Evidence for Chlamydia trachomatis as a Human Papillomavirus Cofactor in the Etiology of Invasive Cervical Cancer in Brazil and the Philippines

Jennifer S. Smith, Nubia Muñoz, Rolando Herrero, José Eluf-Neto, Corazon Ngelangel, Silvia Franceschi, F. Xavier Bosch, Jan M. M. Walboomers, and Rosanna W. Peeling International Agency for Research on Cancer, Lyon, France; Proyecto Epidemiológico Guanacaste, San José, Costa Rica; Universidade de São Paulo, São Paulo, Brazil; Philippine General Hospital, University of the Philippines, Manila; Institu Català d’Oncologia, Barcelona, Spain; Free University Hospital, Amsterdam, The Netherlands; National Laboratory for Sexually Transmitted Diseases, Health Canada, Winnipeg, Canada

Extensive data show that human papillomavirus (HPV) is the central cause of cervical cancer, and high-risk HPV types are associated with a greater disease risk [1]. Only a small proportion of women infected with HPV infection progress to invasive cervical cancer (ICC). Thus, the development of ICC depends on other cofactors acting in conjunction with HPV, such as other sexually transmitted infections (STIs), smoking, hormones, nutritional deficiencies, or host-genetic/immunologic responses.
Clinical observations in the 1970s indicated that genital infection with Chlamydia trachomatis was associated with cervical atypia [2] and cervical neoplasia [3]. Genital C. trachomatis infections may result in chronic cervicitis, pelvic inflammatory disease, and endometritis, whereas ocular C. trachomatis infections may cause trachoma and may result in blindness in some cases. Interest in genital C. trachomatis infection as a potential etiologic factor for ICC has been related to its asymptomatic nature, persistence if left untreated, and induction of metaplasia [4] and chronic inflammation [5].
Two epidemiologic studies examined C. trachomatis in the etiology of ICC and in the context of HPV by using sensitive polymerase chain reaction (PCR)-based assays for HPV DNA detection [6,7]. In a study of cervical neoplasia [6], C. trachomatis seropositivity was a significant risk factor for carcinoma in situ in Spain and Colombia after adjusting for HPV DNA. However, when combined country analyses were restricted to HPV DNA-positive case and control patients, C. trachomatis antibodies were not significantly associated with either in situ carcinoma or ICC. Some misclassification of HPV DNA resulting from the use of a first-generation PCR-based assay could not be ruled out. Another study reported no association between C. trachomatis antibodies and cervical neoplasia after adjusting for HPV DNA among women in Honduras [7]. Results restricted to HPV DNA-positive participants were not presented, although an ELISA known to have C. trachomatis/C. pneumoniae crossreactivity was used. To further examine the role of C. trachomatis infection as a cofactor of HPV in the etiology of ICC, we present results from 2 International Agency for Research on Cancer

Methods
This analysis is based on 2 cervical cancer case-control studies of 499 women with incident ICC (ICC patients) and 539 control patients in São Paulo and Manila, selected as described elsewhere [8,9]. Both studies were part of an IARC multicenter case-control study and used similar protocols and questionnaires for recruitment and data collection.
All women with ICC were newly diagnosed and had received no previous treatment, and their ICC was histologically confirmed. Hospital-based control patients were selected and were frequency age-matched to the case patients by quinquennium of age. Women were not eligible to participate as control patients if they had received any previous cervical cancer treatment or if they had diseases possibly sharing risk factors with cervical cancer (e.g., cardiovascular or cerebrovascular disease, chronic bronchitis, emphysema, or neoplasia of the breast, reproductive or respiratory organs, anus, oral cavity, esophagus, bladder, or liver).
