Comprehensive Analysis of Human Leukocyte Antigen Class I Alleles and Cervical Neoplasia in 3 Epidemiologic Studies

Journal of Infectious Diseases, Sep 2002

To comprehensively explore the relationship between human leukocyte antigen (HLA) class I alleles and cervical neoplasia, a subset of participants from 3 large US and Costa Rican cervix studies were typed for HLA class I alleles. Study subjects were women with cervical cancer or high-grade squamous epithelial lesions (HSILs; n=365) or low-grade squamous epithelial lesions (LSILs; n=275) or who were cytologically normal (control subjects; n=681). Allele-disease associations were assessed by logistic regression analysis. Consistent associations across all studies were observed for HLA-CW*0202 with a combined odds ratio of 0.53 (95% confidence interval [CI], 0.29–0.89) for cancer or HSILs and 0.58 (95% CI, 0.37–1.04) for LSILs, compared with control subjects and adjusted for study. This finding supports the hypothesis that a single allele may be sufficient to confer protection against cervical neoplasia. Given the relationship between HLA-C and its receptors on natural killer (NK) cells, a role is proposed for NK function in human papillomavirus infection and cervical neoplasia

A PDF file should load here. If you do not see its contents the file may be temporarily unavailable at the journal website or you do not have a PDF plug-in installed and enabled in your browser.

Alternatively, you can download the file locally and open with any standalone PDF reader:

https://jid.oxfordjournals.org/content/186/5/598.full.pdf

Comprehensive Analysis of Human Leukocyte Antigen Class I Alleles and Cervical Neoplasia in 3 Epidemiologic Studies

Sophia S. Wang () 2 4 Allan Hildesheim 2 4 Xiaojiang Gao 1 2 Mark Schiffman 2 4 Rolando Herrero 2 11 M. Concepcion Bratti 2 11 Mark E. Sherman 2 4 Willard A. Barnes 2 8 Mitchell D. Greenberg 2 6 Larry McGowan 2 7 Rodrigue Mortel 2 5 Peter E. Schwartz 2 10 Richard J. Zaino 2 5 Andrew G. Glass 2 9 Robert D. Burk 2 3 Peter Karacki 0 2 Mary Carrington 1 2 0 Johns Hopkins University School of Medicine , Baltimore, Maryland 1 Intramural Research Support Program , SAIC- Frederick , National Cancer Institute , Frederick 2 Received 7 February 2002; revised 4 April 2002; electronically published 29 July 2002. Presented in part: 19th International Papillomavirus Conference , Flori- anopolis, Brazil, 1-7 September 2001 (abstract O-94). Informed consent was obtained from all participants in accordance with US Department of Health and Human Services guidelines. The National Institutes of Health and local institutional review boards approved the study. Financial support: National Institutes of Health (contracts CO-12400, CP-21081, and CP-31061; grant CA-78527 to R.D.B.); FUCODOCSA (Costa Rican Foundation for Training in Health Sciences); Caja Costarri- cense de Seguro Social (Costa Rica). Section, Environmental Epidemiology Branch, Div. of Cancer Epidemiology and Genetics, National Cancer Institute , 6120 Executive Blvd., EPS MSC 7234, Bethesda, MD 20892-7234 3 Departments of Pediatrics and Epidemiology and Social Medicine, Albert Einstein College of Medicine , Bronx , New York 4 Interdisciplinary Studies Section, Division of Cancer Epidemiology and Genetics, Environmental Epidemiology Branch, National Cancer Institute , Bethesda 5 Milton S. Hershey Medical Center , Hershey, Pennsylvania 6 Graduate Hospital , Philadelphia 7 Division of Gynecologic Oncology, George Washington University , Washington, DC 8 Lombardi Cancer Center 9 Kaiser Foundation Research Institute , Oakland, California 10 Yale University School of Medicine , New Haven, Connecticut 11 Proyecto Epidemiologico , Guanacaste , Costa Rica To comprehensively explore the relationship between human leukocyte antigen (HLA) class I alleles and cervical neoplasia, a subset of participants from 3 large US and Costa Rican cervix studies were typed for HLA class I alleles. Study subjects were women with cervical cancer or high-grade squamous epithelial lesions (HSILs; n p 365) or low-grade squamous epithelial lesions (LSILs; n p 275) or who were cytologically normal (control subjects; n p 681). Allele-disease associations were assessed by logistic regression analysis. Consistent associations across all studies were