Prospective Study of JC Virus Seroreactivity and the Development of Colorectal Cancers and Adenomas

Infection with JC virus has been proposed as a potential risk factor for cancer, including colon cancer. JC virus is a common polyomavirus infection, with antibody seroprevalence estimates ranging from 44% to 75% among adults in the United States (1). Initial infection with JC virus is asymptomatic and usually occurs in later childhood and adolescence, after which the virus remains latent in the kidneys (2). Reactivation of latent JC virus infection can occur under conditions of severe immunosuppression, causing progressive multifocal leukoencephalopathy in AIDS patients. However, JC virus is detected in 37% to 47% of urine samples from immunocompetent individuals (3-5), suggesting that reactivation of latent infection with active JC virus replication is a common phenomenon among healthy adults. The route of JC virus person-to-person transmission is unclear. JC virus has been detected at high levels in raw sewage from urban areas throughout the world (6) and in shellfish (7), indicating that humans may be exposed to JC virus through food and water contaminated by urine and/or feces (7, 8). JC virus DNA can remain intact at low pH levels (9), supporting its viability as an infection of the gastrointestinal tract, and JC virus viral sequences have been identified from normal human colon mucosa (10).

Several laboratory studies suggest that JC virus play a role in carcinogenesis. JC virus encodes a nonstructural protein called the large tumor (T)-antigen, which initiates viral DNA replication, stimulates host DNA synthesis, and modulates gene transcription (2). Large T-antigen has been shown to impair DNA repair processes (11) and can inhibit apoptosis by binding to and inactivating the tumor suppressor proteins p53 and pRb (12). In addition, recent evidence supports a role for JC virus T-antigen in the disruption of the Wnt signaling pathway implicated in colorectal cancer carcinogenesis (13-15); coexpression of β-catenin and JC virus T-antigen results in increased transcription of c-myc (14), and JC virus–transfected cells exhibit nuclear accumulation of β-catenin and chromosomal aberrations, specifically when JC virus T-antigen is expressed (15).

Given the widespread exposure to JC virus and the oncogenic potential of the virus, several laboratories have investigated the prevalence of JC virus in human colorectal cancer tissues and adenomas. Seven studies detected JC virus DNA sequences in colorectal cancer tissues by PCR, ranging in prevalence from 26% to 96% (13, 16-21). Two of the positive studies observed higher viral loads in colorectal cancer compared with paired adjacent normal mucosa (18, 21), although the absolute viral load in cancer tissues was low (1 copy per 100 cells; ref. 18). Four studies investigated JC virus in adenomas, detecting JC virus DNA in 5% to 82% of samples tested (17, 20-22). JC virus T-antigen expression was detected by immunohistochemistry in 77% of JC virus DNA–positive colorectal cancer tissues (13) and 10% to 20% of JC virus DNA–positive adenomas (20, 22). In contrast to these positive studies, two studies with larger sample sizes (n = 100 and 233) did not detect JC virus DNA in colorectal cancer tissues (23, 24). The reasons for inconsistency across tumor studies may include differences in sample preparation, JC virus detection methods, and/or the underlying patient populations. Only one study has investigated the association between JC virus infection and colorectal cancer risk using serology: among 386 colorectal cancer cases and 386 matched controls selected from male participants in a prospective cohort in Norway, antibodies against JC virus at baseline were not associated with an increased risk of developing colorectal cancer (25).

To further investigate the potential role of JC virus in colorectal cancer, a prospective study of JC virus antibodies and incident colorectal cancer and adenomas was conducted in males and females.

A nested case-control study was conducted within two community-based cohorts in Washington County, Maryland. The two cohorts were established in 1974 (n = 23,951) and 1989 (n = 25,080), and named CLUE I and II, respectively, referring to the recruitment campaign slogan, "Give us a clue to heart disease and cancer." After obtaining written informed consent from all participants, serum (1974) and plasma (1989) were obtained and stored at −70°C. (The term “serum” will be used to describe both types of samples from this point forward.) Participants completed a brief baseline questionnaire at the time of blood donation. Additional follow-up questionnaires were mailed to CLUE II participants in 1996, 1998, 2000, and 2003.

