Introduction

Merkel cell carcinoma is a rare and aggressive neuroendocrine carcinoma of the skin with a 5-year overall survival rate of 40% [1]. The two main risk factor are immunosuppression [2] and sun exposure [1]. Whereas the incidence is still low, with 0.7 cases per 100,000 person-years in the United States in 2013, a dramatic increase of 95% was observed between 2000 and 2013, and a further rise in incidence has been predicted [3].

In 2008, Feng et al. identified a polyomavirus which they found integrated in the genome of Merkel cell carcinoma cells and accordingly named it Merkel cell polyomavirus (MCPyV) [4]. Currently, integration of this virus and the expression of viral oncoproteins named T antigens are established as the primary oncogenic events for approximately 80% of Merkel cell carcinoma cases [4]. The remaining 20% of Merkel cell carcinoma cases lacking MCPyV integration are considered as a distinct tumor subset primarily caused by UV radiation [5, 6]. In line with this notion, substantial differences between the two subsets with respect to morphology [7], and immunohistochemical profiles [8] have been described. Moreover, virus-negative Merkel cell carcinomas are characterized by higher mutational burden with predominant UV signature [5, 6], and a worse outcome than their virus-positive counterparts [9].

In a previous study [10], to assess performance of a set of markers for Merkel cell carcinoma diagnosis, we characterized the expression of nine proteins by immunohistochemistry and determined the MCPyV status by quantitative PCR in a cohort of 118 patients with Merkel cell carcinoma in comparison to 85 with extra-cutaneous neuroendocrine carcinomas. To further exploit this dataset, the current study aimed to (1) compare the morphological and immunohistochemical features of MCPyV-positive and -negative Merkel cell carcinoma and (2) evaluate whether discriminative features could be used as surrogate markers for MCPyV status.

Methods

Patients and samples

Merkel cell carcinoma cases were selected from an historical/prospective cohort of Merkel cell carcinoma patients from 6 French hospital centers. Inclusion criteria for the cohort have been described previously [11]. Briefly, patients had a diagnosis of Merkel cell carcinoma established between 1998 and 2017 (local ethics committee, Tours, France; no. ID RCB2009-A01056-51) [12]. Among the cohort, only cases with sufficient available formalin-fixed paraffin-embedded tumors for representative hematein phloxin saffron (HPS) slide staining and previously determined MCPyV status [10] were included in the present analysis.

Clinical and follow-up data

Age, sex, immunosuppression (HIV infection, organ transplant recipient, hematological malignancies), American Joint Committee on Cancer (AJCC) stage at the time of surgery [13], location of the primary tumor and follow up (recurrence free and specific survival) were collected from patient files.

MCPyV status determination

MCPyV status was determined using real time PCR with a ROC curve validated cut-off as previously described [10]. The Merkel cell carcinoma cell line WaGa (RRID:CVCL_E998) [14] was used as positive control. To note, while immunohistochemical detection of Large T antigen has been proposed as an efficient method to determine MCPyV status [9], our previous study comparing the performances of both procedures to distinguish Merkel cell carcinoma from other neuroendocrine carcinoma, revealed higher sensitivity (83%) and high specificity (97%) of MCPyV genome detection by quantitative PCR [10]. Consequently, LT staining was performed but the classification into MCPyV-positive and negative was based on quantitative PCR results.

Morphologic study

For all specimens, one representative 4 µm thick, Hematein-phloxin-saffron stained section was reviewed by two pathologists (SG, TK) with blinding to diagnosis. Morphologic features were assessed by the following criteria: nuclear shape (0: regular, 1: elongated), presence of nucleoli (0: absent or inconspicuous, 1: present), cell size (0: small: <2 lymphocytic nuclei, 1: moderate: 2 to 3 lymphocytic nuclei, 2: large: >3 lymphocytic nuclei), cytoplasm volume (0: none/inconspicious, 1: abundant), clear cytoplasm (0: no, 1: yes), rosette-like structure (0: no, 1: yes), intraepidermal component (0: no, 1: yes), divergent component (0: no, 1: yes) and associated intraepidermal neoplasia such as actinic keratosis or Bowen disease (0: no, 1: yes). All discordant cases were reviewed collegially.

