Background
Merkel cell carcinoma (MCC) is a neuroendocrine skin tumor with a high potential to metastasize. UV-radiation and Merkel cell polyomavirus (MCV) infection contribute to oncogenesis in MCC. [
1] The viral oncogenic pathway accounts for the large majority of MCC tumors, as 80% of the tumors are MCV-positive. [
2] Cumulative evidence suggests two MCC subgroups with and without MCV infection. [
3‐
7].
Clonal integration of the MCV DNA into the tumor cell genome as well as mutations in the viral Large T Antigen enable oncogenic transformation in MCV-positive MCC. [
2,
8] The non-viral pathway preceding MCV-negative MCC is less understood. Recent studies show that MCV-negative MCC tumors have a much higher mutational burden than MCV-positives. However, none of the earlier studies on driver mutations in MCV-negative tumors have succeeded in identifying prominent mutation. [
9‐
13] While many mutational studies concerning MCC have been carried out in recent years, little is known about the expression of cancer genes at the RNA level.
In the current study, we aspired to examine the RNA expression of 50 known cancer genes in MCC tumors. We used targeted next-generation sequencing to assess the RNA expression profiles of both MCV-positive and MCV-negative tumors. The fundamental aim of our study is to find potential molecular targets to improve the treatment strategies of MCC.
Methods
The Ethics Committee of Helsinki University Hospital approved the study and its plan. The Ministry of Health and Social Affairs granted permission to collect patient data, and the National Authority for Medicolegal Affairs to collect and analyze tissue samples.
From our pool of 270 formalin-fixed paraffin-embedded (FFPE) MCC tumor samples, we chose 13 MCV-negative and 13 MCV-positive tumors based on known MCV status and sufficient tumor sample available. MCC diagnoses were confirmed by morphology compatible with MCC in microscopy and by immunohistochemistry positive for CK-20 and negative for TTF-1. All tumor samples contained at least 50% of tumor tissue.
Patients
Twenty-six patients with a primary MCC tumor were included in this study, 19 females and 7 males. The mean age of the patients was 79 years (range 59–100). The mean tumor size was 33 mm. The study group was divided into two subgroups based on the MCV status of the tumor samples. The MCV status was determined as described in our previous study. [
14] Clinical data of the patients are presented in Table
1.
Table 1
Clinical data and ALK results
1 | P2 | pos | < 80 | Right cheek | |
+
|
+ −
|
−
|
2 | P3 | pos | ≥ 80 | Posterior thigh | 85 |
+
|
+ −
|
+
|
3 | P4 | pos | < 80 | Thorax | 70 |
+
|
+ −
|
−
|
4 | P5 | pos | < 80 | Right knee | 12 |
+
|
+ −
| |
5 | P9 | pos | ≥ 80 | unknown | 20 |
+
|
+ +
|
−
|
6 | N10 | neg | ≥ 80 | Right arm | 50 |
+
|
+ +
|
+
|
7 | N11 | neg | ≥ 80 | Left temple | 15 |
−
|
– –
|
−
|
8 | P12 | pos | ≥ 80 | Forehead | 40 |
+
|
+ −
|
−
|
9 | P13 | pos | < 80 | Right buttock | 34 |
+
|
+ −
|
−
|
10 | P14 | pos | ≥ 80 | Left cheek | 18 |
+
|
+ −
|
−
|
11 | N17 | neg | < 80 | Right cheek | 20 |
+
|
+ −
|
−
|
12 | N18 | neg | < 80 | Right breast | 20 |
+
|
+ −
| |
13 | N19 | neg | < 80 | Calf | 13 |
+
|
+ −
| |
14 | N21 | neg | < 80 | Neck | 25 |
+
|
+ −
|
–
|
15 | P23 | pos | < 80 | Left forearm | 40 |
+
|
+ −
|
−
|
16 | N25 | neg | ≥ 80 | Left back | |
