Merkel cell carcinoma and Merkel cell polyomavirus: evidence for hit-and-run oncogenesis.

TO THE EDITOR In the majority of cases, Merkel cell carcinoma (MCC) harbors Merkel cell polyomavirus (MCV) DNA, which—at least in the tested cases—is clonally integrated at random sites of the host genome and carries characteristic mutations (Feng et al., 2008; Shuda et al., 2008; Sastre-Garau et al., 2009). The latter notion together with our recent reports demonstrating that the expression of the MCV-derived T antigens (TAs) is essential for growth of MCC cell lines in vitro, and in vivo has established MCV as a causative factor in the oncogenesis of MCC (Houben et al., 2010, 2011). The immunological characteristics of MCC as discussed in recent reports corroborates with this hypothesis (Paulson et al., 2011; Touze et al., 2011). For example, Touze et al. (2011) demonstrated that high titers of antibodies against the viral capsid protein VP1 are more frequently detected in MCC patients—although VP1 seems not to be expressed in the tumors—than in a control cohort further supporting MCV infection as a causative factor in most MCC cases. Moreover, Touze et al. (2011) confirm several previous reports that in a substantial fraction of the MCV-positive MCCs the virus is present with less than one copy per cell. For these tumors a role of MCV in their etiology is unclear. In this regard, published data on the impact of MCV presence or viral load on the clinical characteristics of MCC is conflicting. In two recent reports, a worse prognosis for patients with MCV-negative tumors or tumors with low viral copy per cell numbers and lack of large T-antigen (LTA) expression have been reported (Sihto et al., 2009; Bhatia et al., 2010). In others, as well as in our own series, no statistical significant survival differences between patients with MCV-positive or -negative tumors were evident (Handschel et al., 2010; Schrama et al., 2011). In their recent publication, Touze et al. (2011) reported a better progression-free survival for patients with high MCV VP1 antibody titers. Thus, based on their own findings and the reports in the literature, they suggest a model with two subgroups of MCC: one less aggressive with clear MCV etiology, as demonstrated by high viral load, LTA expression, and high titers of MCV antibodies; and a second more aggressive type of MCC with a different etiology, as indicated by low viral load, lack of LTA expression, and MCV antibody titers similar to those observed in the general population. However, although high MCV antibody titers (>10,000) are observed in only 7.3% of the controls, 61.5%, 65%, and 61.1% of the MCC patients with a virus copy per cell of <1, 1–9, and >9, respectively, demonstrated such a high anti-VP1 antibody titers (Touze et al., 2011) arguing against a correlation of viral load and anti-VP1 antibody titer. In contrast, however, VP1 antibody titers in non-MCC patients seem to be correlative to viral presence in the skin (Faust et al., 2011). Consequently, when higher MCV antibodies levels in MCC patients are regarded as an indication for MCV etiology, these data would support the idea that MCV has also been a causal factor for many MCCs even if the MCV genome is not detected in every cell at the time of analysis. It seems conceivable that viral infection initially induces transformation, but that some transformed cells may become independent of MCV. Indeed, adenovirus transformed hamster cells can retain their oncogenic phenotype even though the previously integrated viral DNA is lost (Pfeffer et al., 1999). Overcoming the addiction to the viral oncoproteins is a prerequisite for such a process. Such a loss of oncogene addiction has been observed in animal models: in transgenic mice, the inducible expression of SV40 TAs in the submandibular gland resulted in a severe hyperplasia; silencing of the TA expression within the first 4 months resulted in restitution of the normal state, whereas silencing at later time points was not associated with a regression anymore (Ewald et al., 1996). Importantly, a similar phenomenon might exist in MCC. We recently demonstrated that MCV-positive MCC cell lines strictly depend on the expression of the MCV TAs as TA knockdown leads to growth repression, whereas MCV-negative cell lines were not affected by the TA small hairpin RNA (shRNA; Houben et al., 2010). Notably, however, we also established one MCV-positive cell line (LoKe), which upon TA knockdown—in contrast to seven other MCV-positive cell lines tested—does not display any growth inhibition (Figure 1). LoKe cells were derived from a 65-year-old male patient with histologically confirmed MCC. The LoKe MCV LTA-coding sequence demonstrates the MCC characteristic premature stop codon mutation deleting the helicase domain. The truncated LTA consists of 254 amino acids including the Retinoblastoma protein binding-motif (data not shown; Houben et al., 2011). The integration of the viral genome in chromosome 2 of the tumor cells was detected by applying detection of integrated “polyoma” sequences technology (Sastre-Garau et al., 2009; data not shown). Quantitative PCR revealed an MCV virus load of more than one virus genome per cell and immunocytochemistry demonstrated LTA expression in the vast majority of cells. In addition, serum of the respective patient contained high antibody titers against VP1 (Figure 2) and the TAs (data not shown). All this data together imply the involvement of MCV in oncogenic transformation of the LoKe cells. Although we cannot exclude that the TA knockdown achieved by our shRNA is simply not efficient enough in LoKe cells, we think that it is more likely that these cells have acquired additional aberrations enabling growth even in the absence of T-antigen expression. Thus, in some MCC cases, MCV may only be necessary for tumor initiation, whereas additional mutations during tumor progression render T-antigen expression dispensable for them. This might lead on one hand to the loss of MCV within these tumors and on the other to a different biological behavior including a more aggressive phenotype. The authors state no conflict of interest. This work was supported by the Wilhem-Sander-Stiftung (2007.057.2) and the IZKF Würzburg (B-157). We thank Sonja Hesbacher for excellent technical assistance.