In Brazil, 199 ICC patients (96.6% of eligible subjects) were recruited from 5 public hospitals and 2 cancer hospitals in São Paulo between June 1990 and June 1991. As hospital-based controls, 225 women (94.5% of those eligible) were interviewed and were selected from the 5 public hospitals. In total, 150 women with ICC (75.4% of participants; 137 with squamous and 13 with adenocarcinoma/ adenosquamous cancer) and 173 control patients (76.9%) with chlamydia serology and HPV DNA results were included in this analysis. The main diagnostic categories of Brazilian control patients were diseases of the circulatory system (21.8%), infectious and parasitic diseases (12.9%), diseases of the digestive tract (12.3%), neoplasms (9.5%), and diseases of the nervous system (8.9%).
In the Philippines, 387 ICC patients (100% of those eligible) and 387 control patients (98.7% of those eligible) were identified among patients at the Philippine General Hospital in Manila between April 1991 and April 1993. A total of 349 women with ICC (90.2% of participants: 318 with squamous and 31 with adenocarcinoma/ adenosquamous cancer) and 366 (94.6%) control patients with chlamydia serology and HPV DNA results were included in this analysis. The main diagnostic categories of control patients from the Philippines were urinary tract infections (19.5%), benign disorders of the genital tract (14.3%), menstrual disorders (11.1%), and diseases of the circulatory system (7.9%). Fifty-five healthy women (14.9%) accompanying outpatients, with similar sociodemographic and behavioral characteristics as the hospital-based control patients, were included as control subjects.
Specially trained interviewers administered a standardized questionnaire on sociodemographic characteristics, sexual behavior, and reproductive and contraceptive history. Participating women were asked to provide 10 mL of blood for serologic testing of C. trachomatis and herpes simplex virus (HSV)-2. Blood specimens were processed by centrifugation at the site of collection. The separated serum samples were placed in tubes, frozen at 220 C, and shipped to Lyon, France, for storage. For HPV DNA detection, cervical exfoliated cells from all subjects and cervical biopsy specimens from patients were obtained. Cervical biopsy specimens were kept frozen at 270 C without additives. Histologic slides for cancer diagnosis were reviewed by an expert pathologist. The stage of disease was coded according to International Federation of Gynecology and Obstetrics standards. Cervical exfoliated cells were collected, by sampling the ectocervix with 2 wooden spatulas and the endocervix with 2 cytobrushes, and were placed in tubes with PBS, centrifuged, and stored at 270 C until shipment to a central laboratory for HPV DNA testing.
Laboratory procedures. Serum IgG antibody responses to C. trachomatis were determined by a microimmunofluorescence (MIF) assay, which, at present, is considered to be the most accurate serologic method [10]. The antigen panel consists of purified elementary bodies of C. trachomatis (serovar A and 3 pooled serovar groups of BDE, CJHI, and FGK) and C. pneumoniae [11]. C. pneumoniae was included in the antigen panel to monitor cross-reactive antibody responses and to test for specificity of the C. trachomatis findings. Clinically, C. trachomatis serovars D-K are primarily associated with genital chlamydial infections, whereas serovars A-C are associated with both hyperendemic trachoma and genital infection. Serologic testing was done without knowledge of case-control status.
All serum samples were screened for C. trachomatis at 1:8 dilution and were titered to end point. With the exceptions mentioned below, an IgG titer >1:8 against any of the C. trachomatis serovar groups was considered to be evidence of past infection with C. trachomatis. An IgG titer >1:16 against C. pneumoniae was considered to be evidence of past C. pneumoniae infection. Serum samples that had identical titers for all C. trachomatis and C. pneumoniae species were also tested with C. psittaci (avian strain 6BC) to determine the presence of broad Chlamydia species cross-reactivity. Serum samples with identical titers for all chlamydial species (C. trachomatis, C. pneumoniae, and C. psittaci) were considered to be cross-reactive and were excluded from analyses: 7 Brazilian women (5 ICC patients and 2 control patients) and 4 Philippine women (2 ICC patients and 2 control patients) in this category were excluded. Serum samples from 9 ICC patients and 9 control patients who were seropositive for serovar A and negative for all other C. trachomatis serovars were considered to have evidence of past C. trachomatis ocular trachoma infections and were excluded from analyses.
