Antibody Response to Merkel Cell Polyomavirus Associated with Incident Lymphoma in the Epilymph Case–Control Study in Spain

To date, 9 human polyomaviruses [BKV, JCV, KIV, WUV, Merkel cell polyomavirus (MCV), HPyV6, HPyV7, trichodysplasia spinulosa-associated polyomavirus and HPyV9] have been identified as infectious agents of humans (1). These DNA viruses generally produce persistent asymptomatic infections, commonly acquired during childhood and early adulthood, which can be reactivated under immunosuppression conditions. Although for some polyomavirus, like BKV or JCV, antigens have been shown to have oncogenic properties in experimental studies, a similar role in humans has not been proven (2). Contrary, integrated MCV DNA was isolated from Merkel cell carcinoma (MCC) in around 80% of MCC cases suggesting a potential etiologic role (3). Furthermore, whereas between 46% and 88% of healthy adults are seropositive for antibodies against MCV VP1 protein (4–6), MCC cases show higher seroprevalences (7–9) and higher seroreactivities (6, 9–11). In addition, antibodies to large-T (LT-Ag) and small-T (sT-Ag) antigens are detected almost exclusively in cases (7).

The lymphotropic properties of MCV (2) and the higher risk of chronic lymphocytic leukemia (CLL) in MCC cohorts observed in prospective studies (12) suggest a possible role of MCV in lymphomagenesis. Among subjects with lymphoproliferative disorders, CLL subjects have been shown to have higher MCV DNA detection rates, although prevalences and viral load are low (13–16). LT-Ag detection is almost null in lymphomas but in purified CD5+/CD19+ CLL cells a mutated DNA sequence encoding for MCV LT-Ag has been reported (17–19). Tolstov and colleagues (20) reported on lack of association between MCV VP1 seroreactivity and CLL although the report was based on small numbers.

In this study, we explored the association between seroreactivity to MCV VP1 capsids and risk of lymphoma in the context of an epidemiologic case–control study in Spain (Epilymph study).

Incident lymphoma cases, histologically or cytopathologically diagnosed between 1998 and 2002, in 4 Spanish hospitals were recruited (21). The participant centers were the Catalan Institute of Oncology and Hospital Universitari de Bellvitge from l'Hospitalet de Llobregat, Hospital de Tortosa Verge de la Cinta and Hospital Sant Joan de Reus from Tarragona and Hospital Ramon y Cajal from Madrid. Controls were frequency matched by age, sex, and recruitment centre of lymphoma cases, randomly selected from hospitalised patients after exclusion of those with cancer or immunosuppression-related conditions. A wide range of admission diagnoses for hospital controls were included in the study with none of them accounting for more than 5% of all diagnosis. Lymphoproliferative disorders were classified according to the WHO Classification for Neoplastic Diseases of the Lymphoid Tissues from 2002 (22). Epidemiologic data on demographics, medical and family history, and environmental exposures were obtained by personal interview. All study participants provided written informed consent and study approvals were granted by the Institutional Review Boards of each participating center.

Study subjects with available biological samples for MCV testing consisted of 520 of 700 initial eligible cases and 597 of 655 eligible controls. Reasons for not being included were refusal to participate (n = 28 and n = 23) and no blood sample available (n = 122 and n = 35) for cases and controls, respectively. Among cases, additional reasons were death before the interview (n = 25) and lack of epidemiologic interview (n = 5). Final population for analysis purposes included 468 cases and 552 controls, after exclusion of 28 cases and 44 controls for which antibody measurements could not be done due to loss of serum by evaporation and exclusion of organ graft recipients and HIV positive subjects (25 cases including one subject already excluded due to dilution reliability, and 1 control).

The 468 lymphoma cases included 9 precursor B-cell lymphomas, 83 diffuse large B-cell lymphoma (DLBCL), 34 follicular lymphomas, 108 CLL, 18 lymphoplasmacytic lymphoma (LPL), 64 other B-cell lymphoma, 69 multiple myeloma, 45 classical Hodgkin lymphomas, 6 nonclassical Hodgkin lymphomas, 12 mycosis fungoides, 19 other T-cell lymphomas, and 1 NHL NOS case. In the statistical analysis, 3 levels of categorization of these entities were carried out. On a first level, all cases were considered. On a second level, following the last WHO hematologic guidelines from 2008 (22), mature B-cell lymphoma (B-NHL) category (excluding the 9 precursor lymphoma cases), mature T-cell lymphomas (T-NHL) category (including both mycosis fungoides and other T-cell lymphomas), and Hodgkin lymphomas were considered. At a third level, analysis by mature B-cell lymphoma subtypes was run. In addition, due to recent evidence on different behavior among Hodgkin lymphomas types and its small sample size, 6 cases of nonclassical Hodgkin lymphoma were only included within the all lymphoma category.

