQB1*02), and DRB1*0301-DQB1*02, all compared with subjects who carried neither of the alleles (DRB1*X-DQB1*X)

Results

Allele group and allele-specific associationsUnadjusted ORs of allele groups or specific alleles in case and control subjects are shown in tables 1 and 2, respectively. In this population, there was a significantly elevated risk of SCC associated with 2 DR alleles, DRB1*1001 (OR, 5.6; 95% confidence interval [CI], 1.2–26.1) and DRB1*1101 (OR, 2.4; 95% CI, 1.4–4.2), and 1 DQ allele, DQB1*0301 (OR, 1.6; 95% CI, 1.2–2.2). There were also significantly decreased risks associated with DRB1*0301 (OR, 0.7; 95% CI, 0.5–0.9) and DRB1*13 (OR, 0.6; 95% CI, 0.4–0.9)

Table 1

Relative risk of squamous cell cervical cancer (SCC) and human papillomavirus type 16 (HPV-16)–containing SCC tumors associated with HLA class II DRB1 and DQB1 allele family groups

Table 1

Relative risk of squamous cell cervical cancer (SCC) and human papillomavirus type 16 (HPV-16)–containing SCC tumors associated with HLA class II DRB1 and DQB1 allele family groups

Table 2

Relative risk of squamous cell cervical cancer (SCC) and human papillomavirus type 16 (HPV-16)–containing SCC tumors associated with HLA class II DRB1 and DQB1 alleles

Table 2

Relative risk of squamous cell cervical cancer (SCC) and human papillomavirus type 16 (HPV-16)–containing SCC tumors associated with HLA class II DRB1 and DQB1 alleles

HPV DNA–specific allele associationsThere may be HPV type specificity for some HLA polymorphisms that can increase or decrease the risk of disease. Tissue blocks from 242 of the 315 case subjects were tested for the presence of HPV DNA. HPV DNA was detected in 87.2% of the 242 tumors tested, and HPV-16 was detected in 162 of the HPV-positive tumors (66.9% of tissue blocks tested). The risk estimates for specific alleles associated with SCC were similar when the case group was restricted to HPV-16–positive case subjects (n=162), compared with case subjects who were positive for other HPV types or were HPV DNA negative (n=80) and all control subjects (n=381), for all but 3 polymorphisms shown in tables 1 and 2. There were statistically significantly different risks for HPV-16–containing squamous cell tumors, compared with tumors not containing HPV-16, associated with DRB1*0401 (containing HPV-16: OR, 1.8; 95% CI, 1.2–2.8; not containing HPV-16: OR, 1.0; 95% CI, 0.6–1.9), DRB1*07 (containing HPV-16: OR, 1.5; 95% CI, 1.0–2.3; not containing HPV-16: OR, 0.6; 95% CI, 0.3–1.2), and DQB1*06 (containing HPV-16: OR, 0.7; 95% CI, 0.5–1.0; not containing HPV-16: OR, 1.1; 95% CI, 0.7–1.8)

Haplotype associationsHaplotypes based on inferred linkages frequently found in our population and on a priori associations with cervical cancer are shown in table 3. Elevated risks of SCC were associated with DRB1*0401-DQB1*0301 (OR, 1.7; 95% CI, 1.1–2.7) and DRB1*1101-DQB1*0301 (OR, 2.5; 95% CI, 1.4–4.5). Decreased risks were associated with DRB1*0301-DQB1*02 (OR, 0.7; 95% CI, 0.5–1.0) and DRB1*13-DQB1*06 (OR, 0.6; 95% CI, 0.4–0.9) haplotypes. The 4 statistically significant allele combinations (DRB1*0401-DQB1*0301, DRB1*1101-DQB1*0301, DRB1*0301-DQB1*02, and DRB1*13-DQB1*06) were compared in a model in which there was a common reference group that contained only women who did not carry any of these 4 haplotypes (data not shown). The relative risk estimates were similar to the relative risks presented in table 3

