Introduction

Merkel cell carcinoma (MCC) is a rare tumor of the skin that features an aggressive course, with overall 5-year survival estimated at 40% [1]. MCC occurs essentially in older people and the two main risk factors are sun exposure and immunodeficiency [1]. The diagnosis is based on histology, which reveals high-grade neuroendocrine morphological features close to small cell carcinoma and expression of neuroendocrine markers and/or cytokeratin 20 [2].

In 2008, Moore et al. described Merkel cell polyomavirus (MCPyV) as the major etiological agent of MCC [3]. MCPyV establishes a latent chronic infection in the dermis of most healthy individuals, and the mechanisms leading to oncogenesis are still under investigation. In about 80% of cases, tumor cells harbor integrated virus. Of note, MCPyV large T antigen in tumor cells is characterized by non-sense mutations leading to the loss of replicative abilities of the virus and the lack of late protein synthesis [4]. MCPyV genome expression in tumors is restricted to the early oncogenic proteins (small and large T antigens), and expression of T antigens is required for tumor proliferation [5]. These two viral proteins are probably the main early determinants of MCC oncogenesis.

Whether MCPyV-negative tumors should be considered a specific biological entity is debated. Indeed, MCPyV-negative MCC is thought to be essentially due to UV exposure [6] because the rates of somatic genetic mutations with a predominant UV signature are higher than in MCPyV-positive tumors. Moreover, levels of intratumoral CD8 infiltrates are lower, which suggests decreased immunogenicity [7] and a worse outcome [8].

Combined MCC involves rare MCC variants that represent 5 to 10% of MCC cases [9]. Combined MCC cases are characterized by the association of a main component of MCC with one or more other tumor components harboring non-neuroendocrine differentiation, including an epidermic orthologous component [2] (squamous, basal-cell-like, adnexial, and melanocytic) or a heterologous component [10] (glandular and sarcomatous).

Three main studies detected no MCPyV in combined MCC [9,10,11]. Conversely, a few recent case reports have demonstrated the presence of MCPyV in the MCC component of combined MCC [12], associated with papillomavirus infection in one case [13].

Papillomaviruses and polyomaviruses are closely related double-strand DNA viruses with similar oncogenic abilities. Indeed, beta human papillomaviruses are involved in the oncogenesis of cutaneous squamous cell carcinoma, and alpha-human papillomaviruses involved in cutaneous Bowen disease and combined tumors in the oropharynx (associating neuroendocrine and squamous components) [14]. Hence, human papillomavirus might be an etiological agent of combined carcinomas.

Whether combined MCC belongs to the spectrum of virus-induced tumors or should be considered a non-MCPyV MCC induced by an alternative oncogenetic pathway remains unclear. In this study, we compared the clinical features of conventional and combined MCC and determined the viral status of combined MCC by systematic assessment of MCPyV, 9 other polyomaviruses and 46 papillomaviruses.

Methods

Study period, data, and settings

MCC cases were selected from an ongoing historical/prospective cohort of 223 patients with MCC from six French hospital centers. The diagnosis of MCC was established between 1998 and 2015 (local ethics committee approval, Tours, France, no. RCB2009-A01056-51). The cohort inclusion criteria were previously reported [15]. All tumors were submitted to histological review by an endocrine pathologist (SG), based on the identification of morphological features of high-grade neuroendocrine carcinoma and immunohistochemical expression of epithelial and neuroendocrine markers. Only cases with available formalin-fixed paraffin-embeded (FFPE) samples and sufficient tumor material for tissue microarray inclusion were included in the study (n = 107).

Design of the study

Tumors were classified as combined MCC when the following criteria were met on pathological examination: presence in the same tumor mass of a conventional MCC component and an additional subpopulation of tumor cells showing non-neuroendocrine differentiation in contact with MCC. Both conventional and combined MCC were compared on clinical and immunochemical features.

All cases of combined MCC, 12 randomly selected conventional MCC cases from the cohort and 12 cutaneous squamous cell carcinoma cases from the pathology department of the hospital center of Tours were considered for molecular analysis.

