Improved detection suggests all Merkel cell carcinomas harbor Merkel polyomavirus

A human polyomavirus was recently discovered in Merkel cell carcinoma (MCC) specimens. The Merkel cell polyomavirus (MCPyV) genome undergoes clonal integration into the host cell chromosomes of MCC tumors and expresses small T antigen and truncated large T antigen. Previous studies have consistently reported that MCPyV can be detected in approximately 80% of all MCC tumors. We sought to increase the sensitivity of detection of MCPyV in MCC by developing antibodies capable of detecting large T antigen by immunohistochemistry. In addition, we expanded the repertoire of quantitative PCR primers specific for MCPyV to improve the detection of viral DNA in MCC. Here we report that a novel monoclonal antibody detected MCPyV large T antigen expression in 56 of 58 (97%) unique MCC tumors. PCR analysis specifically detected viral DNA in all 60 unique MCC tumors tested. We also detected inactivating point substitution mutations of TP53 in the two MCC specimens that lacked large T antigen expression and in only 1 of 56 tumors positive for large T antigen. These results indicate that MCPyV is present in MCC tumors more frequently than previously reported and that mutations in TP53 tend to occur in MCC tumors that fail to express MCPyV large T antigen. A human polyomavirus was recently discovered in Merkel cell carcinoma (MCC) specimens. The Merkel cell polyomavirus (MCPyV) genome undergoes clonal integration into the host cell chromosomes of MCC tumors and expresses small T antigen and truncated large T antigen. Previous studies have consistently reported that MCPyV can be detected in approximately 80% of all MCC tumors. We sought to increase the sensitivity of detection of MCPyV in MCC by developing antibodies capable of detecting large T antigen by immunohistochemistry. In addition, we expanded the repertoire of quantitative PCR primers specific for MCPyV to improve the detection of viral DNA in MCC. Here we report that a novel monoclonal antibody detected MCPyV large T antigen expression in 56 of 58 (97%) unique MCC tumors. PCR analysis specifically detected viral DNA in all 60 unique MCC tumors tested. We also detected inactivating point substitution mutations of TP53 in the two MCC specimens that lacked large T antigen expression and in only 1 of 56 tumors positive for large T antigen. These results indicate that MCPyV is present in MCC tumors more frequently than previously reported and that mutations in TP53 tend to occur in MCC tumors that fail to express MCPyV large T antigen.


Introduction
Merkel cell carcinoma (MCC) is a rare skin cancer with high risk for metastasis and death (1). MCC has features of a neuroendo crine tumor, with expression of synaptophysin (SYP) and chro mogranin A (CHGA) (2). In addition, most MCCs express cyto keratin 20 (KRT20; CK20), which distinguishes it from other neuroendocrine tumors (3,4). Risk factors for developing MCC include advanced age, prolonged lifelong UV exposure, and an immunocompromised state from solid organ transplantation or HIV1 infection (5)(6)(7). In addition, there is an increased incidence of MCC in patients with chronic lymphocytic leukemia (CLL) and other hematologic proliferative disorders (8,9).
Recognition of the increased risk for developing MCC in immu nocompromised patients led to a search for a pathogenic cause and the discovery of the Merkel cell polyomavirus (MCPyV). Sequencing of RNA from 4 MCC specimens revealed a novel tran script distantly related to large T antigen of the polyomavirus fam ily (10). Sequencing of the complete MCPyV genome revealed that it had features typical of a polyomavirus, including a circular dou blestranded DNA genome of 5,387 bp; an early region encoding genes for large and small T antigens; a late region with genes for the viral coat proteins VP1, VP2, and VP3; and a regulatory region that contained the viral origin of replication and a bidirectional promoter for the early and late genes.
MCC cells do not typically harbor episomal MCPyV viral DNA. Instead, the MCPyV viral genome is clonally integrated in the tumor chromosomal DNA, as first revealed by Southern blot ting (10). Sequencing of the integrated MCPyV viral genomes from MCC has revealed mutations and deletions of the early region that result in truncation of the large T antigen to encode only the Nterminal 260-455 residues instead of the fulllength 817 residues, while the small T antigen coding sequence remains intact (10)(11)(12). It is not known whether Merkel cells are permissive for MCPyV viral replication. However, expression of fulllength MCPyV T antigen in a MCC cell line has been reported to promote replication of an integrated viral origin of replication (11,13).
