Distinct retinoic gene signatures discriminate Merkel cell polyomavirus‐positive from ‐negative Merkel cell carcinoma cells

Limited molecular knowledge of Merkel cell polyomavirus (MCPyV)‐positive and ‐negative Merkel cell carcinoma (MCC) subsets (MCCP/MCCN) has prevented so far the identification of the MCC origin cell type and, therefore, the development of effective therapies. The retinoic gene signature was investigated in various MCCP, MCCN, and control fibroblast/epithelial cell lines to elucidate the heterogeneous nature of MCC. Hierarchical clustering and principal component analysis indicated that MCCP and MCCN cells were clusterizable from each other and control cells, according to their retinoic gene signature. MCCP versus MCCN differentially expressed genes (n = 43) were identified. Protein–protein interaction network indicated SOX2, ISL1, PAX6, FGF8, ASCL1, OLIG2, SHH, and GLI1 as upregulated hub genes and JAG1 and MYC as downregulated hub genes in MCCP compared to MCCN. Numerous MCCP‐associated hub genes were DNA‐binding/‐transcription factors involved in neurological and Merkel cell development and stemness. Enrichment analyses indicated that MCCP versus MCCN differentially expressed genes predominantly encode for to DNA‐binding/‐transcription factors involved in development, stemness, invasiveness, and cancer. Our findings suggest the neuroendocrine origin of MCCP, by which neuronal precursor cells could undergo an MCPyV‐driven transformation. These overarching results might open the way to novel retinoid‐based MCC therapies.


| INTRODUCTION
Merkel cell carcinoma (MCC) is rare yet aggressive cutaneous neoplasm with a prevalence of 0.1-2.5 cases/100 000/year worldwide. One-third of patients die of the disease. 1  MCPyV DNA integration into the host genome and the viral oncoprotein large T/small T (LT/sT) expression, are key MCCP events. 2 High UV-induced mutational burden is an MCCN hallmark, with RB1 and p53 mutation and MYCL amplification occurring frequently. 1 MCCN is more aggressive than MCCP, in terms of increased development, relapse, and death risks. Despite progresses in the field, both MCC subsets appear to be morphologically similar, and information on distinctive diagnostic/prognostic markers is poor. 3 MCC management includes either platinum-/etoposide-based regimens or surgery followed by adjuvant radio-/chemotherapies. However, their effectiveness is unsatisfactory due to the propensity of MCC to metastasize and develop resistance. 3 Identifying distinctive molecular characteristics for the two MCC subsets could improve diagnostic and therapeutic options.  4 Recent evidence points to a multilinear origin, given the expression of neuroendocrine, epithelial/fibroblast, and B-lymphoid markers. 5 Identifying the MCC origin cell is prevented by the limited molecular information on the two aetiologies. 6 Several retinoic pathway-associated genes are dysregulated in MCC. 5,7 Retinoid responses are mediated by different classes of nuclear receptors, whose signaling activation modulates genes that control morphogenesis, development, and homeostasis. 5,7 Retinoic signaling is dysregulated in carcinoma and components of this pathway are reliable antitumor targets, even for managing cutaneous neoplasms. 8 The impairment of development regulators such as SOX2 and the Hedgehog gene family member SHH has been documented in MCC. 5,7 In particular, SOX2, regulates the embryonic neural crest stem cell maintenance of the Merkel cell lineage, while being reported as expressed in MCC. 9 Similarly, SHH controls stem cell development and neurogenesis. The expression of this gene has been reported in MCC and in association with MCPyV infection. 10 Limited but promising information on retinoic-associated genes in MCC prompted us to explore the heterogeneous nature of MCC by investigating the retinoic gene signature in MCCP and MCCN cells.     were predominantly enriched in gastric cancer, stem cell pluripotency, basal cell carcinoma, retinol metabolism, and chemical carcinogenesis ( Figure 2C). The majority of hub genes resulted to be highly distributed across all identified GO/KEGG pathways (Supporting Information: Table 1). Regarding the remaining hub genes, MYC overexpression is frequent in MCPyV-independent MCC. 3 Despite no information is available on FGF8 and OLIG2 in MCC, they are dysregulated in cancer. 15,16 Both genes can be considered as novel candidate MCCPrelated genes.

| DISCUSSION
The identification of hub MCCP/MCCN differentiating genes might allow the development of novel MCC subset-specific antitumor strategies. Indeed, SOX2, Hedgehog genes, PAX6, OLIG2, ASCL1, and JAG1, are all reliable targets for cancer therapy. 17,18 These genes can, therefore, be considered novel candidate targets for MCCP treatment, and we thus encourage further studies in this direction.
Enrichment analyses were next performed on DEGs to identify critical discriminating pathways. GO analysis indicated that DEGs predominantly consisted of genes involved in embryonic development, which is known to be dysregulated in cancer. 19

| Statistics
Gene expression analysis was performed using the ∆∆ 2 C t method.
Log2 fold change <−2/>2 and p < 0.05 were considered statistically significant. Hierarchical clustering and PCA were performed using ClustVis. GO/KEGG were performed to categorize DEGs. PPI network was constructed using STRING (cut-off > 0.4). Ten PPI hub genes were identified via the cytoHubba plugin in Cytoscape, with maximum neighborhood/edge percolated components and degree algorithms.

CONFLICT OF INTEREST STATEMENT
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.