Cases of colorectal cancer occurring among CLUE cohort members through July 2006 were identified by linkage to the Washington County Cancer Registry, which has been maintained since 1958, and linkage to the Maryland Cancer Registry since 1992. ICD-9 codes (153 and 154) were used to identify colorectal cancer cases diagnosed in 1992-2000, and ICD-10 codes have been in use since 2001 (C18, C19, and C20). Cases were defined as participants of CLUE I or CLUE II who were subsequently diagnosed with colorectal cancer, in which colorectal cancer was their first cancer diagnosis with the possible exceptions of nonmelanoma skin cancer or cervical cancer in situ. Cases had to have been Washington County residents at both the time of baseline blood donation and colorectal cancer diagnosis. A total of 611 colorectal cancer cases were identified among cohort participants, including 118 who participated in both CLUE I and II and contributed blood samples from both timepoints (n = 729 blood samples from 611 colorectal cancer cases). One control was matched to each colorectal cancer case on sex, race, age within 1 y, cohort (CLUE I, CLUE II, or both), and date of blood draw within 2 wk (n = 729 blood samples from 611 matched controls, including 118 controls who participated in both CLUE cohorts). Controls for the colorectal cancer cases were defined as residents of Washington County at the time of blood donation who were not known to have died or developed cancer (except for possibly nonmelanoma skin cancer or cervical cancer in situ) as of the date of diagnosis of the case. Vital status was determined through daily searches of obituaries, monthly reviews of county death certificates, annual reviews of state death certificates, and the National Death Index.

Cases of colorectal adenomas and matched controls were selected for a previous study of inflammation conducted within the CLUE II cohort, and these selection methods have been previously described in detail (26). Briefly, CLUE II participants were asked in the follow-up questionnaires if they had ever undergone a colonoscopy or sigmoidoscopy, and if so, whether a polyp was diagnosed. After obtaining permission, diagnoses were confirmed through medical record review, and cases were restricted to those with a first diagnosis of an adenomatous polyp after cohort enrollment in 1989, with no history of ulcerative colitis. A total of 135 cases of colorectal adenomatous polyps were confirmed, 123 of whom had blood available for the current analysis of JC virus seroreactivity. For each case, adenoma size (in cm) and site (distal colon, proximal colon, or rectum) were abstracted from the endoscopy report, and histology (villous, tubulovillous, or tubular) was obtained from pathology reports. Controls for the adenoma cases were selected from CLUE II participants who reported having an endoscopy after 1989, but also reported that no polyps were detected. Controls could have no history of cancer (except nonmelanoma skin cancer or cervical cancer in situ) or self-reported polyp diagnosis through the end of follow-up. One control was matched to each adenoma case on age, race, sex, date of blood draw, date of endoscopy within 1 y, and region of the colon visualized on endoscopy (i.e. for cases with a polyp in the proximal colon, matched controls had to have had a negative colonoscopy; cases with a polyp in the distal colon could have been matched with controls who had a negative colonoscopy or sigmoidoscopy; ref. 26).

Serum and plasma samples were shipped from the George W. Comstock Center for Public Health Research and Prevention in Hagerstown, MD to the Johns Hopkins School of Medicine in Baltimore, MD for measurement of antibodies to the JC virus capsid protein, VP1. JC virus virus-like particles were produced using a recombinant baculovirus expressing JC virus VP1 (27), the amino acid sequence for which is from the National Center for Biotechnology Information (NCBI) reference genome (accession number NC 001699). Antibodies to virus-like particles were detected using enzyme immunoassays. Serum samples were diluted 1:200 and left to react on antigen-coated plates for 1 h at 37°C, after which antigen-bound immunoglobulin was detected with peroxidase-conjugated antibodies against human IgG. Color development was initiated by the addition of 2,2′-azino-di-(3-ethylbenzthiazoline-6-sulfonate) hydrogen peroxide solution. The reaction was stopped after 20 min by addition of 1% dodecyl sulfate and optical density was measured at 405 nm, with a reference wavelength of 490 nm. Each case and its matched control were maintained as a set to ensure simultaneous processing, and laboratory personnel were masked as to the case-control status of each sample. All samples were tested in duplicate, and the mean value was used in the analysis. Masked quality controls samples were included to confirm the reliability of the assay.