Exposure to MCV (isolate 399) was measured serologically using a MCV VP1 virus-like particle enzyme immunoassay (EIA) as described previously (6) with minor modifications including use of serum at a 1:400 dilution. Serum samples yielding an optical density (OD) value greater than the cut-off value (COV) of OD ≥ 0.220 were considered as seropositive. The COV was established as an OD greater than the mean seroreactivity results plus 4 SDs of serum samples from 63 children aged 2 years old after exclusion of outliers (n = 7).

Possible confounding factors within all available information from the questionnaire were examined. Besides the matching variables, other factors were selected according to previous data on suspected transmission routes used by polyomavirus (oro-fecal, blood) as well as risk factors associated to overall infection (siblings). Known risk factors for lymphoma that could be related to infectious exposure were also included, such as family history of cancer, which could be linked to genetic heritage or exposure to same environmental conditions, and ever lived in a rural area, which could be both related to pesticides or to oro-fecal transmitted infection exposure.

Differences in demographic characteristics between cases and controls as well as differences in viral seroprevalence among control population were assessed using χ² test, or Fisher exact test when applicable. Multinomial logistic regression was used to estimate OR and 95% confidence intervals (CI) of each lymphoma subtype related to MCV seropositivity. Models were adjusted for sex, recruitment area, and age (continuous). Scatterplot of all individual OD values, jittered in order to avoid overlapping of values, were created stratified by lymphoma subtypes and controls. Among the seropositive subjects, the median and interquartile percentiles were displayed. Differences in median seroreactivity between each lymphoma and controls among MCV seropositive subjects were analysed by the Mann–Whitney U test.

To exclude possible effects of treatment on antibody responses, a sensitivity analysis was carried out excluding 73 subjects whose blood sample was taken after start of treatment. Significance level was established at 0.05 and all tests run were 2-sided. Analyses were conducted with Stata software, version 10.1.

After our initial analysis showed an association between DLBCL cases and MCV seroreactivity, additional testing and analyses were carried out. Among the overall control population, for each DLBCL case, we randomly selected 2 control subjects matched by sex, centre, and age within a 5-year window. The final subsample population included 82 cases and 157 control subjects. For 7 cases, either a second-matched control could not be identified or a serum sample from the control was not available. These 239 samples were tested for BKV and JCV seroreactivity using VLP EIA at 1:200 and 1:100 dilutions, respectively as described previously (6, 23–25). COVs used for BKV and JCV were an OD > 0.200 and OD > 0.150, respectively and were determined following the same methodology described above for the MCV EIA.

MCV antibody levels were measured and expressed as EIA units using a previously described method (6). Briefly, quantitation of antibodies was done by incubating serial 4-fold dilutions of test sera and serial 2-fold dilutions of a reference serum in duplicate, starting with a 1:400 dilution, on MCV VLP-coated plates. From the OD values of the serial dilutions of the reference serum, a standard curve was constructed using a 4-parameter equation as implemented in the software package SoftMax Pro. The 4-parameter equation is as follows: A+{(B–A)/[1+((C/OD)⁁D)]} where OD = optical density, A = minimum asymptote of the curve, B = maximum asymptote of the curve, C = the midpoint of the curve, and D = the slope of the curve. The highest dilution of the reference serum that gave an OD value greater than twice the mean of PBS controls was arbitrarily assigned as 1 EIA unit. The operational range of the standard curve was from 1 to 4,096 EIA units. The EIA units of serum samples at each dilution were calculated by interpolating off the standard curve and multiplying by the dilution factor above 1:400. The titer for individual serum samples was the mean of EIA units of 2 or more dilutions of the serum that fell within the operational range of the reference curve (from approximately 5 to 500 EIA units).

Conditional logistic regression was used to study the association of DLBCL with each polyomavirus seroprevalence. Median differences for OD seroreactivity and MCV EIA units were tested using Mann–Whitney U test.

