Killer immunoglobulin-like receptor (KIR) recognition of specific human histocompatibility leukocyte antigen (HLA) class I allotypes contributes to the array of receptor–ligand interactions that determine natural killer (NK) cell response to its target. Contrasting genetic effects of KIR/HLA combinations have been observed in infectious and autoimmune diseases, where genotypes associated with NK cell activation seem to be protective or to confer susceptibility, respectively. We show here that combinations of KIR and HLA loci also affect the risk of developing cervical neoplasia. Specific inhibitory KIR/HLA ligand pairs decrease the risk of developing neoplasia, whereas the presence of the activating receptor KIR3DS1 results in increased risk of disease, particularly when the protective inhibitory combinations are missing. These data suggest a continuum of resistance conferred by NK cell inhibition to susceptibility involving NK cell activation in the development of cervical neoplasia and underscore the pervasive influence of KIR/HLA genetic variation in human disease pathogenesis.

NK cells are an important component of the innate immune system, providing the first line of defense in the early stages of the immune response against viral infections and tumors by production of cytokines or direct cytotoxicity (1, 2). NK cell activity is controlled partially through interactions between killer immunoglobulin-like receptor (KIR) expressed on NK cells and their respective HLA class I ligands expressed on target cells (35). HLA class I loci are characterized by extreme allelic polymorphism, whereas KIR variation stems both from the polygenic nature of KIR haplotypes (i.e., they vary in the number and types of genes present) and from allelic polymorphism at some of the KIR genes (6, 7). Independent segregation of HLA (6p21) and KIR (19q13.4) genes along with KIR specificity for particular HLA allotypes raises the possibility that any given individual may express KIR molecules for which no ligand is present or vice versa, resulting in the lack of NK cell signaling through that specific KIR. Thus, the requirement for specific ligand–receptor interactions at these two polymorphic loci implies an epistatic relationship (i.e., a synergistic relationship involving variation at two or more loci, specific combinations of which are essential for the phenotype to occur) between KIR and HLA in NK cell–mediated immunity.

KIR genes encode molecules that convey either inhibitory or activating signals to NK cells and to a subset of CD8+ T cells. Binding of inhibitory KIR (designated 2DL and 3DL) to specific HLA allotypes has been clearly demonstrated (3, 8), correlating with their ability to inhibit NK cytolysis of target cells bearing those HLA molecules. The inhibitory KIR3DL1 binds HLA-B allotypes that have the Bw4 epitope (determined by amino acid positions 79–83 of the molecule) (9). The ligand for KIR3DS1, an activating KIR that shows 99% amino acid sequence similarity to KIR3DL1 in its extracellular domains, has not been determined, and, indeed, its cell surface expression has not been definitively shown. However, the presence of this gene along with HLA-B alleles encoding Bw4 have an epistatic protective effect on AIDS progression (10), suggesting that, like KIR3DL1, KIR3DS1 might recognize some Bw4 allotypes in HIV+ individuals. Specificity of inhibitory KIR for HLA-Cw allotypes is dictated to a large extent by the presence of asparagine or lysine at position 80 of the HLA-Cw molecule (11, 12). KIR2DL1 recognizes group 2 HLA-Cw molecules that are characterized by Lys80, whereas KIR2DL2 and KIR2DL3 (referred to as KIR2DL2/3 throughout the remainder of the paper) prefer group 1 HLA-Cw molecules containing Asn80. KIR2DL1, KIR2DL2/3, and KIR3DL1 are present in nearly all individuals of European descent, so the presence or absence of their HLA ligands determines the functional significance of these inhibitory KIR in any given individual.

The effect of genetic variation at the KIR locus in combination with genes encoding their HLA ligands on cancer has not been studied previously, even though variation at these genes may be important in the development of cancers that are linked to viruses because KIR genes have been shown to be associated with outcome to viral infections (10, 13). To test the possibility that the KIR genes are involved in risk of developing cervical neoplasia, a cancer caused by human papilloma virus (HPV; reference 14), we examined KIR and HLA in three large cervix studies.

