We have determined the seroprevalence for three recently identified human polyomaviruses (KIV, WUV, and MCV) and confirmed the seroprevalence of two previously known human polyomaviruses (BKV and JCV) and two monkey polyomaviruses (SV40 and LPV) in a human population, using a VP1 capsomere-based ELISA. Our study evaluates both large adult (n
1501) and pediatric populations (n
721) to determine the overall prevalence and age of first exposure to these viruses. Additionally, our assay evaluated potential serocrossreactivity occurring between these viruses.
Previous studies of human Py serology have used a variety of assays, including hemaglutination inhibition (HI) 
and a virus-like particle (VLP) ELISA-based assay 
, which may not be directly comparable. The hemaglutination inhibition assay requires either intact virions or VLPs and only evaluates a subset of antibodies. HI may be less sensitive for determining BKV and JCV VP1 seroreactivity, as compared to enzyme immunoassays 
. Moreover, not all polyomaviruses exhibit hemaglutination (e.g.
, SV40) and HA has not yet been assessed for KIV, WUV and MCV. A VLP-based ELISA may present conformational epitopes and increased specificity over HI. However, while the VLP-based assay may measure only a subset of antibodies, the capsomere assay has the advantage of measuring all VP1-reactive antibodies, and the use of a casein-glutathione conjugate sterically projects the capsomeres from the well surface, allowing their full exposure to the sera. Nonetheless, seroprevalence determinations are likely somewhat dependent on specific conditions of the assay.
Our observed seropositivity for both WUV (69%) and KIV (55%) was high, despite a low reported detection rate in respiratory tract isolates using PCR 
. The PCR data likely represent active infection or ongoing co-infection, rather than overall exposure rates. From the age stratification data, it appears that primary infection with these viruses occurs during early childhood, with 35% positive between ages 1–3 for WU and 32% positive for KIV. There was no cross-reactivity between WUV and KIV, which may have been expected given only 65% amino acid identity between their VP1 proteins 
We found differential seroreactivity to MCV isolates 350 (25%) and 339 (42%). However, these are not true viral isolates but rather PCR amplified sequences, since no infectious virus has yet been characterized for any of the new human polyomaviruses. The PCR amplifications may have detected defective genomes or variants with minor sequence mutations that occurred after viral integration (specific mutational events have been reported for the Merkel large T-antigen protein 
, and VP1 mutations might be anticipated if they affected productive infection). The differential reactivity in our analysis may therefore have resulted from one PCR variant having a “less native” conformation of the recombinant VP1 protein used in the assays (although sufficiently native to generate VP1 pentamers by electron microscopy) or mutated in critical amino acids that affect a specific epitope. Interestingly, there were 164 sera in our population that were seroreactive to the VP1 protein of MCV isolate 350, but not to 339. Also, 560 samples were seroreactive to isolate 339, but not to isolate 350, suggesting that both isolates of MCV may circulate in the human population. It remains to be determined when authentic, infectious isolates are characterized whether there is actual strain variation in serology. However, as discussed below for JCV, strain differences likely will not substantially affect overall seroprevalence data.
If these are actually different strains of MCV, the differential prevalence may indicate differential geographic exposure frequencies to these MCV isolates 
. Also, MCV isolate 339 may contain a sero-dominant MCV VP1 epitope not present in isolate 350. Five amino acid differences occur between the VP1 proteins of MCV 350 and 339. Based on alignment with the VP1 primary amino acid sequence of SV40 and the known structure of SV40 VP1, MCV 339 and 350 do not vary with respect to the surface VP1 exposed variable loop regions thought to comprise the major antigenic determinants, however differences in amino acids occurring close to the surface may indirectly affect loop conformation (Figure S2
). Specifically, H288 of MCV 350 is D288 in isolate 339 (Figure S2
). Based on alignment with known SV40 VP1 structure, this amino acid difference occurs close to the surface near the HI variable loop. It remains to be determined whether the difference in seroprevalence between MCV isolates 350 and 339 is maintained in suspect disease populations. Our data also support previous studies suggesting that the human population has been exposed to an LPV–like virus, antigenically similar to the primate LPV VP1 
, and LPV-like sequences have recently been detected in the white blood cells of immunocompromised individuals 
. Although the recently discovered MCV exhibits a high degree of sequence similarity with LPV () our competition results indicate that MCV and LPV are antigenically distinct viruses.
