|Home | About | Journals | Submit | Contact Us | Français|
We examined host genetic factors to identify those more common in individuals whose human papillomavirus (HPV) infections were most likely to persist and progress to cervical intraepithelial neoplasia grade 3 (CIN3) and cancer.
We genotyped 92 single-nucleotide polymorphisms (SNPs) from 49 candidate immune response and DNA repair genes obtained from 469 women with CIN3 or cancer, 390 women with persistent HPV infections (median duration, 25 months), and 452 random control subjects from the 10,049-woman Guanacaste Costa Rica Natural History Study. We calculated odds ratios and 95% confidence intervals (CIs) for the association of SNP and haplotypes in women with CIN3 or cancer and HPV persistence, compared with random control subjects.
A SNP in the Fanconi anemia complementation group A gene (FANCA) (G501S) was associated with increased risk of CIN3 or cancer. The AG and GG genotypes had a 1.3-fold (95% CI, 0.95–1.8-fold) and 1.7-fold (95% CI, 1.1–2.6-fold) increased risk for CIN3 or cancer, respectively (Ptrend = .008; referent, AA). The FANCA haplotype that included G501S also conferred increased risk of CIN3 or cancer, as did a different haplotype that included 2 other FANCA SNPs (G809A and T266A). A SNP in the innate immune gene IRF3 (S427T) was associated with increased risk for HPV persistence (Ptrend = .009).
Our results require replication but support the role of FANCA variants in cervical cancer susceptibility and of IRF3 in HPV persistence.
Human papillomavirus (HPV) infections are very common, sexually transmitted infections that infrequently persist. It is now understood that persistent infection with 1 of ~15 HPV types is required for the development of cervical cancer and its immediate precursor, cervical intraepithelial neoplasia grade 3 (CIN3) [1, 2]. However, the factors that lead a subset of individuals infected with these oncogenic viruses to have persistent infection and to develop CIN3 or cervical cancer, whereas the vast majority of infected individuals naturally clear their infection(s), are poorly understood.
Host genetic factors are hypothesized to play a role in the pathogenesis of cervical cancer. Efforts to date in cervical cancer etiological research have focused on understanding the role of HPV, but much remains unknown about the role of host genetic factors. Current evidence for the role of host genetics in cervical cancer derives from studies such as those conducted in Scandinavian countries with well-established population registries, where evidence for familial aggregation in cervical cancer incidence has been demonstrated. The risk associations reported were strongest for full relatives, intermediate for half siblings, and lowest for nonbiological relatives, suggesting a genetic effect .
More direct evidence that inherited genetic factors play a role in cervical cancer pathogenesis comes from studies that have shown associations between specific human leukocyte antigens (HLA) and cervical cancer [4, 5]. In brief, HLA is essential for the presentation of viral antigens, and polymorphisms within HLA have been hypothesized to be involved in the pathogenesis of cervical neoplasia via their role in the immunological control of HPV. Most notably, a consistent protective effect has now been demonstrated for the HLA DRB*1301-DQB1*0603 haplotype . A positive association between HLA B7/DQB1*0302 and cervical disease has also been demonstrated in several populations, including the cohort in Costa Rica studied in this analysis . Other than HLA, no other gene polymorphisms have consistently demonstrated an association with cervical cancer. Results from numerous reports on the codon 72 polymorphism in TP53 remain equivocal [6–8], and reports of associations between interleukin (IL)–10 and other candidate genes require further replication [9–15]. To date, a systematic evaluation of multiple gene polymorphisms postulated to play a role in cervical cancer has not been performed.
