Prospective investigation of polyomavirus infection and the risk of adult glioma

Glioma is an aggressive primary tumor of the brain with a poorly understood etiology. We studied the association of 4 human polyomaviruses (HPyV)—JC virus (JCV), BK virus (BKV), human polyomavirus 6 (HPyV6), and Merkel cell polyomavirus (MCPyV) with glioma risk within the Cancer Prevention Study II in the US (CPS-II) and the Janus Serum Bank in Norway. Cohort participants subsequently diagnosed with glioma from the CPS-II (n = 37) and Janus Serum Bank (n = 323), a median of 6.9 and 15.4 years after blood collection, respectively, were matched to individual controls on age, sex, and date of blood draw. Serum antibodies to the major viral capsid protein (VP1) were used to establish infection history for each polyomavirus. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using conditional logistic regression. In the Janus Serum Bank, MCPyV infection was associated with a higher risk of glioma overall (OR: 1.56; 95% CI 1.10, 2.19). A modest, nonsignificant positive association with MCPyV infection was also observed in CPS-II (OR: 1.29; 95% CI 0.54, 3.08). In both cohorts, glioma risk was not significantly related to infection with JCV, BKV or HPyV6. The present study suggests that MCPyV infection may increase glioma risk.

BKV and JCV are classified as possible carcinogens 30 based on "sufficient evidence in experimental animals" but "inadequate evidence of carcinogenicity in humans".
The role of polyomavirus infection in relation to glioma risk in humans is unknown. In the only prospective study to date 31 , antibodies to JCV, BKV, and simian virus 40 (SV40) measured in serum collected 1-22 years before glioma diagnosis were evaluated for association with incident glioma. Glioma cases (n = 44) and age-, gender-and race-matched controls (n = 88) were identified from participants of two specimen banks in Washington County, Maryland. The study detected no association with SV40. A nonsignificantly positive association was found for JCV (OR: 1.46), and an inverse association was found for BKV (OR: 0.66), with suggestively stronger but nonsignificant associations reported when restricting to grade IV glioblastomas (GBM) (ORs of 2.38 and 0.53, respectively).
Using a nested case-control design within two prospective cohort studies with biobanked collected blood, the Janus Serum Bank and the Cancer Prevention Study II (CPS-II) Nutrition cohort, we conducted an exploratory investigation of 4 polyomaviruses, JCV, BKV, HPyV6 and MCPyV, in relation to glioma risk. A multiplex assay was used to detect serum antibodies to the major capsid proteins (VP1) of each virus. To avoid potential bias in results from effects of preclinical disease on serum antibody titers, the study was restricted to cases with blood collected a minimum of 3 years (in the CPS-II) or 5 years (in the Janus Serum Bank) prior to glioma diagnosis.

Methods
Study populations. Data from two cohorts were included in the present study: (1)

the Cancer Prevention
Study-II (CPS-II) Nutrition cohort, a US prospective study 32 ; and (2) the Janus Serum Bank, a population-based prospective study based in Norway 33 . Baseline characteristics of participants from each cohort are shown in Table 1. Incident primary intracranial glioma cases (ICD9 and 10 topography codes: 191 and C71, respectively) were comprised of WHO grade IV glioblastomas (GBM) (ICD-O-3 histology code: 9440-9441), and lower grade gliomas ('nonGBM')(ICD-O-3 histology codes: 9380, 9382, 9400-01, 9410-11, 9420, 9424-25, 9450-9451) [34][35][36] . In CPS-II, among the 32,609 cancer-free participants that provided a blood sample between 1998 and 2001 who were followed through the end of 2013, 37 glioma cases diagnosed a minimum of 3 years after sample collection were included in the present study. For each case, we randomly selected two controls from participants who provided a blood sample, and were alive and had no personal history of cancer (excluding basal and squamous cell carcinoma) as of the diagnosis date of the case. Each control was individually matched to its case on birth date within 1 year, sex, and date of blood draw within 6 months.
