The Biology and Treatment of Merkel Cell Carcinoma: Current Understanding and Research Priorities

Abstract

Merkel cell carcinoma (MCC) is a rare and aggressive skin cancer associated with advanced age and immunosuppression. Over the last decade, it has been discovered that MCC tumors are associated with either integrated Merkel cell polyomavirus, which likely drives tumorigenesis, or a high burden of UV-induced somatic mutations. Both virus-positive and virus-negative MCC tumors are immunogenic and PD-(L)1 checkpoint blockade has proven to be highly effective in treating most patients with metastatic MCC; however, about 50% of treated patients do not experience long term tumor control. Despite these rapid advances in the understanding and management of MCC, many basic, translational, and clinical research questions remain. In March 2018, an International Workshop on Merkel Cell Carcinoma Research was held at the National Cancer Institute where academic, government, and industry thought leaders met to identify high priority research questions. Here we review the biology and treatment of MCC and report the conclusions of the Workshop.

I. Introduction

Our understanding of the biology and management of Merkel cell carcinoma (MCC) has advanced exponentially since its description in 1972. MCC, or primary cutaneous neuroendocrine carcinoma (Figure 1), is named for its ultrastructural and immunophenotypic resemblance to sensory Merkel cells found in the skin. MCC is frequently metastatic and has an estimated 33–46% disease-specific mortality [1, 2]. In most of the world, the majority of MCC tumors is caused by Merkel cell polyomavirus (MCPyV) [3], and the remainder is associated with UV-signature mutations [4, 5]. Therapeutic options have historically been limited for patients with advanced disease; however, new immunotherapeutic approaches are associated with durable responses in a subset of patients [6]. Here, we review the current state of knowledge of MCC biology and treatment, and define key outstanding questions in areas of basic, translational, and clinical MCC research.

Figure 1.

(A) Clinical appearance of MCC, presenting as a rapidly growing nodule on the extremity or head and neck. (B) Microscopic appearance of MCC, demonstrating round cells with scant cytoplasm, neuroendocrine chromatin, and numerous mitotic figures. Trabecular patterning may be prominent. (Hematoxylin and eosin, 400x.)

II. Biological Features of MCC

II.1. Merkel Cell Polyomavirus

MCPyV is a common virus that is the causative agent in most MCC tumors. MCPyV was first identified by digital transcriptome subtraction of MCC tumors [3]. The Polyomaviridae family of small double-stranded DNA viruses, to which MCPyV belongs, includes other human polyomaviruses associated with cutaneous infection (Trichodysplasia spinulosa polyomavirus, human polyomavirus 6, and human polyomavirus 7) or disease of other organ systems (JC polyomavirus, BK polyomavirus, WU polyomavirus, and KI polyomavirus) [7]. To date, MCPyV is the only known human oncovirus in the polyomavirus family; it is currently unknown why MCPyV holds this distinctive status.

The prevalence of subclinical MCPyV infection increases with age, with a seroprevalence in adults of approximately 60–80% [8–16]. The skin appears to be a major site of viral infection, although MCPyV has also been detected in peripheral blood and a range of other organ systems [8, 17–24]. MCPyV infection appears to be asymptomatic [9].

The host cell type for MCPyV infection has remained elusive. Benign Merkel cells are not sufficiently numerous to account for the MCPyV burden detected in skin [25]. A reservoir in peripheral blood monocytes has been proposed [22]. Although MCPyV reporter pseudovirus can enter many cell types including keratinocytes [26, 27], dermal fibroblasts and HEK 293 cells are the only cells where productive viral infection has been demonstrated in vitro [26, 28].

The viral life cycle of MCPyV is similar to other polyomaviruses. The episomal viral genome possesses an early region (ER) and late region (LR), which contain genes encoding proteins that coordinate viral replication and viral capsid proteins, respectively (Figure 2A) [29]. The ER of MCPyV encodes large T antigen (LT), small T antigen (ST), and 57kT antigen transcripts (Figure 2B). The middle T-like overprinting gene ALTO is also proposed to originate from the ER [30–32]. The MCPyV LR encodes VP1 and VP2 capsid proteins [29, 33], and may allow MCV-miR-M1–5p expression [31, 32, 34]. Productive viral infection is associated with cell death rather than oncogenic transformation [28]. MCPyV replication is regulated by E3 ligase targeting of phosphorylated LT as well as feedback inhibition of LT on its own promoter [28] and inhibition of LT production by viral miRNA [34]. These features act together to inhibit virus replication after entry into a host cell so that the virus typically defaults to a latent, non-replicative state after infection [28, 32].

Figure 2.

(A) Merkel cell polyomavirus (MCPyV) is a small double-stranded DNA virus with a 5387 base pair genome that includes a non-coding control region (NCCR), early region coding T-antigen transcripts that coordinate viral replication, and late region coding VP transcripts for virion capsid proteins. (B) Multiple transcripts are generated from the early region by alternative splicing and possibly alternative start sites, including large T antigen (LT), small T antigen (ST), 57k T antigen (57kT), alternative frame of the Large T open reading frame (ALTO), and microRNA MCV-miR-M1. (C) Cellular functions of MCPyV LT. (D) Cellular functions of MCPyV ST. CR1: Conserved region 1. MUR: MCPyV Unique Region. ZNF: Zinc finger. LZ: leucine zipper. LSD: Large T stabilizing domain. NLS: nuclear localization signal. OBD: origin binding domain. ZN: Zinc finger domain. LZ: leucine zipper domain. Rb: retinoblastoma. LSD: Large T stabilization domain.

