No Association between Endogenous Retinoic Acid and Human Papillomavirus Clearance or Incident Cervical Lesions in Brazilian Women

Epidemiologic research has shown that infection with human papillomavirus (HPV) is a cause of cervical cancer (1). Women infected with HPV are more likely to develop cervical abnormalities such as squamous intraepithelial lesions (SIL). Although HPV is the causal agent of cervical cancer, not all HPV infections lead to dysplasia or cervical cancer. Early events following HPV acquisition are critical to the establishment of an HPV infection that evades the immune system, persists over time, and leads to the development of cervical dysplasia and possibly cancer (2). Because the activity of HPV in cervical cells is tightly associated with epithelial cell differentiation (3), regulators of differentiation, such as retinoic acid (RA; ref. 4), have been considered targets for the prevention of HPV persistence and SIL development (5).

RA is the biologically active form of vitamin A, which is obtained from the diet as retinol esters or as provitamin A carotenoids (6). RA isomers (all-trans-RA, 9-cis-RA, and 13-cis-RA) have different biological activities. All-trans-RA and 9-cis-RA bind to nuclear retinoid receptors that regulate gene transcription (7). RA receptor binds to all-trans-RA and 9-cis-RA with equal affinity, whereas retinoid X receptor preferentially binds 9-cis-RA (8). In comparison, 13-cis-RA has no nuclear receptor and is considered an inert form of all-trans-RA (9). In vitro, RA is a potent growth inhibitor of HPV-16 immortalized cervical epithelial cells (10) and downregulates the expression of HPV-16 E6 and E7 (11–14). In epidemiologic studies, RA precursors (vitamin A and provitamin A carotenoids) have been inconsistently associated with a decreased risk of cervical neoplasia (15). Regression of cervical intraepithelial neoplasia (CIN) II following treatment with all-trans-RA has been reported (16); however, subsequent chemoprevention trials using all-trans-RA, 13-cis-RA, or RA-derivatives did not observe significant differences between treatment and placebo groups for CIN II or CIN III regression (17–21). There has been one published report suggesting a decrease in serum all-trans-RA with increasing cervical lesion severity, which did not take into consideration HPV infection (22). Overall, the association between RA and HPV-associated neoplasia remains undetermined.

Until recently, epidemiologic studies did not have access to a sufficiently sensitive laboratory assay for the measurement of circulating RA, which occurs at very low concentrations. We have previously demonstrated the feasibility of measuring endogenous RA, identified low within-person variability of total RA and individual isomers (all-trans, 13-cis-, and 9-cis-RA), and established that serum RA measures correlate with RA precursors, both circulating and dietary (23). Serum RA seems to be a reliable measure of dietary RA for use in larger-scale epidemiologic studies. Using this novel measure of circulating RA, we evaluated if serum RA levels were associated with early events in cervical carcinogenesis, specifically type-specific HPV clearance and risk of incident SIL. This study of 643 women was part of a larger study investigating antioxidant nutrient and HPV outcomes nested within the Ludwig-McGill Cohort in São Paulo, Brazil (24, 25).

Participants of the Ludwig-McGill Cohort Study were recruited from patients who attended a comprehensive maternal and child health maintenance program between 1993 and 1997 that served low-income families in São Paulo, Brazil (26). Cohort participants were followed every 4 months during the first year and twice yearly thereafter over 9 years. During each study visit, participants underwent a gynecologic exam, provided a blood sample, and were interviewed using a structured questionnaire that was specific for each visit (26). Each study participant signed an approved informed consent document. The institutional review boards and ethical committees of all institutions with which the authors are affiliated approved all study documents and procedures.

During each participant visit, samples of ectocervical and endocervical cells were collected by use of an Accelon biosampler (Medscand, Inc.) for Pap cytology and HPV testing. Pap smears were prepared on a glass slide and fixed in 95% ethanol. The biosampler containing exfoliated cells was swirled in Tris-EDTA buffer (pH 7.4), delivered to the Ludwig Institute laboratory at 4°C, and then archived at −20°C until used for HPV testing. Pap smears were evaluated for cytologic diagnosis according to the Bethesda system [negative for abnormality, atypical squamous or glandular cells of undetermined significance (ASCUS/AGUS), low-grade SIL (LSIL), high-grade SIL (HSIL), cancer, or inconclusive; ref. 27].