We evaluated the association of 92 candidate single-nucleotide polymorphisms (SNPs) in 49 immune response and DNA repair genes—selected on the basis of previous evidence of functional consequence or reported association with cervical cancer, HPV infection, or other infections—with risk of HPV persistence and progression to CIN3 or cervical cancer in the population-based Guanacaste cohort in Costa Rica (table 1). We hypothesized that immune response gene polymorphisms would affect risk for HPV persistence and progression to cancer by modulating the immune response. We also hypothesized that DNA repair genes would play a role in progression through their function in identifying and repairing DNA damage caused by HPV or HPV cofactors. We evaluated selected genetic variants on the basis of prior laboratory evidence that suggested functional consequences for an allele or associations with cervical cancer or viral infection in previous studies.
The present study was nested within a population-based cohort study of 10,049 women in Guanacaste, Costa Rica. In brief, the Guanacaste Study involved a population-based cohort of 10,049 women recruited over an 18-month period in 1993–19944 and followed for 7 years. For cohort participants, cervical cells were available for HPV testing, as described elsewhere [16, 17], and buffy coat specimens were available for host-gene polymorphism testing.
For the present analysis, we selected the following women from the cohort study: (1) all women histologically confirmed to have prevalent or incident CIN3 or cancer (n = 184; median age, 36 years [range, 18–86 years]); (2) all women with evidence of HPV persistence, defined as women who tested positive for the same HPV type at 2 consecutive visits (n = 432; median duration of persistence, 25 months [range, 5–93 months]); and (3) a random selection of control subjects from the population-based cohort (n = 492). We note that 403 (82%) of the random control subjects were HPV negative at study enrollment, 54 (11%) were positive for a single type, and 34 (7%) were positive for >1 HPV type. We also identified additional individuals with CIN3 or cancer from Guanacaste who received a diagnosis of CIN3 or cancer during the period in which our cohort study was conducted (hereafter referred to as “supplemental case patients”). These supplemental case patients were initially identified from review of the Costa Rican National Tumor Registry and review of cytology listings from the National Cytology Laboratory in Costa Rica, followed by review of hospital and/or pathology records to verify that they had had CIN3 histologically confirmed. Of 448 women identified as eligible, 56 (13%) were deceased, 18 refused (4%), 39 (8%) could not be found, and 4 (1%) were sick or pregnant, with the result that 331 (74%) of the women were included as supplemental case patients (median age, 42 years; range, 20–89 years). We note that although a third of the case patients in the cohort study had cancer, half of the supplemental case patients had cancer, resulting in a slightly higher median age. Twenty milliliters of peripheral blood were collected from the enrolled supplemental case patients, and DNA was extracted from one 10-mL tube. The study was approved by both the National Cancer Institute and Costa Rican institutional review boards, and all subjects provided signed informed consent in accordance with US Department of Health and Human Services guidelines.
DNA was extracted from buffy coats with PureGene purification kits (Autopure; Gentra Systems) at Sera-Care (Frederick, Maryland). For the supplemental case patients, DNA extraction was done at the University of Costa Rica.
For specimens from the cohort study, cervical cells collected with cervix brushes and stored in standard transport medium were used for polymerase chain reaction–based HPV DNA testing with L1 MY09/MY11 consensus primer methods [16, 18, 19]. Supplemental case patients did not have cervical cells collected for HPV testing.
Of the 1439 women selected, 454 random control subjects, 390 women with HPV persistence, 149 women with CIN3 or cervical cancer from the cohort study, and 322 supplemental case patients with CIN3 or cervical cancer had sufficient buffy coat DNA for genotyping. We selected SNPs with ≥5% prevalence in the control group and evidence of functional consequence or hypothesized association with cervical cancer, HPV infection, or other viral infections. Genotyping was conducted at the National Cancer Institute Core Genotyping Facility (Gaithersburg, Maryland) by use of Taqman (Applied Biosciences) or Epoch Biosciences platforms. Sequence data and assay conditions are provided online (Cancer Genome Anatomy Project SNP500Cancer Database, http://snp500cancer.nci.nih.gov) . For quality control (QC), we included duplicate samples from 100 participants, which were interspersed for all assays and to which the laboratory was blinded. Agreement for QC duplicates was 99% for all assays except IRF3 (rs2304204), which was excluded from further analysis because of a 9.6% discordance among QC specimens. For each plate of 368 samples, the following 4 genotype-specific QC samples were also included: homozygote wild-type (WT), heterozygote, homozygote variant, and DNA-negative controls.