Immunohistochemistry

We extracted data for the expression of the following markers from a previous study [10]: cytokeratin 20, thyroid transcription factor 1 (TTF- 1), atonal homolog 1 (ATOH1), neurofilament (NF), special AT-rich sequence-binding protein 2 (SATB2), paired box protein 5 (PAX5), terminal desoxynucleotidyl transferase (TdT), CD99, epithelial membrane antigen (EMA) referred as MUC1 in the previous study, and large T antigen (CM2B4).

In addition, we analyzed cytokeratin 7 and p53 expression as well as the staining pattern (dot, diffuse or mixed) for cytokeratins 8, 18 and 20. All antibodies are available in supplementary method Table S1. A Benchmark platform was used for staining, except for cytokeratins 8, 18 and CM2B4 stainings, which were manually performed. p53 expression was evaluated according to the Allred score whereas 0, 7 and 8 are considered as abnormal expression, indicating loss of active p53 [15]. For all immunohistochemical analyses, the number of uninterpretable samples (mainly due to failure of tissue microarray inclusion) is mentioned in the corresponding figures.

Statistical analyses

Continuous data are described by medians (Q1–Q3) and categorical data with number and percentage of interpretable cases. Associations were assessed by Mann–Whitney and two-tailed Fisher’s exact tests for continuous and categorical data, respectively. P < 0.05 was considered statistically significant. MCPyV status was determined by qPCR with a previously validated cut-off (MCPyV load >1.2 copies/cell) [10]. Categories and thresholds of immunohistochemical markers were derived from previous studies [10, 16,17,18]. Since no thresholds were previously determined for cytokeratins 8, and 18, the same categories as for cytokeratin 20 were applied. The diagnostic accuracy of immunohistochemical markers to determine MCPyV status was compared with the reference standard (quantitative PCR) by using the positive likelihood ratio as a measure of accuracy combining sensitivity and specificity. Recurrence-free survival and specific survival related to patient MCPyV status were analyzed by log-rank test and presented as Kaplan-Meier curves. Univariate and multivariate Cox proportional-hazards regression was used to identify factors associated with Merkel cell carcinoma recurrence and death, estimating hazard ratios (HRs) and 95% confidence intervals (CIs). Specific deaths were considered events. Covariates were identified as potential prognostic confounders with p ≤ 0.25 on Cox univariate regression analysis and then included in the multivariate Cox analysis. Statistical analysis involved use of XL-Stat-Life (Addinsoft, Paris, France).

Results

Patient characteristics and clinical outcome

For 101 Merkel cell carcinoma cases corresponding to 80 MCPyV-positive and 21 MCPyV-negative, sufficient material for morphologic examination allowed inclusion (Fig. 1/Flow Chart). To underline common and distinctive features of the two groups, clinical data are compared in Table 1. Virus-positive and -negative Merkel cell carcinoma did not differ with respect to age, sex, immune status, stage (American Joint Committee on Cancer), and location of the primary tumor. By contrast, Merkel cell polyomavirus-positivity was significantly associated with lower risk of recurrence (HR 0.36 CI 0.18–0.74, P = 0.005) and specific death (HR 0.37 CI 0.15–0.89), P = 0.03) on univariate analysis (Table 1/supplementary data S1) and was also statistically significant in multivariate Cox analysis (supplementary data S1). These results confirm virus-negative status as a negative prognostic marker for Merkel cell carcinoma [9].

Fig. 1
figure 1

Flow chart of the study

Full size image
Table 1 Clinical, morphologic and immunohistochemical features of MCPyV-negative (n = 21) and -positive MCC (n = 80) cases. Characteristics of the two subsets were compared by non-parametric Mann Whitney and Fisher’s exact tests for quantitative and qualitative data, respectively. For survival analyses, log rank test was used. Positive likelihood ratios for MCPyV positivity were assessed for markers that differed between the two groups (p < 0.05), by MCPyV status, by real-time PCR as the gold standard and predetermined cutoffs (underlined criteria vs rest) [4, 6, 11]
Full size table