A blinded reproducibility study was conducted by retesting a random sample of 10% of the serum specimens (n ¼ 105) twice by MIF. C. trachomatis serovars A, BDE, CJHI, and FGK and C. pneumoniae had the following percentages of agreement: 83.9%, 87.5%, 84.8%, 83.9%, and 87.5%, respectively. The k agreement for C. trachomatis positivity for the 2 repeat tests was 0.75 overall: 0.65 for case patients and 0.80 for control patients. The k agreement for C. pneumoniae positivity was 0.6 for the 2 repeat tests overall: 0.56 for case and 0.65 for control patients.
Screening for HPV DNA was done by a PCR assay based on GP5/ 6/TS-PCR and G5 + /6 + primers, as described elsewhere [8,9]. In brief, amplification of a fragment of the b-globin gene served as an internal control for sufficiency of each specimen for amplification. Specimens were labeled "HPV X" for HPV-positive specimens when a specific HPV type could not be determined. GP5/6/TS-PCR primers detect sequenced HPV types 6, 11, 16, 18, 31, and 33, and unsequenced HPV types at subpicogram levels. G5 + /6 + primers detect over 30 [9]. All case and control patient specimens labeled as HPV X were retested with G5 + /6 + primers. HPV types considered as high risk included HPV 16,18,26,31,33,35,39,45, 51-53, 56, 58, 59, 66, 68, 73, IS39, and W13b. To further investigate HPV positivity among subjects, we retested all HPV-negative biopsy specimens and a sample of negative cell pellets from control patients with GP5 + /6 + or E7 HPV PCR primers [12]. With these additional G5 + /6 + and E7 results, HPV DNA prevalence among squamous ICC patients increased in both Brazil (from 88.7% to 92.7%) and in the Philippines (from 93.8% to 95.3%). We took special precautions to minimize contamination, as described elsewhere in detail [12]. To control for any potential effect of HSV-2, HSV-2 antibodies were detected by use of the Gull HSV-2 gG ELISA with Western blot confirmation of all positive, equivocal, and borderline negative ELISA results, as described elsewhere [13].
Because this case-control study was designed to investigate C. trachomatis as an HPV cofactor, only participants having both valid C. trachomatis serology and HPV DNA results could be included in this analysis. The main reason for exclusion was unavailability of HPV results due to b-globin negativity. Participating and nonparticipating women had similar sociodemographic and sexual behavior characteristics (data not shown).
Statistical analyses. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated by multiple logistic regression [14]. ORs for the association were adjusted for 4 age groups (,40, 40-49, 50-59, and >60 years), country of residence, and other confounding factors as stated. Missing values wereexcluded from statistical analyses. Trend tests for categorized data were conducted by assigning sequential scores to the ordered categories.

Results
In total, we studied 455 ICC patients with squamous ICC, 44 with adenocarcinoma/adenosquamous ICC, and 539 agematched control patients from Brazil and the Philippines (table  1). C. trachomatis seroprevalence was significantly higher among all squamous ICC patients (47.7%) than among control patients (22.1%;P< .0001)butwasnotsignificantlyhigheramongadenocarcinoma/adenosquamous ICC patients (29.6%) than among control patients (22.1%; P ¼ :3). The difference in seroprevalence between squamous ICC patients and control patients increased with increasing C. trachomatis antibody titer. Among control patients, C. trachomatis seroprevalence was similar in Brazil (20.2%) and the Philippines (23.0%). C. pneumoniae seropositivity did not significantly differ between ICC patients and control patients but was lower among control participants in Brazil (68.2%) than in the Philippines (82.8%; P , :001). HPV DNA positivity was significantly higher among all ICC patients (93.8%) than among control patients (11.3%; P , :001).