Results And Discussion

To test the possibility that the KIR genes are involved in risk of developing cervical neoplasia, a cancer caused by HPV (14), we grouped HLA-B (HLA-B Bw4) and HLA-Cw (groups 1 and 2) class I alleles according to KIR ligand specificity using HLA genotypes determined previously from participants in three large cervical studies and compared the HLA group frequencies in cytologically normal versus cervical intraepithelial neoplasia 3 (CIN3)/cancer patients. Using a dominant model, both HLA-Cw group 2 and HLA-B Bw4 allelic groups were significantly associated with a decreased risk of disease when data from the three studies were merged (odds ratio [OR] = 0.71, P = 0.02, and OR = 0.70, P = 0.015, respectively; Fig. 1 and Table S1), a trend that was reflected in each individual study (although not significantly in most comparisons) and either consistent or more pronounced when restricted to HPV+ controls (Cw group 2, OR = 0.77, 95% confidence interval = 0.53–1.12; Bw4, OR = 0.56, 95% confidence interval = 0.38–0.81). Although we cannot entirely rule out the possible association between these alleles and HPV infection, the observation of similar associations in HPV-restricted controls supports a role for these alleles in cervical cancer pathogenesis. The presence of both Bw4 and Cw group 2 showed a somewhat stronger effect than the presence of an allele or alleles from only one of these groups (OR = 0.56, P = 0.002 for both, vs. OR = 0.65, P = 0.024 for only one; Table I). Alternatively, the presence of at least one copy of Cw group 1 was significantly associated with increased risk of disease in the Eastern U.S. study (OR = 2.14, P = 0.02) but not in the Costa Rica or Portland studies (Table S1). However, Cw group 1 in the absence of both Cw group 2 and Bw4 did associate with susceptibility in each of the three studies (OR = 1.47–2.09), although, again, significance was reached only in the Eastern U.S. study (for merged data, OR = 1.6, P = 0.0066). Overall, these data suggest that one mode of influence conferred by HLA class I on cervical neoplasia may stem from their capacity as ligands for inhibitory KIR molecules (inhibitory KIR genes that encode products recognizing Cw group 1, Cw group 2, and Bw4 are present in >95% of individuals studied here). Specifically, effector cell inhibition mediated by KIR2DL1 and KIR3DL1 (the receptors for Cw group 2 and Bw4, respectively) may participate in decreasing the risk of cervical neoplasia.

Unlike several of the inhibitory KIR, the activating KIR are present on only a fraction of KIR haplotypes. Given the protective influence of Bw4 and Cw group 2 alleles, which encode the high-affinity ligands for two inhibitory KIR, we hypothesized that the presence of activating KIR may confer a susceptibility effect. Two of the three cervix studies from which DNA was available (Eastern U.S. and Costa Rica) were typed for the presence or absence of six activating and five inhibitory KIR genes, which, along with three anchor KIR loci that are always present, constitute the known assortment of KIR haplotypes (Table S2). KIR3DS1, an activating receptor that shows high sequence similarity to KIR3DL1, was observed significantly more frequently in CIN3/cancer cases than in controls (OR = 1.64, P = 0.009). The activating KIR genes, KIR2DS1 and KIR2DS5, showed a weak association with increased risk of CIN3/cancer that bordered on significance (Table S2), possibly because of their strong linkage disequilibrium with KIR3DS1 (15, 16). Thus, activating and inhibitory KIR seem to have opposing effects on the risk of developing cervical neoplasia.