Clinical and histological data

The following data were collected from patient files: age, sex, tumor site, tumor extension at the time of diagnosis (reported in accordance with American Joint Committee on Cancer [AJCC] staging), immunosuppression (HIV infection, organ transplant recipients, hematological malignancies) [16], and follow-up data. On histology, the characteristics of the non-MCC component were noted: histological type confirmed by immunochemistry, notable expression of squamous cell carcinoma markers, degree of cytological atypia, and presence of keratinization. In addition, the relation between the two tumoral components was investigated: the tumor was considered admixed when one component surrounded multiple small foci of another component and distinct when the two components were located in two distinct areas of the tumor. Finally, the characteristics of the epidermis (connection with the tumor or not, presence of an ulceration) were noted.

Immunohistochemistry

FFPE tumor samples were included in a tissue microarray. Briefly, representative areas were selected on hematoxylin/eosin-stained sections (representative of the two tumor components in combined MCC), extracted by using a 1-mm tissue core and mounted by using a semi-motorized tissue arrayer (MTA booster OI v2.00, Alphelys). For each patient, five tumor cores were placed adjacent to each other on the tissue microarray.

Tumors were screened with a panel of antibodies including conventional MCC markers used for diagnosis (pan-cytokeratin AE1-AE3, chromogranin A, synaptophysin, cytokeratin 20), several markers rarely expressed by MCC (thyroid transcription factor 1 (TTF-1), cytokeratin 7), squamous differentiation markers (cytokeratin 5/6, p40), MCPyV large T antigen, and p53. Antibodies and dilutions are available in Supplemental Table S1. Staining was performed on a Benchmark platform, except for CM2B4 staining, which was manually performed, as previously described [5]. Immunohistochemical viral status was interpreted by using the Allred score [8]: intensity and percentage of positive cells were assessed by an 8-point semi-quantitative score. A score > 2 was considered MCPyV-positive. In the same way, p53 expression was evaluated according to the Allred score, considering scores 0, 7, and 8 as abnormal expression, predictive for loss of active p53 [17]. Intratumor CD8 infiltrate was scored as previously described [7]. For all immunohistochemical analyses, the number of uninterpretable samples (mainly due to failure of tissue microarray inclusion) is mentioned in the figures.

DNA extraction

Three 10-μm-thick FFPE sections of representative tumor areas were used for molecular analysis. In addition, for MCPyV-positive combined MCC cases showing “distinct patterns” (defined as a large distinct area of both components), each component underwent specific sample coring, followed by a morphological control on HE slides. Genomic DNA was isolated from FFPE tissue samples by using a Maxwell 16 instrument (Promega) with the Maxwell 16 FFPE Plus LEV DNA purification kit (Promega) according to the manufacturer’s instructions.

VP1 gene PCR assay

MCPyV VP1 coding sequence was detected by nested PCR as previously described [4]. Primer sequences are listed in Supplemental Table S2.

Large T antigen quantitative PCR assay

Quantitative PCR assay was performed as reported previously [18]. Briefly, 100 ng DNA was mixed with 0.2 μM primers (Supplemental Table S2), 0.1 μM DNA probe, and 2xTaqman Universal PCR Master Mix (Applied Biosystems) in a final volume of 25 μl. PCR reaction involved use of Applied Biosystems 7500 Real-Time PCR Systems programmed for 50 °C × 2 min with an initial denaturation at 95 °C × 15 min, followed by 45 cycles at 95 °C × 15 s and 60 °C × 60 s. Normalization was to human albumin gene level under the same conditions. DNA range was determined by using the MKL-1 cell line as a reference (6 points: 100 copies to 10,000,000 copies).