In the Janus Serum Bank, glioma cases and controls were identified from a previous nested case-control study of glioma 12 ; a total of 323 glioma cases diagnosed a minimum of 5 years after blood collection were included in the present analysis. Each control was individually matched to its case on birth date within 2 years, sex, date of blood draw within 12 months, and county of residence, and was alive and free from any cancer except basal and squamous cell carcinoma on the date of glioma diagnosis of the case subject. All subjects were of European ancestry.
All research was performed in accordance with the human subjects' protection principles (Declaration of Helsinki), and the study was approved by the ethical review board of the Moffitt Cancer Center and from both participating cohorts (CPSII-NC: Emory University IRB #00045780; Janus: Regional Committees for Medical and Health Research Ethics, Application 9821, #2017/2140/ REK-s-o-B). All participants of the CPSII-NC provided written informed consent at the time of blood draw. As described previously 37 , Janus Serum Bank donors gave broad verbal consent (1973)(1974)(1975)(1976)(1977)(1978)(1979)(1980)(1981)(1982)(1983)(1984)(1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992) or provided explicit written informed consent (1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004) for use of samples for cancer research, in accordance with Norwegian legal requirements over time. The Norwegian Data Protection Authority has approved use of Janus data and samples collected between 1972 and 2004, while requiring that donors are free to unconditionally withdraw their consent at any time. All research projects using Janus specimens and data from the Cancer Registry of Norway also need approval from a Regional Committee for Medical and Health Research Ethics. www.nature.com/scientificreports/ Multiplex serological assays. Multiplex serological analyses to measure antibodies to the viral proteins were conducted at the German Cancer Research Center (Deutsches Krebsforschungszentrum, DKFZ) in Heidelberg, Germany. This method allows the simultaneous measurement of antibody responses to multiple proteins in biologic samples 38 . Polyomaviruses lead to strong and reproducible antibody responses to the major capsid protein VP1 39,40 , a sensitive cumulative marker of past or present infection. For the current study, a prediagnostic plasma sample was obtained from each case and control from both participating cohort studies and shipped to the lab in Heidelberg for the analysis of serological markers of infection. Serological analyses were performed by fluorescent bead-based multiplex serology as previously described [38][39][40][41] . Briefly, the method is based on a glutathione S-transferase (GST) capture assay combined with fluorescent-bead (Luminex) technology. Proteins of interest are fused with an N-terminal GST domain. These recombinant GST-VP1 fusion proteins are affinity-purified in situ through binding to glutathione casein-coated fluorescence labelled polystyrene beads. Each fusion protein is bound to a spectrally distinct bead set (SeroMap, Luminex). The bead mix is then incubated in 96-well plates with participant plasma. Bound antibodies were detected with biotinylated triple (IgG, IgA, IgM) goat anti-human secondary antibody and streptavidin-R phycoerythrin reporter conjugate. A Luminex analyzer was used to identify the internal color of the individual beads and to quantify their reporter fluorescence (expressed as median fluorescence intensity (MFI)) as a measure of viral antibody titer. GST only (i.e. without viral antigen fusion component) served for background determination (i.e., negative control). Auto-fluorescence of each bead set and background reactions resulting from binding of secondary reagents were determined in one well per plate without human serum. Mean background values were subtracted from raw antibody output to determine antigen-specific reactivity (i.e., unit of statistical analysis). The serological analyses included the viral major capsid protein 1 (VP1) of four human polyomaviruses (JCV, BKV, HPyV6 and MCPyV). All samples from both cohorts were analyzed in a single laboratory batch avoiding measurement error otherwise introduced by the need for a bridging panel to normalize results across studies. Laboratory personnel were blinded to casecontrol status.

Statistical analysis.