Since the original report, several studies have confirmed the pres ence of the virus in approximately 80% of MCC (14)(15)(16)(17). In most studies, MCPyV DNA has been detected by PCR amplification of DNA extracted from formalinfixed, paraffinembedded (FFPE) tissue samples. This approach has led to the recognition that the number of copies of viral DNA integrated in a tumor can vary from less than one to several hundred copies per cell (18)(19)(20). In addition to detection of viral DNA in MCC, immunohistochemistry with antibodies specific for MCPyV large T antigen and small T anti gen has been reported (20,21). Similarly, immunohistochemistry studies of FFPE sections of MCC report that approximately 80% of MCCs express MCPyV large T antigen (17,22).
Based on homology to other polyomaviruses, the MCPyV large and small T antigens are predicted to be oncogenic and contribute directly to the carcinogenesis of MCC. Given that approximately 80% of MCCs contain MCPyV viral DNA and express MCPyV large T antigen, the question arises whether the remaining MCC can be distinguished by any feature other than the absence of MCPyV. If an MCC tumor does not contain MCPyV DNA or express T anti gens, then it would be predicted that these tumors contain muta tions in oncogenes or tumor suppressor genes that contribute to oncogenesis. To date, few mutations in known oncogenes or tumor suppressors have been reported in MCC. Inactivating muta tions in the p53 tumor suppressor gene (TP53) have been reported to occur rarely in MCC (23,24). In addition, activating mutations in PIK3CA have also been reported in a few MCC tumors (25,26).
This study was performed to determine whether the sensitivity of detection of MCPyV in MCC could be improved, with the ulti mate goal of enabling highconfidence discrimination between viruspositive and negative tumors. To accomplish this goal, we analyzed 75 archival FFPE MCC tumor specimens obtained from 60 patients. We performed immunohistochemistry staining with a newly developed mouse monoclonal antibody specific for MCPyV large T antigen. We also developed several quantitative PCR (qPCR) primer and probe sets to determine the viral copy number per tumor cell. Furthermore, we performed mass spectrometry based genotyping of 112 oncogenes and tumor suppressor genes from DNA extracted from these same tumor samples.

Results
Cases. FFPE specimens corresponding to 60 patients presenting with MCC to a referral specialty clinic were included in this study (Supplemental Table 1; supplemental material available online with this article; doi:10.1172/JCI64116DS1). The patients included 25 females and 35 males who ranged in age from 45 to 90 years (median, 74). Fourteen patients had a prior history, comorbidity, or medication consistent with an immunocompromised state from malignancy, autoimmune disease, or solid organ transplantation.
From this cohort, a total of 75 archival MCC samples were available for study that included 44 primary cutaneous, 12 meta static lymph node, and 6 metastatic visceral tumors obtained at initial presentation and 13 recurrent tumors obtained after ini tial therapy (Supplemental Table 1). Review of clinical pathol ogy reports indicated that 59 of 60 MCC patient specimens were positive for CK20 immunostaining. Only one specimen, from unique patient identifier (UPI) 31, had no reported CK20 stain ing. DNA was extracted from all samples, and FFPE sections from 58 patients corresponding to 72 samples were available for immunohistochemistry staining.
Western blotting of MCPyV large T antigen. The CM2B4 mouse monoclonal antibody was generated against a peptide corre sponding to residues 116-129 of MCPyV large T antigen con tained within the second exon of large T antigen (20). For addi tional antibodies against MCPyV T antigens, we generated a mouse monoclonal antibody (Ab3) against a fragment of MCPyV large T antigen corresponding to the Nterminal 260 residues  (i.e., 1-260) produced in bacteria ( Figure 1A). Additional experi ments confirmed that Ab3 recognized large T antigen and not small T antigen, indicating that residues between 79 and 260 were required for binding. Epitope mapping with an overlapping series of linear peptides corresponding to the entire large T antigen coding region indicated that Ab3 and CM2B4 could bind to the same peptides, suggesting that they recognize a similar epitope on MCPyV large T antigen (Yanis L. Tolstov, Yuan Chang, and Patrick S. Moore, University of Pittsburgh Cancer Institute, Pitts burgh, Pennsylvania, USA, personal communication).