Table 1 shows the characteristics of the study population. No differences were detected by frequency-matching recruitment characteristics, that is, sex, age, and center of recruitment, indicating a balanced study design. No differences were detected between cases and controls except for family history of cancer, which was more frequently observed among cases. Detailed descriptive data by lymphoma subtype are provided at supplemental Table 1. Main differences among subtypes and the group of controls were observed in the age distribution and educational level. The associations identified in the table were always introduced in the analysis to evaluate potential confounders. Sensitivity analysis restricting control population by age range for each lymphoma subtype did not modify the presented data.

Characteristics of MCV seropositivity in control population are detailed in Table 2. Seroprevalence for MCV was 81%. A higher seroprevalence was observed among subjects aged 57 to 67 years compared with younger subjects, although no linear trend was observed with increasing age (P value = 0.5). No differences were observed with other potential confounders.

Distribution of individual MCV OD data displayed by lymphoma subtype is shown in Fig. 1. A wide spectrum of OD values were observed in the lymphoma subtypes and controls. For DLBCL and CLL, values were more concentrated at high OD values, whereas in multiple myeloma they were grouped at lower OD values.

Table 3 shows the OR for association of each lymphoma subtype with MCV seropositivity. MCV seroprevalence was similar in all lymphoma category (85.5%) as compared with controls (81%). By lymphoma subtypes, MCV seroprevalence was significantly higher only in DLBCL cases (96.4%; OR = 6.10, 95% CI = 1.88–19.75) compared with controls. Although not statistically significant, MCV seroprevalence values were higher for follicular lymphoma, CLL, LPL, Hodgkin lymphoma, and T-NHL whereas lower values were observed for multiple myeloma and other B-NHL compared with controls. Exclusion of subjects that had already started the treatment at the time of blood sampling did not modify the presented results and the association described for DLBCL remained significant. Furthermore, in order to rule out a chance finding due to the cut point used alternative values were considered (OD > 0.30, > 0.40, > 0.50, and > 0.60) however results remained unchanged.

When, among seropositive subjects, median OD values (Fig. 1) for each lymphoma subtype were compared with control population (median = 1.59), no significant differences were observed except for multiple myeloma with a median OD value of 1.07 (P value = 0.003). Further analysis in seropositive subjects by categorized data in tertiles only provided significant results for multiple myeloma (data not shown).

Within the subsample of DLBCL and matched controls, seroprevalences for BKV and JCV among control population were of 84% and 70%, respectively. There was no association between DLBCL cases and seroprevalence to BKV (74.4%; OR = 0.52, 95% CI = 0.25–1.06) or JCV (75.6%; OR = 1.35, 95% CI = 0.75–2.41). However, a higher association was obtained for MCV within this restricted subsample (OR = 7.07, 95% CI = 2.11–23.76). Furthermore, the median level of MCV VP1 capsid antibody expressed in EIA units were higher in DLBCL cases (374.2 EIA units) than controls (232.6 EIA units) among those previously defined as seropositive subjects, but differences did not reach statistical significance (P value = 0.086).

In this multicentric lymphoma case–control study, MCV seroprevalence was significantly higher in DLBCL cases than among the control population.

To our knowledge, there is no previous seroepidemiologic study focussing on MCV and a detailed evaluation of lymphoma subtypes. Two previous studies checked for MCV DNA presence in lymphoma tissues, including DLBCL among others. Toracchio and colleagues (13) did not identify viral DNA in 52 DLBCL cases, whereas Shuda and colleagues (14) detected viral DNA in 2 of 65 DLBCL although at lower DNA copy numbers than those quantified in MCC cases. Similar to MCC seroepidemiology, higher MCV VP1 seroprevalences were observed in cases compared with controls (7–9). Although some studies have also reported higher levels of VP1 capsid seroreactivity in MCC cases compared with controls (6, 9–11) we found a higher level of MCV VP1 antibodies in DLBCL cases compared with controls but the difference was of borderline significance. The divergence in results may be explained because of the methodology used and a potential limited ability to produce a strong immune response by DLBCL cases.

No statistically significant results at the 5% level were observed for other lymphoma entities although higher seroprevalences were observed for some of them. These results are consistent with a previous serologic study from Tolstov and colleagues (20) on 18 CLL cases where they found no median differences when compared with 18 acute lymphoblastic leukemia subjects. Our serologic data are also in agreement with previous studies where almost null viral DNA and/or antigen were detected (13–16). An exception to the above is a report from Pantulu and colleagues (17) and complemented by Haugg and colleagues (18) describing the detection of DNA encoding a mutated MCV LT-Ag in CLL cells. Our results do not refute a specific role in CLL although it is intriguing that such a wide range of lymphoma entities with different biological behaviors, such as follicular lymphoma or T-cell lymphoma, shows higher seroprevalences than control populations. These findings suggest a broader and weaker role of the virus as either an etiologic agent or as a secondary player in lymphoproliferative malignancies other than DLBCL.