Odds Ratios (OR) and 95% Confidence Intervals (CI) for associations of PyV infection and glioma risk were calculated using conditional logistic regression. All models were inherently adjusted for matching factors. We considered the association of glioma with seroprevalence (i.e., the presence or absence of VP1 antibodies) of each of the four HPyVs, with seropositivity for each HPyV defined as virus-specific VP1 antibody level ≥ 250 MFI 41

Results
The analyses were based on 37 glioma cases (27 GBMs) and 74 controls from CPS-II and 323 glioma cases (196 GBMs) and 323 controls from the Janus Serum Bank (Table 1).
In the CPS-II, approximately half of the cases and controls were male (46%) and the median age at blood collection was 71 years (range 59-77 years). Year of blood collection ranged from 1999 to 2001 (median: 2000). Glioma cases were diagnosed between 2002 and 2013. The median interval between blood collection and glioma diagnosis was 6.9 years and ranged from 3.4 years to 13.0 years. In the Janus Serum Bank, a majority of the cases were male (67%) and the median age at blood collection was 41 years (range 19-67 years), with a tight distribution on age (interquartile range in age: 39.8-42.6 years). Year of blood draw ranged from 1972 to 1991 (median: 1983). Glioma cases were diagnosed between 1980 and 2009. The median interval between blood collection and glioma diagnosis was 15.4 years and ranged from 5.1 to 34.9 years. In both studies, controls were similar to cases on age at blood draw, gender, and year of blood draw as these were matching factors. In CPS-II, the median age and interquartile range (IQR) at glioma diagnosis was 79 (75-81) and 76 (73)(74)(75)(76)(77)(78), in males and females, respectively. In the Janus Serum Bank, in which participants were younger at enrollment, the median age and IQR of cases at glioma diagnosis for males and females were 57.1 (50.4-63.3) and 56.3 (52.1-60.3), respectively.
In the Janus Serum Bank, infection with MCPyV was associated with a significantly increased risk of all gliomas combined (OR: 1.56; 95% CI 1.10-2.19) ( www.nature.com/scientificreports/ modest and non-significant positive association with MCPyV infection was also observed in CPS-II (Table 2) for all gliomas combined (OR: 1.29; 95% CI 0.54-3.08).
We examined cross-reactivity between the 4 examined PyVs 41 in the Janus Serum Bank and found modest correlations in antibody titers ranging from r = 0.082 (p-value = 0.035) for the correlation of BKV with HPyV6, to r = 0.167 (p-value < 0.001) for the correlation of MCPyV with HPyV6. Therefore, in the Janus Serum Bank we examined associations of glioma with seroprevalence of each PyV adjusted for seroprevalence of the other PyVs and all results were essentially unchanged (not shown). For example, the odds ratio for MCPyV infection was 1.56 (95% CI 1. 10-2.19) in the unadjusted model and 1.61 (95% CI 1.13-2.28) in the model adjusted for seroprevalence of BKV, JCV and HPyV6.