When effects of KIR3DS1 were evaluated jointly with Cw group 2 (Fig. 2, left) or Bw4 (Fig. 2, right), a significant gradient of effects from susceptibility to protection was observed in the following order (Fig. 2 and Table S3): (a) presence of KIR3DS1 in the absence of ligand for the inhibitory KIR2DL1 or KIR3DL1 (OR = 1; baseline), (b) absence of both activating KIR3DS1 and inhibitory KIR ligand (OR = 0.69–0.81), (c) presence of both KIR3DS1 and inhibitory KIR ligand (OR = 0.60–0.67), and (d) absence of KIR3DS1 in the presence of inhibitory KIR ligand (OR = 0.44–0.47). These trends were evident for KIR3DS1 in combination with Cw group 2 (ligand for KIR2DL1; P trend = 0.0026) as well as Bw4 (ligand for KIR3DL1; P trend = 0.0025), supporting a scenario in which the risk of developing cervical neoplasia is, to some extent, a function of the level of NK cell activation contributed by KIR.

The HLA-B and -C loci are in strong linkage disequilibrium, being separated by only 81.5 kb. Indeed, HLA-B Bw4 and HLA-Cw group 2 are in significant but not complete linkage disequilibrium (D′[Q2] = 0.413, P < 0.0001). Thus, it can be difficult to distinguish between disease associations involving one or both of these loci in the context of their function as inhibitory KIR ligands, particularly because the receptors that recognize these ligands, KIR3DL1 and KIR2DL1, are present in 95–98% of all individuals (Table S2). To address this issue with regard to the KIR3DS1/HLA-Cw group 2 and KIR3DS1/HLA-B Bw4 associations in cervical disease, we tested for independent and additive effects of these compound genotypes, which were suggested in the analysis of the HLA groups where KIR genotype was not considered (Table I). A significant test for trend was observed for odds ratios determined in a progressive manner from genotypes expected to result in the most activation to those resulting in the most inhibition as follows (P trend = 0.0012; Fig. 3 and Table S4): (a) absence of both Bw4 and Cw group 2, presence of KIR3DS1; (b) presence of either Bw4 or Cw group, presence of KIR3DS1; (c) absence of Bw4 and Cw group 2, absence of KIR3DS1; (d) presence of Bw4 and Cw group 2, presence of KIR3DS1; (e) presence of Bw4 and/or Cw group 2, absence of KIR3DS1. Although numbers are limited, the data suggest that having both Bw4 and Cw group 2 may be more protective than having only one of these inhibitory ligands when KIR3DS1 is present (compare the second and fourth sets in Fig. 3). However, in the absence of KIR3DS1, there did not seem to be an additive effect of having both Bw4 and Cw group 2 relative to having only one of these ligands (Bw4 alone, OR = 0.48, P = 0.06; Cw group 2 alone, OR = 0.43, P = 0.04; Bw4 and Cw group 2, OR = 0.44, P = 0.02). These statistics also indicate independent protective effects of both Bw4 and Cw group 2, as opposed to a single protective locus that is simply marked by a second locus through linkage disequilibrium.

Previous experimental data are consistent with the hypothesis that KIR2DL1 provides a stronger inhibitory signal than do KIR2DL2 and KIR2DL3 (17), and previously reported genetic epidemiological data support this model (13, 18). We suggest that the data reported herein present another illustration of the distinction among KIR2DL receptors in effector cell signaling, in which KIR2DL1 binding to Cw group 2 ligands may send a stronger inhibitory signal to the effector cells than does KIR2DL2 or KIR2DL3 upon binding to Cw group 1 ligands on target cells. The potentially strong inhibition mediated by KIR2DL1–Cw group 2, along with that conferred by KIR3DL1–Bw4 interactions, may avert to some extent the immune activation, which we propose is linked to cervical pathogenesis. A model for the gradation of responses from most activating (susceptible) to most inhibitory (protective) is shown in Fig. 4.

