Introduction
Primary leptomeningeal melanocytic neoplasms (LMNs) are infrequent tumors, forming a spectrum ranging from benign or low-grade malignant melanocytomas to frankly malignant melanomas [
1]. These tumors share molecular features with uveal melanomas (UMs). In contrast to cutaneous melanomas (CMs), both LMNs and UMs frequently carry mutations in the G protein encoding genes
GNAQ and
GNA11, whereas mutations in
BRAF and in the
TERT promoter are infrequent [
2‐
8]. This situation reflects the heterogeneous molecular background of different groups of melanoma and has important implications for targeted therapy.
In the past years, inactivating mutations in the tumor suppressor gene
BAP1 (BRCA-associated protein 1) were shown to be implicated in UM [
9,
10]. The
BAP1 gene is located on chromosome 3p21.1 and encodes a nuclear ubiquitinase involved in epigenetic modulation of chromatin [
11]. Somatic
BAP1 mutations are predominantly present in UMs with monosomy 3 (~85 %), the latter being a strong predictor for metastatic disease [
9,
10,
12]. In this setting
BAP1 functions as a tumor suppressor gene, with loss of one copy of chromosome 3 and mutation in the other
BAP1 allele representing the two hits causing inactivation of this gene. Indeed, in UMs with disomy 3 (and a good prognosis), mutations in
BAP1 are rare [
6,
9,
10]. A small proportion of patients with UM (~2–3 %) harbor a germline mutation in
BAP1 [
13]. These patients suffer from the
BAP1 hereditary cancer syndrome and have an increased risk of developing cutaneous melanocytic tumors as well as a spectrum of non-melanocytic neoplasms including mesothelioma, renal cell carcinoma, meningeoma, and adenocarcinoma of the lung [
14,
15]. Very recently, it was suggested that primary leptomeningeal melanoma is part of this cancer predisposition syndrome as well [
16].
Furthermore, recurrent hotspot mutations in the
SF3B1 gene (mainly at codon 625) and mutations of the
EIF1AX gene (spread over exon 1 and 2) were recently reported in UMs, especially in tumors with disomy 3 (up to 30 and 50 % of disomy 3 tumors, respectively) [
17‐
19]. These mutations in UMs appeared to be largely mutually exclusively with
BAP1 mutations, while in CMs these mutations were found to be very infrequent (~1 %) [
20].
It is currently unknown whether somatic mutations in BAP1, SF3B1 and EIF1AX also characterize primary LMNs. Using Sanger sequencing, we searched for mutations in hotspot regions of SF3B1 (exon 14 and 15) and exon 1 and 2 and flanking intronic regions of EIF1AX in a series of 24 primary LMNs. Additionally, we performed immunohistochemistry for the detection of BAP1 protein loss as a surrogate marker for identification of inactivating BAP1 mutations.
Discussion
In this study we investigated whether primary LMNs share genetic alterations with UMs in addition to
GNAQ and
GNA11 mutations. Recently, a role for
BAP1 in primary melanoma of the CNS was suggested based on the identification of a
BAP1 germline mutation in a patient with primary CNS melanoma with monosomy 3 and a family history of UM and meningioma [
16]. In our study, we chose for BAP1 immunohistochemistry as a surrogate marker for the identification of an underlying inactivating
BAP1 mutation as it offers an economical and faster alternative to sequence analysis of all 17 exons of
BAP1 [
12,
26,
27]. Typically, complete loss of nuclear BAP1 expression is found in
BAP1-mutated UMs [
12,
26‐
28]. None of the samples in our series showed such complete loss of nuclear BAP1 staining, suggesting absence of underlying
BAP1 mutations. In three cases (including the single patient with monosomy 3 in the LMN and the patient with liver metastases but disomy 3 in the LMN),
BAP1 mutation status was available through Sanger sequencing analyses and confirmed absence of mutations. However, our study has some limitations. In some cases,
BAP1 mutations may still have been missed as BAP1 immunohistochemistry was reported to have a sensitivity of ~ 88 % [
12]. Also, very rarely, heterogeneous (‘mosaic’) BAP1 nuclear immunostaining has been described in UM cases with a
BAP1 mutation showing loss of nuclear staining in only 20 % of nuclei [
12]. In our series, two cases showed nuclear BAP1 staining in about 80 % of tumor cells and an underlying
BAP1 mutation can thus not completely be ruled out. Furthermore, as
BAP1 is mainly implicated in metastatic UM with monosomy 3, there might be a selection bias in our patient group as it mainly concerns (relatively) low-grade tumors with disomy for chromosome 3. A larger number of primary LMNs with monosomy 3 should thus be investigated to further explore the role of
BAP1 in these neoplasms. This is also important for therapeutic reasons as epigenetic modulators such as histone deacetylase (HDAC) inhibitors were shown to reverse the biochemical effects of
BAP1 mutations in UM cells by inducing growth arrest and differentiation. Clinical trials are now evaluating HDAC inhibitors as a therapeutic option in UM patients [
13,
29].
Recurrent mutations in the
SF3B1 gene have been detected in several types of cancer such as UMs, breast and pancreatic carcinoma, and hematological diseases like CLL and MDS [
18,
30‐
33]. These mutations affect hotspot codons, the hotspot being associated with cancer type. For example, codon 700 mutations are frequently present in CLL and MDS, while in UMs codon 625 is much more frequently involved [
18,
34]. Especially in low-grade UMs with disomy for chromosome 3, heterozygous mutations in
SF3B1 are present in 10 to 30 % of UM [
6,
17‐
19]. The fact that in UMs these
SF3B1 mutations are almost mutually exclusive with
BAP1 mutations suggests different pathways in the oncogenesis and/or malignant progression of these neoplasms.