observed for HLA-CW*0202 with a combined odds ratio of 0.53 (95% confidence interval [CI], 0.29-0.89) for cancer or HSILs and 0.58 (95% CI, 0.37-1.04) for LSILs, compared with control subjects and adjusted for study. This finding supports the hypothesis that a single allele may be sufficient to confer protection against cervical neoplasia. Given the relationship between HLA-C and its receptors on natural killer (NK) cells, a role is proposed for NK function in human papillomavirus infection and cervical neoplasia. - HLA alleles involved in presenting foreign antigens to immune cells are important in host immune responses to viral and other pathogens. Among the most polymorphic human genes [1], HLA polymorphisms result in variations of the peptide-binding cleft and influence specificity of the antigens bound and presented to T cells. Class I HLA molecules (HLA-A, -B, and -C) present foreign antigens to CD8 cytotoxic T lymphocytes (CTL), and class II molecules (HLA-DR, -DQ, and -DP) present antigenic peptides to CD4 T helper cells [2, 3]. Although the importance of HLA class II genes in cervical neoplasia pathogenesis has been demonstrated over the past decade [411], class I allele associations with cervical neoplasia have not been widely documented. However, CTL responses to viral infections (specifically to human papillomavirus [HPV] infection) have been well documented [12, 13]. In addition, downregulation, as well as complete loss of class I antigen expression, has been reported in cervical cancer and its immediate precursors. Such alterations in class I antigen expression enable HPVinfected cells to escape detection by the immune system by becoming nonimmunogenic [1417]. Although the role that HLA class I molecules play in cervical neoplasia is well established, the importance of individual HLA class I alleles has not been fully explored. Development of high-resolution genotyping allowed us to complete HLA typing for 3 large cervical neoplasia studies in the United States and Costa Rica. To identify HLA alleles of importance in cervical neoplasia, we typed HLA class I and class II alleles from the cervical neoplasia studies of 2 ethnically distinct populations. HLA class II allele findings have been published from our 24,000-woman cohort in Portland, Oregon [18], and from our 10,077-woman cohort in Guanacaste, Costa Study population Guanacaste, Costa Rica Portland, Oregon Eastern United States Total Study type Population-based cohort Kaiser Permanente cohort Six-center case-control Cancer or HSILsa LSILsb Control subjectsc Rica [19]. As in the HLA class II analyses, the extreme polymorphism of HLA class I alleles and their fairly even distribution result in low frequencies of individual alleles, making single allele-disease associations difficult to observe. Here, we present study-specific and combined results for HLA class I alleledisease associations from all 3 studies. We examined HLA class I involvement in the development of cervical neoplasia and identified consistent associations in the distinct study populations. Subjects and Methods Study population. Participants were selected from 3 studies sponsored by the National Cancer Institute: a 10,077 woman population-based cohort in Guanacaste, Costa Rica [20]; a 24,000 woman cohort in Portland, Oregon [18]; and a 750 woman multicenter study of histologic subtypes of cervical neoplasia in the eastern United States [21]. Details of the study designs have been described elsewhere [18, 20, 21]. The Costa Rican cohort is an ethnically admixed population, whereas the 2 US study groups are predominantly white. In the Costa Rican cohort of women diagnosed with cancer (n p 40), high-grade squamous intraepithelial lesions (HSILs; n p 130), and low-grade squamous intraepithelial lesions (LSILs; n p 106) and in population control subjects (n p 250), a subset of 24 (60%) patients with cancer, 100 (77%) patients with HSILs, 79 (75%) patients with LSILs, and 118 (47%) control subjects were typed for HLA class I alleles. In the Portland cohort of women with HSILs (n p 141) and LSILs (n p 212) and cytologically normal control subjects (n p 368), a subset of 124 (88%) women with HSILs, 196 (92%) women with LSILs (99 HPV-16 positive and 97 HPV negative), and 350 (95%) cytologically normal control subjects (HPV-16positive women were oversampled) were typed for HLA class I alleles. In the eastern US study of 234 in situ and invasive squamous cell carcinomas and 307 population-based control subjects, a subset of 117 (50%) women with cancer and 213 (69%) cytologically normal population-based control subjects were typed for HLA class I alleles. The 166 adenocarcinomas that were part of the eastern US study were not included in the present analysis. Final analytic groups. For the present study, 1321 women were typed for HLA class I alleles, including 321 from the Costa Rican cohort, 670 from the Portland cohort, and 330 from the eastern US study. Our final analytic group consisted of 365 women with cancer or HSILs, 275 with LSILs, and 681 with normal cytologic test results (control subjects; table 1). HLA testing. HLA class I loci were molecularly typed with DNA extracted from buffy coat [20, 21] or cervicovaginal lavage samples [18] collected from each participant. HLA class I genes were typed by using polymerase chain reaction (PCR) and singlestranded oligonucleotide probe-based protocols developed by the 13th International Histocompatibility Workshop (http://www.ihwg .org/protocols/protocol.htm). HPV testing. Cervicovaginal samples were tested by PCR for HPV DNA, as described elsewhere [18, 20, 21]. In the Portland cohort, HPV was typed with MY09/11 consensus primers via dot blot. In the eastern US case-control study, HPV was also typed with MY09/11 consensus primers, but by strip technology. In the Costa Rican cohort, HPV typing was done by both PCR (with MY09/11) and the hybrid-capture tube test. Statistical methods. A case-control analysis was conducted in which patients with cancer or HSILs and patients with LSILs were compared with population control subjects. We combined women with invasive cancer with our HSIL group for the final analysis. However, because HSILs comprised cervical intraepithelial lesion (CIN) 3 as well as CIN2, we also conducted separate analyses for CIN3, CIN2, and cancer. Because there was no notable difference between the independent groups with regard to allele frequency, we combined HSILs and cancer for the final analysis. HLA-A, -B, and -CW allele frequencies were initially calculated, and statistical differences between case patients and control subjects were identified by x2 test for significance or by Fishers exact test when there were !5 subjects per cell. Further analyses for alleles found to be significantly different between case and control groups were conducted. For these alleles, odds ratios (ORs) and 95% confidence intervals (CIs) were calculated to determine the magnitude and statistical significance of associations [22, 23]. Although we used logistic regression analysis to examine allele-disease associations, while adjusting for other alleles whose frequencies were significantly different between case patients and control subjects (e.g., where all alleles undergoing further analyses are placed in the same regression model), these adjustments did not alter the results. Thus, for study-specific ORs, unadjusted risk estimates are shown. For combined analyses where data for case patients and control subjects from the 3 studies were merged to obtain an overall OR, the es HLA-A Study control subjects possessing specified HLA class I alleles. Costa Rica (n p 118)a Eastern United States (n p 213)b Portland, Oregon (n p 350)c Percentage of population Percentage of population Percentage of population NOTE. The alleles listed have frequencies 5% in at least one study control population. a Control subjects were typed by loci: HLA-A (n p 112), HLA-B (n p 102), and HLA-CW (n p 95). b Control subjects were typed by loci: HLA-A (n p 203), HLA-B (n p 208), and HLA-CW (n p 206). c Control subjects were typed by loci: HLA-A (n p 333), HLA-B (n p 322), and HLA-CW (n p 330). d The allele frequency was statistically significantly different (x2 test) from the eastern US study control population. e The allele frequency was statistically significantly different (x2 test) from the Portland study control population. f The allele frequency statistically was significantly different (x2 test) from the Costa Rican study control population. timates were adjusted by study (each study as a dummy variable) to account for potential