A cutoff for JC virus seropositivity was calculated by comparing the distribution of absorbance values in the study population with the distribution of absorbance values obtained from children ages 1 to 5 y (n = 47). Young children are considered a low-prevalence population for exposure to JC virus, given that initial infection occurs in late childhood to early adolescence. We used an iterative statistical approach that excluded outliers in the distribution of children's test results until no remaining value was greater than three SDs above the mean optical density or a maximum of three iterations was reached. Seropositivity was then defined as four SDs above the final mean optical density (minus children outliers; absorbance value = 0.067). Using this approach, the seroprevalence in children ages 1 to 5 y was 10.6%.

Baseline characteristics were compared between cases and matched controls using McNemar's test for two-level categorical variables and Bowker's test of symmetry for three-level categorical variables. Factors assessed at baseline in both cohorts [sex, age and smoking status (current versus never; former versus never) and use of nonsteroidal anti-inflammatory drugs within the last 48 h] were compared between JC virus–positive and JC virus–negative controls using generalized linear models to account for intra-individual correlations in the subset of controls (n = 118) who participated in both cohorts and contributed two observations to the analysis. Baseline body mass index (BMI) and recent hormone use in women were compared between JC virus–positive and JC virus–negative controls using Fisher's exact test, as these variables were ascertained only for CLUE II participants. Associations with baseline JC virus seropositivity were also compared between individuals with and without a family history of colorectal cancer, which was assessed in the CLUE II follow-up questionnaires.

Continuous JC virus antibody levels were first compared between colorectal cancer cases and matched controls and colorectal adenoma cases and matched controls using the Wilcoxon Mann-Whitney test. Individuals were then classified as JC virus–positive or JC virus–negative based on the binary cutpoint described above. The associations between JC virus seropositivity and colorectal cancer or adenomas were evaluated by calculating odds ratios (OR) and 95% confidence intervals (95% CI) using conditional logistic regression models with robust sandwich estimates of the covariance matrix (28). Associations between JC virus and colorectal cancer were similar between the two cohorts; therefore, all observations were combined in the final analyses. For colorectal cancer, ORs were estimated overall and by sex, site (colon versus rectal cancer), stage at diagnosis, and time between blood draw and diagnosis (<1-9, 10-19, and 20-31 y), the latter for which participants in both cohorts contributed two observations in different categories. For adenomas, ORs were estimated by number of adenomas (one versus multiple), site (rectum, distal, or proximal colon), histology (tubular or tubulovillous/villous), and size (< or ≥0.55 cm, the median, which was used as the cutpoint due to small numbers of cases with adenomas >1 cm). If a case had multiple adenomas, the largest adenoma and the adenoma with the worst histology were selected for analyses conducted by size and histology, respectively. For the analysis by site, cases with multiple adenomas contributed one observation for each site at which the case had an adenoma. Matched ORs and 95% CIs were first obtained from a model that included JC virus serostatus as the only independent variable. Among CLUE II participants, a multivariable model was used to further adjust for two factors associated with JC virus infection, smoking status and BMI, the latter of which was available only for CLUE II participants. Interactions between JC virus seropositivity and age or sex in relation to colorectal cancer or adenoma risk were evaluated by including an interaction term in the conditional logistic regression model, the coefficient for which was evaluated by the Wald test. All statistical tests were two-sided. Analyses were conducted using SAS, version 9.1 (SAS Institute, Inc.).

Baseline characteristics of 611 colorectal cancer case-control pairs and 123 adenoma case-control pairs are presented in Table 1. Cases were more likely to be current or former smokers than their matched controls, but the differences were not statistically significant (P = 0.18 for colorectal cancer, P = 0.08 for adenomas). Nonsteroidal anti-inflammatory drug use at baseline was less common among colorectal cancer cases versus controls (P = 0.05) but did not differ between adenoma cases versus controls (P = 0.18). No case-control differences were observed in BMI, although family history of colorectal cancer was statistically significantly positively associated with both colorectal cancer and adenomas (Table 1).

No difference in JC virus seroprevalence was observed between males and females (Table 2). JC virus seroprevalence tended to decrease with age, although this trend was not statistically significant. Former smokers were less likely to have antibodies to JC virus than current and never smokers. No difference in JC virus seroprevalence was observed by nonsteroidal anti-inflammatory drug use at baseline. Within the CLUE II cohort, JC virus seroprevalence decreased with increasing BMI at baseline (P = 0.05). JC virus infection was not associated with family history of colorectal cancer or recent hormone use among women (Table 2).