We compared the Western blot signal obtained with Ab3 and CM2B4 using lysates prepared from 4 MCC cell lines, including MKL1, MKL2, MS1, and WaGa. Two cell lines, UISO and U2OS, without MCPyV as well as a U2OS derivative containing a cDNA for MCPyV large T antigen served as controls ( Figure 1B). Both Ab3 and CM2B4 detected a signal corresponding to MCPyV large T anti gen in the 4 MCC cell lines but not in UISO or U2OS cells. Notably, the size of the large T antigen varied in the 4 MCC cell lines, reflect ing differences in the truncating mutations (27). The U2OS cell line with large T antigen expressed fulllength as well as a smaller version likely corresponding to the 57kDa alternatively spliced form (11).
Immunohistochemistry staining of MCC specimens. Sections from MCC tumor specimens were stained with Ab3 and CM2B4 (20). Both antibodies were optimized for signal strength and back ground staining using MCPyVpositive sections of MCC. Sec tions of MCC corresponding to 58 unique patients were available for immunostaining with Ab3 and 57 with CM2B4. Scoring was based on the intensity of signal, with 3+ indicating strong staining; 2+, moderate; 1+, weak; and 0, absent.
In most cases, both antibodies detected a nuclear signal sur rounded by a small rim of unstained cytoplasm ( Figure 3A, UPI 26). Nuclear and cytoplasmic signals were detected in a few samples that strongly stained with Ab3 and CM2B4 ( Figure 3A, UPI 18). In several cases, staining with higher concentrations of Ab3 increased the strength of the signal. For example, specimens from UPI 37 stained weakly with Ab3 at 0.6 μg/ml and moderately well with Ab3 at 2.4 μg/ml but negative when stained with CM2B4 ( Figure 3B).
Samples from a total of 38 cases stained strongly (3+) or moder ately (2+) with CM2B4, and each of these also stained strongly or moderately with Ab3 ( Figure 2A and Supplemental Table 1). In con trast, Ab3 stained a total of 50 cases strongly or moderately. Sam ples from a total of 6 cases stained weakly with Ab3, and 2 of these also stained weakly with CM2B4, including UPI 27 ( Figure 3A) and UPI 14. Ab3 detected a specific signal in 9 samples that was not detected with CM2B4 (for example, UPI 30 and UPI 37; Figure 3, A and B). Conversely, there were no specimens that stained posi tive with CM2B4 that were negative with Ab3. Overall, the staining intensity for large T antigen was higher with Ab3 compared with CM2B4 in 27 of 57 (47%) samples. Importantly, when different specimens from the same patient were stained, we observed con sistent staining intensity between them. For example, the primary tumor and a metastatic lymph node from UPI 30 ( Figure 3A) both stained positive with Ab3 but negative with CM2B4. Samples from only two cases, UPI 15 and UPI 42, were not positive when stained with either Ab3 or CM2B4 ( Figure 3C). Overall, Ab3 staining inten sity was significantly correlated with CM2B4 staining (Table 1).