No SNPs failed genotyping. Successful genotyping was achieved for 96%–100% of DNA samples for all SNPs. One SNP (IL6 [rs1800795]) was not in Hardy Weinberg equilibrium (P < .01) among control subjects; genotype assignments and QC data from replicates were thus rechecked for this SNP and the accuracy of this assay was confirmed, in accordance with the sequence and assay specifications on the SNP500 Web site.
We evaluated a total of 1312 women: 470 women who had received a diagnosis of CIN3 or cancer, 390 women with persistent HPV infection, and 452 random control subjects for whom genotyping results were obtained.
We calculated odds ratios (ORs) and 95% confidence intervals (CIs) for each genotype with respect to each disease outcome, using the homozygous WT genotype as the reference group. We first compared individuals with CIN3 or cervical cancer to random control subjects. Of the gene variants that were statistically significantly associated with CIN3 or cervical cancer (P < .05), we further evaluated whether their associations were consistent for HPV persistence and/or disease progression with the following respective comparisons: women with HPV persistence versus random control subjects and women with CIN3 or cervical cancer cases versus women with HPV persistence. For SNPs for which no differences in genotype frequencies were identified between women with HPV persistence and women with CIN3 or cervical cancer, we also compared the combined group of women with CIN3 or cervical cancer and women with HPV persistence (n = 860) to random control subjects (n = 452). Similarly, for those SNPs for which no difference in genotype frequencies were identified between women with HPV persistence and random control subjects, we also compared women with CIN3 or cervical cancer (n = 470) to the combined group of women with HPV persistence and random control subjects (n = 842) for increased analytic power.
We conducted both crude and analyses adjusted for age (<30 years, 30–49 years, and ≥50 years). For each outcome, we calculated the Ptrend value on the basis of the 3-level ordinal variable (0, 1, and 2) of homozygote WT, heterozygote, and homozygote variant in a logistic regression model. For the evaluation of associations with HPV persistence, we also conducted analyses in which HPV persistence in the 390 women was restricted to the following groups: (1) women infected with any oncogenic HPV strain (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and/or 68) (n = 180), (2) women with persistent HPV-16 infection (n = 41), and (3) women who had persistent HPV infection for ≥2 years (n = 199; there were 82 women infected with an oncogenic strain of HPV in this group). In analyses restricted to infection with HPV-16, we also evaluated women with CIN3 or cervical cancer who tested positive for HPV-16 (n = 25).
In addition, we conducted analyses that restricted random control subjects to women who self-reported being sexually active; however, because the 11 virgins excluded in these analyses subsequently reported sexual activity during study follow-up and because results of analyses that excluded them were virtually identical to results of analyses that included them, we present data that includes all control subjects. All logistic regression models were unconditional and conducted using SAS software (version 8.2; SAS Institute).
Because some of our results could be false-positive findings due to chance, we calculated the probability that our findings were false-positives, using the Benjamini-Hochberg method to calculate the false discovery rate (FDR) , which reflects the expected ratio of false-positive findings to the total number of significant findings. We applied the FDR method to the Ptrend values, allowing for the fewest comparisons (degrees of freedom) to assess the additive model. We considered an FDR value of <0.2 as notable. Because the FDR does not consider prior probability, we also calculated the false-positive report probabilities (FPRP) . We used a criterion of 0.2, as suggested by Wacholder et al. . Using this criterion, we expected that 20% of tests with FPRP below 0.2 would have false-positive results, if our prior probabilities were correct. Prior values were determined on the basis of the weight of the scientific evidence for the outcomes of interest.