MCPyV-positive and -negative cases harbor distinct morphologic features

To evaluate the morphologic differences between the virus-positive and -negative Merkel cell carcinoma cases, assessment of nine microscopic criteria was conducted (Table 1/Fig. 2). We identified nuclear roundness to be associated (p = 0.005) with virus-positivity, while elongated nuclei—similar to the spindle-shape variant of small cell lung cancer—were observed more frequently in virus-negative cases. Moreover, the latter samples more frequently displayed a “large cell neuroendocrine carcinoma” phenotype, with larger cell size (p = 0.026), abundant cytoplasm (p = 4×10−7) and clearly visible nucleoli (p = 0.002) (Table 1/Fig. 2). The combination of elongated nuclei and abundant cytoplasm was observed in 19% (n = 4) and 1% (n = 1) of virus-negative and -positive cases, respectively. Furthermore, rosette-like structures (supplementary data S2) and clear cytoplasm (Fig. 2) were also associated with absence of Merkel cell polyomavirus (p = 0.02 and 2×10−6 respectively). To note, 100% of cases with clear cytoplasm have also an extended cytoplasmic size. Finally, only virus-negative Merkel cell carcinoma harbored intraepidermal Merkel cell carcinoma components and displayed Bowen-associated disease or divergent differentiation (supplementary data S2).

Fig. 2
figure 2

Morphologic features of MCPyV-negative and -positive Merkel cell carcinoma cases (bar = 100 µm): on standard examination, MCPyV-negative cases are characterized by more irregular, elongated nuclei as observed in small cell lung cancer Additionally, some cases had large cell neuroendocrine carcinoma features with abundant and clear cytoplasm and predominant nuclei. By contrast, MCPyV-positive cases are composed of medium to small round tumor cells with scant cytoplasm and round nucleus. MCC: Merkel cell carcinoma

Full size image

These results confirmed that many Merkel cell polyomavirus-positive and -negative Merkel cell carcinoma can be distinguished based on morphological criteria [7, 18, 19] which probably reflects significant biological differences between the two groups. However, we also identified difficult-to-classify MCPyV-negative cases lacking prototypic morphologic features (supplementary data S2).

MCPyV-positive and -negative cases feature distinct immunohistochemical profiles

To determine whether also immunohistochemistry could discriminate between virus-positive and -negative Merkel cell carcinoma, we compared the two groups with respect to expression of a panel of diagnostic markers (Fig. 3/Table 1). Positivity for TTF1 and cytokeratin 7, lack of or overexpression of p53, and frequent lack of expression of neurofilament were hallmarks of the virus-negative cases (Table 1). By contrast, virus-positive cases not only featured large T antigen-positivity but also high EMA expression and more frequently a dot like staining pattern for the cytokeratins 8, 18 and 20, as well as for CD99. These findings demonstrate substantial variations in the immunohistochemical profiles of virus-positive and -negative Merkel cell tumors and additionally suggest a possible impact of the T antigens on cytoskeletal organization.

Fig. 3
figure 3

Immunophenotypes of MCPyV-negative and -positive Merkel cell carcinoma cases (bar = 100 µm): Immunohistochemical investigation of CD99 (a), cytokeratins 7 (b), 8 (c), 18 (d) and 20 (e), EMA (f), p53 (g) and TTF-1 (h) revealed several distinctive features between the two groups: MCPyV-positive status was associated with cytokeratins 8, 18 and 20 and CD99 dot-like expression pattern as well as high EMA expression, whereas MCPyV-negative cases frequently displayed expression of cytokeratin 7 and TTF-1 and abnormal p53 expression. MCC: Merkel cell carcinoma

Full size image

CD99 dot-like pattern as a marker of MCPyV-positive Merkel cell carcinoma

To compare the performances of all investigated markers to predict virus status, positive likelihood ratios were determined by using previously described cut-offs [7, 10, 16, 18] (Table 1). These analyses identified CD99 dot-like expression pattern as most highly associated with virus-positivity of Merkel cell carcinoma (sensitivity = 81% [95% CI: 70–89], specificity = 90% [95% CI: 68–99], positive predictive value = 97% [95% CI: 89–99], negative predictive value 55% [95% CI: 43–66], positive likelihood ratio = 8.08 [95% CI: 2.16–30.21]). In line with this, such CD99 dot-like pattern was found in 86% (n = 49/57) of the cases which demonstrated large T antigen-expression in immunohistochemistry, as compared with only 35% (n = 12/34) of large T antigen-negative cases (supplementary data S3). Of interest, 10 MCPyV-positive cases lacking immunohistochemical large T antigen expression still showed a CD99 dot-like expression pattern (positive and negative predictive values of CD99 dot pattern for MCPyV status determination in the Large T non expressing cases: 83% [95% CI: 56–95] and 71% [95% CI: 56–82] respectively). These results suggest that the CD99 expression pattern might serve as an additional indicator to evaluate the Merkel cell carcinoma virus status.