Cervical cells were treated with 100 μg/mL proteinase K for 3 hours at 55°C, followed by organic extraction and ethanol precipitation of DNA. HPV DNA was detected by PCR amplification of a 450-bp segment in the L1 viral gene using MY09/11 consensus primers (28, 29), as previously reported (26). Each PCR included positive and negative controls and DNA quality was assessed using amplification of a 268-bp fragment of the β-globin gene. HPV typing was done by dot-blot hybridization with individual oligonucleotide probes specific for 27 HPV genital types (ref. 30; types 6/11, 16, 18, 26, 31, 33, 35, 39, 40, 42, 45, 51-59, 66, 68, 73, 82, 83, and 84). The PCR amplification products were further tested by RFLP analysis of the L1 fragment (31) to resolve ambiguous results from the dot-blot hybridization and to distinguish additional HPVs (32, 34, 44, 61, 62, 67, 69-72, and 89). All HPV analyses were done at the Ludwig Institute for Cancer Research, São Paulo, Brazil.

Nested within the Ludwig-McGill Cohort, we identified a sample of 818 participants for an investigation of antioxidant nutrients and HPV persistence and risk of SIL (24, 25). Selection criteria for the antioxidant nutrient study included available HPV results from all year 1 study visits, archived blood samples available, and normal or ASCUS cervical cytology at enrollment. HPV-positive women were frequency matched on age (±5 years) and enrollment date (±180 days) to HPV-negative women during the same follow-up period. For the current study, valid serum RA results were available for 643 women where values were within ±3 SD of the mean RA. Among the 643 women, 364 were HPV positive, 277 were HPV negative, and 78 developed a LSIL or HSIL as the most severe cytologic grade during follow-up (65 LSIL and 13 HSIL). Among women with an incident SIL lesion, 96% (75 of 78) tested positive for HPV at least once before or at the same visit as the lesion was detected.

All nonfasting blood samples (∼10 mL) were collected by venipuncture into vacutainers by a trained nurse at the time of the clinic visit. The samples were centrifuged within 6 to 8 hours of collection. Aliquots (1 mL) of serum were stored in 1.8-mL Nunc cryovials at −20°C in a non-frost-free freezer until shipped for analyses.

RA analyses were conducted on 500 μL of serum from samples obtained at the first two visits (months 0 and 4), as previously reported (23), using a modification of the method initially reported by Horst et al. (32). All analytic procedures were done under UV-protected lights. Linear calibration curves were prepared consisting of three concentrations of RA isomers that spanned serum physiologic levels. Quantification was done by external standard calibration using peak area ratios. In-house quality control samples were analyzed at the beginning and end of each sample queue. The relative SD of analytes in the quality control samples ranged from 10% to 15%. Quality control samples were the basis of determining the quality and inclusion of results in the final analysis. For statistical purposes, samples with values below the detectable limit of the assay (0.3 ng/mL or 5 pmol/L) were assigned a value halfway between zero and the lower limits of detection, which consisted of 32 (2.5%) all-trans-RA samples of the 1,286 samples analyzed from the 643 women. No 9-cis-RA or 13-cis-RA samples were below the detectable limit. The mean RA level was calculated from each woman for the two measures and then categorized into quartiles based on the HPV-positive group.

For person-level analyses, women infected with single or multiple oncogenic HPV types (HPV types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 66; ref. 33) and women infected with both oncogenic and nononcogenic HPV types were categorized as having an oncogenic infection. Unclassified HPV types were excluded from the analysis of type-specific HPV clearance. Women infected with a single or multiple nononcogenic types were categorized as having a nononcogenic infection. Of the 643 women in this analysis, 248 women were classified as having oncogenic infections and 116 as having nononcogenic infections. In infection-level analyses, women with multiple infections contributed multiple outcome events. Of the 755 HPV infections, 396 (52.5%) were oncogenic and 359 (47.5%) were nononcogenic. HPV-16 was also evaluated as a separate outcome among 91 HPV-16 infections.

Time to HPV clearance for each type-specific HPV infection was measured as the time from the first visit positive for a specific HPV type until the first of two consecutively negative HPV tests (clearance event), or until the last visit if the participant remained positive throughout the study period (censored event). Two consecutively negative tests defined clearance to reduce the possibility of a false-negative test result (e.g., +, −, +), which occurred in 29 women (8% of HPV positive). HPV status was assumed to remain unchanged across single missing visits, but infections spanning two or more consecutive missing visits after the time of infection were excluded (n = 2). HPV infections that became negative at a woman's last study visit were censored. To ensure comparability of clearance assumptions for censored and noncensored infections, we added 6 months to person-time contributed by HPV infections present at a woman's last study visit.