Table 3

Risk of squamous cell cervical cancer associated with haplotype combinations

Table 3

Risk of squamous cell cervical cancer associated with haplotype combinations

HPV DNA–specific haplotype associationsAs with the single-allele associations, HLA haplotypes associations with SCC were assessed according to the presence or absence of HPV-16 in the tumor. Two haplotypes of the 11 examined in this study were associated with a significantly increased risk of HPV-16–containing tumors, compared with non–HPV-16–containing tumors, by simultaneous comparison of the case groups by polytomous regression (table 3). The risk of SCC associated with DRB1*07-DQB1*0303 was confined to HPV-16–positive case subjects (OR, 2.0; 95% CI, 1.1–3.8) (P=.08 for comparison with HPV-16–negative case subjects; OR, 0.8; 95% CI, 0.3–2.4). There was a suggestion that the risk of SCC associated with DRB1*1401-DQB1*0503 was higher for HPV-16–positive case subjects (OR, 1.8; 95% CI, 0.8–4.0) (P=.04 for comparison with HPV-16–negative case subjects; OR, 0.3; 95% CI, 0.0–2.4); however, only 1 HPV-16–negative case subject was included in the analysis

To investigate whether associations were due to HPV infection or to the development of cancer, an analysis restricted to HPV-16 serologically–positive control subjects (n=65) and HPV-16 DNA–positive case subjects (n=162) was conducted. There was again a suggestion of an increased risk of HPV-16–containing SCC tumors with DRB1*07 -DQB1*0303 (OR, 2.8; 95% CI, 0.8–9.8). In similar analyses, the risk of HPV-16–containing tumors was significantly associated with DRB1*0401-DQB1*0301 (OR, 3.7; 95% CI, 1.2–10.9) and marginally associated with DRB1*1101-DQB1*0301 (OR, 2.6; 95% CI, 0.7–9.2) and DRB1*0301-DQB1*02 (OR, 0.7; 95% CI, 0.3–1.3). There was not a significant relationship between HPV-16–containing tumors and DRB1*13-DQB1*06 in this analysis (OR, 0.7; 95% CI, 0.3–1.7)

HomozygosityWe were interested in exploring whether a specific allele or combination of alleles, rather than the inferred haplotype, was responsible for the significant associations with SCC. To examine this question, the risk of SCC associated with heterozygosity and homozygosity for an allele was compared with that for subjects with no copies of the allele. This analysis was restricted to those 7 alleles that made up the significant haplotype associations in table 3. For the 4 DRB1 alleles, DRB1*0301, DRB1*0401, DRB1*1101, and DRB1*13, only 30 subjects were homozygous. Although the number of homozygotes is small because of the large number of possible alleles at the DRB1 locus, it is noteworthy that there was a suggestion of an inverse relationship for DRB1*13 heterozygosity (OR, 0.7; 95% CI, 0.5–1.0) that became stronger for DRB1*13 homozygosity, because all 8 homozygous subjects were control subjects (table 4). For the 3 DQB1 homozygous allele combinations, the risk estimates were stronger for homozygous alleles than for heterozygous alleles. There was an increased risk associated with DQB1*0301 heterozygosity (OR, 1.5; 95% CI, 1.1–2.1) that was somewhat stronger for homozygosity (OR, 2.1; 95% CI, 0.9–4.9)

Table 4

Risk of squamous cell cervical cancer associated with number of copies of alleles

Table 4

Risk of squamous cell cervical cancer associated with number of copies of alleles

Constituent allelesIn table 5, the importance of specific alleles, compared with carriership of inferred haplotypes, was directly compared for those haplotypes that were significantly related to SCC in table 3. For example, we estimated the risk of SCC associated with DRB1*0401-DQB1*X (where X is any DQB1 allele other than DQB1*0301), DRB1*1101-DQB1*X (where X is any DQB1 allele other than DQB1*0301), DRB1*X-DQB1*0301 (where X is any DRB1 allele other than DRB1*1101 or DRB1*0401), and the DRB1*0401-DQB1*0301 and DRB1*1101-DQB1*0301 haplotypes, compared with subjects who carried none of the alleles of interest (DRB1*X-DQB1*X). As in table 3, there were higher risks in table 5 for the haplotypes DRB1*1101-DQB1*0301 (OR, 3.0; 95% CI, 1.7–5.5) and DRB1*0401-DQB1*0301 (OR, 1.8; 95% CI, 1.2–2.9), and there was no increase in the risk associated with any of the alleles by themselves (e.g., DRB1*X-DQB1*0301 (OR, 1.0; 95% CI, 0.7–1.6)

Table 5

Risk of squamous cell cervical cancer associated with haplotypes and their constituent alleles

Table 5

Risk of squamous cell cervical cancer associated with haplotypes and their constituent alleles