The prevalence of C. trachomatis antibodies did not differ significantly by clinical cancer stage in either Brazil or the Philippines (data not shown) or for ICC patients from the 2 countries combined (combined countries): stage I, 40.2% (n ¼ 87); stage II, 47.8% (n ¼ 180); stage III, 47.5% (n ¼ 200); and stage IV, 50.0% (n ¼ 4; P trend = . 3). Table 2 shows that C. trachomatis seropositivity was significantly associated with sexual behavioral factors among control patients in the age-adjusted analyses. In the multivariate model,  C. trachomatis was positively associated with a woman's report of her husband's sexual behavior and HSV-2 seropositivity. Women ,40 years old had higher C. trachomatis seropositivity than older control patients. After controlling for sexual behavior, Filipino control participants were more likely than Brazilian control participants to be C. trachomatis seropositive. Elevated C. trachomatis antibody titers (.128) among control patients were clearly associated with sexual behavioral factors (data not shown). C. trachomatis seropositivity did not significantly differ by any specific diagnostic category among control participants (data not shown). C. pneumoniae seropositivity was not associated with sexual behavior among control patients.
To determine whether past C. trachomatis infection may increase the risk of being a carrier of HPV infection, we examined associations between C. trachomatis antibodies and HPV DNA among control patients (table 3). Results were similar in Brazil and the Philippines (data not shown). Neither C. trachomatis seropositivity (OR, 1.4; 95% CI, 0.7-2.7) nor high C. trachomatis antibody titers (OR, 1.6; 95% CI, 0.7-3.7) were significantly associated with HPV DNA detection. C. pneumoniae seropositivity was not associated with HPV DNA positivity among control patients.
To more effectively control for the strong effect of HPV, analyses were restricted to squamous ICC patients and control patients with HPV DNA (table 4). Among all HPV DNA-positive par-ticipants, C. trachomatis-seropositive women had a 2-fold increased risk of squamous ICC (OR, 2.1; 95% CI, 1.1-4.0). Results were similar in both countries. Elevated C. trachomatis antibody titers were associated with an increased risk of squamous cervical cancer, with a significant trend (P trend = .01). Further analyses controlling for age at first intercourse, number of sex partners, parity, cytologic screening, smoking, or oral contraceptive use did not significantly affect risk estimates (data not shown). After excluding the additional GP5 + 6 + or E7 HPV PCR results for all HPV-negative biopsy specimens and a sample of negative cell pellets from control patients, recalculated ORs for association between C. trachomatis and ICC were not significantly different from those presented in table 4 (data not shown). The presence of C. trachomatis antibodies was associated with an increased (2-fold) risk for squamous cancer among women with high-risk HPV DNA in the combined country analysis (OR, 2.4; 95% CI, 1.1-5.4; data not shown). The effect of C. trachomatis among HPV-negative participants (OR, 3.4; 95% CI, 1.4-8.3), based on 25 HPV-negative women, was consistent with that among HPV-positive participants (data not shown).
C. pneumoniae infection was not associated with squamous or adenosquamous ICC in either Brazil or the Philippines (data not shown) or in the combined analysis of both countries. The associ-  ations between C. trachomatis and squamous cervical cancer were similar for C. pneumoniae-seropositive and -seronegative women (data not shown).
On the basis of data from 44 ICC patients, C. trachomatis seropositivity was not associated with adenocarcinoma/adenosquamous ICC risk in Brazil or the Philippines (data not shown) or in the combined country analysis (OR, 0.8; 95% CI, 0.3-2.2) among HPV DNA-positive participants after controlling for age, HSV-2 seropositivity, and a woman's reported history of her husband's sexual behavior.
Analyses were conducted for each C. trachomatis serovar group (A, BED, CHIJ, and FGK). The association between C. trachomatis and squamous or adenocarcinoma/adenosquamous ICC was similar in all serovar groups (data not shown).