Multivariable Cox proportional hazard modeling was conducted to estimate hazard ratios (HR) and 95% confidence intervals (95% CI) of clearance events adjusted for potential confounders. For infection-level analyses, multivariable Cox proportional hazard modeling was adjusted for clustered data. Mean RA level categorized into quartiles was included as a fixed covariate within the models. In the person-level analyses, women who developed an incident SIL (LSIL or HSIL) during follow-up (cases) were compared with women who did not develop LSIL or HSIL over the same time period (controls). Logistic regression was done to estimate the magnitude of association (odds ratio and 95% CI) between RA isomers and incident SIL, controlling for HPV status or restricting to women that were HPV positive (data not shown).

All multivariable models evaluated factors that could potentially confound or mediate the associations with persistence of HPV infection, SIL, and serum RA levels (23), including age, race, income, education, marital status, smoking, oral contraceptive use, number of pregnancies, age at first sexual intercourse, and number of sexual partners (lifetime or last 5 years). Factors that altered the risk estimate by 10% were retained in the final model. In the infection-level analyses, adjustment factors previously associated with circulating RA and HPV were forced in the model. Tests for trends were done by treating categorized RA variables as continuous in the final models. All statistical tests done were two-sided and declared at the 5% significance level. Statistical analyses were done with SAS (version 9.1.3, SAS Institute) and Intercooled STATA (Stata, Release 9.0, Stata Corp.).

The demographic and risk factor distributions for this sample of women from the Ludwig-McGill Cohort study have been previously published (23, 25). Women were between the ages of 18 and 57 years, with a median age of 32 years. The concentration of RA isomers in this sample is presented in Table 1. The medians (range) of the RA isomer were 1.455 ng/mL (0.505-4.895 ng/mL) for 13-cis-RA, 1.240 ng/mL (0.360-3.225 ng/mL) for 9-cis-RA, and 0.715 ng/mL (0.150-2.625 ng/mL) for all-trans-RA (Table 1). RA isomer levels did not differ between HPV-negative and HPV-positive women, nor did levels significantly differ between SIL cases and controls.

Table 2 presents the risk (HR) of type-specific HPV clearance among the 248 women positive for at least one oncogenic HPV type and the 116 women positive for nononcogenic infection. The probability of clearing an oncogenic HPV infection was not significantly different across RA isomer quartiles. There was a significant trend of increased rate of nononcogenic HPV clearance with increasing all-trans-RA (P = 0.048); however, individual quartiles of all-trans-RA were not significant. In the infection-level analyses, serum RA isomers were not associated with individual oncogenic HPV clearance (Supplementary Table S1). Among the 359 nononcogenic HPV infections, women in the highest quartile of all-trans-RA were 1.55 (95% CI, 1.02-2.36; P-trend = 0.01) times more likely to clear a nononcogenic HPV infection compared with women in the lowest quartile (Supplementary Table S1). Consistent with the analysis that included both prevalent and incident infections, no significant associations were observed when restricting analyses to incident HPV infections in both the woman-level and infection-level analyses (Supplementary Table S2 and S3). Clearance of HPV 16 was significantly less likely among women with total RA above the median (adjusted HR, 0.58; 95% CI, 0.35-0.98) and marginally significant among women with 9-cis-RA level above the median (adjusted HR, 0.63; 95% CI, 0.38-1.06; Table 3). There were no significant associations observed between clearance of HPV 16 and 13-cis-RA and all-trans-RA.

Table 4 presents the relative risk of incident SIL according to endogenous RA isomer levels, controlling for HPV status. The risk of incident SIL was not significantly associated with serum RA isomer level. Results did not differ when analyses were restricted to HPV-positive cases (n = 75) and controls (n = 277; data not shown). To further explore the relationship between endogenous RA and incident SIL, we conducted separate analyses for lesions that occurred within 2 years of RA measurement (n = 46) and those that occurred ≥2 years after RA measurement (n = 32). Risk estimates did not differ between these two approaches (data not shown).

There were no consistently significant associations between serum RA isomers and oncogenic or nononcogenic HPV clearance observed in this study. However, increasing all-trans-RA was marginally associated with increased rate of nononcogenic HPV clearance (P-trend = 0.05). In addition, serum RA levels were not associated with risk of SIL. Based on these findings and the results from completed trials of RA, we conclude that there is no association between RA and duration of an HPV infection or SIL detection.