Similarly, the relative risk estimates for SCC appear to be driven by the haplotypes and not the specific alleles for DRB1*13-DQB1*06, because neither DRB1*13 nor DQB1*06 in other haplotype combinations was significantly associated with disease. In addition, because there was no elevated risk of SCC associated with DRB1*1501-DQB1*0602, which was the inferred haplotype for 88.3% of the subjects with DRB1*1501, it does not seem that the DQB1*0602 allele confers a decreased risk. For DRB1*0301-DQB1*02, there were no subjects with DRB1*0301 who did not have DQB1*02. When those subjects with DRB1*X-DQB1*02 were examined further, all but 1 control and 2 case subjects had the DRB1*07-DQB1*02 haplotype (OR, 1.0; 95% CI, 0.6–1.5). Again, the significantly decreased risk for SCC was associated with the inferred haplotype

Discussion

This study supports the hypothesis that certain HLA class II DR and DQ allele combinations affect the risk of invasive cervical cancer. Specifically, DRB1*0401-DQB1*0301 (OR, 1.7; 95% CI, 1.1–2.7) and DRB1*1101-DQB1*0301 (OR, 2.5; 95% CI, 1.4–4.5) increased the risk of SCC. There was also evidence for a decreased risk of SCC associated with DRB1*0301-DQB1*02 (OR, 0.7; 95% CI, 0.5–1.0) and DRB1*13-DQB1*06 (OR, 0.6; 95% CI, 0.4–0.9) haplotypes

The combination of high-resolution HLA typing of the alleles at the DR and DQ loci and the large number of subjects in this study allowed us to examine all inferred haplotypes that are present at a frequency of ⩾5% among case or control subjects. Of the 4 haplotypes in this study that were significantly associated with SCC, the 2 that conferred an increased risk of SCC, DRB1*1101-DQB1*0301 and DRB1*0401-DQB1*0301, have been reported elsewhere in studies of cervical intraepithelial neoplasia [45] and SCC [20] in white women. Two studies of HLA and HPV-16E variants also found an increased risk of cervical cancer containing variant HPV types associated with DRB1*1101-DQB1*0301 and DRB1*0401-DQB1*0301 [26, 46]. It may be that viral factors, such as type and variant, and virus load are important promoters of cervical carcinogenesis in the context of the host’s HLA type

Two haplotypes seemed to confer a significant decreased risk of SCC in this population, DRB1*0301-DQB1*02 and DRB1*13-DQB1*06. Only DRB1*13-DQB1*06 has been reported in previous studies; nonsignificant decreased risks for HPV-16–positive SCC tumors associated with DRB1*1301-DQB1*0603 and DRB1*1302-DQB1*0604 were reported in a study by Apple et al. [21]. It may be that the many studies that do report associations for the DRB1*13 allele underestimate the effect of the allele combinations by focusing on individual alleles. Studies have reported decreased risks associated with DRB1*13, including studies of 3 white populations (Hispanic Americans [21], Dutch [29], and French [28]), a Costa Rican population [22], and a Senegalese population [23]. The association with invasive cervical cancer was not significant in studies of British [20] and Brazilian [24] populations, although both studies had a lower prevalence of DRB1*13 among control subjects than among case subjects. Although these studies all had similarly good HLA typing techniques, it may be that low power or differences in the underlying distribution of HLA alleles in the population would result in such disparate findings. Other differences, such as the particular HPV variant contracted or other protective or detrimental exposures, such as diet, cigarette smoking, parity, or hormone use, may also contribute to the difference between estimates in these studies

To better understand the mechanism by which HLA class II alleles may affect SCC risk, the relative contributions of the inferred haplotypes versus specific alleles that make up the haplotypes were directly compared in table 5. In this study, when the specific allele and various allele combinations were analyzed, the decreased risk associated with DRB1*13 (OR, 0.6; 95% CI, 0.4–0.9) appeared to be attributable to either the haplotype DRB1*13-DQB1*06 or the individual alleles. There was some support for the potential importance of the specific allele from the suggestion of a dose response with the number of alleles homozygous for DRB1*13 (table 4), because all 8 subjects with 2 DQB1*13 alleles were control subjects. However, 6 of the 8 control subjects also had 2 DQB1*06 alleles, making it unclear whether the DRB1*13 allele or the haplotype was more important. DQB1*06 does not have the consistently inverse associations found with DRB1*13 (e.g., the lack of association between SCC and DRB1*1501-DQB1*0602 in this study; table 5), so it does not seem likely that DQB1*06 is sufficient to protect against SCC. There may also be modulation of the effect of 1 allele in the presence of other nearby alleles that was not detected, including alleles that were not measured in this study