+
|
+ −
|
+
|
17 | P26 | pos | ≥ 80 | Right shoulder | 24 |
+
|
+ −
|
−
|
18 | P28 | pos | ≥ 80 | Left arm | 75 |
+
|
−
| |
19 | N29 | neg | ≥ 80 | Back | 75 |
−
|
– –
|
−
|
20 | N31 | neg | < 80 | Left foot | 10 |
+
|
+ +
|
−
|
21 | N32 | neg | ≥ 80 | Right breast | 23 |
−
| (+) –
|
−
|
22 | N33 | neg | ≥ 80 | Front of left ear | 18 |
−
| (+) –
|
−
|
23 | P34 | pos | < 80 | Flank | 20 |
+
|
+ +
|
+
|
24 | N35 | neg | < 80 | Upper abdomen | 25 |
+
|
+ −
|
−
|
25 | P36 | pos | < 80 | Right buttock | 30 |
+
|
−
| |
26 | N37 | neg | ≥ 80 | Right cheek | 30 |
+
|
+ −
|
+
|
The total RNA was extracted from MCC tumor samples and two normal control skin samples. Extraction was performed according to the manufacturer’s manual using the miRNeasy mini Kit (Qiagen, Hilden, Germany). Qubit 2.0 Fluorometer (Thermo Fisher Scientific) and 2200 TapeStation System in combination with RNA ScreenTape assay (Agilent Technologies, Santa Clara, CA, USA) was used to measure the quantity and the quality of the RNA.
Targeted next-generation sequencing
Quantitative RNA expression analysis was performed by amplicon-based next-generation sequencing (NGS) using Ion Torrent technology (Thermo Fisher Scientific, Waltham, MA, USA). The Ion AmpliSeq™ RNA Library Kit (Thermo Fisher Scientific, Waltham, MA, USA) was used to construct the libraries from 20 ng of RNA. RNA was reverse transcribed, and targeted regions of RNA were PCR amplified using the Ion AmpliSeq™ RNA Cancer Panel (Thermo Fisher Scientific, Waltham, MA, USA) consisting of specific primers sets to amplify 50 target genes. (Additional file
1: Table S1) The Amplicons were then partially digested, and barcode adapters were ligated with the Ion Xpress™ Barcode Adapter Kit (Thermo Fisher Scientific, Waltham, MA, USA) to yield a barcoded library. Library concentrations were measured using the Qubit 2.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, USA).
Templates for sequencing were prepared using the Ion PGM™ Hi-Q™ OT2 Kit (Thermo Fisher Scientific) and Ion OneTouch™ 2 System (Thermo Fisher Scientific, Waltham, MA, USA). Ion Sphere™ particles were enriched with Ion OneTouch ES (Thermo Fisher Scientific, Waltham, MA, USA) and loaded onto an Ion 318™ Chip (Thermo Fisher Scientific, Waltham, MA, USA).
Sequencing was performed on the Ion Torrent PGM™ System using the Ion PGM™ Hi-Q™ Sequencing Kit (Thermo Fisher Scientific, Waltham, MA, USA).
Data analysis
Sequencing data was processed using Torrent Suite™ Software. The Coverage Analysis plugin was used to create amplicon counts. The mean length of amplicons was 150 bp. The amplicon count data created was imported into a Chipster [
15] (
http://chipster.csc.fi/index.shtml) for further differential gene expression analysis. Differential expression analysis, to compare expression differences between tumor and normal skin tissue and between MCV-positive and MCV negative tumors, was performed on read count data carried out with DESeq2. Differently expressed genes were determined from adjusted
p-values and log
2 fold change. To control for false positives, the
p-values were corrected for multiple testing and the adjusted
p-value or FDR (false discovery rate) calculated using Benjamini-Hochberg correction. An average of 344,458 reads (after quality check) were obtained for each sample with an average of 95% on target.