MCPyV viral status determination

Because MCPyV is a ubiquitous virus of the skin infecting a large part of the population and the papillary dermis is the site of replication of wild-type episomal MCPyV [19], low viral load detection may be expected in the dermis of healthy people in the absence of MCC and in non-MCC skin neoplasms when using ultrasensitive methods [19,20,21]. To avoid detection of wild-type episomal MCPyV of the dermis unrelated to MCC tumors, a first validation step of the MCPyV detection procedures (VP1 PCR, large T antigen quantitative PCR) was performed with 12 conventional MCC and 12 non-MCC tumor samples as positive and negative controls, which confirmed the high sensitivity and specificity of the procedures. Therefore, MCPyV-positive status was retained only in cases positive with both validated molecular procedures.

Type-specific multiplex genotyping assays

In total, 76 different polyomaviruses and papillomaviruses were investigated by type-specific multiplex genotyping (TS-MPG) assay, a validated, highly sensitive procedure [22] designed to detect low load of episomal viruses in the skin [23] and the environment [24].

The procedure combines multiplex PCR and bead-based Luminex technology (Luminex Corp., Austin, TX, USA), as previously described [22, 25]. Multiplex type-specific PCR involved use of specific primers for detecting 9 polyomaviruses (BKV, KIV, JCV, WUV, TSV, HPyV6, HPyV7, HPyV9, and SV40), 19 high-risk alpha-human papillomaviruses (types 16, 18, 26, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68a, 68b, 70, 73, and 82), 2 low-risk alpha-human papillomaviruses (types 6, 11), and 46 beta human papillomaviruses (types 5, 8, 9, 12, 14, 15, 17, 19, 20, 21, 22, 23, 24, 25, 36, 37, 38, 47, 49, 75, 76, 80, 92, 93, 96, 98, 99, 100, 104, 105, 107, 110, 111, 113, 115, 118, 120, 122, 124, 143, 145, 150, 151, 152, 159 and 174). Two primers for amplifying beta-globin were added to provide a positive control for determining quality of the template DNA. Of note, the original set of primers included MCPyV sequences, which were excluded from the present study because of detection of the episomal virus in healthy skin as well as non-MCC tumors [21] as described previously.

Statistical analysis

Continuous data are described as median and range and categorical data as number and percentage of cases for which data were available. Proportional analysis was assessed by two-tailed Fisher’s exact test. Continuous variables were compared by non-parametric Mann-Whitney U test. Recurrence-free survival and overall survival related to patient characteristics were analyzed by log-rank test and represented by Kaplan-Meier curves. Univariate and multivariate Cox proportional-hazards regression was used to identify factors associated with MCC recurrence and death, estimating hazard ratios (HRs) and 95% confidence intervals (CIs). Overall deaths were considered as events and living patients were censored on the date of last follow-up. AJCC stage [26], immunosuppression [16] and covariates with p ≤ 0.20 on Cox univariate analysis were included in the multivariate Cox analysis as potential prognostic confounders. Statistical analysis involved use of XL-Stat-Life (Addinsoft, Paris, France). p < 0.05 was considered statistically significant.

Results

Characterization of the MCC population

Among 128 MCC cases with available FFPE samples, 8 (6%) met combined MCC criteria (Fig. 1), exhibiting squamous cell carcinoma differentiation in all cases. Characteristics of these cases are shown in Table 1. The 120 other tumors were classified as conventional MCC; the 8 combined MCC and 99 conventional cases could be included in the tissue microarray and were considered in this study.

Fig. 1
figure 1

Flow chart of cases in the study. MCC Merkel cell carcinoma, FFPE formalin-fixed paraffin-embedded, TMA tissue microarray

Full size image
Table 1 Features of combined Merkel cell carcinoma (MCC) cases
Full size table

Comparison between combined and conventional MCC

Clinical and immunohistochemical data of both groups are summarized in Table 2.

Table 2 Clinical, immunochemical, and virological features of combined and conventional MCC cases
Full size table

Combined MCC cases were more often diagnosed at localized stages (stage I: n = 5, 63%; stage II: n = 3, 37%) than conventional MCC cases (stage I: n = 23, 28%; stage II: n = 26, 32%, stage III: n = 29, 36%; stage IV: n = 3, 4%; localized (stages I–II) vs metastatic diseases (stages III–IV), p = 0.046). In addition, most combined MCC cases (n = 5, 62%) occurred in the head and neck area, with no significant difference from conventional MCC cases (n = 27, 34%) (p = 0.3).