Discussion

With respect to tumor cell morphology and immunophenotype, several differences between virus-positive and –negative Merkel cell carcinoma were assessed in the present study. In accordance with previous reports [7, 18], several distinctive microscopic features were observed between the two groups. Moreover, the virus-negative tumors differed from the others by a so called “aberrant” immunohistochemical profile [8] with reduced expression of the Merkel cell carcinoma marker i.e., neurofilament and a more prevalent positivity of those normally observed in extra-cutaneous neuroendocrine carcinomas such as TTF-1 and cytokeratin 7. Interestingly, dot-like expression patterns of cytokeratins and CD99 were more frequent in virus-positive cases and accordingly, CD99 expression pattern was identified as suitable additional marker for the determination of the Merkel cell polyomavirus status of Merkel cell carcinoma.

Two viral oncoproteins i.e., small T and Large T are expressed in Merkel cell carcinoma tumor cells and are considered as the main oncogenic triggers for the development of virus-positive Merkel cell carcinoma [20, 21]. In contrast, UV-induced mutations are thought to drive carcinogenesis of virus-negative Merkel cell carcinoma [5, 6, 22, 23]. Targeting of the same oncogenic pathways (RB1 and p53) either by T antigens or somatic mutations, may account for the common neuroendocrine phenotype [24,25,26] observed in virus-positive and negative Merkel cell carcinomas. Nevertheless virus-negative tumors are now considered as a subset genetically distinct from the others [27] and characterized by a very high mutational burden (10 mutations per Mb) while very low mutation frequencies (0.4 mutations per Mb) were detected in Merkel cell polyomavirus-positive Merkel cell carcinoma [5, 28,29,30]. Interestingly genomic complexity and cancer mutation burden have been demonstrated to correlate with microscopic features of tumor cells such as nuclear pleomorphism [31] and cytological atypia [32]. Indeed, in soft tissue tumors [33], “simple karyotype” neoplasias such as Ewing sarcoma with recurrent EWSR1 rearrangement display uniform, regular cytology, while “complex karyotype” sarcoma feature more pronounced cytological atypia. Similarly, the degree of differentiation was directly related to genomic alteration level in cutaneous squamous cell carcinoma [34].

In line with such observations, substantial morphologic differences were observed between virus-positive and -negative Merkel cell carcinoma. Indeed, Katano et al. [19] reported that the 6 Merkel cell polyomavirus-positive cases investigated were characterized by round and vesicular nuclei with fine granular chromatin and small nucleoli whereas, by contrast, most of the five virus-negative samples had polygonal nuclei with clear cytoplasm. Applying morphometry, Kuwamoto et al. [7, 18] confirmed that virus-negative cases had more irregular nuclei and more abundant cytoplasm in a set of 26 Merkel cell carcinoma cases. In our series, investigating 101 Merkel cell carcinoma cases we provide further confirmation of these microscopy-studies.

Indeed, we found MCPyV-positive cases to be characterized by uniform round-ovoid nuclei, scant cytoplasm, and frequently displaying a morphology close that of Burkitt lymphoma or Ewing sarcoma, both neoplasias either induced by virus or chromosomal translocation [33, 35]. By contrast, more heterogeneous cytological features with marked atypia were observed in virus-negative Merkel cell carcinoma which, in our view, exhibited close similarities with extra-cutaneous neuroendocrine carcinoma. Indeed some virus-negative cases appeared as a dense proliferation of spindle cells with elongated dark nuclei similar to the spindle shape variant of small cell lung cancer [36] while others cases, comparable to the tumor previously reported as large cell neuroendocrine carcinoma of the skin [37], feature abundant cytoplasm and prominent nuclei as shown in Fig. 2. To note, intermediary phenotypes were also observed.