Results for the association between JC virus infection and colorectal cancer are presented in Table 3. Based on 729 pairs of observations from 611 case-control matched pairs, JC virus seropositivity was not associated with an increased risk of developing colorectal cancer (OR, 0.90; 95% CI, 0.79-1.03). The risk estimate was almost identical after adjustment for smoking and BMI among the 257 case-control pairs in the CLUE II cohort (OR, 0.94; 95% CI, 0.72-1.21). After stratification by gender, no association between JC virus seropositivity and colorectal cancer was observed among men, with or without adjustment for smoking and BMI. JC virus seropositivity was associated with a statistically significant decreased risk of colorectal cancer among women, although this association was attenuated after adjustment for smoking and BMI among CLUE II participants. When analyses were conducted separately for colon versus rectal cancer, no association with JC virus seropositivity was observed for colon cancer, whereas a nonstatistically significant inverse association was observed for rectal cancer. No clear patterns in JC virus–associated colorectal cancer risk were observed across categories of stage at cancer diagnosis or time between blood draw and cancer diagnosis (Table 3). When JC virus antibody levels were treated as a continuous variable, there were no case-control differences for colon cancer (P = 0.46 for CLUE I, P = 0.85 for CLUE II) or rectal cancer (P = 0.31 for CLUE I, P = 0.57 for CLUE II; data not shown).

Results for the association between JC virus seropositivity and colorectal adenoma are presented in Table 4. Overall, JC virus seropositivity at baseline was not associated with an increased risk of adenoma development in the subsequent 15 years of follow-up. Increased risk estimates were observed for developing multiple adenomas (OR, 1.81; 95% CI, 0.88-3.74) and adenomas ≥0.55 cm (OR, 1.65; 95% CI, 0.86-3.19), but neither of these associations were statistically significant. No clear patterns in adenoma risk were observed by location within colon or histology. Adenoma results are stratified by gender in Table 5. JC virus seropositivity at baseline was associated with a >2-fold increased risk of subsequent adenoma among men with adjustment for smoking and BMI (OR, 2.31; 95% CI, 1.20-4.46). Conversely, a statistically significant decreased risk of adenoma was associated with baseline JC virus seropositivity among women (OR, 0.31; 95% CI, 0.14-0.67; P < 0.001 for interaction between men and women). The positive association between JC virus seropositivity and adenoma risk observed among men was particularly strong for those who developed multiple adenomas (OR, 6.71; 95% CI, 1.34-33.60) and for men whose largest adenoma was ≥0.55 cm (OR, 3.83; 95% CI, 1.20-12.25), the median adenoma size in this study population. Among women, JC virus seropositivity was consistently associated with decreased risks of adenoma development across adenoma number and size. No clear differences in adenoma risks were observed by location within the colon versus rectum or by histology among men or women (Table 5). When JC virus antibody levels were treated as a continuous variable, female adenoma cases had statistically significantly lower levels than controls (P = 0.02). Male adenoma cases had higher JC virus antibody levels than controls, although the difference was not statistically significant (P = 0.11).

JC virus seropositivity was not associated with increased risk of developing colorectal cancer up to 31 years later in either men or women. A positive association was observed between JC virus seropositivity and adenoma in men, whereas an inverse association was observed in women. Our findings are consistent with the only other serologic study of JC virus infection and colorectal cancer, a prospective study conducted among men in Norway, which observed no increased risk of colorectal cancer among men who were JC virus seropositive at baseline (25). This present study is the first analysis of JC virus antibodies in relation to colorectal cancer in women and adenomas in both men and women.