The haplotype structure for FANCA (rs2239359, rs7190823, and rs7195066) was examined using Haploview (version 3.11) . We estimated haplotypes by using the expectation-maximization algorithm . By use of the statistical package HaploStats in R (version 2.0.1; CRAN) , overall differences in haplotype distribution between study groups were assessed with the global score test . Risk estimates were estimated from the additive model, which fitted a logistic regression model and used posterior probabilities of the haplotypes as weights to estimate the regression coefficients in an iterative manner, adjusting for age.
We identified polymorphisms in 4 DNA repair genes (FANCA, EXO1, CYBA, and XRCC1) and 2 immune response genes (IRF3 and TLR2) that were statistically significantly associated (P < .05) with CIN3 or cervical cancer when compared with the genotype distribution of those polymorphisms in random control subjects (table 2). Specifically, the IRF3 S427T (rs7251), EXO1 T439M (rs4149963), CYBA 3′UTR (rs7195830), and FANCA G501S (rs2239359) polymorphisms demonstrated increased risk for CIN3 or cervical cancer for each additional variant allele with Ptrend values of .01, .02, .04, and .008, respectively. The TLR2 S450S (rs3804100) and XRCC1 Q399R (rs25487) polymorphisms demonstrated decreased risk with each additional variant with Ptrend values of .02 and .03, respectively. Haplotype analyses of the 3 FANCA polymorphisms that were genotyped (FANCA G809D [rs7195066], G501S [rs2239359], and T266A [rs7190823]) demonstrated that, consistent with the individual SNP results, the haplotype with only the FANCA G501S (A-G-G) variant conferred the highest risk for CIN3 or cervical cancer (OR, 1.8 [95% CI, 1.4–2.5]). We note, however, that the haplotype with variants in the 2 other FANCA polymorphisms also yielded statistically significant risk elevation (OR for G-A-A, 1.6 [95% CI, 1.1–2.2]) for CIN3 or cervical cancer, when compared with the most common haplotype (A-A-G). We further evaluated SNPs in these 6 genes to determine whether their effects were more pronounced for HPV persistence or for progression to CIN3 or cervical cancer.
Polymorphisms in IRF3 S427T and XRCC1 Q399R were significantly associated with HPV persistence with Ptrend values of .04 and .03, respectively, when women with persistent HPV infection were compared with random control subjects (table 3). With the added power of combining the group of women with CIN3 or cervical cancer and the group of women with persistent HPV infection, compared with random control subjects, we observed a 1.3-fold risk increase (95% CI, 1.0–1.7) for IRF3 and a 1.5-fold risk increase (95% CI, 1.1–2.1) for the CG and CC genotypes, respectively (referent, GG) (Ptrend = .009). For XRCC1 Q399R, we observed decreased risk for the AG and AA genotypes with odds ratios of 0.81 (95% CI, 0.64–1.0) and 0.61 (95% CI, 0.41–0.91), respectively (Ptrend = .009).
Results for the persistence of infection with oncogenic strains of HPV were consistent, although they were statistically significant only for the XRCC1 Q399R polymorphism (women with persistent HPV infection and women with CIN3 or cervical cancer versus random control subjects, OR for AG, 0.70 [95% CI, 0.49–1.0] and OR for AA, 0.49 [95% CI, 0.26–0.93]; Ptrend = .009). In analyses restricted to women with persistent HPV infection for ≥2 or more years, associations with IRF3 were pronounced, with odds ratios of 1.8 (95% CI, 1.2–2.7) and 2.0 (95% CI, 1.2–3.2) for the CG and CC genotypes, respectively (Ptrend = .005); these results were consistent and statistically significant for women with HPV persistence due to oncogenic strains and women with persistent infection due to nononcogenic strains. Results from analyses restricted to women infected with HPV-16 were consistent with the overall results but not statistically significant (data not shown).