On immunohistochemistry, combined MCC cases showed frequent expression of TTF1 (n = 5, 62%), which was rarely expressed in the other tumors (n = 10, 11%; p = 2.10−3). Furthermore, all combined MCC cases but only 15 (16%) conventional MCC cases demonstrated abnormal p53 staining, possibly reflecting p53-inactivating mutation (p = 3.10−6). Intratumoral CD8 infiltrates were absent in 6 (75%) combined MCC cases and was brisk (score 1) in the other cases. In contrast, CD8 intratumoral cells were observed in 64 (67%) conventional MCC cases (p = 0.03), with high density (scores 2–5) in 12 (13%). Representative illustrations of immunohistochemical staining are in Fig. 2. Only one combined MCC case presented weak large T antigen positivity (Fig. 3), whereas the viral protein was detectable by immunohistochemistry in 56 (62%) conventional MCC cases (p = 0.01).

Fig. 2
figure 2

Representative microphotographs of the immunohistochemical study for cytokeratin 20 (a–c), TTF-1 (d–f), p53 (g–i), and CD8 immune infiltrate (j–l) in conventional (a, d, g, j) and combined MCC (MCC (b, e, h, k) and squamous (c, f, i, l) component. To note, both components of combined cases harbored same immunochemical p53 profile as illustrate in (i): overexpresion of p53 in a small islet of squamous carcinoma (black star) surrounding by the MCC component

Full size image
Fig. 3
figure 3

Immunohistochemical detection of Merkel cell polyomavirus (MCPyV) large T antigen in a combined MCC sample: a moderate nuclear large T antigen expression in the MCC component and b lack of large T antigen expression in the squamous carcinoma component

Full size image

Follow-up data were available for 86 patients including 7 combined MCC cases. Median duration of follow-up was 17 months (ranges 2–209) and 36 recurrences and 33 deaths were reported during follow-up. On univariate analysis, combined MCC patients harbored a trend towards increased risk of recurrence (HR 2.44, 95% CI 0.95–6.29, p = 0.065) as shown in Fig. 4. Only male sex was associated with a decreased risk of recurrence (HR 2.83, 95% CI 1.44–5.56, p = 0.002) (Supplemental Table S3) whereas male sex and older age were associated with death (HR 2.27, 95% CI 1.13–4.58, p = 0.022 and HR 2.07, 95% CI 1.02–4.19, p = 0.043, respectively).

Fig. 4
figure 4

Kaplan-Meier survival curves for the Merkel cell carcinoma population by the MCC type (combined vs conventional cases): a recurrence-free survival and b overall survival

Full size image

A multivariate Cox analysis model including age, sex, immunosuppression, and AJCC stage (Table 3) revealed increased risk of recurrence (HR 4.15, 95% CI 1.37–12.57, p = 0.012) and death (HR 4.15, 95% CI 1.22–14.16, p = 0.023) with combined MCC.

Table 3 Multivariate Cox proportional-hazard analysis of factors associated with MCC recurrence and death
Full size table

MCPyV genome detection

The preliminary validation step of the MCPyV detection procedures (VP1 PCR, Large T antigen quantitative PCR) allowed us to detect MCPyV in all 12 conventional MCC cases except one by quantitative PCR (Table 2). All non-MCC tumors were negative for MCPyV with the two procedures, which led to the validation of this bimodal strategy for MCPyV status characterization.

This bimodal approach revealed MCPyV-positive status in 4/8 combined MCC cases, as shown in Table 2. Median MCPyV load was lower in combined MCC than MCC control cases (5.7 [range 0.13–28] vs 58 [4–313] copies/cell) (p = 0.04). Three of the four MCPyV-positive combined tumors were eligible for specific sampling independently targeting the two tumor components, and the remaining case consisted of closely intermixed components, which ruled out reliable specific separation. MCPyV DNA could be detected in the MCC component of the three tumors but was consistently absent in the other non-MCC part of the tumor.