Oncogenic transformation by MCPyV is hypothesized to require two events: integration of the viral genome into the host genome, and truncation of LT to render the viral genome replication deficient (Figure 3) [3, 35]. Viral integration into the host genome may occur by accidental genome fragmentation during MCPyV replication; the integration site appears relatively random, without consistent involvement of cellular tumor suppressor genes or oncogenes [3, 36–40]. In VP-MCC tumors, mutations in LT disrupt the DNA binding domain and the helicase domain distal to the RB binding motif (Figure 2C). The resulting truncated LT retains its ability to bind the RB protein and promote cell cycle progression [35], but cannot mediate virus replication [41, 42]. The integrated and mutated virome no longer produces infectious virus. The very low probability of this required combination of events may explain why MCC is rare despite the ubiquity of MCPyV infection.

Figure 3.

Proposed tumorigenesis pathways for MCC in the presence or absence of MCPyV. TSG: tumor suppressor gene. LT: large T antigen. VN-MCC: virus-negative MCC. VP-MCC: virus-positive MCC.

LT and ST exhibit diverse activities that may contribute to oncogenesis. There is evidence that cultured VP-MCC cells are dependent upon both viral T antigens for proliferation and survival [43–45]. Similar to SV40 Large T antigen, MCPyV LT directly binds and inactivates the RB protein (Figure 2C). This has been shown to enhance expression of the survivin oncoprotein, which may be an important therapeutic target in VP-MCC tumors [46]. Unlike the SV40 polyomavirus large T antigen, MCPyV LT has not been shown to directly bind p53 [47]. Additional functions of LT, which are likely lost after truncation, include DNA binding, bromodomain protein-4 (BRD4) binding, helicase activity, and cell growth inhibition [29]. MCC tumor-associated mutations in MCPyV also likely disrupt the ALTO gene [30].

Multiple lines of evidence suggest a critical role for ST in VP-MCC oncogenesis. ST expression is sufficient to transform rat-1 fibroblasts in culture [48]. Transgenic mouse models indicate that ST expression is transformative in various organ systems, including the epidermis [49–52]. In mouse models, MCPyV T antigens alone are not sufficient to induce neuroendocrine tumors [50]; however, co-expression of ST with the transcription factor ATOH1 generates intraepidermal MCC-like proliferations [52] and, with the loss of p53, undifferentiated anaplastic tumors [49].

The cellular activities of ST are diverse (Figure 2D). Similar to other small T antigens, the MCPyV ST has a PP2A region that binds and inhibits various components of the serine/threonine phosphatase complexes PP2A and PP4 [53]. ST interaction with PP4 may mediate inhibition of the NF-κB pathway activity, as well as exert pro-motility effects on the actin and microtubule cytoskeleton [54–57]. The PP2A binding domain is dispensable for cell transformation by MCPyV ST [29], indicating that other domains of ST are crucial for oncogenesis.

MCPyV ST possesses a distinct domain known as the LT stabilizing domain (LSD) (Figure 2D) that has been proposed to inhibit E3 ubiquitin ligases including FBXW7 and CDC20 [58, 59]. Co-expression of ST and LT leads to stabilization of LT, which may be mediated via inhibition of FBXW7 by ST to prevent degradation of LT (although neither protein encodes the classic high-affinity phosphodegrons for FBXW7 binding). Other oncogenic proteins including Myc may be stabilized by ST by a similar mechanism [58]. Interaction of the LSD domain with CDC20 has been reported to promote cap-dependent translation via phosphorylation of the eukaryotic translation initiation factor 4E-binding protein 1 (4E-BP1) [59]. The LSD domain is required for oncogenic activity by ST in vitro and in vivo, supporting a critical role for this domain in MCPyV oncogenesis [50, 52, 58].

ST has been found to bind L-Myc to regulate the EP400 histone acetyltransferase and chromatin remodeling complex (Figure 2D) [60]. Additional activities described for ST include increased aerobic glycolysis, possibly via MYC and NF-κΒ pathways [61].

Detection of Tumor-Associated MCPyV

Methods for detection of MCPyV in tumors include immunohistochemistry (IHC), PCR, RNA or DNA in situ hybridization, and next generation sequencing [37, 62–66]. These assays vary significantly in sensitivity and specificity for detection of tumor-associated MCPyV.

IHC for expression of T antigen proteins is a common approach for MCPyV detection. The most broadly used antibody clone for detecting MCPyV LT, CM2B4, is commercially available, and has approximately 88% sensitivity and 94% specificity (compared to multimodal approaches combining PCR and IHC) [67]. Nonspecific staining by CM2B4 may occur in tonsillar and lymphoid tissues [67, 68]. Other antibodies recognizing LT, ST, or the common T antigen region have been reported, and may have additional utility in detecting the viral proteins [67].

PCR for MCPyV is another common detection method. The 5’ end of the 2^nd exon of LT is frequently targeted [64]. Sensitivity is improved by additional PCR reactions targeting other amplicons in the early region, including ST [64]. Quantitative PCR allows for estimation of the number of integrated MCPyV copies per host cell genome, relative to the reference MCC cell line MKL-2. Copy number estimation by this method may range from extremely low (<1 MCPyV copy per 100 cells) to thousands of copies per cell [64]. Detection of extremely low copy number is incompletely understood, but may be due to technical factors including inefficient PCR amplification caused by mutations in the integrated MCPyV genome, low tumor purity, or detection of infectious wild type MCPyV in background skin [69]. In cases with low tumor purity, copy number estimates of tumor-associated MCPyV may overlap with levels observed due to background signals in skin and non-MCC tumors [63, 69]. Because qPCR does not allow for visual confirmation that positive results are associated with tumor cells, background infection cannot be excluded in MCC tumors with low signal [63]. A multimodal approach incorporating PCR and IHC results may be the most sensitive and specific method for confirming MCPyV status by commonly used assays [67].