TLR2 S450S, EXO1 T439M, CYBA 3′UTR, and FANCA polymorphisms were associated with the risk of progression to CIN3 or cervical cancer. Risk for each additional variant allele was elevated for EXO1, CYBA, and all 3 FANCA polymorphisms and decreased for TLR2 S450S when women with CIN3 or cervical cancer were compared to women with HPV persistence only (table 4). Again, with the added power of comparing women with CIN3 or cervical cancer to women with HPV persistence combined with random control subjects, we observed more significant Ptrend values for EXO1 (Ptrend = .007), CYBA (Ptrend = .01), FANCA G501S (Ptrend = .01), and TLR2 S450S variant allele (Ptrend = .03). Notably, the FANCA haplotype with the FANCA G501S variant and the haplotype with both the FANCA G809D and T266A variants demonstrated increased risk for CIN3 or cervical cancer (OR for A-G-G, 1.6 [95% CI, 1.2–2.2] and OR for G-A-A, 2.4 [95% CI, 1.6–3.6]), compared with the common A-A-G haplotype.
In analyses that compared women with CIN3 or cervical cancer to women with persistent infection due to oncogenic HPV types, associations for TLR2 S450S, EXO1 T439M, FANCA G809D and FANCA T266A remained statistically significant with Ptrend values of .02, .04, .02, and .03, respectively. All also remained statistically significant in analyses comparing women with CIN3 or cervical cancer to women persistently infected with oncogenic HPV types ≥2 years (table 5, which is only available in the electronic version). In addition, results were consistent in analyses restricted to women persistently infected with HPV-16 (data not shown).
We note that the all FDR values based on the Ptrend were above our predefined notable threshold of 0.2 after taking into account all SNPs tested. By the FPRP, we found that the association between FANCA G501S and CIN3 or cervical cancer (assigning a prior probability of 0.05) resulted in a FPRP value of 0.17 for an additive model with an OR of 1.5, suggesting a <20% chance of being a false-positive, given our prior probability. No other results were found notable by FPRP.
In our evaluation of selected immune response and DNA repair gene variants and their association with HPV persistence and progression to cervical cancer, we report that common variants in genes influencing DNA damage were associated with both HPV persistence and progression to CIN3 or cervical cancer and genes influencing immune response were associated with HPV persistence.
Most notably, polymorphisms within the DNA repair gene FANCA were associated with CIN3 or cervical cancer but not with HPV persistence. FANCA is 1 of 12 groups of genes within the Fanconi anemia pathway and is thought to play a role in the recognition of DNA damage and the repair of DNA damage by homologous recombination. FANCA is the major gene implicated in Fanconi anemia (FA) with FANCA mutations accounting for ~70% of all FA cases. FA patients are characterized by increased apoptosis in hematopoietic cells, chromosome instability, sensitivity to DNA cross-linking damage, DNA damage from oxidative stress and/or reactive oxygen species, and telomere shortening . FA patients are susceptible to cancer, including cervical cancer and other HPV-associated tumors , and our results, which implicate variants in FANCA with disease progression, add to the current understanding of FA and cervical cancer. Our data suggest that, in addition to FA mutations, FA variants may be an important host event involved in susceptibility to cervical cancer.
We also identified associations for DNA repair genes EXO1 and CYBA with progression to CIN3 or cervical cancer and associations for XRCC1 with HPV persistence. The association we observed between EXO1 and CYBA and disease progression supports the involvement of DNA repair in cervical pathogenesis and progression to CIN3 or cervical cancer [29, 30]. XRCC1 plays a role in base-excision repair of spontaneous and induced DNA damage [31–34], and its association with HPV persistence (both overall and of oncogenic HPV strains) was not expected. Although it is possible that the increased susceptibility to DNA damage among women with the XRCC1 R399Q variant facilitates HPV persistence, our findings require replication and further investigation.