To rule out the possibility of nonspecific staining with Ab3, we performed staining on sections of discarded tonsillar tissue (5 unique specimens) and reactive lymph nodes (2) from unrelated cases. Staining with Ab3 revealed no detectable signal from these 7 different normal tissue specimens. Representative stains with Ab3 of normal lymph node and tonsil are shown in Figure 4A. We tested 10 unique GI neuroendocrine tumors and detected no signal by immunohistochemistry staining with Ab3 (Supplemen tal Table 2). Two representative cases are shown in Figure 4B. In addition, immunostaining of 8 unique small cell lung carcinomas with CM2B4 and Ab3 were negative (Supplemental Table 2). Two representative cases are shown in Figure 4, C and D. Sections from an MCC case (UPI 36) served as a positive control for Ab3 and C2B4 staining performed at the same time as staining of the small cell lung carcinoma specimens ( Figure 4E). qPCR detection of MCPyV DNA. We developed a qPCR assay to determine the copy number of viral genomes per tumor cell. Two different qPCR primer sets were adapted from a previous report, and 3 additional quantitative PCR primer and probe sets were generated (15). All primer and probe sets were designed to amplify sequences corresponding to nucleotides 196-1,407 of MCPyV encoding small T antigen and the Nterminal 260 residues of LT ( Table 2). This region of MCPyV genome was selected because it was present within all sequenced MCPyV variants obtained from MCC tumors. We designed 3 primer sets corresponding to the second exon (LT2, Set6, and Set7) of large T antigen and 2 primer sets (LT3 and Set9) corresponding to the unique region of small T antigen ( Figure 1A). These primer sequences were identical to most MCPyV sequences and were not crossreactive to any other DNA sequence in the NCBI database. Primers specific for the human TPO gene were used to normalize for diploid gene copy number, since this gene was reported to have infrequent copy number alterations in a study of MCC using array comparative genomic hybridization (28).
To validate the PCR primers, we tested them against DNA extracted from the 4 MCPyVpositive MCC cell lines and the U2OS cell line containing the cDNA for MCPyV large T antigen ( Figure 1C). As expected, no amplification with any MCPyV primer set was observed in DNA extracted from the parental U2OS cell line. In contrast, the U2OS cell line containing the large T anti gen cDNA was amplified with the 3 primer sets specific for large T antigen (LT2, Set6, and Set7) but not with primer sets LT3 and Set9 specific for small T antigen ( Figure 1C).
All 5 primer sets amplified DNA from the 4 MCC cell lines tested. The MKL2 cell line contained the lowest copy number, with each of the MCPyV primer sets yielding Ct values approximately equal to the values corresponding to the TPO gene. Based on this obser vation, we normalized the values obtained from the MKL2 cell line to be equivalent to a single copy of the viral genome per cell to compare with all other samples (Supplemental Table 1). The MS1 cell line contained 3-5 copies of the viral genome per cell depending on the results from the 5 individual primer sets, while the MKL1 cell line contained an average of 21 copies per cell. The WaGa cell line showed some variation in copy number, ranging from 12 to 52 copies per cell depending on the primer used. Of note, the relative MCPyV copy number in the 4 MCC cell lines corresponded roughly to the relative strength of the large T anti gen signal by Western blotting, with the highest levels detected in WaGa and MKL1 and the lowest in MKL2 cells ( Figure 1B).
Using these validated primer and probe sets, we performed qPCR with DNA extracted from the same FFPE tumor blocks used for immunohistochemistry. To conserve tumor DNA samples, we initially tested 3 MCPyV primer/probe sets: LT2, LT3, and Set7, together with TPO in triplicate for all samples. The copy number for each primer set was compared with the TPO value and normal ized to the values obtained from the MKL2 cell line (Supplemen tal Table 1). When the results indicated that the viral copy number per cell was less than 1.0 for any of the LT2, LT3, and Set7 prim er sets, then Set6 and Set9 were also tested. Overall, the MCPyV genome copy number obtained from 75 specimens corresponding to 60 unique patients using the 5 primer/probe sets varied from less than 0.001 to 3,203 copies per cell.  Most samples showed good agreement in copy number when dif ferent specimens from the same patient were tested (Supplemental Table 1). Furthermore, there was good agreement for a given speci men when different primer sets were tested (Table 1). However, there were several specimens that showed agreement between 2 or more primer sets but not with other sets. For example, UPI 59 showed copy number values between 5 and 15 with LT2, LT3, Set6, and Set9 but less than 0.1 for Set7. Remarkably, UPI 11 revealed values more than 2,000 with LT3, Set7, and Set9 but less than 0.1 with LT2 and Set6.