Cervical neoplasia is the third disease with which KIR3DS1 has been associated (10), underscoring the need to determine whether this receptor is expressed, and, if so, to identify its ligand. Rather than being directly involved in disease processes, KIR3DS1 may simply mark a neighboring gene that is directly involved in pathogenesis of cervical neoplasia. Indeed, all individuals in this study with KIR3DS1 also had KIR2DS1. However, four individuals who had KIR2DS1 were missing KIR3DS1, and all four were disease-free, accounting for the weaker association between KIR2DS1 and disease (Table S2) and suggesting that KIR2DS1 may not independently increase risk of this disease. Given the strong linkage disequilibrium among the KIR genes, it is difficult to rule out the possibility that both KIR2DS1 and KIR3DS1 mark another locus that directly influences the risk of developing cervical neoplasia.

The existence of functionally opposing KIR molecules suggests that the health benefits derived from any given activating or inhibitory KIR may be coupled to certain risks, so that a KIR gene (or haplotype) conferring protection against one disease may actually predispose to another. This model is corroborated by genetic epidemiological data, which show that specific KIR/HLA compound genotypes expected to result in an activating phenotype are associated with protection against some infectious diseases (10, 13) but are also associated with susceptibility to autoimmune pathogenesis (1921). In this regard, cervical neoplasia may more closely resemble autoimmune disease, in spite of its infectious etiology.

HPV infection is necessary, but not sufficient, for development of cervical cancer (22). Nevertheless, HPV infections rarely progress to cervical cancer; most HPV infections cause no or only mild cytologic abnormalities that subsequently regress to normalcy. Cervical inflammation does not seem to correlate significantly with HPV infection but has been associated with high-grade lesions in oncogenic HPV-infected women (23). Thus, although acquired immune responses to HPV may deter development of high-grade lesions and cancer, inappropriate localized hyperresponsiveness, mediated in part by KIR–ligand interactions, may actually increase the risk of disease progression. The findings described here indicate that further studies of the inflammatory process in the progression to cervical neoplasia and of the role of NK cells (and CTLs) in exacerbating this process are justified.

Materials And Methods

Study population.

Informed consent was obtained from all participants in accordance with U.S. Department of Health and Human Services guidelines. The study protocol was approved by the Institutional Review Board of the National Institutes of Health and local review boards. Participants were selected from three cervix studies sponsored by the National Cancer Institute and for whom specimens were available for the current investigation: a 24,000-woman cohort in Portland, Oregon (24); a 10,077-woman population-based cohort in Guanacaste, Costa Rica (25); and a 570 woman multicenter study of histologic subtypes of cervical neoplasia in the Eastern U.S. (26). In brief, all cervical cancer and CIN3 cases were included from the three studies; all controls from the U.S.-based case control study were included, and a selected group of population-based controls were included from the two cohort studies as previously described (24, 27, 28) for a nested-case control design. Although HLA class I typing was completed in a subset of all three studies, samples were available for KIR typing only in the latter two studies. Our final analytic group for HLA analysis included 257 women with cancer (squamous cell carcinoma) or CIN3 and 675 normal controls. Because KIR typing was not conducted in the Portland cohort (because of lack of DNA), analysis of combinatorial KIR/HLA effects included 196 CIN3/cancer cases and 330 controls.

HPV testing.

Cervicovaginal samples were tested by PCR for HPV DNA, as described elsewhere (2426). In the Portland cohort, HPV was typed with MY09/11 consensus primers using dot blot. In the Eastern U.S. case-control study, HPV was also typed with MY09/11 consensus primers, but by strip technology. In the Costa Rica cohort, HPV typing was done by both PCR (with MY09/11) and the hybrid-capture tube test.

HLA genotyping.

HLA class I genotypes had been determined previously for all three cohorts (28). Briefly, protocols involving sequence-specific oligonucleotide probe hybridization developed by the 13th International Histocompatibility Workshop (www.ihwg.org/protocols/protocol.htm) were employed.

KIR genotyping.