differences among the 3 populations. We calculated the Pearson correlation coefficients between alleles to identify alleles in possible linkage disequilibrium (LD; defined as r 0.9). Although LD could only be inferred on the basis of a high correlation coefficient, this was done to ensure that independent associations between these alleles and disease could be calculated. To identify HLA alleles associated with disease progression, we conducted HPV-restricted analyses in which patients with cancer or HSILs were compared with patients with LSILs and HPV-positive control subjects, thus identifying HLA class I alleles associated with progression from low-grade to high-grade disease. In addition, we conducted HPV-16restricted analyses to determine whether HPV type specificity for class I allele-disease associations existed, as shown for class II alleles [4]. The HPV-16restricted analysis was conducted in the Portland cohort because of the study design, which oversampled HPV-16positive control subjects; however, these analyses were not done for the eastern US and Costa Rican studies because of the minimal number of HPV-16positive population control subjects. Statistical analyses were done with SAS software (version 8.12; SAS Institute). All tests of statistical significance are 2-sided. Results Allele frequencies (by individual subjects) in the 3 population control groups were first compared. Alleles with frequencies of 5% in any study control group are summarized in table 2. As expected, statistically significant differences were observed mostly between the Costa Rican admixed population and the 2 predominantly white US populations. For example, although HLA-B*4002 had a relatively high allele frequency of 19% among the Costa Rican control population, its allele frequency was 2% among the eastern US study control population and 1% among the Portland control population. In addition, some alleles of relatively high frequency in the US studies (e.g., HLACW*0701 at 26% in the eastern US study and 27% in the Portland study) had low allele frequencies in Costa Rica (5%). Even consistently high-frequency alleles, such as HLA-A*0201, were statistically significantly different between the studies (30% in Costa Rica, 41% in the eastern US study, and 45% in the Portland study). Nevertheless, a number of allele frequencies were consistent across all 3 populations (e.g., HLA-B*3501 at 11%13%, CW*0702 at 21%26%, and HLA-CW*0202 at 8% 10%). Alleles that are statistically significantly different from the other 2 study populations are indicated in table 2. The allele frequency among patients with cancer or HSILs and those with LSILs was significantly different (P ! .05) from that of control subjects in any one study or in combined analyses for the following alleles: HLA-A*0206, A*0301, A*3101, A*3103, A*3303, A*3402, A*6803, B*1402, B*1508, B*1512, B*2705, B*3503, B*3517, B*3901, B*3908, B*4901, B*5301, B*8101, CW*0202, CW*0305, CW*0401, CW*0801, and CW*0802. For these alleles, we conducted further analyses. First, we calculated the Pearson correlation coefficients for these alleles with each other and with all other HLA class I alleles. No combination with these alleles had a correlation coefficient of r 0.9, denoting possible LD, nor were any of these alleles moderately correlated. We therefore proceeded to assess independent associations between these alleles with disease. Table 3 shows the disease ORs for these alleles and for cancer or HSILs by study. For a single allele, HLA-CW*0202, the allele frequency among patients with cancer or HSILs was significantly lower than that among control subjects (merged OR, 0.53; 95% CI, 0.290.89). This risk decrease was consistent across all 3 studies with an OR of 0.75 (95% CI, 0.351.62) in Portland, 0.22 (95% CI, 0.060.80) in Costa Rica, and 0.62 (95% CI, 0.241.63) in the eastern United States. Other alleles that appeared to decrease the risk for cancer or HSILs were HLA-B*2705 and HLA-CW*0401. However, although a statistically significant decrease in risk for cancer or HSILs was observed for both alleles in the eastern US study (HLA-B*2705: OR, 0.29; 95% CI, 0.080.99; HLA-CW*0401: OR, 0.52; 95% CI, 0.290.94), the decreases in risk were not statistically significant in Costa Rica and they were not observed in the