Discussion
We believe that our study is the largest of C. trachomatis, HPV, and ICC to date that takes into account the strong effect of HPV and type-specific C. trachomatis serology results. These data indicate a moderate, but significant, association between C. trachomatis infection and ICC in the presence of HPV DNA. C. trachomatis seropositivity was consistently associated with an increased risk (2-fold) of squamous ICC among HPV DNA-positive participants, after adjusting for confounding factors. A pattern of increasing squamous ICC risk with increasing C. trachomatis titers was observed, which suggests a doseresponse effect. C. trachomatis was not clearly associated with adenocarcinoma/adenosquamous carcinoma, although this observation was based on a small number of cases.
The positive associations of squamous ICC and C. trachomatis infection in this study (OR, 2.1) are consistent with results from case-control studies of ICC that have used less sensitive methods than PCR for HPV detection, namely, Southern blotting (OR, 4.8) [15] and HPV serology (OR, 1.7-2.2) [16,17]. Nested casecontrol studies have primarily found significant positive associations between C. trachomatis serum antibodies and ICC (OR, 1.5-2.5) [18][19][20], although one smaller study found a nonsignificant 3-fold increase in ICC risk (OR, 3.0; 95% CI, 0.7-13.4) [21]. However, residual confounding due to HPV in these studies cannot be excluded, because less sensitive HPV serologic assays were used to assess HPV status [22].
Due to the notable association between HPV and ICC, a prerequisite to evaluating an HPV cofactor is to accurately assess HPV infection. The PCR-based assay used is one of the most accurate methods to assess cervical HPV infection [23]. Given the notably high prevalence of HPV DNA in cervical cancer worldwide (99%) [12], we concentrated on analyses restricted to HPV DNA-positive women to reduce the likelihood of residual HPV confounding, so as to assess C. trachomatis as an HPV cofactor. However, an association between C. trachomatis antibodies and squamous ICC also emerged among HPV DNA-negative case and control patients.
Our data did not show a significant association between C. trachomatis seropositivity and HPV DNA positivity, which may be due to the different nature of these 2 markers. Whereas HPV DNA indicates both current or persistent HPV infection, C. trachomatis seropositivity represents a more cumulative measure of exposure to C. trachomatis infection.
The use of species-specific MIF serologic testing allows the ascertainment of past C. trachomatis infection while differentiating C.pneumoniae antibodies fromC. trachomatisserovar groupings [11]. Unlike C. pneumoniae, C. trachomatis antibodies were significantly associated with sexual behavior and with squamous ICC. Results from our blinded reproducibility test showed a high agreement between seropositivity to the different C. trachomatis serovars (.83%) and "substantial" k agreement for C. trachomatis positivity [24] for both case and control participants. C. trachomatis IgG antibodies may persist for years among women with acute infections [25], and antibody persistence and elevated titers are considered to be related to longer, more severe, and recurrent chlamydial infections [26]. Reliable data on the natural history and long-term persistence of C. trachomatis antibodies are, however, lacking, although the disappearance of IgG C. trachomatis serum antibodies is considered to be rare in women [10,27].
Among the potential limitations of our study is the use of hospital-based control patients. This could have led to biased results if C. trachomatis seroprevalence among the control patients was not representative of the population source for the ICC patients. Control participants in these studies, however, had a wide range of diagnostic categories and were recruited in tertiary public hospitals with wide reference populations. C. trachomatis seropositivity did not differ significantly by any major diagnostic category among control patients. Although the case-control design did not elucidate the temporal association between C. trachomatis seropositivity and ICC, all ICC patients in our study were newly diagnosed and had not received previous cervical cancer treatment and the frequency of C. trachomatis seropositivity did not vary by the clinical stage of ICC, thus reducing the likelihood that our findings may be attributed to the development of C. trachomatis antibodies following the onset of invasive disease.