Findings of a lack of association between serum RA isomers and HPV clearance and incident SIL are consistent with previous reports. Observational studies have suggested that precursors of RA (vitamin A and provitamin A carotenoids; ref. 15) were inversely associated with HPV persistence (24, 34), cervical dysplasia (35–37), and invasive cervical cancer (36, 38). However, epidemiologic findings have been inconsistent, and several reports find no association between vitamin A or carotenoids and HPV persistence (25) or CIN. Of the five placebo controlled chemoprevention trials published (16–18, 20, 21), only one reported a significant reduction of CIN after treatment with all-trans-RA (16). A recent Cochrane review concluded there was no evidence of an effect of retinoids on regression of CIN III and only a possible, but statistically insignificant, effect on regression of CIN II (39). In a reanalysis of one trial (16), all-trans-RA treatment was marginally associated with regression of CIN II (odds ratio, 0.5; 95% CI, 0.25-1.02; P = 0.06; ref. 39). Only one study has evaluated the association between serum all-trans-RA level and cervical lesions. Grace and colleagues, who reported higher levels of all-trans-RA than found in our study, observed a significant inverse relationship between all-trans-RA and increasing cervical lesion severity (22).

The Ludwig-McGill Cohort Study offered a unique opportunity to assess the association between endogenous RA levels and HPV clearance/SIL lesion development in a sample of 634 women. This large well-characterized cohort had up to 9 years of follow-up for the measurements of type-specific HPV infection and diagnosis of SIL lesions. The primary definition of HPV clearance included both prevalent and incident HPV infections; however, the conclusions did not change when analyses were restricted to incident infections. We applied a stringent definition of clearance (two consecutively negative HPV tests) to reduce misclassification and considered both the woman and the virus as the statistical unit of analysis. Due to the fact that the outcome ascertainment was based on cytologic analysis, a potential limitation of this study is misclassification of lesion outcome. However, all cytologic assessments were carefully conducted in a reference laboratory following a strict quality control protocol. The Ludwig-McGill cohort opted for an intensive, expert cytologic review of all subjects in the study every 4 to 6 months and referred all instances of HSIL for colposcopy. This approach reduced the likelihood of unnecessary biopsies, which may interfere with the natural history of early lesions (40). Nevertheless, the occurrence of false-negative Pap tests could have resulted in underestimates (or shorter estimates) of regression time and in either overestimates or underestimates of progression time, depending on whether these test results occurred at lesion outset or during the sojourn period.

A possible limitation of this study is the misclassification of circulating RA, due either to degradation over time or to intraindividual variability. As previously reported, the circulating RA levels measured in this population were similar to previous studies and had low intraindividual variability (23). To further limit the misclassification of RA due to intraindividual variability, the current analyses used the mean of two measures from the same woman. Finally, we acknowledge that cervical RA levels may have a greater effect on the natural history of HPV compared with circulating levels.

Although this is the largest study to evaluate the association between serum RA and HPV events, one may argue that our null findings may be due to a lack of statistical power to detect differences. For each RA isomer, we estimated the minimum mean difference in RA concentration detectable between SIL cases and controls. Allowing for a 5% type I error rate, our study with 78 cases and 565 controls had 80% power to detect a mean difference in total RA of 0.32 ng/mL, assuming a SD of 0.96 ng/mL for each group. This translates into a 9% difference, assuming a mean in the control group of 3.50 ng/mL. Similarly, for each specific RA isomer measured, we had the statistical power to detect between a 10% and a 13% difference by SIL group (data not shown). Overall, we had the statistical power to detect small percent differences in mean RA isomer level by case and control groups, if differences did exist. The power estimates presented do not take into consideration possible misclassification of RA levels and may have overestimated the power of our study. It is possible that a much larger sample size might have found a statistically significant weak effect; however, the clinical significance of a significant weak effect is unclear.

Based on the consistency of a lack of an association between HPV and SIL with a sample size large enough to detect ≥9% difference in means, we conclude that there is no association between serum RA levels and HPV events. Our findings support the results from chemoprevention trials that did not find an association between treatment with RA and regression of cervical dysplasia. In conclusion, the role of RA in inhibiting HPV-induced carcinogenesis, as shown in vitro, lacks confirmatory evidence within epidemiologic studies.

No potential conflicts of interest were disclosed.

We thank Dr. Ji-Hyun Lee for providing statistical expertise during manuscript preparation; Maria L. Baggio and Lenice Galan for management of the patients and specimen collections; and Silvaneide Ferreira and Raquel Hessel for data entry, sample retrieval, and shipment and laboratory analysis.

Grant Support: National Cancer Institute grants CA70269 and CA81310, National Cancer Institute Cancer Prevention and Control Predoctoral Fellowship grant R25CA078447, Canadian Institutes of Health Research (CIHR) grants MA-13647 and MOP-49396, and an intramural grant by the Ludwig Institute for Cancer Research.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.