While virus-positive Merkel cell carcinoma is almost exclusively located in the dermis and subcutis, involvement of the epidermis has been mostly reported for virus-negative cases. Indeed, detection of an associated intra-epidermal neoplasia as well as a divergent differentiation are predictive of MCPyV-negative status [7, 38,39,40]. Although, intra-epidermal spreading of Merkel cell carcinoma was only observed in two cases in our study, both of them were virus-negative suggesting that epidermotropism as an additional –although rare- characteristic of virus-negative Merkel cell carcinoma.

Whereas cytokeratin 20-positivity and TTF-1 negativity are currently used in routine practice to distinguish Merkel cell carcinomas from extra-cutaneous carcinomas [10, 41], our study confirms the prevalence of so called “aberrant” immunohistochemical profiles in virus-negative cases [8]. Indeed, these latter differed from the viral induced tumors by a more frequent negativity of the Merkel cell carcinoma markers i.e., neurofilament [8, 42, 43], and more frequent positivity of TTF-1 [44] and cytokeratin 7 [8, 45] again underlining the phenotypic similarities between virus-negative Merkel cell carcinoma and extra-cutaneous neuroendocrine carcinomas. In addition, abnormal p53 expression probably due to loss of protein function by somatic mutations was frequently observed in UV-induced tumors as previously reported [46, 47].

Interestingly, cytokeratins and CD99 “dot-like” patterns were associated with virus-positivity of Merkel cell carcinoma. In the interfollicular epidermis under physiological conditions, cytokeratins 8, 18 and 20 expression is restricted to the Merkel cell lineage [48,49,50] and accordingly, frequent positivity of Merkel cell carcinoma for these cytokeratins is observed [27]. In contrast to non-neoplastic Merkel cells that show a diffuse cytokeratin expression, Merkel cell carcinoma cells often harbor cytokeratins arranged in paranuclear dots. Of note, expression of cytokeratins 8 and 18 either in a diffuse or in a dot-like pattern can also be observed in extra-cutaneous high grade neuroendocrine carcinomas [51, 52]. Interestingly such cytokeratins dot-like pattern in virus-positive Merkel cell carcinoma not only renders this feature a useful additional marker for diagnosis but also suggests a potential involvement of the T antigens in cytokeratin “dot like” relocation. Indeed, using a transgenic mouse model, Verhaegen et al. [53] previously obtained similar cytokeratins 8 and 20 dot-like pattern upon ectopic expression of small T in Merkel cells and in line with such findings, capability of T antigens to disrupt cellular cytoskeletal organization has previously been emphasized [54, 55]. A possible explanation for these dots is entrapment of the Golgi apparatus in the cytokeratin aggregates which might explain why not only cytokeratins but also the membrane protein CD99 can be found in paranuclear dots. Indeed, CD99 is a transmembrane protein involved in a broad spectrum of physiological and pathological conditions such as cell migration and intracellular trafficking [56]. In Epstein Barr virus-related Hodgkin lymphoma, direct downregulation of CD99 by the latent membrane protein 1 (LMP1) affects tumor cell differentiation and contributes to immune escape via downregulation of major histocompatibility complex class 1 [57]. Therefore CD99 sequestration in cytoplasmic “dot”, might reduce the protein membrane delivery, and consequently contribute to the aggressiveness of virus positive-Merkel cell carcinoma.

While determination of the Merkel cell polyomavirus status is not yet recommended in the Merkel cell carcinoma guidelines, virus-negative cases constitute a subset of tumors phenotypically [8, 18] and genetically [5, 6] distinct from the others as described above and as underscored by the present study. In particular, increased aggressiveness and worse outcome [9] suggest a potential value of routine determination of the virus status in Merkel cell carcinoma patients. Although further confirmation in independent cohorts are needed, our results suggest that CD99 expression—with testing already available in pathology laboratories—might be used as a surrogate or associated with large T antigen immunohistochemistry to predict MCPyV status in clinical practice.

To conclude, our results confirm that MCPyV-positive and -negative Merkel cell carcinoma cases are characterized by distinct morphological and immunohistochemical features that imply a significant difference in tumor biology and behavior. Importantly, we identified a dot-like pattern in CD99 expression as a relevant marker associated with MCPyV status.