Genetic variability of VP1 proteins among polyomaviruses.
While the seroprevalence of BKV in our study population is consistent with previous reports 
, the JCV seroprevalence (39%) was somewhat lower 
. There may be several reasons for this finding: 1) epidemiologic evidence suggests that JCV exposure may differ geographically 
, 2) JCV VP1- specific antibodies may not have as high affinity for the JCV VP1 proteins compared to BKV VP1-specific antibodies, resulting in an overall decreased sensitivity of the JCV assay, 3) although only the MAD-11 JCV genotype has been found to be serologically distinct 
, our assay used only genotype 2B and antibodies other than against genotype 2B may bind that VP1 protein with lower affinity, reducing the sensitivity of the assay.
BKV VP1 IgG antibodies previously have been observed to cross-react with SV40 VP1 
. Using recombinant VP1 proteins to compete potential seroreactivity between BKV and SV40, we found the seroprevalence of SV40 to be approximately 2% ( and Table S1
). In the population of individuals who were seropositive for SV40 and either BKV and/or JCV (n
195), SV40 seroreactivity could be competed with VP1 pentamers of both BKV and JCV in 43% of the coincident population. SV40 seroreactivity could be competed with BKV VP1 capsomeres alone in 32% of coincident samples, while 3% of coincident samples could be competed only with JCV VP1 capsomeres. If there was no specific seroreactivity we did not assign seropositivity to SV40, since we assume that some antibodies must be specific to SV40 to indicate infection. Indeed, we identified 48 individuals with “specific” SV40 antibodies, supporting our assumption that the SV40 serological response is not restricted only to cross-reactive antibodies with BKV or JCV. It is interesting to speculate that BKV infection might confer protection against SV40, an attribute that may have had a selective advantage at one time. Although we cannot explain the residual SV40 serology of 2%, it is possible that a yet unidentified human polyomaviruses may account for this reactivity.
The external surface variable loop domains of the VP1 proteins may present the dominant epitopes of these viruses. For example, comparison of the loop regions of SV40 and BKV reveals a high degree of similarity. Specifically, the BC loops of SV40 and the BK strain of BKV share 53% identity, while the DE and HI loops exhibit 71% and 80% identity, respectively. Comparison of the loops of LPV and MCV, and those of KIV and WUV, reveals fewer similarities (); LPV and MCV exhibit 21% identity in the BC loop, and 26% and 20% in their respective DE and HI loops. KIV and WUV share 33% and 38% respective identity in the BC and DE loops, and only 20% identity in the HI loop. These differences may account for the lack of crossreactivity between these viruses.
Our pediatric population exhibits similar seroprevalence values compared to the adults, indicating that primary infection with polyomaviruses occurs in childhood (). We found seropositivity to BKV, WUV, KIV, LPV, and MCV appears in early childhood whereas that for JCV occurs in pre-adolescence (). Additionally, our data suggest there may be an age-related waning of BKV VP1 specific antibodies, however, our data do not indicate an age-related waning for any of the other 6 polyomaviruses assayed (). While the adult samples are likely representative of the demographics in the Denver area, and reflect healthy individuals within the guidelines for blood donation, the pediatric samples were obtained from inpatients and outpatients at The Children's Hospital Denver and although exclusion criteria were employed, some individuals may have had co-morbid or intercurrent illnesses.
Although animal models have suggested that polyomaviruses could be human tumor viruses 
, no substantive cause-effect relationships have yet been established. The debate over SV40 induction of human tumors remains controversial 
, and our data do not support a permissive human infection with SV40, such as seen with the other human polyomaviruses. If SV40 infects humans, it is very limited in scope. Furthermore, if a specific tumor type is presumed attributable to SV40 infection, serological validation now would be an essential factor in the analysis. The discovery of MCV integrated in a specific tumor 
is a possible indication of the ability of these viruses to contribute to tumorigenesis, but perhaps only in a subset of cell types and only in immunosuppressed individuals. The identification of the human counterpart to LPV may also reveal a connection to human cancer. However, given the high seroprevalence of these viruses, serological support of etiologic connections to specific diseases may be problematic. Nonetheless, these ubiquitous viruses appear to be significant human pathogens in immunosuppressed populations.