ALK Immunohistochemistry
The sections (3 μm) were stained with fully automated immunostainer Ventana Benchmark XT (Roche/Ventana, Tucson, AZ, USA). For both ALK antibodies we used heat- induced epitope retrieval buffer Cell Conditioning 1 (Roche/Ventana, 950–124) for 64 min in 98 °C. The dilutions and incubation times for ALK antibodies were: clone 5A4 (Novocastra™, Leica Biosystems, Wetzlar, Germany) 1:100 for 40 min/36 °C and clone D5F3 (Ventana/Roche, Tucson, AZ, USA) 28 min/36 °C. The three- step, multimer based detection kit, OptiView (760–700, Roche/Ventana), was used to detect the antibodies. Amplification step was added for both protocols by using separated amplification kit (Roche/Ventana, 760–099). The slides were finally stained with hematoxylin (Mayer, S3099, Dako, Glostrup, Denmark). For the control of the staining quality we used skin, appendix and known ALK positive and ALK negative tumor tissue. The stained slides were examined by researchers TB, TV, MK and VKS.
Fluorescence in situ hybridization (FISH)
FISH was performed on 2 μm thick FFPE tumor sections. The sections fixed on microscopic slides were de-paraffinized, pre-treated with protease and hybridized with Vysis LSI ALK Dual Color Break Apart FISH probes according to the vendor’s guidelines (Abbott Molecular Inc., Des Plaines, IL, USA) and as described previously. [
16] Results were checked under a fluorescence microscope. The criteria for considering a cell to be
ALK gene rearrangement positive was: presence of at least one green and orange signal pair split apart by ≥2 signal diameters (pair-signal type fusion), or a single orange without corresponding green signal (single-signal type fusion). The cells were considered
ALK fusion negative if they had fused or if they had close orange and green signal.
Discussion
In this study, we analyzed the RNA expression of 50 cancer-related genes in 26 MCC tumors by targeted next-generation sequencing and compared their expression with normal, non-cancerous skin samples. Among the 50 cancer-related genes, we identified eight genes (Table
2) that had differential expression in tumor tissue. Further confirmation of the results with quantitative PCR in larger tumor cohort and comprising more genes is required.
We recorded overexpression of cancer related genes including
ALK,
CDKN2A, and
EZH2 compared with normal skin. Among the under-expressed genes, we identified
EGFR, in concordance with our earlier study showing negative EGFR expression by immunohistochemistry in MCC tumors. [
9] Earlier studies have found inactivating
RB1 and
TP53 mutations driving MCV-negative tumors, [
10] however we did not find different expression of
RB1 or
TP53 in MCV-negative tumors compared to MCV-positive tumors.
In this study, we recorded for the first time
ALK overexpression in tumors of MCC patients.
ALK expression was seen in all patients with high levels in 22/26 tumor samples (Table
1, Fig.
2).
ALK overexpression at the mRNA level seen in our results fits well with the protein expression results of the only ALK IHC study on MCC [
17], whereby they demonstrated ALK expression in 30/32 MCC tumors by one (clone D5F3) of the three antibodies used in the study. They however noted that the frequency of ALK positive cases for MCC tumors depended a lot on the antibody used and found ALK positivity in 4/32 with the ALK1 clone and 28/32 with clone 5A4. In our IHC analysis we used the clones D5F3 and 5A4. Positive results were seen in 22/24 (Clone D5F3) and 4/26 (Clone 5A4) tumors, and only in those that had a high RNA expression. IHC staining was more even and vivid with clone D5F3 (Fig.
3). Therefore, our results correspond well with the previous ALK IHC study and it seems that clone D5F3 is the most sensitive clone in detecting ALK expression in MCC tumors.
ALK is normally expressed predominantly in central nervous system and it likely functions in development of the brain. ALK is however, known mainly for its role in various types of cancer. One common mechanism of ALK activation in tumors is
ALK gene rearrangement leading to fusion protein like NPM-ALK in anaplastic large cell lymphomas [
18] and EML4–ALK in non-small-cell lung cancer. [
19]. In order to study whether the mechanism behind the mRNA overexpression is a fusion gene or gene amplification, we performed the FISH analysis. As no fusions or high level amplification (only 2–3 times gain in 5 cases) were seen by FISH, the mechanism behind
ALK overexpression might possibly be epigenetic or as a result of over-activation of a transcription factor. Similar to our results, no rearrangement or other cytogenetic aberration of the
ALK gene have been reported in MCC. [
17].