DNA was available for KIR typing in participants of the Costa Rica cohort and the study of cervical neoplasia in the Eastern U.S. (see Study populations). Genomic DNA was genotyped for presence or absence of the following genes: 2DL1, 2DL2, 2DL3, 2DL4, 2DL5, 2DS1, 2DS2, 2DS3, 2DS4, 2DS5, 3DL1, 3DL2, 3DL3, 3DS1, and 2DP1. Genotyping was performed using PCR amplification with two pairs of locus-specific primers (PCR-SSP) as previously described (19). Internal control primers that amplify a 796-bp fragment of the third intron of DRB1 were also included in each PCR to confirm robust PCR amplifications. Amplification was performed in a volume of 5 μl containing 200 μM dNTP, 500 nM primer, 1.5 mM MgCl2, 20 mM Tris-HCl, pH 8.4, 50 mM KCl, and 0.5 U of Platinum Taq DNA polymerase (Invitrogen). Cycling was performed as follows: 2 min at 94°C; 5 cycles of 94°C for 5 s, 65°C for 15 s, 72°C for 30 s; 21 cycles of 94°C for 5 s, 60°C for 15 s, 72°C for 30 s; 4 cycles of 94°C for 5 s, 55°C for 1 min, 72°C for 2 min; and a final extension step of 10 min at 72°C. PCR products were electrophoresed in 2% agarose gels containing ethidium bromide, and predicted size products were visualized under ultraviolet light.

Statistical methods.

A case-control analysis was conducted in which patients with cancer or CIN3 were compared with population control subjects. HLA-B Bw4, HLA-Cw group 2, and KIR genotype frequencies were initially determined for cases and controls; ORs and 95% CIs were calculated to determine the magnitude and statistical significance of associations (29). In merged analyses, the analyses were adjusted by study to account for potential differences among the populations. Additional adjustment by HPV cofactors for CIN3/cancer endpoints (smoking, parity, oral contraceptive use) did not affect our risk estimates. Because these HPV cofactors were not associated with KIR/HLA genotypes, we present results from the more robust, parsimonious study-adjusted model. Statistical analyses were performed using SAS software (version 8.12; SAS Institute). All tests of statistical significance are 2-sided. Linkage disequilibrium between the group of HLA-B Bw4 and the Cw group 2 alleles was tested by computing D′ (30).

Online supplemental material

Table S1 shows the raw data for Fig. 1. Table S2 shows independent associations between all KIR genes tested and CIN3/cancer. Table S3 shows the raw data for Fig. 2 with the middle genotypes considered separately (Table S3, A) and grouped together (Table S3, B). Table S4 shows the raw data for Fig. 3 with the middle genotypes considered separately (Table S4, A) and grouped together (Table S4, B).

Acknowledgments

We are indebted to E. Frazier for her assistance in the illustration of Fig. 4.

This work was supported with federal funds from the U.S. National Cancer Institute, National Institutes of Health under contract no. NO1-CO-12400. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does the mention of trade names, commercial products, or organizations imply endorsement by the U.S. government.

The authors have no conflicting financial interests.