Portland study. No single allele had a statistically significant or consistent increase in risk for cancer or HSILs. Although possession of HLA-B*1508 appeared to increase the risk for cancer or HSILs in all 3 studies, the risk estimates were not statistically significant in the studies or in the merged analyses (OR, 9.31; 95% CI, 0.7958.5). Furthermore, although possession of the HLAA*3303, -B*3503, or -B*3901 alleles statistically significantly increased the risk for cancer or HSILs in the Portland study, with risk estimates of 3.35 (95% CI, 1.268.88), 2.95 (95% CI, 1.088.03), and 3.43 (95% CI, 1.1310.4), respectively, the increases in risk were not observed in the Costa Rican or eastern US study. Table 4 shows the disease ORs for alleles and LSILs. The eastern US study was not included in this analysis because it did not include patients with LSILs. Consistent with findings for patients with cancer or HSILs, HLA-CW*0202 was associated with a decreased risk for developing LSILs in the merged analyses (OR, 0.58; 95% CI, 0.371.04) and by study (Portland: OR, 0.52; 95% CI, 0.251.07; Costa Rica: OR, 0.73; 95% CI, 0.262.02), albeit not statistically significant. In addition, although possession of the HLA-B*4901 allele consistently reduced the risk for developing LSILs in Portland (OR, 0.48; 95% CI, 0.131.74) and Costa Rica (OR, 0.37; 95% CI, 0.04 3.33), the risk estimates were not statistically significant nor were they statistically significant in merged analyses (OR, 0.45; 95% CI, 0.151.36). Portland, Oregon Costa Rica Eastern United States No. (%) of No. (%) of case patients control subjects (n p 124) (n p 350) No. (%) of No. (%) of case patients control subjects (n p 124) (n p 118) No. (%) of No. (%) of case patients control subjects (n p 117) (n p 213) 95% CI 95% CI 95% CI 95% CI 95% CI Portland, Oregon No. (%) of case patients (n p 196) No. (%) of control subjects (n p 350) No. (%) of case patients (n p 78) Costa Rica CI, confidence interval; OR, odds ratio. A statistically significant increase in risk for LSILs in the merged analyses was observed for HLA-A*3101 (OR, 2.02; 95% CI, 1.073.80), which was consistent in Portland (OR, 2.27; 95% CI, 1.055.33) and Costa Rica (OR, 1.55; 95% CI, 0.56 4.32). An increase in risk for developing LSILs in the merged analyses was also observed for HLA-CW*0802 (OR, 1.73; 95% CI, 1.022.93), with study-specific ORs of 1.48 (95% CI, 0.80 2.74) in Portland and 2.71 (95% CI, 0.947.71) in Costa Rica. These increases in risk observed for LSILs (HLA-A*3101 and -B*0802) were not observed for patients with cancer or HSILs. Although possession of the HLA-B*1512 allele also appeared to increase risk for LSILs, this allele was only present in the Portland population. In addition, although not statistically significant in the merged analyses (OR, 1.95; 95% CI, 0.944.07), possession of the HLA-B*1402 allele did statistically significantly increase risk for LSILs in Costa Rica (OR, 3.66; 95% CI, 1.0912.3). This finding was consistent, but not statistically significant, in the Portland study (OR, 1.26; 95% CI, 0.473.36). To evaluate the association between HLA class I alleles and HPV progression, we conducted analyses restricted to HPVinfected women (table 5). Thus, we compared women with cancer or HSILs with those with LSILs and HPV-positive control women and elevated the baseline group to women at risk for progression. Findings significant in merged analyses included a decreased risk for HLA-CW*0802 (OR, 0.52; 95% CI, 0.28 0.96), which was consistent, albeit not statistically significant, across the studies (Portland: OR, 0.51; 95% CI, 0.211.26; Costa Rica: OR, 0.57; 95% CI, 0.211.57; eastern US: OR, 0.42; 95% CI, 0.101.86). In addition, although possession of HLAB*3901 did not increase risk for progression in the Costa Rica study, the risk estimate was significant in the Portland study (OR, 3.48; 95% CI, 1.1510.6). Likewise, possession of HLACW*0401 statistically significantly decreased risk for progression only in the eastern US study (OR, 0.26; 95% CI, 0.100.72). Finally, although possession of HLA-CW*0202 appeared to decrease risk in merged analyses (OR, 0.68; 95% CI, 0.351.31) and in all 3 individual studies, none was statistically significant. Only in the Portland study were we able to conduct analyses to assess