When examining risk according to tertiles of antibody levels to BKV, JCV, and HPyV6 in the Janus Serum Bank, no significant trends in risk emerged for all gliomas combined (p-value trends = 0.761, 0.579 and 0.798, respectively) ( Table 3), or among GBMs and nonGBMs when considered separately (data not shown). For MCPyV, a positive association of borderline significance was observed with increasing titer in gliomas overall (p-value trend = 0.052) ( Table 3). The result for MCPyV differed according to glioma subtype (Supplemental Table S1): for GBM, ORs rose with increasing tertile (OR for tertile 3 vs seronegativity: 1.80 (95% CI 1.03-3.15) with a significant trend (p-value trend = 0.048), whereas for nonGBM, the OR was significantly elevated only in the first tertile (  www.nature.com/scientificreports/ and > 45 years, respectively), though no individual OR was statistically significant. Infection with JCV, BKV and HPyV6 were unrelated to glioma risk regardless of elapsed time between blood collection and glioma diagnosis and subject age at blood collection. In Janus, blood samples were collected over a 20 year period (between 1972 and 1991) and results were similar for samples collected in the first half (i.e.,1972-1981) versus the second half (i.e., 1982-1991) of the interval for each of the 4 HPyVs that we studied (Supplemental Table S2

Discussion
This is the first large prospective study to examine JCV, BKV, MCPyV and HPyV6 as potential causal agents in glioma. We found no consistent evidence supporting a role for BKV, JCV or HPyV6. In contrast, infection with MCPyV was positively associated with the risk of glioma. In Janus, the association with MCPyV was observed regardless of glioma subtype and gender. To our knowledge, this is the first study to implicate MCPyV infection in the development of a CNS tumor in humans, following evidence that a phylogenetically closely related virus (RacPyV) causes glioma-like tumors in raccoons 29,47 . One previous prospective study 31 based on a small set of incident cases (44 gliomas) examined glioma risk in relation to BKV and JCV infection. The authors reported a suggestive positive association with JCV and inverse association with BKV and findings that were more pronounced among GBMs (ORs of 2.38 and 0.53, respectively). The current study yielded nonsignificant results for both BKV and JCV, consistent with a lack of conclusive evidence of a role for these viruses in the development of human tumors 48 .
The present findings for MCPyV are of interest in view of the established association of this virus with Merkel cell carcinoma (MCC), an aggressive neuroendocrine tumor of the skin 28,49 . The etiology of MCC has been linked to ultraviolet light (UV) exposure and to infection with MCPyV. The attributable risk of each exposure depends on geography: in northern latitudes, the majority of MCC cases are of viral etiology, whereas in areas with high UV exposure, more cases are related to UV-mediated carcinogenesis 50 . UV exposure is thought to be involved in both viral-mediated and non-viral-mediated carcinogenesis by contributing to immunosuppression or DNA damage, respectively. With the exception of our report, to our knowledge no other tumor in humans has been linked to MCPyV exposure.
A biologic mechanism linking MCPyV infection to the risk of glioma is unknown. Whereas MCPyV is not specifically known to cross the blood-brain-barrier 14 , other human PyVs such as JCV are known to do so 26 , as does RacPyV 29 . In MCC, viral DNA is found integrated within the tumor genome in a clonal pattern suggesting that MCPyV infection and integration precede clonal expansion of tumor cells 27 . Whereas studies to date have not demonstrated MCPyV viral protein expression in human glial tumors 51,52 , MCPyV may contribute to glioma via a range of mechanisms 16 such as neuroinflammation 53,54 , insertional mutagenesis [55][56][57] , or effects on gene expression 58 that would not leave traces of viral integration in the tumor genome. Of note, RacPyV is considered the causal agent of neuroglial olfactory tract tumors in free-ranging raccoons 29 ; the virus has not been found integrated in tumor DNA, suggesting a mechanism of PyV-mediated oncogenesis not requiring direct host cell integration 47 . MCPyV is a latent infection that occurs most often in early childhood 21,42 and lies dormant but may reactivate in the context of immunosuppression to cause disease (i.e. BKV-related nephropathy in transplant patients and JCV-related PML in AIDS patients); therefore, IgG capsid antibodies reflect infection occurring decades prior to blood collection. The present association of MCPyV with glioma, including the suggestion of increased risk with advancing age at blood collection in the Janus series, is consistent with reactivation of latent virus possibly triggered by immune senescence with advancing age 59,60 contributing to glioma onset. Why a common exposure like MCPyV might lead to the development of glioma is unknown. However, it is notable that MCC is also a rare tumor caused by exposure to this virus. Furthermore, a similar pattern is observed for Epstein-Barr virus, a ubiquitous Herpesvirus that is considered a necessary cause of nasopharyngeal carcinoma and Burkitt lymphoma, which are both rare cancers 61 . The association between EBV and rare malignancies suggests that as-yet unknown co-factors influence host susceptibility to this virus 62 . Analogously, MCPyV may contribute to neuro-oncogenesis only in the presence of host immune or genetic co-factors, a hypothesis that should be examined in future studies.