Other newer approaches for MCPyV detection have been less extensively investigated, but may have advantages over immunohistochemistry and PCR. RNA in situ hybridization may have similar sensitivity to PCR, and allows for visual correlation with tissue morphology to exclude background infection [63]. Next generation sequencing can be effective in detecting MCPyV sequences, including tumor-specific truncating mutations. Next generation sequencing by a hybrid capture approach can further demonstrate viral integration sites [37] and therefore provides the greatest specificity for confirming the presence of tumor MCPyV. However, currently the time, expense, and expertise required for hybrid capture next-generation sequencing make this approach impractical for many diagnostic and research laboratories.

II.2. Non-Viral Changes in MCC

Similar to other malignancies, MCC tumors display genomic aberrations including chromosomal copy number changes and mutations. However, there are significant differences between VP-MCC and virus-negative (VN)-MCC tumors in the patterns of these changes. Standard terminology for designating the viral status of MCC tumors has not been established. We propose that “VP-MCC” and “VN-MCC” for virus-positive and virus-negative MCC, respectively, represent useful standard abbreviations for this purpose.

Most MCC tumors harbor chromosomal copy number variations (CNV), regardless of viral status [70]. VP-MCC tumors display fewer CNVs compared to VN-MCC tumors [70]. MCC tumors are heterogeneous with regard to patterns of chromosomal changes [70, 71]. However, certain alterations are recurrent in a minority of tumors. Gains of chromosome 1p34, including the MYCL oncogene, may occur in approximately 39% of tumors, and may be observed in both VP-MCC and VN-MCC tumors [70]. Deletions affecting RB1 are also frequent, and also occur in both VP-MCC and VN-MCC [70, 72, 73].

VP-MCC and VN-MCC have distinct somatic mutation patterns. VN-MCC displays a high mutational burden, UV mutational signature, and highly recurrent inactivation of tumor suppressors, including TP53, RB1, and NOTCH family genes (Figure 3) [5, 74–76]. In contrast, VP-MCC tumors tend to display low mutational burden, no definitive mutational signature, and absence of TP53 and RB1 mutations [5, 74, 76]. Mutations in VP-MCC are predominantly subclonal, suggesting that the majority of these mutations are late events rather than founder events [71], possibly arising due to transcription-coupled damage or in the setting of double-stranded DNA breaks [36]. In contrast, TP53 and RB1 mutations in VN-MCC are clonal, and shared by primary tumors and matched metastases [71]. Hotspot activating mutations of oncogenes are observed in both VP-MCC and VN-MCC, including HRAS, KRAS, and PIK3CA [4, 5, 76–79]. Other hotspot activating mutations have only been described in VN-MCC tumors, including mutations of KNSTRN, RAC1, AKT1, and EZH2 [74, 76, 77]. Most VN-MCC tumors demonstrate activation of at least one known proto-oncogene. Functional studies of the oncogenic pathways in VN-MCC have been confounded by lack of transgenic mouse models [80] and the atypical biology of several MCPyV-negative cell lines [81, 82].

Epigenetic deregulation can also contribute to tumor aggressiveness. Epigenetic deregulation may include DNA modifications (especially promoter silencing by CpG island hypermethylation) and histone modifications. Promoter hypermethylation in MCC affects genes including DUSP2, CDKN2A, and the RASSF family [83–86]. The Polycomb group complex, which includes the oncogene EZH2 and mediates gene silencing via histone H3 lysine 27 trimethylation, may also be deregulated in MCC [77, 87, 88]. Epigenetic silencing of HLA genes by histone deacetylases may contribute to immune evasion in MCC [89–91].

II.3. MCC Cell of Origin

Identifying the cell type from which a neoplasm arises has implications for tumor initiation mechanisms, experimental modeling, and possibly therapeutic susceptibilities. MCC is a poorly differentiated neuroendocrine carcinoma that lacks a recognized benign or dysplastic precursor. In addition, MCC tumors are often found in the dermis but can arise in any layer of the skin (from intraepidermal to subcutaneous)[2]. Thus, fundamental clues for the MCC cell of origin are lacking. Although MCC cells have immunophenotypic and ultrastructural similarities to benign Merkel cells, Merkel cells are post-mitotic in vivo, and the anatomic regions with the highest Merkel cell density are not the most frequent sites of MCC [2, 29]. Based on these observations, most investigators do not consider mature Merkel cells to be a likely candidate for the MCC cell of origin [7]. Instead, MCCs are likely derived from a cell population that recapitulates the Merkel cell differentiation pathway either before or during neoplastic transformation. In addition, given that VN-MCC displays UV-signature mutations while VP-MCC does not, these tumor types may arise from distinct cells of origin with different levels of baseline photodamage [80].

Multiple candidates for the MCC cell of origin have been proposed. A portion of MCCs express B-cell and lymphoid markers and harbor immunoglobulin rearrangements [92], raising consideration for an early B-cell origin [93]. Although fibroblasts can support productive MCPyV infection [26], MCC lacks similarity to classic fibroblastic tumors. Other dermal populations may represent more probable candidates for the MCC cell of origin. Dermal mesenchymal stem cells have remarkable plasticity, with potential for differentiation into lineages including neurons and keratinocyte-like cells [94]. Recently, transgenic mouse models demonstrating oncogenic capability of ST within epidermis [50–52] have provided support for an epidermal origin. Mouse models that more fully recapitulate MCC tumors, possibly combined with lineage tracing experiments, may provide further evidence for the MCC cell of origin.