References

1
Lodoen, M.B., and L.L. Lanier.
2005
. Viral modulation of NK cell immunity.
Nat. Rev. Microbiol.
3
:
59
–69.
2
Biron, C.A., and L. Brossay.
2001
. NK cells and NKT cells in innate defense against viral infections.
Curr. Opin. Immunol.
13
:
458
–464.
3
Long, E.O., and S. Rajagopalan.
2000
. HLA class I recognition by killer cell Ig-like receptors.
Semin. Immunol.
12
:
101
–108.
4
Vales-Gomez, M., H. Reyburn, and J. Strominger.
2000
. Interaction between the human NK receptors and their ligands.
Crit. Rev. Immunol.
20
:
223
–244.
5
Vivier, E., J.A. Nunes, and F. Vely.
2004
. Natural killer cell signaling pathways.
Science.
306
:
1517
–1519.
6
Carrington, M., and P.J. Norman. 2003. The KIR gene cluster. National Library of Medicine (US), National Center for Biotechnology Information, Bethesda, MD. Available at: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=Books.
7
Hsu, K.C., S. Chida, B. Dupont, and D.E. Geraghty.
2002
. The killer cell immunoglobulin-like receptor (KIR) genomic region: gene-order, haplotypes and allelic polymorphism.
Immunol. Rev.
190
:
40
–52.
8
Reyburn, H., O. Mandelboim, M. Vales-Gomez, E.G. Sheu, L. Pazmany, D.M. Davis, and J.L. Strominger.
1997
. Human NK cells: their ligands, receptors and functions.
Immunol. Rev.
155
:
119
–125.
9
Gumperz, J.E., V. Litwin, J.H. Phillips, L.L. Lanier, and P. Parham.
1995
. The Bw4 public epitope of HLA-B molecules confers reactivity with natural killer cell clones that express NKB1, a putative HLA receptor.
J. Exp. Med.
181
:
1133
–1144.
10
Martin, M.P., X. Gao, J.H. Lee, G.W. Nelson, R. Detels, J.J. Goedert, S. Buchbinder, K. Hoots, D. Vlahov, J. Trowsdale, et al.
2002
. Epistatic interaction between KIR3DS1 and HLA-B delays the progression to AIDS.
Nat. Genet.
31
:
429
–434.
11
Biassoni, R., M. Falco, A. Cambiaggi, P. Costa, S. Verdiani, D. Pende, R. Conte, C. Di Donato, P. Parham, and L. Moretta.
1995
. Amino acid substitutions can influence the natural killer (NK)-mediated recognition of HLA-C molecules. Role of serine-77 and lysine-80 in the target cell protection from lysis mediated by “group 2” or “group 1” NK clones.
J. Exp. Med.
182
:
605
–609.
12
Mandelboim, O., H.T. Reyburn, M. Vales-Gomez, L. Pazmany, M. Colonna, G. Borsellino, and J.L. Strominger.
1996
. Protection from lysis by natural killer cells of group 1 and 2 specificity is mediated by residue 80 in human histocompatibility leukocyte antigen C alleles and also occurs with empty major histocompatibility complex molecules.
J. Exp. Med.
184
:
913
–922.
13
Khakoo, S.I., C.L. Thio, M.P. Martin, C.R. Brooks, X. Gao, J. Astemborski, J. Cheng, J.J. Goedert, D. Vlahov, M. Hilgartner, et al.
2004
. HLA and NK cell inhibitory receptor genes in resolving hepatitis C virus infection.
Science.
305
:
872
–874.
14
Franco, E.L.
1995
. Cancer causes revisited: human papillomavirus and cervical neoplasia.
J. Natl. Cancer Inst.
87
:
779
–780.
15
Cook, M.A., P.A. Moss, and D.C. Briggs.
2003
. The distribution of 13 killer-cell immunoglobulin-like receptor loci in UK blood donors from three ethnic groups.
Eur. J. Immunogenet.
30
:
213
–221.
16
Norman, P.J., H.A. Stephens, D.H. Verity, D. Chandanayingyong, and R.W. Vaughan.
2001
. Distribution of natural killer cell immunoglobulin-like receptor sequences in three ethnic groups.
Immunogenetics.
52
:
195
–205.
17
Winter, C.C., J.E. Gumperz, P. Parham, E.O. Long, and N. Wagtmann.