ALK expression without any rearrangement or amplifications of the
ALK has, been reported in hepatocellular carcinoma associated with poor prognosis and occurrence of micrometastases [
20]. While no clear driving pathway in MCC has been identified, ALK is expressed in MCC tumors at both the RNA and protein level, and therefore, studying the mechanism and significance of this overexpression remains intriguing. Studying the correlation between ALK expression and survival in a larger MCC tumor cohort would be of future interest.
CDKN2A was another gene overexpressed in MCC tumors compared to normal skin. Our results are in concordance with our group’s previous study showing an expression of p16 (encoded by
CDKN2A) by IHC in 97.7% of 88 MCC tumors. [
21]
CDKN2A is frequently mutated or deleted in a wide variety of tumors, including malignant melanoma and MCC and is considered to be a tumor suppressor gene. [
9,
22,
23] p16 overexpression in malignant tumors is thought to be a mechanism to overcome proliferation resulting from the failure of the RB1 pathway due to viral infection, genetic/epigenetic silencing of
RB1 gene or other mechanism. [
24] p16 overexpression is additionally reported in HPV infection associated cervical dysplasia and carcinoma as well as in cervical neuroendocrine tumors. [
25‐
27] In MCC,
CDKN2A RNA expression and p16 protein expression were, however, independent from the tumors’ MCV-status, although the role of
RB1 in MCC is reported earlier. [
28].
The expression profile of MCC tumors also showed overexpression of
EZH2 that codes for an enzyme important in heterochromatin formation via DNA methylation. Mutations or overexpression of
EZH2 are seen in many forms of malignancies. [
29,
30] It is thought that overexpression of EZH2 inhibits the transcription of tumor suppressors, thus promoting malignant transformation. Inhibitors of EZH2 are in development, and their role in cancer treatment is being studied. In malignant melanoma, Zingg et al. recently found that EZH2 expression correlated with poor survival and promotes the initiation and progression of melanoma in mouse models as well as human cell cultures. Using both an RNA interference and preclinical drug GSK305 to target EZH2, they showed that EZH2 could be a promising therapeutic target in the treatment of melanoma patients. [
31] To our knowledge,
EZH2 expression is not previously reported in MCC tumors. It could be that EZH2 transcriptionally silences important tumor suppressors in MCC, similarly to its function in melanoma. Studying the role of the overexpression of
EZH2 in MCC oncogenesis would be of interest as a drug target.
Comparison of gene expression among the MCV –negative and positive groups showed a higher expression of
MET, SMO, FGFR3 and NOTCH1 in MCV-negative tumors compared to MCV positive tumors.
MET is a known target of microRNA-34a, and in our previous study, we had reported under-expression of this miRNA in MCV-negative MCC. [
7]. Notably, two genes,
JAK3 and
NPM1 were overexpressed in MCV-positive tumors compared to MCV-negative.
JAK3 is predominantly expressed in immune cells. To our knowledge,
JAK3 mutations have not been reported in MCC. Activating
JAK3 mutations are seen in hematological malignancies. Deficiency/defects in
JAK3 leading to low amounts of functional protein are associated with immune dysfunction/immunodeficiency. [
32]
NPM1 is involved in many processes and is a fusion partner with many genes, especially
ALK. Overexpression of
NPM1 is reported in many tumors including HCC, colon cancer and glioblastoma. [
33‐
36].
NPM1 associates with viral proteins of different viruses and is implicated in various stages of viral infection. [
37], and this might be a reason for its higher levels seen in MCV-positive tumors. The significance of the genes differentially expressed between the MCV-negative and positive groups is however not clear and needs further investigation.