Other virus-associated malignancies may be informative when considering candidates for the MCC cell of origin. In cervical squamous cell carcinoma, tumor phenotype and precursor lesions support human papilloma virus (HPV)-infected epithelial stem cells as the cells of origin [95]. For Kaposi sarcoma herpesvirus (KSHV)-associated primary effusion lymphoma, recent observations have raised the possibility that mesothelial cells may represent the cell of origin via mesenchymal-to-lymphoid transformation [96]. Therefore, tumor phenotype may not be a reliable predictor of the cell of origin for virus-driven malignancies.

III. Clinical Features of MCC

III.1. Epidemiology

As of 2013, the annual United States incidence of MCC was 0.7 per 100,000 [97]. MCC incidence almost doubled in the United States from 2000–2013 and is expected to exceed 3,000 cases per year by 2025 [97], with similar increases in many, but not all, European countries [98]. The basis for this increasing incidence is unclear, but may be related to an aging population and increased diagnostic recognition [97]. The frequency of MCC is higher closer to the equator, and much lower in non-Caucasian populations [98, 99], suggesting an association with UV radiation. Australia has the highest incidence of MCC [99], notably associated with high environmental UV exposure and tumors that are more frequently virus-negative [100–102]. MCC displays higher incidence in immunosuppressed populations [103]. Other than immunosuppressed individuals, MCC arises almost exclusively in individuals of advanced age. Unlike other cancers where disease incidence peaks then declines with age, MCC incidence continues to increase as patients enter the 8^th and 9^th decades of life, possibly due to immune senescence [97]. Thus, MCC affects populations that frequently have significant comorbidities that may complicate management.

III.2. Clinical Presentation and Pathologic Diagnosis

The classic clinical presentation of MCC is a rapidly growing red or violaceous nodule on sun-exposed skin of an elderly, fair-skinned individual. In a fraction of cases, MCC presents in a lymph node without an identifiable cutaneous tumor, presumably reflecting metastatic disease with regression of the primary skin tumor [2, 7].

Confirming the diagnosis of MCC requires evaluation of histopathologic and immunohistochemical findings. The histopathologic differential diagnosis may include lymphoma, small cell melanoma, metastatic small cell lung cancer to the skin, and other small round cell neoplasms involving the skin [2]. MCC may also be misdiagnosed as basal cell carcinoma [104]. Immunohistochemical stains are necessary for diagnostic confirmation. MCCs express neuroendocrine markers such as chromogranin A and/or synaptophysin, although these are not specific markers. Cytokeratin 20 (CK20) is expressed focally or diffusely in most MCC tumors. Keratin staining in MCC displays a paranuclear dot-like pattern, cytoplasmic/membranous pattern, or both. Neurofilament, another intermediate filament, can also show a paranuclear dot pattern. CK20 expression and the paranuclear-dot pattern of intermediate filament staining are relatively specific for MCC [2]. Synoptic reporting of newly diagnosed lesions facilitates treatment decisions and prognostic studies, and at a minimum should include tumor size (cm), peripheral and deep margin status, lymphovascular invasion, and extracutaneous extension (bone, muscle, fascia, cartilage) [105].

MCC is morphologically indistinguishable from metastatic small cell lung carcinoma. Immunohistochemical stains useful for this distinction include CK20, TTF1, MCPyV LT, and neurofilament proteins. The MCC immunophenotype is CK20⁺, LT^+/−, neurofilament⁺, and TTF1⁻. Metastatic SCLC to the skin is CK20⁻, LT⁻, neurofilament⁻, and TTF1⁺ [2]. However, expression patterns vary, and no single marker is sufficiently sensitive or specific for distinction of MCC from metastatic small cell lung carcinoma. Thus, a panel of markers is necessary for challenging cases such as CK20-negative MCC. In addition, certain non-lung small cell carcinomas, such as parotid and cervical primaries, also frequently express CK20 and may be therefore more challenging to distinguish from MCC, especially in the setting of metastatic MCC from an unknown primary site [2]. As next generation sequencing analyses enter more widespread use, MCPyV detection combined with mutational signature analysis may be useful for distinguishing MCC (MCPyV-positive or UV-signature mutations) from metastatic SCLC (MCPyV-negative and tobacco signature mutations) [77].

MCC has the potential to develop distant cutaneous metastases. In the rare patient who presents with a second cutaneous MCC tumor that is spatially and temporally separated from the initial primary MCC, such that the tumor is clinically designated a second primary, clonality studies based on copy number alterations, mutations, or MCPyV sequencing may be useful in distinguishing metastatic disease from a second primary MCC [71, 106].

III.3. Prognostic Findings

The most significant prognostic parameters for MCC include tumor size and the presence of locoregional or distant metastases. These factors form the basis of the American Joint Committee on Cancer staging system for MCC [107, 108]. Although increasing primary tumor size correlates with increased risk of metastatic disease, MCC tumors of any size have significant risk of occult metastasis, supporting the use of sentinel lymph node biopsy for all cases [109]. Additional features of the primary tumor such as lymphovascular invasion and tumor growth pattern may also have prognostic significance [110]. Clinically detectable nodal disease is associated with worse outcome than microscopic metastases [107]. Other findings associated with a worse prognosis include sheet-like involvement in lymph node metastases, and an increasing number of lymph nodes with metastatic involvement [108, 111]

Due to the difficulties of predicting aggressive disease based on clinical and morphologic findings alone, prognostic biomarkers for MCC have been an area of intense investigation. Studies of the prognostic role of MCPyV status have had mixed results, predominantly finding either worsened prognosis in VN-MCC tumors or no difference [112]. Recently, the largest study to date to incorporate immunohistochemical and PCR evaluation of MCPyV status found improved prognosis associated with VP-MCC [67]. However, both VP-MCC and VN-MCC may display aggressive and fatal courses.