Other than age, there are no known risk factors for primary infection with JC virus. Analysis of JC virus antibody data from the controls in this study suggested that smoking status and BMI could also be related to JC virus infection. However, adjustment for these factors did not change the association between JC virus seropositivity and colorectal cancer. C-reactive protein levels, a marker of inflammation, were available from a previous case-control study conducted within this cohort (26, 29). Adjustment for C-reactive protein levels in the subset of participants for whom data were available did not change the risk estimates for JC virus seropositivity and colorectal cancer or adenomas (data not shown). JC virus reactivation is common in pregnancy (30), and parity has been associated with a decreased risk of colorectal cancer in some studies (31). However, only 8% of women had never been pregnant at baseline, and no association was observed with JC virus seropositivity (data not shown). Therefore, the inverse associations observed between JC virus seropositivity and colorectal cancer/adenomas among women were not likely due to negative confounding by parity. Alternative explanations for the observed differences in JC virus–associated adenoma risk by gender are unclear. Increased BMI has been shown to be more strongly associated with colorectal cancer among men (32), and postmenopausal women with no recent use of hormone replacement therapy (33), suggesting estrogen status may be a modifying factor. However, there were no differences in JC virus–associated adenoma risk observed between women who reported hormone use at baseline versus those who did not (data not shown).

It has been suggested that JC virus may contribute to colon cancer development through a “hit-and-run” mechanism (34) whereby JC virus infection is involved in the early stages of colorectal cancer carcinogenesis through disruption of the Wnt signaling pathway but is not needed for tumor progression. Specifically, coexpression of JC virus T-antigen and β-catenin can result in increased transcription of c-myc (14), leading to chromosomal instability, which can progress in the absence of JC virus T-antigen expression (15). Disruption of the Wnt signaling pathway can result in chromosomal instability within normal colon mucosa and transformation to early adenoma (35), whereas subsequent progression of early adenoma through intermediate and late stages and onto carcinoma is dependent upon subsequent genetic alterations. This model could explain the positive association between JC virus antibodies and adenomas among men, juxtaposed with the null association with colorectal cancer. However, if JC virus is involved at the earliest stages of colorectal cancer carcinogenesis, then the association between JC virus seroreactivity and colorectal cancer would most likely be evident in blood samples obtained decades prior to colorectal cancer diagnosis, and no increased risk of colorectal cancer was observed in association with JC virus antibodies measured up to 31 years prior to diagnosis among men (data not shown).

JC virus infection elicits the formation of several types of antibodies in humans. The present study measured IgG antibodies to the JC virus capsid, which are produced in response to initial asymptomatic infection with JC virus, usually occurring in late childhood. The presence of JC virus IgG capsid antibodies does not protect against reactivation of infection (3), although high versus low levels of JC virus capsid antibodies can distinguish people shedding virus in their urine from nonshedders, and thus, may serve as a more specific marker of JC virus reactivation (25). We compared continuous JC virus IgG antibody levels between cases and controls and observed no association with colorectal cancer in males or females. However, measurement of other classes of antibodies to the JC virus capsid and/or antibodies to the T-antigen may provide additional information about the association between JC virus infection and cancer. Ideally, one would investigate the full signature of JC virus by measuring DNA sequences and protein expression in tumor tissue, in addition to circulating antibodies (36). However, tumor tissues were not available from the cases in this study.

The present study has some additional limitations, including the initial assessment of adenomas through self-report. The adenoma cases and controls were ascertained among respondents to the CLUE II follow-up questionnaire(s), with response rates ranging from 62% to 70%. It is unlikely, however, that respondents differed from nonrespondents with respect to JC virus serostatus, and therefore, selection bias did not likely result. Only those adenoma cases that were verified with a pathology report were included in the current analysis. However, pathology reports were not obtained from respondents who reported having had a colonoscopy or sigmoidoscopy without an adenoma diagnosis. Therefore, it is possible that some adenoma “cases” were misclassified as “controls,” potentially biasing the observed results toward the null.

Prospective studies with long duration of follow-up are important for investigations of risk factors that may be involved in the early stages of colorectal cancer carcinogenesis, such as JC virus infection. To our knowledge, the present study was the first to investigate JC virus infection and colorectal cancer among women, and the first seroepidemiologic study of JC virus antibodies and adenomas. The sample size for the analysis of adenomas was smaller than that for colorectal cancer, and future studies are needed to replicate the findings for adenomas and to investigate mechanisms that could differ by gender. If JC virus infection is indeed confirmed as a risk factor for adenomas among men, then it could be a target for novel colorectal cancer prevention strategies. However, given the inconsistencies in tumor studies, and the limited data from epidemiologic studies, more information is needed to evaluate the association.

No potential conflicts of interest were disclosed.

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.

We thank Barbara Clayman for the laboratory assistance and all who participated in the CLUE cohorts.