SF3B1 encodes subunit 1 of splicing factor 3b, which is a component of the spliceosome that participates in splicing of pre-mRNA. It was shown that
SF3B1 mutations are associated with differential alternative splicing of several protein encoding genes in UMs [
19]. Moreover,
SF3B1 mutant cell lines were found to be sensitive to the SF3b complex inhibitor spliceostatin A, suggesting a new therapeutic target in tumors carrying this mutation [
32]. Up to now, it was unknown whether
SF3B1 mutations also occur in primary LMNs. In our series, in three out of 24 cases (13 %) an
SF3B1 mutation affecting codon 625 or 634 was detected, this is at the lower end of the range of the frequency of
SF3B1 mutations reported in UMs (10–30 %) and in contrast to the very low frequency reported for CMs (~1 %) [
6,
17‐
20]. As far as could be assessed, all three LMNs showing an
SF3B1 mutation were tumors with disomy for chromosome 3. Furthermore, like in UMs, in two cases the
SF3B1 mutation co-occurred with a
GNAQ or
GNA11 mutation, while in a third case (with neurocutaneous melanocytosis) an
NRAS mutation was present. The
GNAQ,
GNA11 and
NRAS genes are now thought to play a role in the initiation of tumorigenesis, while mutation in
SF3B1 (or
BAP1) would then occur in a later phase of the oncogenic process [
18]. The co-occurrence of an
NRAS and an
SF3B1 mutation in the CNS melanoma of the neurocutaneous melanocytosis patient in our study is interesting in this respect as the
SF3B1 mutation was absent in the congenital melanocytic nevus of this patient. Patients with neurocutaneous melanocytosis have a large and/or multiple congenital melanocytic nevi of the skin in association with a primary LMN [
1]. Instead of
GNAQ or
GNA11 mutations, these patients frequently demonstrate identical
NRAS mutations in both the congenital melanocytic nevus and in the LMN [
35,
36]. This is thought to be the result of an
NRAS-mutated clone of melanocyte precursors migrating to skin and CNS early in embryogenesis [
37]. The observation that the
SF3B1 mutation was not present in the melanocytic nevus of this patient suggests that it indeed plays a role later on in tumorigenesis. In addition, all three mutations in
SF3B1 in our study were found in tumors which clinically showed aggressive behavior. However, the number of patients and the follow-up in our study are too limited to allow for firm conclusions about a prognostic role of
SF3B1 in this setting. Of note, in one of the three patients the
SF3B1 mutation (c.1900G > A (p.(Val634Ile))) was present in a neoplasm diagnosed as melanocytoma, suggesting that
SF3B1 mutations are not necessarily associated with worrisome histology.
Several recent studies have reported mutations in
EIF1AX in different cancer types, including melanoma and thyroid and ovarian cancer [
17,
38,
39]. In UMs, heterozygous mutations in exon 1 and 2 of
EIF1AX have reported to occur especially in tumors with disomy for chromosome 3 (up to 48 % of these tumors) [
17]. Mutations may occur in different loci of these exons and lead to amino acid substitutions or short deletions. Rarely splice site mutations have been reported [
6,
10,
17].
EIF1AX encodes the eukaryotic translation initiation factor 1A (eIF1A), which is involved in initiation phase of translation of eukaryotic cells by stabilizing the formation of the functional ribosome around the AUG start codon. The exact role of
EIF1AX mutations in tumorigenesis is currently not well understood but it has been suggested that mutations in
EIF1AX could diminish the rate of bulk translation [
17,
40]. In our series of primary LMNs we found a relatively high frequency of
EIF1AX missense mutations (21 %) which is in the range reported for UM (19–48 %) [
6,
17]. In contrast,
EIF1AX mutations are very rare in CM (5/231, ~2 %) [
20]. As in UMs, as far as could be assessed in our series, these mutations occurred in primary LMNs with disomy for chromosome 3 and were mutually exclusive with
SF3B1 mutations, but co-occurred with
GNAQ or
GNA11 mutations. In addition (and like
SF3B1 mutations), the
EIF1AX mutations in our cases occurred both in melanocytomas as well as melanomas, suggesting that they are not necessarily associated with worrisome histology, but the prognostic implications of these mutations remain to be elucidated. Finally, the
EIF1AX and
SF3B1 mutations in LMNs occurred in hotspot regions of these genes, and in UMs such mutations were shown to be somatic in origin [
17,
18]. However, as non-neoplastic tissue of the patients with LMNs was not available for further testing, strictly speaking we cannot rule out the possibility that in some of these cases it concerned a germline mutation.
Competing interests
The authors declare that they have no competing interest.
Authors’ contribution
HK, CP, WB, PW, BK, and PG contributed substantially to conception and design of the study. All authors contributed substantially to acquisition of data, and analysis and interpretation of data. HK wrote and edited the manuscript. All authors contributed in drafting the manuscript and revising it for intellectual content. MJ carried out the molecular genetic analyses. CP supervised the molecular genetic analyses. DC, VW, WB, BK, and HK supervised and scored the BAP1 immunohistochemistry. HK revised histology. All authors read and approved the final version of the manuscript.