HPV-16 type specificity, because of the original study design, which oversampled control subjects among HPV-16 positive women. For this population, none of the previously mentioned alleles had statistically significant allele-disease associations and thus did not support HPV-16 type specificity for these associations (data not shown). The association for HLACW*0202 with HPV-16positive patients with HSILs, compared with HPV-16positive control subjects, in merged analysis was 0.93 (95% CI, 0.412.11); for HLA-CW*0802, the merged risk estimate was 0.68 (95% CI, 0.261.75). Discussion We found a reduction in disease risk with a single allele, HLACW*0202. Unlike other alleles examined, the association between HLA-CW*0202 and cancer or HSILs was statistically Portland, Oregon Costa Rica Eastern United States No. (%) of No. (%) of case patients control subjects (n p 124) (n p 355) No. (%) of No. (%) of case patients control subjects (n p 111) (n p 95) No. (%) of No. (%) of case patients control subjects (n p 60) (n p 45) 95% CI 95% CI 95% CI CI, confidence interval; OR, odds ratio. significant in combined analyses and was consistent across all 3 populations and in the 2 predominant ethnic groups (admixture in Costa Rica and whites in the United States). Although not reaching statistical significance for the individual US studies, the decrease in risk for cancer or HSILs was statistically significant in the Costa Rican study, where the prevalence of HLA-CW*0202 was also the highest (10%). This association was also consistent, albeit not statistically so, for LSILs in merged and study-specific analyses. These findings support the hypothesis that possession of a single protective HLA allele is sufficient for protection from HPV infection and subsequent cervical neoplasia. HLA-CW*0802 was associated with a decreased risk for disease progression but was associated with an increased risk for LSILs (with no association observed for cancer or HSILs). We also found an increased risk for LSILs for women with HLAA*3101 but not for cancer or HSILs. From the current analyses, it is not clear why these alleles might increase risk for LSILs but not for HSILs/cancer or why HLA-CW*0802 might simultaneously be associated with an increased risk for LSILs but a decreased risk for progression. From our analyses, it is difficult to conclude whether these differential associations indicate potentially different roles for each allele (e.g., in initial infection by HPV or immunosurveillance of infected cells). In addition to a chance finding, another possible explanation may reside in the unmeasured alleles with which these alleles are in LD. On the basis of the current data, it is difficult to draw conclusions about the roles that HLA-CW*0802 and HLA-A*3101 play in cervical neoplasia. Limitations in the present study include the potential for confounding by ethnicity, also known as population stratification [24], in merged analysis. Although we accounted for potential confounding due to population stratification by adjusting for study in our merged analyses, the theoretical possibility of residual confounding within studies remains. However, we believe that within-study confounding by ethnicity is not likely to strongly affect our results because, in Costa Rica, the population is highly admixed, and, in our US-based studies, the vast majority of subjects were of white ethnicity. Limitations with merged analyses also include the potential for missing associations that may exist if LD with another allele (whether HLA or another critical gene) is needed for an association to be present. As alluded to earlier, it is plausible that inconsistent findings between the different studies may be real and due to possible LD with another critical allele that we have not measured and which may not be present in another population; such associations would not be detected in our present analyses. Future directions would include identification of complete HLA class I and II haplotypes and assessing their relationships with disease. It is also plausible that some of the inconsistent findings in this study may be attributed to false-positive results from the multiple comparisons of HLA alleles. Because of these limitations, criteria for significance consisted not only of statistical