The power of our analyses examining C. trachomatis as an HPV cofactor in the etiology of cervical cancer is also limited because of the small number of HPV-positive control patients. Persistent HPV DNA infections have been associated with an increased risk of cervical neoplasia, particularly among women with high-risk HPV types [28]. In this study, a cross-sectional measurement of HPV DNA was taken among case and control patients. The extent to which this HPV DNA measurement among control patients .45 years old represents a persistent HPV infection merits further investigation. However, C. trachomatis antibodies were associated with an increased squamous cancer risk when analyses were restricted to women who had more persistent, high-risk DNA types.
The increasing risk of squamous cervical cancer with increasing C. trachomatis antibody titers gives further support to the results found. High antibody titers may be a marker of persistent C. trachomatis infection, since women with long-term complications of chlamydial infection, such as pelvic inflammatory disease or tubal infertility, have significantly higher levels of MIF antibody than women with cervical chlamydial infection [29].
No specific C. trachomatis serovar group (A, BDE, CJHI, or FGK) was associated with a higher squamous cell cancer risk. The microbial risk factor associated with squamous cancer risk may not be serovar specific. Alternatively, the statistical power to detect meaningful differences between the associations of different C. trachomatis serovar groups and ICC by use of serologic testing may be limited because of common subspecies cross-reactivity and because women may have been infected with multiple serovars. The main shortcoming of this study is that, by use of BDE, CJHI, and FGK serovar group data in the MIF assay, it was not possible to distinguish clearly between exposure to C. trachomatis genital infections and hyperendemic ocular trachoma infections. Notwithstanding, it is difficult to believe that our results can be attributed to an effect of trachoma infection, because both studies were conducted in urban centers without a history of hyperendemic trachoma, and high C. trachomatis antibody titers were significantly associated with sexual behavior. Furthermore, when analyses were stratified by factors potentially associated with trachoma infection (i.e., place of residence or educational status), similar associations were found between C. trachomatis antibodies and ICC, which indicates that the associations found are not likely to be an effect of trachoma infection.
In this study of 44 patients with ICC, adenocarcinoma/adenosquamous carcinoma patients did not have a significantly higher prevalence of C. trachomatis antibodies than control patients. Koskela et al. [19] found similar results among 32 adenocarcinoma patients, which were contrary to expectations because of the tropism of C. trachomatis for endocervical cells. These results require confirmation with a larger sample size.
In terms of biologic plausibility, genital C. trachomatis infections are clinically characterized by cervical atypia and inflammation, as determined by cytologic, histopathologic, and colposcopic examinations [30,31]. C. trachomatis infections may induce immature metaplasia [5], and both HPV and C. trachomatis may infect metaplastic tissue of the squamocolumnar junction, where cervical neoplasia arise [32]. The chronicity of C. trachomatis infection in conjunction with HPV may be a more pertinent factor mediating ICC risk.
Although a carcinogenic interaction between C. trachomatis and HPV has not been directly demonstrated, in vitro data show that C. trachomatis may inhibit cell apoptosis [33], a contributory element for carcinogenesis. Alternatively, inflammatory cytokine responses during a chlamydial infection may produce reactive oxygen species that might cause DNA damage or modification, providing a mechanistic link between chronic inflammation and malignant transformation [34]. Other bacterial or parasitic infections causing chronic inflammation have also been implicated in human cancer, such as Helicobacter pylori with stomach cancer and Schistosoma haematobium with bladder cancer.
Our results, based on a large number of newly diagnosed ICC patients, consistently indicate a potential etiologic role for C. trachomatis infection as an HPV cofactor in the development of squamous ICC. Further epidemiologic studies are needed to clarify the role of C. trachomatis in the etiology of cervical cancer. Additional prospective data are needed on the induction of inflammation by C. trachomatis and other STIs and on the effect of their relative timing, in conjunction with HPV infection, on the risk of cervical neoplasia, in addition to the effect of the treatment of C. trachomatis infection on the progression of cervical neoplasia.