For patients with VP-MCC, MCPyV serology may be informative for prognosis and guiding management in MCC. Higher anti-VP1 antibody titers and presence of anti-ST serum antibodies at diagnosis have been associated with better outcome [113, 114]. Persistent or re-emergent anti-ST serum antibodies have been associated with poor prognosis and recurrence [115].

Interestingly, patients presenting with nodal or presumably metastatic MCC, where no skin lesion can be identified have better outcomes than same-stage patients with known primary tumor [107, 116, 117]. An anti-tumor immune response has been proposed to underlie both the primary tumor regression and improved patient outcomes in such cases. Analogously, multiple studies have confirmed the importance of immune competence as a determinant of MCC prognosis. Immunosuppressed patients (including those with chronic lymphocytic leukemia) not only have a higher incidence of MCC, but also display markedly worse survivals [103, 118].

Immune markers have been extensively investigated in MCC. Patterns of immune infiltrates may reflect both MCPyV status and disease outcomes. Relative to VN-MCC, VP-MCC are more likely to be associated with evidence of a brisk inflammatory response and increased number of CD8+ T-cells [119–121]. Immunologic findings that have been correlated with improved prognosis in MCC include increased CD8+ T-cells (either tumor-infiltrating or at the tumor periphery), tumor-infiltrating MCPyV-specific T-cells, and tumor PD-L1 expression [89, 120–124].

Additional markers proposed to have prognostic significance in MCC include detection of p63, EZH2, survivin (nuclear pattern), CD34 (vascular density), vascular endothelial growth factor, vascular E-selectin, Sonic Hedgehog, phospho-STAT5B, CADM1, and MAL1, among others [2]. TP53 mutations has also been shown to have prognostic significance [5], although it is unknown whether this effect is independent of MCPyV status.

III.4. Management of MCC

Excision with 1–2 cm margins and radiotherapy are the mainstays of management for primary MCC tumors. Adjuvant radiotherapy to the primary tumor site is often recommended; however, the morbidity of radiation may be avoided with low local recurrence rates in a subset of patients (i.e. tumors smaller than 2 cm in size without other adverse prognostic factors) [125].

Due to the risk of occult nodal disease, sentinel lymph node biopsy is recommended by the National Comprehensive Cancer Network for patients without clinically detectable metastatic disease [6]. Any size of metastatic deposit is currently considered positive with regard to N staging; therefore, immunohistochemistry for broad-spectrum cytokeratin and/or CK20 is routinely used to improve detection of micrometastases in sentinel lymph nodes [2, 126]. The National Comprehensive Cancer Network recommends management of clinically detectable or occult nodal disease with imaging for distant metastases, followed by lymph node dissection and/or radiotherapy to the nodal basin [105].

Systemic therapies have traditionally included chemotherapy such as platinum-based therapies, taxanes, anthracyclines, and etoposide. Due to lack of durable response and no impact on survival, chemotherapy is currently considered to have a palliative role [127].

Recently, the anti-PD-L1 antibody avelumab was approved as therapy for metastatic MCC by the United States Food and Drug Administration, the European Medicines Agency, Swissmedic, and the Japanese Ministry of Health, Labor and Welfare [128]. Other immunotherapies for MCC have also demonstrated effectiveness in clinical trials. The rationale and scope of immunotherapy for MCC is discussed in greater detail below. The success of immune-based therapies represents a milestone in the management of advanced MCC. However, not all patients experience durable response to immune-based therapies. Furthermore, patients requiring immunosuppression in the setting of solid organ transplants or those with autoimmune disease may not be optimal candidates for immune-based therapy. Therefore, predicting and improving response to immunotherapy, and identifying alternative therapies for cases in which immunotherapy is contraindicated or ineffective, represent current research priorities in MCC.

IV. Investigational Therapies in MCC

Several observations predicted that immune-based therapy would be effective in MCC. Spontaneous regression of MCC is an uncommon but well-documented event presumably due to immune activation. T-cell infiltrated MCC tumors with lymphocyte infiltrates demonstrate better outcomes than tumors lacking infiltrating lymphocytes. The abscopal effect of irradiation has also been observed in MCC, suggestive of a modulated immune response [129]. MCC tumors have a higher incidence and worse prognosis in the setting of immune compromise. Finally, tumors express either MCPyV antigens or UV-mutation associated neoantigens, providing putative targets for antitumor immunity (Figure 3) [7].

Both the innate and adaptive immune systems can target tumors. Adaptive antitumor immunity is mediated by antigen-presenting cells activating effector T-cells and is modulated by stimulatory signaling (including OX40/OX40L) and inhibitory signals (CTLA4, PD-1/PD-L1)(Figure 4) [130]. Of these, the PD-1/PD-L1 immune checkpoint has been intensely investigated in MCC. Several open-label phase II clinical trials of immune checkpoint inhibitors have demonstrated durable responses in a subset of MCC patients. In 88 patients with metastatic MCC who had previously received chemotherapy, the anti-PD-L1 antibody avelumab showed a 32.8% objective response rate (ORR), including 9.1% complete response (CR), with durable response [131] and associated improvement in quality of life [132]. In a study with 26 patients, the anti-PD-1 antibody pembrolizumab demonstrated 56% ORR, including 15.4% CR, as first-line systemic therapy [133]. The anti-PD-1 antibody nivolumab has also shown promise [7, 134, 135]. In all of these studies, tumors responded to immune checkpoint inhibitors regardless of viral status, although cohort sizes were too small for rigorous comparison between VP-MCC and VN-MCC. On the basis of clinical trial results, avelumab recently became the first and, at this time, the only therapy approved by the US Food and Drug Administration for metastatic MCC [128]. Investigation of avelumab and nivolumab as adjuvant therapy for high-risk disease is ongoing.