significance in one study or in merged analysis but also of consistent findings across studies, in addition to statistical significance in the merged analysis. Nevertheless, although associations consistent in all 3 studies benefited from the statistical significance gained by merging the studies, the lack of statistical significance within the individual studies remains a limitation. Strengths of our study included extensive high-resolution HLA genotyping by one laboratory. This allowed identification of individual allele-disease associations in HLA class I alleles, which has not been reported previously. Furthermore, our findings are from 3 large independent studies plus analyses that merged all 3 studies when results were consistent across studies. Merging the different studies enhanced the sample size for assessing individual allele-disease associations plus data from 3 studies provided sufficient strength to ensure consistency in findings across populations and ethnic groups. We emphasize that this ability to identify consistency across populations and ethnic groups is the real strength of the present analyses. For alleles with adequate frequencies in each study (5% frequency in control subjects), such as for HLA-CW*0202, the pooling of data provided robust analyses. For alleles with widely varying frequencies between studies, we emphasize the need for consistent findings across individual studies. This need is due to the unique control definitions for each study; although all control subjects are defined as cytologically normal, the Costa Rican and eastern US study control subjects were population based, whereas the Portland study control subjects were oversampled for HPV-16positive women. As noted, the importance of HLA class I molecules in cervical neoplasia is well established. Down-regulation of HLA class I antigens affects immune surveillance of viral infection and affects effective elimination of infected cells [25, 26]. For other immune-related disorders, such as ankylosing spondylitis, psoriasis, and AIDS progression, HLA class I alleledisease associations have been established. To our knowledge, our study is the first to comprehensively assess HLA class I alleles and cervical neoplasia. The observation consistent in all 3 studies and across the 2 ethnic groups is the decreased risk for developing cancer or HSILs and LSILs observed for women with the HLA-CW*0202 allele. However, some women with cancer or HSILs possess the HLA-CW*0202 allele, suggesting that the importance of additional factors involved in the disease process and emphasizing that the role of HLA molecules is only one of many factors involved in disease development. Future studies should include the complete assessment of HLA class I and class II molecules (haplotype analyses), assessment of HPV type specificity (to identify protective or risk alleles specific to oncogenic HPV types), and the exploration of innate immunity. On the basis of the present results, we believe that an exploration of the role of NK cells is warranted because of the role HLA-C molecules play in stimulating cellular immune responses via their recognition by NK cells [27, 28]. It is plausible that the protective association observed for disease and disease progression with HLA-C alleles is indicative of the involvement that CTL and NK cells play in the host response to viral infections and in recognizing and destroying human tumor cells [28]. The lack of HPV type specificity observed for HLACW*0202 with disease is also consistent with a proposed role for NK cells, since NK cell responses are not antigen specific. We suggest there may be a possible involvement of NK function in HPV infection and subsequent cervical neoplasia.


This is a preview of a remote PDF: https://jid.oxfordjournals.org/content/186/5/598.full.pdf

Sophia S. Wang, Allan Hildesheim, Xiaojiang Gao, Mark Schiffman, Rolando Herrero, M. Concepcion Bratti, Mark E. Sherman, Willard A. Barnes, Mitchell D. Greenberg, Larry McGowan, Rodrigue Mortel, Peter E. Schwartz, Richard J. Zaino, Andrew G. Glass, Robert D. Burk, Peter Karacki, Mary Carrington. Comprehensive Analysis of Human Leukocyte Antigen Class I Alleles and Cervical Neoplasia in 3 Epidemiologic Studies, Journal of Infectious Diseases, 2002, 598-605, DOI: 10.1086/342295