One of the major constraints in the interpretation of these data is the potential reverse causality due to disease-related immunosuppression. On one hand, the disease itself can show hypogammaglobulinemia or an impaired lymphocyte function, such as in multiple myeloma, resulting in a decreased immune system (26). The deficiency in antibody production characteristic of multiple myeloma may explain the lower seroprevalence and seroreactivity observed in our study multiple myeloma subjects. Furthermore, a decreased seroprevalence to other infectious agents associated with multiple myeloma has been also observed in the same study samples (21, 27–29). However, on the other hand, several lymphoma subtypes showed an increased seroprevalence. Although we cannot provide a clear explanation for this observation, this may indicate that in response to an intrinsic disruption of the normal lymphocyte maturation among lymphoma subjects, a generalized reactivation of common viral DNA infections can be induced. However, antibodies to BKV and JCV tested in the DLBCL subsample within this study showed a different behavior than the one observed for MCV. Although our observations were not affected by prior treatment status or by categorised stage of disease, we cannot preclude a reverse causality effect in our data.

Suspected oncogenic mechanisms for MCV have been based on SV40 studies, whose oncoproteins are involved in several pathways related to cell proliferation, apoptosis, integrin signaling, mitosis, and DNA reparation (2). Principal known targets include p53 and pRb family proteins and PP2A enzymes. However, the region from SV40 LT-Ag that interacts with p53 is commonly mutated in MCV LT-Ag and the induced transformation ability of MCV sT-Ag is associated to high hyperphosphorilation of the 4E-BP1 complex but independent from interaction with PP2A (30). The phosphorylated expression of 4E-BP1 in lymphomas ranges from a high variability depending on lymphoma subtype (31) to almost ubiquitous in AIDS-related lymphoma (32). However, few studies have been published on specific MCV oncogenic mechanisms, which makes it difficult to provide a clear explanation on how this virus and lymphomas are related.

One of the strengths of the study is its epidemiologic study design. Previous studies used cross-sectional data from diverse lymphoma cases, whereas our case–control design allowed us to compare exposure in cases versus controls and adjust for possible confounding factors obtained by questionnaire. Recruited cases were incident lymphoma subjects recently diagnosed, most of them with blood samples taken before treatment initiation. Furthermore, immunosuppressed subjects due to HIV or organ draft recipients were excluded from the study.

We conclude that serologic evidence of exposure to MCV is associated with DLBCL and to a lesser extent to other lymphoma entities. Given the oncogenic nature of this polyomavirus, its possible etiologic role in DLBCL deserves further investigation, both at molecular as well as epidemiologic levels. Prospective studies are needed to preclude a reverse causality effect of these associations

No potential conflicts of interest were disclosed.

Conception and design: R.P. Viscidi, S. de Sanjose

Development of methodology: R.P. Viscidi, S. de Sanjose

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Robles, A. Poloczek, D. Casabonne, Y. Benavente, R.P. Viscidi, S. de Sanjose

Writing, review, and/or revision of the manuscript: C. Robles, A. Poloczek, D. Casabonne, E. Gonzalez-Barca, R. Bosch, Y. Benavente, R.P. Viscidi, S. de Sanjose

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Robles

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Poloczek, E. Gonzalez-Barca, R. Bosch, R.P. Viscidi, S. de Sanjose

Study supervision: S. de Sanjose

The authors would like to thank Laura Costas and Sara Tous for their helpful comments and suggestions, Barbara Silver for assistance in performing ELISA assays and all those subjects who took part in this study, providing questionnaire data and blood samples.

This work was partially supported by public grants from the European Commission (grant references QLK4-CT-2000-00422 and FOOD-CT-2006-023103); from the Spanish Ministry of Health (grant references CIBERESP, FIS 08-1555, FIS 11 08110, RCESP C03/09, RTICESP C03/10, and RTIC RD06/0020/0095), from the Marató de TV3 Foundation (grant reference 051210), and from the Agència de Gestió d'Ajuts Universitaris i de Recerca – Generalitat de Catalunya (grant reference 2009SGR1465) who had no role in the data collection, analysis or interpretation of the results.

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.