Figure 4.

Signaling pathways that modulate antitumor immunity and represent potential therapeutic targets.

Several immune-based therapies that act through mechanisms other than PD-1/PD-L1 inhibition are under investigation for MCC (Table 1). Therapeutic combinations including anti-CTLA4 checkpoint inhibitors are an area of active investigation [6]. Utomilumab, which may stimulate anti-tumor immunity by acting as an agonist of the CD137/4–1BB receptor, has been shown in a phase I trial to be well-tolerated, with preliminary evidence for anti-tumor activity [136]. Adoptive cell transfer using in vitro expanded autologous antitumor T-cells in combination with interferon and radiotherapy was associated with tumor regression in one case [137]. A phase I/II trial combining adoptive cell transfer with interleukin-2 therapy and checkpoint inhibition is underway (Table 1) [6, 138]. In animal models, DNA vaccines were effective against B16 murine melanoma lines expressing LT or ST [139, 140]. The cutaneous location of primary and satellite/in transit MCC renders these tumors amenable to injections. Intratumoral administration of interleukin-12 (IL-12) cDNA via electroporation was investigated in a phase II trial of MCC (n=15) with some objective responses Finally, phase II trials are recruiting patients to determine whether intratumoral injection of modified herpesvirus (Talimogene Laherparepvec) is effective in tumor oncolysis and possibly stimulation of antitumor immunity [138].

Table 1.

Investigational agents in MCC.

Intervention	Proposed Mechanism	Clinical Trial Phase(s)	Clinical Trial Number(s)
PD-1/PD-L1 checkpoint inhibitors
Nivolumab	Anti-PD-1 antibody	Multiple phases 1 and 2 (Recruiting)	NCT02196961, NCT03071406, NCT03071757
Pembrolizumab	Anti-PD-1 antibody	Phase 2 (Active)	NCT02267603, NCT03304639
Avelumab	Anti-PD-L1 antibody	Expanded access, multiple phases 1–3 recruiting	NCT00655655, NCT02155647, NCT02584829, NCT03167164, NCT03271372
ABBV-181	Anti-PD1 antibody	Phase 1 (Recruiting)	NCT03000257
Durvalumab	Anti-PD-L1 antibody	Phase 1/2 (Recruiting)	NCT02643303
CK-301	Anti-PD-L1 antibody	Phase 1 (Recruiting)	NCT03212404
CTLA4 inhibitors
Ipilimumab	Anti-CTLA4 antibody	Phase 2 (Recruiting)	NCT02196961
Tremelimumab	Anti-CTLA4 antibody	Phase 1/2 (Recruiting)	NCT02643303
Other immune-based therapies
INCAGN01876	Anti-GITR agonistic monoclonal antibody	Multiple phase 1/2 (Recruiting)	NCT03277352, NCT03126110
INCAGN01949	Anti-OX40 agonist antibody	Phase 1/2 (Recruiting)	NCT03241173
TTI-621	SIRPa-Fc fusion protein	Phase 1 (Recruiting)	NCT02890368
Adoptive immunotherapy	MCPyV TAg-specific polyclonal autologous CD8-positive T cells	Phase 1/2 (Recruiting)	NCT02584829
Interferon Beta		Phase 1/2 (Recruiting)	NCT02584829
Poly ICLC	Stimulation of cytotoxic cytokines	Phase 1/2 (Recruiting)	NCT02643303
ALT-803	IL-15 agonist	Phase 1/2 (Not yet recruiting)	NCT03167164
GI-6301	Brachyury-expressing yeast vaccine	Phase 1/2 (Not yet recruiting)	NCT03167164
NK cell transfer	NK cell transfer	Multiple (phase 1/2 and 2)	NCT02465957, NCT03167164
GLA-SE	Toll-like receptor 4 agonist	Phase 1 (Completed)	NCT02035657
TavokinogeneTelseplasmid	IL-12 plasmid (CD8+ T -cell and NK cell stimulation)	Phase 2 (Completed)	NCT01440816
F16IL2	F16-IL2 fusion protein for NK cell stimulation	Phase 2 (Unknown status)	NCT02054884
Talimogene Laherparepvec (TVEC)	Oncolytic HSV-1 encoding GM-CSF	Multiple phase 2 (Recruiting)	NCT02819843
ABBV-368	Immunotherapy (not further specified)	Phase 1 (Recruiting)	NCT03071757
ETBX-051	Vaccine (not further specified)	Phase 1/2 (Not yet recruiting)	NCT03167164
ETBX-061	Vaccine (not further specified)	Phase 1/2 (Not yet recruiting)	NCT03167164
mTOR inhibitors
Temsirolimus	mTOR inhibitor	Phase 1 (Completed)	NCT01155258
MLN0128	Dual inhibitor of raptor-mTOR and rictor-mTOR	Phase 1/2 (Recruiting)	NCT02514824
Everolimus	mTOR inhibitor	Multiple phase 1 (Active and completed)	NCT00655655, NCT01204476
Somatostatin analogs
Octreotide Acetate	Somatostatin analog	Phase 1 (Completed)	NCT01204476
Pasireotide	Somatostatin analog	Phase 1 (Completed)	NCT01652547
Lanreotide	Somatostatin analog	Phase 2 (Completed)	NCT02351128
PEN-221	Somatostatin analog	Phase 1/2 (Recruiting)	NCT02936323
Bevacizumab	Anti-VEGF	Phase 1/2 (Not yet recruiting)	NCT03167164
Receptor tyrosine kinase inhibitors
Imatinib mesylate	KIT inhibitor	Phase 2 (Completed)	NCT00068783
Cixutumumab	Anti-IGF1R antibody	Phase 1 (Completed)	NCT01204476
Cabozantinib	Inhibits multiple receptor tyrosine kinases	Phase 2 (Active)	NCT02036476
Other targeted therapies
Oblimersen sodium	BCL2 antisense oligonucleotide	Phase 2 (Completed)	NCT00079131
BB-10901	Anti-CD56 antibody conjugated to microtubule assembly inhibitor	Phase 1 (Completed)	NCT00346385
Conventional chemotherapeutics
Cyclophosphamide, fluorouracil,methotrexate		Phase 2 (Completed)	NCT00003549
Irinotecan hydrochloride		Phase 2 (Completed)	NCT00004922
Capecitabine, temozolomide		Phase 2 (Completed)	NCT00869050
Carboplatin, etoposide		Phase 2 (Active)	NCT01013779