For 48 of 60 cases (80%), at least one primer set detected greater than 2 copies per cell, consistent with the clonal integration of MCPyV DNA in every tumor cell (Supplemental Table 1 and Figure 2B). Of the remaining 12 cases with less than 2 copies of MCPyV genome detected, 7 cases had viral copy number greater than 0.1 detected with at least one primer set. For 2 of these cases, UPI 15 and 42, immunohistochemistry staining for large T anti gen was negative with both antibodies ( Figure 3C). An additional 5 cases (UPI 27, 47, 49, 57, and 60) yielded viral copy number val ues greater than 0.03 but less than 0.1. Despite the low copy num ber values, samples from all 5 of these cases were stained positive with Ab3, including 3 that stained moderately (2+). In addition, samples from 2 of these cases also stained positive with CM2B4.
To rule out nonspecific amplification of DNA using MCPyV specific primers from tissues, we extracted DNA from FFPE tissue blocks of 3 tonsillar and 3 reactive lymph nodes from unrelated cases. We used DNA extracted from the MKL2 cell line as a posi tive control. From all control tissues, we were able to amplify DNA using the primer and probe set for the TPO gene with Ct values similar to MKL2 DNA. In contrast, we found no signal for MCPyV with any of the 5 primers and probe sets from the tonsillar or reac tive lymph node specimens. These results indicate that viral copy numbers as low as 0.03 (MCC UPI 60) were specific for detection of MCPyV DNA. Furthermore, the absence of staining with Ab3 in normal tonsil and lymph node sections as well as in GI neuroendo crine and small cell lung carcinoma specimens support the conclu sion that positive staining in MCC was specific for the presence of MCPyV large T antigen.
We prepared a nonredundant list of patient values using the high est reported value for each assay including immunohistochemistry and qPCR and then performed a pairwise Spearman rank cor relation. As shown in Table 1, we observed a significant correla tion between every combination of antibody staining and qPCR primer sets. The highest degree of correlation was between Ab3 and CM2B4 staining intensity, as well between qPCR values for LT3 and Set7, LT3 and Set9, and LT2 and Set7. No differences were observed in overall survival or time to disease progression in patients with high versus low MCPyV copy number or large T anti gen expression (data not shown).
OncoMap. We employed a mass spectrometry-based genotyp ing method to detect specific point substitution mutations in 112 commonly mutated oncogenes and tumor suppressor genes (29,30). This approach was designed to detect more than 1,000 polymorphic and mutated alleles corresponding to these genes. DNA isolated from the 60 unique patients encompassing a total of 75 samples plus the 4 MCC cell lines were evaluated. After ini tial screening, all potential mutations were validated for the entire tumor set by homogenous mass extension (hME) (29,30).
From 75 tumor specimens, only 4 tumor specimens con tained a validated point mutation in any of the 112 oncogenes and tumor suppressor genes evaluated (Supplemental Table 1). Two samples with negative staining for MCPyV large T antigen by immunohistochemistry with Ab3 and CM2B4 and PCR copy number less than 0.7 for all 5 primer sets contained an inactivat ing point mutation, R248Q, in TP53 (UPI 42 and 15). In addition, a tumor specimen (UPI 18) with strong staining for large T anti gen with both Ab3 and CM2B4 and copy number between 2 and 4 contained a different inactivating point substitution mutation, R273C, in TP53. The presence of inactivating point mutations in p53 in both tumors without large T antigen expression and in only 1 of 58 specimens that expressed large T antigen is highly signifi cant (Fisher's exact P = 0.0017). A fourth tumor specimen (UPI 38) contained a point substitution in ATM P604S (31). Notably, none of the other tumor specimens nor the 4 MCC cell lines showed the presence of any additional point mutations or polymorphisms tested in this panel.

Discussion
The discovery of MCPyV and the detection of viral DNA and T anti gens in MCC provided an important new insight into the onco genesis of this highly aggressive cancer (10,20). Since this initial discovery, many centers from around the world have confirmed the presence of MCPyV in approximately 80% of MCCs. A vari ety of methods have been used to detect the presence of the virus, including Southern blot analysis (10), PCR amplification of viral DNA, DIPS (detection of integrated polyomavirus sequences), in situ hybridization (16), DNA sequencing by hybrid capture (12), RTPCR, and immunohistochemistry of the viral T antigens (20).