Clinical, histological, and immunochemical features of MCPyV-positive and MCPyV-negative combined MCC cases are summarized in Table 1, and representative histological features of both subgroups are in Fig. 5. MCPyV was detected in combined MCC cases with basaloid squamous cell carcinoma (n = 2) and conventional squamous cell carcinoma (n = 2).

Fig. 5
figure 5

Representative microphotographs of MCPyV-positive (a–d) and MCPyV-negative (e–h) combined MCC samples. The neuroendocrine component is characterized by sheet of small uniform cells with high nucleocytoplasmic ratio (white star). It is associated with another carcinomatous component (black star): basaloid squamous cell carcinoma (a, b), conventional squamous cell carcinoma (c, d, e, h), and Bowen disease (f, g)

Full size image

Other polyomaviruses and papillomaviruses

Beta papillomavirus DNA was detected in 5/8 combined MCC cases (63%), 9/12 conventional MCC cases (75%), and 9/12 (75%) squamous cell carcinoma cases. No recurrence considering genotypes was observed in the combined MCC group, except for human papillomavirus 98, which was present in 2 cases (Supplemental Table S4). In addition, we detected human polyomavirus 6 in one conventional MCC case and human papillomavirus 16 in 2 squamous cell carcinoma cases.

Discussion

We identified 8 cases of combined MCC in a series of 128 MCC tumors. Combined MCC often expressed TTF1, which is almost always absent in conventional MCC cases and often showed p53-aberrant expression. In addition, multivariate analysis revealed an increased risk of recurrence and death in this population. Using two independent validated molecular procedures, we detected MCPyV in half of our combined MCC cases, only in the MCC component. Moreover, combined MCC cases showed lower MCPyV load as compared with conventional MCC tumors and frequent large T antigen negativity on immunochemistry. Finally, beta papillomaviruses were frequent in combined and conventional MCC and non-MCC tumor samples.

In accordance with previous reports, our series shows that combined MCC is a rare tumor, representing 6% of our cohort. Combined MCC tumors occurred preferentially on the head, the main sun-exposed area in our country. On immunohistochemistry, the neuroendocrine component of these tumors harbored the same phenotypical markers as conventional MCC, notably cytokeratin 20, but with unusual expression of TTF1 [27]. In addition, aberrant expression of p53, suggesting p53 mutation [9, 28], was common. This finding agrees with genomic analysis performed by Pullitzer et al. reporting higher somatic gene mutation rate in combined than conventional MCC [29].

In accordance with the Martin et al. study, which found 61% mortality3, we found combined MCC an aggressive neoplasm with decreased recurrence-free survival (p = 0.012) and overall survival (p = 0.023) as compared with conventional MCC, which highlights the importance of this morphological distinction in current practise.

Three studies [9,10,11] by Busam (in 2009), Martin (in 2013), and Carter (in 2017) reported the absence of MCPyV in combined MCC (7 and 15 cases, respectively). As a result, the authors proposed that combined MCCs should be included in the MCPyV-negative cutaneous tumor group. However, two case reports have given contradictory results, identifying MCPyV in two combined MCC cases by molecular biology [13, 30]. In addition, MCPyV was identified in the neuroendocrine component of a combined MCC case by high-sensitive immunofluorescence staining [12] and in another case by conventional immunochemistry [31]. Here, using a multimodal approach for MCPyV detection, we confirm these findings in a large series of combined MCC.

Weak expression of MCPyV large T antigen was detectable by immunochemistry in only one case with high MCPyV viral load (11 copies/cell), and MCPyV DNA was identified in half of the cases with a lower viral load. Variations in viral load offer an explanation for the previous discordant results. Indeed, immunochemical procedures used in most studies have a sensitivity of 80% as compared with amplification techniques, which probably reflects the influence of viral load on the sensitivity of the different identification approaches [32].