Combinations of immune-based therapies with more traditional treatments may also be effective for MCC. Downregulation of MHC-1 expression in MCC can be reversed by radiotherapy, interferon, and chemotherapies, potentially increasing tumor sensitivity to immunotherapy approaches [90, 137]. The histone deacetylase inhibitor vorinostat in combination with mithramycin A chemotherapy can also upregulate MHC-1 expression in MCC cell lines and xenografts [91]. Radiotherapy also has the potential to induce release of tumor antigens and increase inflammation, further synergizing with immunotherapy [6].

Alternatives to immune-based therapies are needed in patients with advanced MCC who are immunosuppressed or whose disease does not respond to immunotherapy. Several types of targeted inhibitors have been investigated in MCC cell lines and xenograft models. Targeting of the antiapoptotic protein survivin by the small molecule inhibitor YM155 was found to potently initiate cell death in VP-MCC cell lines and xenografts [46, 141]. A small molecule inhibitor of the BCL2 family, ABT-263, induced apoptosis in the majority of MCC cell lines tested, regardless of MCPyV status [142] and was effective in xenograft studies [143]. A small phase II trial (n=12) of the Bcl-2 antisense agent G3139 failed to show benefit [144]. MCC tumors express somatostatin receptors and somatostatin analogues are being tested for the molecular imaging and treatment of MCC [6, 145, 146]. Although Hedgehog signaling plays a role in the development of normal Merkel cells [147–149], MCC cell lines do not display sensitivity to Hedgehog inhibitors [150].

In line with the observation of recurrent activation of the PI3K pathway in MCC, preclinical and limited clinical data support efficacy of PI3K/mTOR pathway inhibition in MCC. MCC cell lines are sensitive to inhibition of PI3K and mTOR [78, 79, 151, 152]. A complete response to the PI3Kδ inhibitor, idelalisib, was reported in one clinical case, although the response may have been confounded by concurrent use of radiotherapy [153]. Several clinical trials are addressing mTOR inhibition in MCC (Table 1) [138]. In VP-MCC with wild type p53, MDM2 inhibitors may be useful [154].

In general, activating tyrosine kinase mutations are lacking in MCC, and thus far there has been little evidence that tyrosine kinase inhibition is an effective approach for MCC therapy. KIT is expressed in MCC, but activating mutations have not been reported. Although a case of MCC remission with imatinib has been reported [155], a phase II trial failed to show benefit [156]. The multikinase inhibitor pazopanib inhibits kinases including VEGR, PDGR, FGFR, and KIT. Pazopanib has been associated with clinical benefit in case reports, and was the subject of a trial in the UK [6, 157]. A clinical trial of the multikinase inhibitor cabozantinib as an angiogenesis inhibitor in patients with advanced, platinum-failure MCC was terminated prematurely due to poor tolerability and lack of responses in the first 8 patients [158]. Of note, evidence for ERK activation in MCC is lacking [159], and HRAS mutations do not render MCC cell lines responsive to MEK inhibition [76].

V. Summary

On March 5–6, 2018 over 50 participants gathered at the National Cancer Institute, NIH for the International Workshop on Merkel Cell Carcinoma Research (IWMCC). Academic, government, and industry scientists along with regulatory and patient representatives came together to discuss the state of MCC research, identify high priority research directions, and develop research strategies. The Workshop addressed basic, translational, and clinical research with the goal of improving outcomes for patients with MCC. Consensus statements and top research priorities identified by the IWMCC Working Group are presented below. An extended list of research questions considered at the Workshop can be found in Supplemental Table 1.

V.1. Basic Science Consensus Statement

Key basic research questions for MCC relate to the MCC cell of origin, distinctions between VP-MCC and VN-MCC, and MCPyV biology. The cell of origin for MCC remains unknown, and may be of significantly different lineage than suggested by the tumor phenotype. Furthermore, VP-MCC and VN-MCC may have distinct cells of origin.

Fundamental questions remain regarding the biology of MCPyV in human infection and tumorigenesis. The cellular reservoir(s) and transmission mechanisms for infection by wild-type MCPyV remain unknown. T antigens and other viral gene products have been shown to have diverse functions involving numerous cellular partners. In particular, the numerous activities of ST are incompletely elucidated. It is not well understood which viral protein(s) are most critical for mediating viral latency, replication, transformation, and maintaining an oncogenic phenotype. Identification of the most crucial tumor-promoting pathways has implications for effective therapeutic targeting of VP-MCC, and may suggest an explanation for the apparent unique status of MCPyV as the only polyomavirus known to be a significant human oncovirus.