One goal of this study was to determine whether detection of the MCPyV and viral gene products could be improved. A new mono clonal antibody, Ab3, was developed that has markedly increased sensitivity in detecting MCPyV large T antigen (97%) in MCC com pared with CM2B4 (80%) (20). In addition to its improved sen sitivity in detecting MCPyV large T antigen expression in MCC, Ab3 does not stain lymphocytes, even when used at higher con centrations or when staining tonsillar and reactive lymph node tissue specimens. The lack of nonspecific staining may be espe cially useful in clinical specimens, since MCC tumors frequently contain infiltrating lymphocytes (32). Furthermore, Ab3 did not stain tumor specimens from GI neuroendocrine tumors or small cell lung cancer, supporting its usefulness as a clinical biomarker for MCC. It should be noted that CK20 staining was positive in 59 of the 60 cases tested, and the CK20 status of the remaining case was not tested. Therefore, it is not known whether Ab3 can detect MCPyV large T antigen in the rare cases of CK20negative MCC. Our study is consistent with an earlier report that the CM2B4 antibody could not detect MCPyV large T antigen in every case of MCC (33). It was noted that the CM2B4 antibody failed to detect large T antigen in several cases of MCC but that many of these negative specimens stained positive with the monoclonal antibody CM5E1 specific for MCPyV small T antigen (33). The combined use of CM2B4 and CM5E1 led to detection of MCPyV T antigens in 47 of 51 (92%) cases of MCC, whereas CM2B4 alone detected large T antigen in 75% of cases (33). Our study is com patible with this small T antigen expression study, where use of CM2B4 alone underestimated the fraction of MCC cases that express the MCPyV T antigens.
To date, every report has detected MCPyV DNA in a significant proportion of MCC tumor specimens in their cohorts. Using PCR and related methods, most centers have reported that approximate ly 80% of MCCs contain detectable amounts of MCPyV viral DNA. In our study, more than 0.1 viral genome per cell was detected in 55 of 60 cases (92%), consistent with clonal integration of MCPyV DNA. The slightly higher than expected frequency of MCPyV DNA could be related to the use of 5 unique primer sets, which there by increased the chances of detection. In the 5 remaining cases, MCPyV DNA was detected in a range from 0.03 to 0.08 copies per cell. These values exceeded those obtained from qPCR amplifica tion of normal lymph node and tonsil tissue. This is consistent with a prior study that observed 60foldhigher levels of MCPyV DNA in MCC specimens compared with the levels found in normal tissues (19). Therefore, in this series of MCC, there was evidence for specific detection of MCPyV DNA in every case. Further evidence supporting this conclusion is that in all 5 cases of MCC with less than 0.1 copy viral genome per cell, expression of large T antigen was detected with Ab3, and 2 also stained with CM2B4.
It should be noted that studies with extracted DNA from fresh frozen MCC samples have detected a higher frequency of MCPyV than studies with extracted DNA from FFPE tumor blocks. For example, one study detected MCPyV in all 12 cases of MCC when DNA was isolated from fresh frozen tumor specimens and in only 9 of 20 cases when DNA was extracted from FFPE specimens (34). It is likely that the formalin fixation process, longterm storage, and harsh extraction methods of archival MCC specimens reduce the likelihood of detection of MCPyV DNA by qPCR. Furthermore, the frequent mutations and polymorphisms observed in MCPyV DNA obtained from MCC reduce the efficiency of detection by qPCR (10).
The results of qPCR amplification revealed a large variation in MCPyV genome copy number per cell in MCC. Peak values ranged from 0.03 to several thousand copies per cell. Notably, we esti mated copy number based on the assumption that the MKL2 cell line had a single copy of MCPyV DNA per cell, since the Ct values for MCPyV were similar to those obtained for TPO. This assump tion may have underestimated the copy number of MCPyV in the MKL2 cell line, thereby systematically reducing the estimate of viral copy number in all other specimens. For example, if the MKL2 tumor cell line contains 2 copies of MCPyV DNA per cell, then each tumor specimen would contain twice as many viral cop ies than we have reported here. In addition, it should be noted that normalization of each tumor specimen to its corresponding TPO DNA content may not be accurate in all cases, since some tumors may have undergone copy number alterations of this gene (28). We observed a significant correlation in the intensity of immuno histochemistry staining with Ab3 and CM2B4 relative to the copy number of MCPyV per cell. Given that MCPyV integration occurs randomly in the host cell DNA, expression of the MCPyV T anti gens is likely dependent on the local chromatin accessibility to transcription factors and RNA polymerases. Expression of MCPyV T antigens is likely to be higher when the viral DNA is integrated into open chromatin regions than into heterochromatin regions even when the viral copy number may be very high.