Several studies did report a decreased risk of cervical cancer associated with the DRB1*13 haplotypes. An association with DRB1*13 and DRB1*13-DQB1*0603 was found in studies of Hispanic women (OR, 0.3; 95% CI, 0.1–0.8) [21] and white French women (OR, 0.2; 95% CI, 0.1–0.6) [28], respectively. The strongest haplotype association that included DRB1*13 in our study was for DRB1*13-DQB1*06 (OR, 0.6; 95% CI, 0.4–0.9)

There was an increased risk for DRB1*1001 (OR, 5.6; 95% CI, 1.2–26.1) in this study that was not reported in any of the published studies. The prevalence of the polymorphism in the Seattle population indicates that there would be a small attributable risk for this allele, and its low prevalence could explain why associations were not detected in smaller studies. All subjects with DRB1*10 had the inferred haplotype DRB1*1001-DQB1*0501. The DQB1*0501 allele was not associated with risk of SCC in the most commonly occurring haplotype that included the allele: DRB1*0101-DQB1*0501 (OR, 1.2; 95% CI, 0.8–1.8)

There was a nearly 2-fold increase in the risk of SCC associated with DRB1*11 (OR, 1.8; 95% CI, 1.2–2.9), which was most prevalent as variant DRB1*1101 in this study. A Dutch study [29] reported an increased risk with DRB1*1101 that was confined to HPV-16–positive cancers, and a Senegalese study [23] found no increased risk with DRB1*11 but did see an increased risk with the DRB1*1101-DQB1*0301 haplotype. This haplotype was found for all but 4 subjects with DRB1*1101 in the present study

Some polymorphisms may more readily bind certain HPV type sequences and present them to the immune system; therefore, polymorphisms that do not bind HPV peptides well may increase risk and those that do present processed HPV to CD4+ cells may decrease risk [47]. A recent report by Odunsi and Ganesan [48] found that the affinity of the binding between HPV-16 epitopes and HLA molecules was lower for risk-conferring than risk-reducing alleles. This lack of affinity may underlie the increased risk associated with DRB1*0401-DQB1*0301 and DRB1*1101-DQB1*0301 haplotypes

In this study, the case group was split between those with tumors containing HPV-16, the most common oncogenic HPV type, and those with tumors that did not contain HPV-16 to assess HPV-16–specific associations. Because the comparison of interest is that of the case groups with each other, we used polytomous regression to test for the difference between the case groups rather than attempting to compare 2 ORs with overlapping CIs with each other without a formal statistical test. The relative risks for HPV-16–containing SCC tumors appeared to be different (although not quite significantly) from that for tumors containing other high-risk HPV types associated with DRB1*0401 (containing HPV-16: OR, 1.8; 95% CI, 1.2–2.8; not containing HPV-16: OR, 1.0; 95% CI, 0.6–1.9; P=.07). However, this difference was less clear when the risk was examined separately for HPV-16–containing and non–HPV-16–containing tumors associated with the DRB1*0401-DQB1*0301 haplotype (containing HPV-16: OR, 2.1; 95% CI, 1.3–3.5; not containing HPV-16: OR, 1.4; 95% CI, 0.7–2.9; P=.26). The only other haplotype that contains DRB1*0401, DRB1*0401-DQB1*0302, showed a lack of association with either HPV-16–containing or non–HPV-16–containing SCC tumors (containing HPV-16: OR, 1.1; 95% CI, 0.6–2.1; not containing HPV-16, OR, 0.6; 95% CI, 0.2–1.6; P=.19). There may be an important difference between HPV-16–containing and non–HPV-16–containing tumors that cannot be definitively detected in this study because of small case groups, or it may be that the effects of HLA are directly toward other tumor-associated antigens, rather than HPV peptides, or that the relevant epitopes are conserved across HPV types