Of the immune response genes evaluated, a variant in the innate immunity gene IRF3 was associated with HPV persistence, and a TLR2 variant was associated with progression to CIN3 or cervical cancer. Our results for IRF3 are consistent with our hypothesis about immune response genes involved in persistent infection and consistent with the growing literature on IRF3 in viral infections, including herpes simplex virus 1 infection and hepatitis C persistence [35–37]. We note that the association with HPV persistence was further pronounced when persistence was limited to women persistently infected for ≥2 years, but the associations were equally significant for persistence for ≥2 years of infection with either oncogenic or nononcogenic HPV strains, consistent with IRF3’s role in innate immunity. Finally, Toll-like receptors are essential for response to bacterial infections and inflammatory response [38, 39], and our observed association between the TLR2 S450S variant and CIN3 or cervical cancer support the potential role of innate immune response genes and inflammatory response in progression from HPV persistence to CIN3 and cervical cancer.
Study limitations include potential survival bias, as supplemental case patients diagnosed outside the Guanacaste cohort were retrospectively ascertained, and DNA could not be obtained from deceased individuals. However, since half of supplemental case patients had CIN3, survival bias is unlikely to affect our results. Because of our limited number of patients with invasive cancer, the use of CIN3 or cancer as a case group may have obscured associations specific to invasion. Finally, although we targeted predefined genes and intended for our evaluation to be hypothesis generating, we cannot exclude the possibility that some of our results are false-positives (or false-negatives). Although we did not find FDR values above our predefined notable threshold of 0.2 after taking into account all SNPs tested, we note that the association between FANCA G501S and CIN3 or cervical cancer (assigning a prior probability of .05) was notable by FPRP (0.17), indicating a <20% chance of being a false-positive, given our prior probability. Finally, study limitations also include our evaluation of a relatively small proportion of these genes, as permitted by our candidate SNP selection process, which was based largely on available biological evidence and validated assays. For example, full coverage of the FANCA gene using tag SNPs would require an additional 18 SNPs. Thus, the fact that we did not select tag SNPs or conduct other, more comprehensive analyses within each candidate gene may have decreased our ability to identify significant SNPs related to disease. We therefore cannot discount the possibility that some of the genes evaluated could be associated with disease but are not thus identified in our analyses because of the limited number of SNPs evaluated.
Study strengths include our population-based study design, which approximated a case-cohort design because the proportion of women in the cohort with evidence of CIN3 or cancer was small. In our analysis, we have excluded women with a CIN2 diagnosis from both the case and control groups, because it is frequently misclassified. The final 3 comparison groups (women with CIN3 or cervical cancer, women with HPV persistence, and random control subjects) allowed evaluation of both HPV persistence and disease progression.
In summary, our results require replication but, to our knowledge, we are the first to report potential host genetic variants relevant for HPV persistence and those relevant for progression to CIN3 or cervical cancer. If replicated, functional studies to determine the biological relevance of confirmed variants should be pursued. Improved gene coverage of the implicated genes (e.g., FANCA and IRF3) and evaluation of additional genes within these DNA repair and immune response pathways can help refine and develop the findings reported here, if real. Finally, future efforts should include evaluating the interplay between viral and host genetics along with HPV cofactors in determining the risk of HPV persistence and of progression to CIN3 or cervical cancer.
We are grateful to Sabrina Chen from the Information Management Services for data management and programming support. We are indebted to Robert Welch for his management of this genotyping effort at the National Cancer Institute Core Genotyping Facility and for his scientific contributions to our research efforts. We are also grateful to Jill Koshiol for her careful review of this manuscript.
This work was supported by the National Cancer Institute Intramural Program and National Institutes of Health (contracts CO-12400, CP-21081, and CP-31061 to R.H.; grant CA-78527 to R.D.B.); FUCODOCSA (Costa Rican Foundation for Training in Health Sciences); Caja Costarricense de Seguro Social (Costa Rica).
Potential conflicts of interest: none reported.
Presented in part: 24th International Papillomavirus Conference and Clinical Workshop, Beijing, China, 3–9 November 2007 (abstract 340).