A key strength of the present study was the prospective design and investigation of prediagnostic blood samples for detection of viral antibodies. In the Janus Serum Bank, glioma cases were diagnosed over a broad interval following blood collection (5-35 years, with a median of 15 years) allowing for examination of potential latency of viral exposures on glioma risk. Furthermore, all analytic cases were diagnosed presumably outside of the preclinical window (3 or more years following blood collection in CPS-II, and 5 or more years in Janus) reducing potential influence of preclinical tumor on viral markers.
We acknowledge several limitations. Power to detect associations by gender and according to latency and antibody titer, were limited in the Janus Serum Bank and these evaluations could not be performed in CPS-II due to small numbers. For the same reason, we were unable to examine associations according to specific lower grade glioma histologic subtypes in Janus. Furthermore, IDH1 mutation status, which is an important molecular classifier of low grade gliomas and secondary GBM 63 , was not available in either study to examine whether associations vary according to IDH1 status. Lack of a statistically significant association for MCPyV in CPS-II could be explained by the small number of incident glioma cases or potentially the older age at onset of glioma in CPS-II cases when compared to Janus cases, although data from Janus Serum Bank suggests if anything a stronger association of the virus with glioma in older persons. Blood samples were collected in some cases many years before serologic testing for HPyV antibodies and long storage times could potentially have affected assay results. However, we note that antibodies are considered robust biomolecules 64  www.nature.com/scientificreports/ storage periods at low temperatures (at or below − 20 °C) 65 . In our study, we were able to detect a range of values for all HPyV antibodies, suggesting that the titer distribution at the time of measurement reflected the titer range at blood draw 40 . Furthermore, in studies of HPV 66 , antibody titers have been shown to be comparable in serum samples from Janus versus other cohorts with shorter storage times; as HPV antibody titers after natural infection are much lower than for HPyV (and therefore less robust in serological assays), these results provide further assurance that long storage times did not affect the present results. We lacked information on lifestyle factors such as smoking and obesity which are known to impact immune surveillance [67][68][69][70][71][72] and may confound or alter strength of associations between viruses and cancer. Immunity is known to wane with age 59,73 and we had a limited ability to consider associations according to age given the tight age distribution in the Janus Serum Bank. We also could not investigate associations by race/ethnicity. The study was based on a single blood measurement and infections that occurred after blood collection would have introduced misclassification of exposure; however, as studies have shown that seroprevalence of PyVs within individuals is stable over a decade or more in adults 40 , and that most individuals are infected with PyVs in childhood or adolescence 21,42 , we expect that misclassification of PyV infection status was limited among the middle aged adults in these cohorts. Finally, we did not consider antibodies to the viral T-antigens in MCPyV which are the key oncogenic drivers for MCC 74 ; demonstation of a more prominent association of the T-antigen in MCPyV with glioma risk, as has been demonstrated in MCC 74 , would strengthen the case for a causal association. In summary, in this pilot investigation we found the first evidence linking glioma risk specifically to infection with MCPyV but not with the 3 other PyVs examined. Investigations based on larger, diverse study populations with long-standing biorepositories are needed to clarify the role for MCPyV and other PyVs that infect humans 21,43 as potential causal agents in glioma.

Data availability
For access to data from the Janus Serum Bank approval must be obtained from The Regional Ethics Committee of Southern Norway and from the Cancer Registry of Norway (in charge of the Janus Serum Bank), according to national and European regulations of data protection. Data underlying the CPSII-NC results are available upon request by following the procedures outlined here: https:// www. cancer. org/ conte nt/ dam/ cancer-org/ resea rch/ epide miolo gy/ cancer-preve ntion-study-data-access-polic ies. pdf