A study by Apple et al. [21] reported increases in the risk of invasive SCC associated with DRB1*1501-DQB1*0602 of nearly 3-fold for all case subjects (OR, 2.9; 95% CI, 1.3–6.7) and almost 5-fold for HPV-16–positive case subjects (OR, 4.8; 95% CI, 1.9–11.8) among Hispanic women. There were not enough cancers positive for other HPV types in that study (n=26) to compare risks by HPV type, but there is a suggestion of an increased risk specific to HPV-16–positive cancer. Likewise, Cuzick et al. [20] reported an increased risk for SCC associated with this haplotype among white women in Britain that was restricted to HPV-16–positive case subjects (OR, 1.8; 95% 1.0–3.3) and was not found among HPV-16–negative case subjects (n=33). The present study of white women in the northwestern United States did not find an increased risk associated with DRB1*1501-DQB1*0602 among all case subjects (OR, 1.0; 95% CI, 0.7–1.4) or HPV-16–positive case subjects (OR, 0.9; 95% CI, 0.6–1.4), but there was a slightly elevated risk associated with non–HPV-16–containing SCC tumors (OR, 1.4; 95% CI, 0.8–2.3; P=.22). The prevalence of this haplotype among the control subjects in this study was 26.2%, similar to the 21.3% reported for white British women [20] and much higher than the 6.4% reported among Hispanic women in the United States [21]

In contrast to an increased risk of invasive disease associated with DRB1*1501-DQB1*0602 in the Hispanic American and British studies, Hildesheim et al. [49] reported a decreased risk of HPV-16–containing high-grade squamous intraepithelial cervical lesions associated with DRB1*1501-DQB1*0602 (OR, 0.2; 95% CI, 0.1–0.6) from a US study of white women in the Portland, Oregon, area. The recent study by Wang et al. [22] conducted in Costa Rica also reported a decreased risk of high-grade squamous intraepithelial lesions (OR, 0.4; 95% CI, 0.2–0.8). These studies suggest that, regardless of HPV type or racial group, there may be a reduced risk of high-grade squamous intraepithelial lesions associated with this haplotype that is not seen with invasive disease. However, this haplotype then either has no influence on risk or increases the relative risk of invasive disease. It may be that viral gene products expressed in high-grade squamous intraepithelial cervical lesions but not in cancers are the immune system targets that trigger clearance in association with this haplotype

In this study, a population-based cancer registry was used to identify cases. This setting avoids patient selection by pattern of practice within or outside an institution. The use of a population-based control group, which reflects the population that gave rise to the cases and thus strengthens the internal validity of the study, also makes it more likely that the associations described are generalizable to the local population

The extensive polymorphism of HLA genes complicates studies that focus on the role of HLA in disease, because this heterogeneity may prevent moderate associations from reaching significance, even in a large study such as this. Furthermore, the number of polymorphisms in HLA genes increases the probability that chance associations will be found because of the large number of comparisons made during analysis. In this study, overall likelihood ratio tests for each locus showed a difference between case and control subjects associated with the alleles for both loci, and the data were presented without corrections for multiple comparisons. The most stringent test of the hypotheses generated in this study will be made when the analysis is repeated in another study that is now in progress in this same population

The present study was able to support some, but not all, previous findings of increased and decreased risks of SCC associated with DQB1 and DRB1 polymorphisms. Some of the differences between our study and others may be due to the different distribution of allele combinations or alleles present in the populations studied, chance associations, or linkage with unknown susceptibility genes. Furthermore, haplotypes that are reported as separate alleles without respect to their linkage may mask associations that are important when the haplotype is considered. These associations need to be examined in large populations with use of high-resolution HLA typing to explore interactions with HPV types or subtypes. It is important to study the relationship between HLA polymorphisms and cervical cancer in large, well-characterized populations, so that the pivotal role of cell-mediated immunity in HPV regression and the implications for therapeutic and prophylactic vaccines can be clarified

Acknowledgments

We thank the women who participated in this study, for their time; the study staff, the interviewers, and the registry staff, for their dedication; and Barbara McKnight and Jackie Starr, for their contributions to data analysis