Determination of point substitution mutations in 112 onco genes and tumor suppressor genes revealed the surprising result that most tumor samples (56 of 60; 93%) did not contain any point substitution mutations in the genes surveyed. In addition, we did not detect any mutations in 4 MCC cell lines that contain MCPyV. Consistent with previous studies, we did not find activating muta tions in PDGFRA or KIT (23,(35)(36)(37)(38). We did identify 3 tumors with p53 point substitutions, most notably in 2 specimens with absent expression of MCPyV large T antigen. Notably, the MKL1 cell line had been reported to express wildtype p53 transcript (39). Other studies have reported infrequent p53 mutations (24,40,41). A pre vious study found p53 mutations in 11 of 23 MCC tumors that had no detectable large T antigen expression by immunohistochemistry using CM2B4 antibody and absence of p53 mutations in 16 large T antigen-positive MCC tumors (42). The OncoMap genotyping approach does not detect large deletions, and therefore it is pos sible that some MCC tumors in our study contained deletions of the TP53 gene. It should be noted that cervical cancers with evi dence for human papilloma virus (HPV) typically do not contain p53 mutations, while HPVnegative cervical cancers often contain p53 mutations (43). Thus, while all MCC may contain MCPyV viral DNA, if the expression of the viral T antigens is absent, tumors may be selected for the presence of inactivating p53 mutations.
It should be noted that integration of MCPyV DNA into MCC typically results in truncation of the large T antigen viral onco gene. The mutations preserve the retinoblastoma protein binding or LXCXE motif but disrupt the helicase domain of large T antigen (11). The helicase domain of the related polyomavirus SV40 large T antigen binds specifically to p53 (44). It has not been reported whether wildtype, fulllength MCPyV large T antigen can bind to p53 in a manner similar to other wellknown polyomaviruses including SV40, BKPyV, and JCPyV (45). In any case, integration and mutation of MCPyV would be expected to delete this potential p53binding domain of large T antigen. Therefore, if MCPyVposi tive MCC tumors disrupt p53 signaling, then this is accomplished by another mechanism, one that does not involve the C terminus of large T antigen. It is possible that MCPyV small T antigen con tributes to functional inactivation of p53 in a manner similar to mouse polyoma virus (46). We did not address the possibility that episomal viral DNA was present in these tumor specimens. How ever, mutation of the large T antigen helicase domain is expected to disable its ability to support episomal viral replication.
The significance of the point substitution P604S in ATM is unclear, since it is not within the catalytic domain. This may rep resent a normal polymorphism, although normal tissue was not available from this patient to exclude this possibility (31). Two recent reports have identified point substitution mutations in the PIK3CA gene in MCC specimens (25,26). In a study of 60 MCC archival FFPE specimens, Nardi and colleagues identified TP53 mutations in 3 cases and PIK3CA mutations in 6 cases. Notably, 2 of the 3 cases with TP53 mutations and only 1 of the 6 cases Table 2 qPCR primer and probes with PIK3CA mutations contained MCPyV DNA according to PCR amplification. In the study by Becker and colleagues, 2 of 38 MCC specimens contained PIK3CA mutations (15). Both tumors with the PIK3CA mutation tested positive for MCPyV by PCR amplifi cation of viral DNA. Notably, this group did not find mutations in the PIK3CA gene in the MKL1 and WaGa cell lines.
In summary, this study revealed the presence of MCPyV DNA in all MCC tumors analyzed. Furthermore, the development of a new monoclonal antibody enabled the detection of MCPyV large T antigen expression in 97% of MCC cases tested. The two cases without detectable levels of large T antigen contained an inactivat ing mutation in TP53. We conclude that the presence of MCPyV in MCC is more common than previously reported and that improved detection methods may reveal that all MCC specimens contain viral DNA. The results of this study support the model that the MCPyV T antigens contribute to the pathogenesis of MCC and when their expression is absent, somatic mutations in TP53 and potentially other genes may be selected during oncogenesis.