1998
. Direct binding and functional transfer of NK cell inhibitory receptors reveal novel patterns of HLA-C allotype recognition.
J. Immunol.
161
:
571
–577.
18
Hiby, S.E., J.J. Walker, K.M. O'Shaughnessy, C.W.G. Redman, M. Carrington, J. Trowsdale, and A. Moffett.
2004
. Combinations of maternal KIR and fetal HLA-C genes influence the risk of preeclampsia and reproductive success.
J. Exp. Med.
200
:
957
–965.
19
Martin, M.P., G. Nelson, J.H. Lee, F. Pellett, X. Gao, J. Wade, M.J. Wilson, J. Trowsdale, D. Gladman, and M. Carrington.
2002
. Cutting edge: susceptibility to psoriatic arthritis: influence of activating killer Ig-like receptor genes in the absence of specific HLA-C alleles.
J. Immunol.
169
:
2818
–2822.
20
Yen, J.H., B.E. Moore, T. Nakajima, D. Scholl, D.J. Schaid, C.M. Weyand, and J.J. Goronzy.
2001
. Major histocompatibility complex class I-recognizing receptors are disease risk genes in rheumatoid arthritis.
J. Exp. Med.
193
:
1159
–1167.
21
van der Slik, A.R., B.P. Koeleman, W. Verduijn, G.J. Bruining, B.O. Roep, and M.J. Giphart.
2003
. KIR in type 1 diabetes: disparate distribution of activating and inhibitory natural killer cell receptors in patients versus HLA-matched control subjects.
Diabetes.
52
:
2639
–2642.
22
Walboomers, J.M., M.V. Jacobs, M.M. Manos, F.X. Bosch, J.A. Kummer, K.V. Shah, P.J. Snijders, J. Peto, C.J. Meijer, and N. Munoz.
1999
. Human papillomavirus is a necessary cause of invasive cervical cancer worldwide.
J. Pathol.
189
:
12
–19.
23
Hawes, S.E., and N.B. Kiviat.
2002
. Are genital infections and inflammation cofactors in the pathogenesis of invasive cervical cancer?
J. Natl. Cancer Inst.
94
:
1592
–1593.
24
Hildesheim, A., M. Schiffman, D.R. Scott, D. Marti, T. Kissner, M.E. Sherman, A.G. Glass, M.M. Manos, A.T. Lorincz, R.J. Kurman, et al.
1998
. 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.
7
:
1035
–1041.
25
Herrero, R., M.H. Schiffman, C. Bratti, A. Hildesheim, I. Balmaceda, M.E. Sherman, M. Greenberg, F. Cardenas, V. Gomez, K. Helgesen, et al.
1997
. Design and methods of a population-based natural history study of cervical neoplasia in a rural province of Costa Rica: the Guanacaste Project.
Rev. Panam. Salud Publica.
1
:
362
–375.
26
Lacey, J.V. Jr., L.A. Brinton, F.M. Abbas, W.A. Barnes, P.E. Gravitt, M.D. Greenberg, S.M. Greene, O.C. Hadjimichael, L. McGowan, R. Mortel, et al.
1999
. Oral contraceptives as risk factors for cervical adenocarcinomas and squamous cell carcinomas.
Cancer Epidemiol. Biomarkers Prev.
8
:
1079
–1085.
27
Wang, S.S., C.M. Wheeler, A. Hildesheim, M. Schiffman, R. Herrero, M.C. Bratti, M.E. Sherman, M. Alfaro, M.L. Hutchinson, J. Morales, et al.
2001
. Human leukocyte antigen class I and II alleles and risk of cervical neoplasia: results from a population-based study in Costa Rica.
J. Infect. Dis.
184
:
1310
–1314.
28
Wang, S.S., A. Hildesheim, X. Gao, M. Schiffman, R. Herrero, M.C. Bratti, M.E. Sherman, W.A. Barnes, M.D. Greenberg, L. McGowan, et al.
2002
. Comprehensive analysis of human leukocyte antigen class I alleles and cervical neoplasia in 3 epidemiologic studies.
J. Infect. Dis.
186
:
598
–605.
29
Breslow, N.E., and N.E. Day.
1980
. Statistical methods in cancer research. The analysis of case-control studies.
IARC Sci Publ
.
1
:
5
–338.
30
Lewontin, R.C.
1988
. On measures of gametic disequilibrium.
Genetics.
120
:
849
–852.

Supplementary data