References

1.
IARC Working Group on the Evaluation of Carcinogenic Risks to Humans
Human papillomaviruses
IARC Monogr Eval Carcinog Risks Hum
 , 
1995
, vol. 
64
 (pg. 
277
-
82
)
2.
Holowaty
P
Miller
AB
Rohan
T
To
T
Natural history of dysplasia of the uterine cervix
J Natl Cancer Inst
 , 
1999
, vol. 
91
 (pg. 
252
-
8
)
3.
Schlecht
NF
Kulaga
S
Robitaille
J
, et al.  . 
Persistent human papillomavirus infection as a predictor of cervical intraepithelial neoplasia
JAMA
 , 
2001
, vol. 
286
 (pg. 
3106
-
14
)
4.
Walboomers
JM
Jacobs
MV
Manos
MM
, et al.  . 
Human papillomavirus is a necessary cause of invasive cervical cancer worldwide
J Pathol
 , 
1999
, vol. 
189
 (pg. 
12
-
9
)
5.
zur Hausen
H
Papillomaviruses in anogenital cancer as a model to understand the role of viruses in human cancers
Cancer Res
 , 
1989
, vol. 
49
 (pg. 
4677
-
81
)
6.
Bouwes Bavinck
JN
Berkhout
RJ
HPV infections and immunosuppression
Clin Dermatol
 , 
1997
, vol. 
15
 (pg. 
427
-
37
)
7.
Sillman
FH
Sentovich
S
Shaffer
D
Anogenital neoplasia in renal transplant patients
Ann Transplant
 , 
1997
, vol. 
2
 (pg. 
59
-
66
)
8.
Tindle
RW
Frazer
IH
zur Hausen
H
Immune response to human papillomaviruses and the prospects for human papillomavirus–specific immunization
Human pathogenic papillomaviruses
 , 
1994
Heidelberg, Germany
Springer Verlag
(pg. 
217
-
53
)
9.
Frisch
M
Biggar
RJ
Goedert
JJ
Human papillomavirus–associated cancers in patients with human immunodeficiency virus infection and acquired immunodeficiency syndrome
J Natl Cancer Inst
 , 
2000
, vol. 
92
 (pg. 
1500
-
10
)
10.
Vardy
DA
Baadsgaard
O
Hansen
ER
Lisby
S
Vejlsgaard
GL
The cellular immune response to human papillomavirus infection
Int J Dermatol
 , 
1990
, vol. 
29
 (pg. 
603
-
10
)
11.
Hilders
CG
Houbiers
JG
Krul
EJ
Fleuren
GJ
The expression of histocompatibility‐related leukocyte antigens in the pathway to cervical carcinoma
Am J Clin Pathol
 , 
1994
, vol. 
101
 (pg. 
5
-
12
)
12.
Erb
P
Grogg
D
Troxler
M
Kennedy
M
Fluri
M
CD4+ T cell–mediated killing of MHC class II–positive antigen‐presenting cells. I. Characterization of target cell recognition by in vivo or in vitro activated CD4+ killer T cells
J Immunol
 , 
1990
, vol. 
144
 (pg. 
790
-
5
)
13.
Han
R
Breitburd
F
Marche
PN
Orth
G
Linkage of regression and malignant conversion of rabbit viral papillomas to MHC class II genes
Nature
 , 
1992
, vol. 
356
 (pg. 
66
-
8
)
14.
Wank
R
Thomssen
C
High risk of squamous cell carcinoma of the cervix for women with HLA‐DQw3
Nature
 , 
1991
, vol. 
352
 (pg. 
723
-
5
)
15.
Wank
R
Schendel
DJ
Thomssen
C
HLA antigens and cervical carcinoma
Nature
 , 
1992
, vol. 
356
 (pg. 
22
-
3
)
16.
Nawa
A
Nishiyama
Y
Kobayashi
T
, et al.  . 
Association of human leukocyte antigen–B1*03 with cervical cancer in Japanese women aged 35 years and younger
Cancer
 , 
1995
, vol. 
75
 (pg. 
518
-
21
)
17.
Helland
A
Borresen
AL
Kaern
J
Ronningen
KS
Thorsby
E
HLA antigens and cervical carcinoma
Nature
 , 
1992
, vol. 
356
 (pg. 
22
-
3
)
18.
Gregoire
L
Lawrence
WD
Kukuruga
D
Eisenbrey
AB
Lancaster
WD
Association between HLA‐DQB1 alleles and risk for cervical cancer in African American women
Int J Cancer
 , 
1994
, vol. 
57
 (pg. 
504
-
7
)
19.
Duggan‐Keen
MF
Keating
PJ
Stevens
FR
, et al.  . 
Immunogenetic factors in HPV‐associated cervical cancer: influence on disease progression