Methods
Immunohistochemistry. Immunohistochemistry was performed using 4μm thick FFPE tissue sections. Briefly, slides were soaked in xylene, passed through graded alcohols, and put in distilled water. Slides were then pre treated with 1.0 mM EDTA, pH 8.0 (Zymed), in a steam pressure cooker (Decloaking Chamber, Biocare Medical) as per the manufacturer's instruc tions, followed by washing in distilled water. All further steps were per formed at room temperature in a hydrated chamber. Slides were pretreated with Peroxidase Block (Dako) for 5 minutes to quench endogenous peroxi dase activity. Mouse monoclonal antibodies Ab3 and CM2B4 (Santa Cruz Biotechnology Inc.) were applied at the indicated final concentration (0.6 or 2.4 μg/ml for Ab3; 2 μg/ml for CM2B4) in Dako diluent for 1 hour. Slides were washed in 50 mM TrisHCl, pH 7.4, and detected with antimouse Envision+ kit (DAKO) as per the manufacturer's instructions. After further washing, immunoperoxidase staining was developed using a DAB chromo gen (Dako) and counterstained with hematoxylin. Images were acquired using an Aperio ScanScope Digital Slide Scanner and ImageScope Viewer.
DNA extraction and quantification. FFPE tumor blocks were sectioned and stained with hematoxylin and eosin, and tumors were identified. DNA was extracted from 7 to 10μm sections of macrodissected tumors using QIAGEN QIAamp DNA FFPE following the manufacturer's instructions, with the following exceptions: CitriSolv (Fisher) was used for deparaf finization and DNA samples were treated with RNAse A after Proteinase K digestion was performed. DNA concentration was determined with Pico Green dsDNA Quantitation Reagent (Life Technologies).
Purification of total DNA from cell lines was performed by using a DNeasy Blood & Tissue kit (QIAGEN) following the manufacturer's instructions. DNA concentration was determined with a NanoDrop spectro photometer (NanoDrop Technologies).
The qPCR assays were performed with 96well optical grade plates and caps (Agilent Technologies) with a Stratagene Mx3005P realtime PCR machine programmed for 50°C × 2 minutes with an initial denaturation at 95°C × 15 minutes, followed by 45 cycles at 95°C × 15 seconds and 60°C × 60 seconds. Reactions were performed in triplicate with 15 ng input DNA.
MCPyV copy number was calculated with the assumption that the MKL2 cell line contained 1 copy of viral DNA per cell, and results were analyzed using the 2 -ΔΔCt method, where ΔΔCt = (Ct MCPyV primer set -Ct TPO) -(Ct MCPyV primer set -Ct TPO) MKL2 (47).
OncoMap. OncoMap v3.0 (DanaFarber Cancer Institute) genotyping was performed on 112 genes corresponding to 1,047 mutations tested, as previ ously described (30). Briefly, DNA extracted from FFPE tumor specimens and MCC cell lines was subjected to whole genome amplification, followed by multiplex PCR to amplify regions of interest. After denaturation, PCR products were incubated with specific oligonucleotides, and a primer extension reaction was performed in the presence of chainterminating dideoxynucleotides. Primer extension products were analyzed by MALDI TOF mass spectrometry to determine the mutation status. The validation chemistry (hME) was performed on unamplified DNA to preclude any arti fact introduced by the whole genome amplification process.
Statistics. A nonredundant list of patient values using the highest reported value for each immunohistochemistry and copy number assay was pre pared, and then a pairwise Spearman rank correlation was performed ( Table 1). P values less than 0.05 were considered significant.
Study approval. This study was conducted according to the principles expressed in the Declaration of Helsinki. Work on human tissue was con ducted on excess archival human material after obtaining approval from the DanaFarber Cancer Institute Institutional Review Board and informed consent from each subject. The patients' medical records were reviewed to collect demographic data, clinical history, presentation, and staging.