Considering the high prevalence of the MCPyV in normal skin, the use of highly sensitive molecular biology procedures might represent a pitfall in the identification of an MCPyV-related tumor, introducing a risk of false-positive results, which justifies our multimodal approach. Although MCPyV is considered a hallmark of MCC, episomal MCPyV can be detected with very low load in the papillary dermis of normal skin in many healthy adults [19] and in non-MCC skin tumors [20]. With the lack of an international gold standard for MCPyV quantification, arbitrary cutoffs have been used [33] to distinguish MCPyV-negative from MCPyV-positive tumors. To avoid this bias, we introduced a first step of calibration of our molecular MCPyV detection test in the study, using conventional MCC and non-MCC skin tumors as controls. Indeed in our setting, the sensitivity of the MCPyV detection was carefully managed to detect only significant MCPyV loads associated with MCC.

The significance of MCPyV detection in tumors and its relation with somatic mutations is still unclear. According to the “hit and run” phenomenon [34], MCC could feature loss of large T antigen dependence. Indeed, accumulation of somatic mutations—notably p53 mutations—could cause an empowerment process and might finally cause loss of MCPyV sequence by a selection process. In this way, the low viral load that we detected in combined MCC cases could be due to the presence of a minor MCPyV-positive tumor subpopulation. The divergent differentiation component that we found was squamous cell carcinoma in all cases. Because squamous cell carcinoma is related to papillomavirus infection [35], papillomaviruses could represent a possible etiological agent for combined MCC. Beta human papillomaviruses were detected in 63% of our combined MCC cases and 75% of conventional MCC cases, which rules out the possibility of a specific association between human papillomavirus and combined MCC. However, the frequent detection of cutaneous human papillomavirus in MCC in our study raises the question of their potential impact as a co-carcinogen.

Because specific morphological features—notably non-keratinizing basaloid morphology—have been described in virus-induced tumors [36], we compared morphological features of MCPyV-positive and MCPyV-negative tumors (data not shown). Although we found no statistically significant difference because of the small number of cases, we detected MCPyV in the two tumors containing squamous components with basaloid morphology, which suggests possible interactions between viral status and morphology.

MCPyV integration has been found in the main oncogenic event in MCC, but the nature of the cell in which this integration occurs is unknown. Of note, in contrast to conventional MCC, combined MCC with squamous differentiation showed an epidermal connection in all of our cases. The development of combined tumors implicates a common progenitor cell; as previously mentioned, Bowen’s disease is an epidermic intraepithelial neoplasm often present in combined MCC. Considering these two findings, combined MCC might originate from an epidermal cell. In addition, keratinocytes have been suggested as precursors of Merkel cells [37]. These considerations led us to hypothesize that MCPyV infection in a keratinocyte could be the first oncogenic event in combined MCC. Premature occurrence and accumulation of UV-induced somatic mutations could secondarily lead to the loss of the virus in a portion of the cell population and to squamous differentiation in the intraepithelial component. At the same time, the MCC component could invade the dermis and further undergo progressive loss of MCPyV. Hence, advanced combined MCC would feature very low load or lack of MCPyV, high somatic mutation rate, lack of T cell response and impaired outcome. In this way, our results do not rule out that combined MCC shares the same genetic background and behavior as other MCC types with negative-status or low-load MCPyV but may improve our understanding of their oncogenesis.

Our study has some limitations, owing to the low number of combined MCC cases, which is inherent to the low incidence of the tumor. The frequent MCPyV negativity found on immunostaining in combined MCC—currently considered the main tool for viral status determination [8]—led us to use molecular detection procedures, which implies sensitivity bias management, as discussed previously. Nonetheless, our results were validated by bimodal molecular procedures and positive and negative relevant controls.

To conclude, we detected MCPyV in half of our combined MCC cases, with a lower viral load than with conventional MCC, which suggests a shared oncogenesis between both MCC variants. The impact of MCPyV on the oncogenesis and behavior of the tumor remains to be determined.