Eur J Immunogenet
 , 
1996
, vol. 
23
 (pg. 
275
-
84
)
20.
Cuzick
J
Terry
G
Ho
L
, et al.  . 
Association between high‐risk HPV types, HLA DRB1* and DQB1* alleles, and cervical cancer in British women
Br J Cancer
 , 
2000
, vol. 
82
 (pg. 
1348
-
52
)
21.
Apple
RJ
Erlich
HA
Klitz
W
Manos
MM
Becker
TM
Wheeler
CM
HLA DR‐DQ associations with cervical carcinoma show papillomavirus‐type specificity
Nat Genet
 , 
1994
, vol. 
6
 (pg. 
157
-
62
)
22.
Wang
SS
Wheeler
CM
Hildesheim
A
, et al.  . 
HLA class I and II alleles and risk of cervical neoplasia: results from a population‐based study in Costa Rica
J Infect Dis
 , 
2001
, vol. 
184
 (pg. 
1310
-
4
)
23.
Lin
P
Koutsky
LA
Critchlow
CW
, et al.  . 
HLA class II DR‐DQ and increased risk of cervical cancer among Senegalese women
Cancer Epidemiol Biomarkers Prev
 , 
2001
, vol. 
10
 (pg. 
1037
-
45
)
24.
Maciag
PC
Schlecht
NF
Souza
PSA
Franco
EL
Villa
LL
Petzl‐Erler
ML
Major histocompatibility complex class II polymorphisms and risk of cervical cancer and human papillomavirus infection in Brazilian women
Cancer Epidemiol Biomarkers Prev
 , 
2000
, vol. 
9
 (pg. 
1183
-
91
)
25.
Beskow
AH
Josefsson
AM
Gyllensten
UB
HLA class II alleles associated with infection by HPV16 in cervical cancer in situ
Int J Cancer
 , 
2001
, vol. 
93
 (pg. 
817
-
22
)
26.
Terry
G
Ho
L
Cuzick
J
Analysis of E2 amino acid variants of human papillomavirus types 16 and 18 and their associations with lesion grade and HLA DR/DQ type
Int J Cancer
 , 
1997
, vol. 
73
 (pg. 
651
-
5
)
27.
Apple
RJ
Becker
TM
Wheeler
CM
Erlich
HA
Comparison of HLA DR‐DQ disease associations found with cervical dysplasia and invasive cervical carcinoma
J Natl Cancer Inst
 , 
1995
, vol. 
87
 (pg. 
427
-
36
)
28.
Sastre‐Garau
X
Loste
MN
Vincent‐Salomon
A
, et al.  . 
Decreased frequency of HLA‐DRB1*3 alleles in Frenchwomen with HPV‐positive carcinoma of the cervix
Int J Cancer
 , 
1996
, vol. 
69
 (pg. 
159
-
64
)
29.
Krul
EJT
Schipper
RF
Schreuder
GMT
Fleuren
GJ
Kenter
GG
Melief
JM
HLA and susceptibility to cervical neoplasia
Hum Immunol
 , 
1999
, vol. 
60
 (pg. 
337
-
42
)
30.
Stern
PL
Brown
M
Stacey
SN
, et al.  . 
Natural HPV immunity and vaccination strategies
J Clin Virol
 , 
2000
, vol. 
19
 (pg. 
57
-
66
)
31.
Greenstone
HL
Nieland
JD
de Visser
KE
, et al.  . 
Chimeric papillomavirus virus–like particles elicit antitumor immunity against the E7 oncoprotein in an HPV16 tumor model
Proc Natl Acad Sci USA
 , 
1998
, vol. 
95
 (pg. 
1800
-
5
)
32.
Steller
MA
Schiller
JT
Human papillomavirus immunology and vaccine prospects
J Natl Cancer Inst Monogr
 , 
1996
, vol. 
21
 (pg. 
145
-
8
)
33.
Daling
JR
Madeleine
MM
McKnight
B
, et al.  . 
The relationship of human papillomavirus–related cervical tumors to cigarette smoking, oral contraceptive use, and prior herpes simplex type 2 infection
Cancer Epidemiol Biomarkers Prev
 , 
1996
, vol. 
5
 (pg. 
541
-
8
)
34.
Hankey
BF
Ries
LA
Edwards
BK
The surveillance, epidemiology, and end results program: a national resource
Cancer Epidemiol Biomarkers Prev
 , 
1999
, vol. 
8
 (pg. 
1117
-
21
)
35.
International classification of diseases for oncology. 2nd ed
 , 
1990
Geneva
World Health Organization
36.
Hartge
P
Brinton
LA
Rosenthal
JF
Cahill
JI
Hoover
RA
Waksberg
J
Random digit dialing in selecting a population‐based control group
Am J Epidemiol