Despite molecular differences, VP-MCC and VN-MCC are almost indistinguishable clinically and in their response to immune checkpoint inhibitors. Common putative driver mutations (RB1, TP53) are present in VN-MCC. Studies comparing VN-MCC to VP-MCC may ultimately determine a minimum set of signaling pathways that must be disrupted—either by mutation or by virus infection—to initiate MCC tumorigenesis. Parallels may be drawn from head and neck SCC, which also has a mixed viral and non-viral etiology. Mouse modeling of MCC may help to address questions including the cell of origin, which viral T antigens (and T antigen functional domains) are sufficient to generate VP-MCC, and which genomic changes are sufficient to generate VN-MCC.

Key Basic Science Questions in MCC

• What is the cell of origin for MCC?

• How does the MCC tumor microenvironment contribute to pathogenesis and immune evasion?

• What are the relative contributions of ST and LT to tumor initiation and maintenance?

• How do we model MCC (VP-MCC and VN-MCC) in mice, including features such as immune response/evasion, invasion, and metastasis?

• What cell types harbor productive MCPyV infection?

V.2. Translational Consensus Statement

While great progress has been made for about half of patients with advanced MCC based on benefit from PD-(L)1 blockade, many patients have primary or acquired resistance to these agents. Combination immune therapy approaches must be explored, but MCC therapy may also benefit from targeted approaches that synergize with immune therapy or could be used in checkpoint-refractory patients. Unbiased small molecule screens for agents that selectively affect VP-MCC or VN-MCC cell lines are ongoing. The wild type status of p53 in most VP-MCC tumors suggests potential efficacy for a small molecule that activates the p53 pathway, such as a nutlin-like agent [160], which could synergize with other therapies. Survivin or Bcl-2 targeted therapies may also take advantage of the biology of most MCC tumors. These studies could be facilitated by the use of patient derived xenografts (PDXs), which are more readily established from MCC tumors than are cell lines. Because many patients with VP-MCC tumors do not respond to immune checkpoint agents, it is likely that the number, diversity and avidity of MCPyV-specific T cells is not optimal. Future investigations may examine therapeutic vaccines as neoadjuvant therapy for MCC, such as using vaccinia virus to deliver inactivated viral oncoproteins, as has been demonstrated for HPV-driven premalignant disease [161]. Although a role for MCPyV status in guiding clinical management of MCC has not yet been established, translational investigations of MCC would benefit from routinely determining the MCPyV status of tumors, given the striking molecular differences between VP-MCC and VN-MCC.

Key Translational Questions in MCC

• What are the potential therapeutic targets in VN-MCC and VP-MCC?

• Do genomic alterations—either global (e.g. mutation burden) or specific (e.g. PIK3CA mutation)—have implications for prognosis or therapeutic response?

• What is the optimal detection method for MCPyV?

• What is the clinical value of distinguishing VP-MCC and VN-MCC?

V.3. Clinical Consensus Statement

The consensus on the clinical knowledge gaps and research priorities in MCC clusters around three major themes: therapeutics, biomarkers and infrastructure.

The current armamentarium for management of MCC includes surgery, radiation therapy, cytotoxic chemotherapy, immune checkpoint inhibition with PD-(L)1 inhibitors, and clinical trials. The advent of PD-(L)1 inhibitors has been of great benefit for some MCC patients. However, it is too early to determine long-term outcomes. Moreover, there are subsets of patients who present refractory to PD-(L)1 inhibitors or develop resistance over time. Immunosuppression, solid or hematologic transplantation, and hematologic malignancies are known risk factors for MCC. These patients have been routinely excluded from clinical trials and the optimal therapeutic strategy for these individuals is unclear. Finally, conventional cytotoxic chemotherapy and radiation therapy are active therapies that predate PD-(L)1 agents, and the best use of these agents in conjunction with immunotherapy remains to be determined.

The fact that MCC is either polyomavirus positive (VP-MCC) or negative (VN-MCC) presents both unique opportunities and a therapeutic conundrum. At this time, MCC viral status does not help segregate patients into those likely to respond to any therapy given the present state of our knowledge and pharmacopeia offerings. We anticipate that targetable oncogenic drivers, at least in the VP-MCC group, are within reach. VN-MCC tumors possess a molecular signature similar to that of the prototypical pattern seen in melanoma, characterized by UV-damage to the DNA and the presence of neoantigens. Thus, clinical development may likewise follow the path taken by other immunotherapy-responsive tumors such as melanoma.

Advancing the clinical management of MCC would be enhanced by the creation of a collaborative infrastructure that allows for formal information sharing and rapid clinical trial design and implementation. This is critical in a disease with low prevalence. Current understanding of MCC’s natural history and outcomes related to therapeutic intervention is scant, fractionated amongst individual investigators, and lacks common standards.

To further our knowledge of the natural history of MCC, prognostic factors, predictors of drug response, implications of MCPyV status in therapeutic response, as well as rapid clinical trial design and implementation, a collaborative multi-institutional effort is needed.

Key Clinical Questions in MCC

• What alternatives to immunotherapy are effective for non-responders and ineligible patients?

• What markers or tumor characteristics predict response to immune checkpoint inhibitors?

• What treatment combinations will enhance the efficacy of immunotherapy for MCC?

• What is the optimal role for radiotherapy in MCC management?

• How can clinical trials be optimized to maximize collaboration and minimize competition for this rare tumor?

V.4. Closing Remarks

Substantial advances have been made in MCC biology, diagnosis, and therapy. However, significant research challenges remain. There is a striking molecular dichotomy between VP-MCC and VN-MCC, but the biological and clinical consequences of this difference—including critical pathways for tumorigenesis—are incompletely understood. The MCC cell of origin remains unknown. Finally, continued progress in improving patient outcomes will require the validation of additional therapeutic options to augment gains provided by surgery, radiotherapy, and immunotherapy.