 , 
1984
, vol. 
120
 (pg. 
825
-
33
)
37.
Harlow
BL
Davis
S
Two one‐step methods for household screening and interviewing using random digit dialing
Am J Epidemiol
 , 
1988
, vol. 
127
 (pg. 
857
-
63
)
38.
Mickelson
EM
Smith
A
McKinney
S
Anderson
G
Hansen
JA
A comparative study of HLA‐DRB typing by standard serology and hybridization of non‐radioactive sequence‐specific oligonucleotide probes
Tissue Antigens
 , 
1993
, vol. 
41
 (pg. 
86
-
93
)
39.
Smith
AG
Nelson
JL
Regen
L
, et al.  . 
Six new DR52‐associated DRB1 alleles, three of DR8, two of DR11, and one of DR6, generate polymorphism in the MHC
Tissue Antigens
 , 
1996
, vol. 
48
 (pg. 
118
-
26
)
40.
Begovich
AB
McClure
GR
Suraj
VC
, et al.  . 
Polymorphism, recombination, and linkage disequilibrium within the HLA class II region
J Immunol
 , 
1992
, vol. 
148
 (pg. 
249
-
58
)
41.
Clayton
J
Lonjou
C
Charron
D
Anthropology tables: allele and haplotype frequencies for HLA loci in various ethnic groups
Proceedings of the 12th International Histocompatibility Workshop (St. Malo, France)
 , 
1997
Paris
EDK
(pg. 
665
-
775
)
42.
Gjertson
DW
Lee
SH
Gjertson
DW
Terasaki
PI
HLA‐A/B and ‐DRB1/DQB1 allele level haplotype frequencies
HLA 1998
 , 
1998
Lenexa, KS
American Society for Histocompatibility and Immunogenetics
(pg. 
365
-
450
)
43.
Beckmann
AM
Sherman
KJ
Myerson
D
Daling
JR
McDougall
JK
Galloway
DA
Comparative virologic studies of condylomata acuminata reveal a lack of dual infections with human papillomaviruses
J Infect Dis
 , 
1991
, vol. 
163
 (pg. 
393
-
6
)
44.
Farewell
VT
Dahlberg
S
Some statistical methodology for the analysis of HLA data
Biometrics
 , 
1984
, vol. 
40
 (pg. 
547
-
60
)
45.
Odunsi
K
Terry
G
Ho
L
Bell
J
Cuzick
J
Ganesan
TS
Susceptibility to HPV‐associated cervical intra‐epithelial neoplasia is determined by specific HLA DR‐DQ alleles
Int J Cancer
 , 
1996
, vol. 
67
 (pg. 
595
-
602
)
46.
Zehbe
I
Tachezy
R
Tmyilineous
J
, et al.  . 
Human papillomavirus 16 E6 polymorphisms in cervical lesions from different European populations and their correlation with human leukocyte antigen class II haplotypes
Int J Cancer
 , 
2001
, vol. 
94
 (pg. 
711
-
6
)
47.
Diepolder
HM
Scholz
S
Pape
GR
Influence of HLA alleles on outcome of hepatitis C virus infection
Lancet
 , 
1999
, vol. 
354
 (pg. 
2094
-
5
)
48.
Odunsi
K
Ganesan
T
Motif analysis of HLA class II molecules that determine the HPV associated risk of cervical carcinogenesis
Int J Mol Med
 , 
2001
, vol. 
8
 (pg. 
405
-
12
)
49.
Hildesheim
A
Schiffman
M
Scott
DR
, et al.  . 
Human leukocyte antigen class I/II alleles and development of human papillomavirus–related cervical neoplasia: results from a case‐control study conducted in the United States
Cancer Epidemiol Biomarkers Prev
 , 
1998
, vol. 
7
 (pg. 
1035
-
41
)
Presented in part: 19th International Papillomavirus Conference, Florianópolis, Brazil, 1–7 September 2001 (abstract 0-130)
Informed consent was obtained from all subjects, and the human experimentation guidelines of the US Department of Health and Human Services and those of the authors’ institutions were followed in the conduct of the research
Financial support: National Institutes of Health (CA-42792); Cancer Surveillance System of the Fred Hutchinson Cancer Research Center (funded by contract CN-67009); Fred Hutchinson Cancer Research Center