Introduction
Uveal melanoma (UM) is the most common primary intraocular malignancy in Caucasian adults and may lead to metastatic disease in up to 50% of patients [
1,
2]. Current treatments are hardly ever effective against metastases [
3], and hence, most research efforts are focused on the development of targeted therapies or immunotherapeutic approaches, such as treatments with immune checkpoint inhibitors, vaccination, or adoptive T cell therapy [
4‐
8].
Some UM express increased levels of human leukocyte antigen (HLA) class I and are infiltrated by macrophages and lymphocytes, and this is known as an inflammatory phenotype [
9‐
12]. The presence of this inflammatory phenotype has been correlated with a specific genetic aberration, which is the loss of one copy of chromosome 3 (monosomy 3) [
13]. Other chromosomal abnormalities frequently occur in chromosomes 1, 6, and 8 [
14‐
17]. Following an initiating mutation in either GNAQ or GNA11, gain of 8q is thought to be one of the earliest genetic aberrations, followed by loss of one chromosome 3 [
18,
19]. Gain of 8q and monosomy 3 and are both associated with the development of UM metastases and a poor prognosis [
16,
20]. Similarly, gene expression analysis has been used to divide UM into two major classes, 1 and 2, which are good predictors of prognosis [
21,
22]. Moreover, we recently showed that, in our hands, class II tumors can be subdivided into IIa and IIb: while class IIa tumors are composed of highly homogeneous tumor cells and class IIb tumors contain a larger percentage of non-tumor cells which are likely to be immune cells. Interestingly, class IIa and IIb tumors differed in their numbers of chromosome 8q copies [
19].
Chromosome 3 contains the gene for
BRCA1-associated protein 1 (BAP1). In UM, inactivating hemizygous mutations in this gene have been found [
23,
24], which are associated with loss of BAP1 protein expression and a high metastatic risk [
23‐
26]. However, monosomy 3 and loss of BAP1 may occur independently, as tumors with a normal chromosome 3 status with lack of BAP1 expression have been identified, as well as tumors with monosomy 3 that still express BAP1. Such atypical tumors show high-risk clinico-pathological features and convey an increased metastatic risk [
25]. BAP1 is a member of the ubiquitin-carboxy-terminal hydrolase (UCH) family [
27], and is also known as ubiquitin carboxyl-terminal hydrolase L2 (UCHL2). Another member of the UCH family, ubiquitin carboxyl-terminal hydrolase L1 (UCHL1), is associated with suppressed production of pro-inflammatory chemokines and cytokines in keratinocytes [
28]. We, therefore, hypothesized that loss of BAP1 expression might be related to macrophage and/or T cell infiltration.
To assess whether genetic alterations affect immune cell infiltration in UM, we studied the presence and type of tumor-infiltrating immune cells in UM subtypes consisting of typical cases of UM (e.g. disomy 3 tumors with BAP1 expression and monosomy 3 with loss of BAP1 expression) and atypical cases of UM (e.g. disomy 3 tumors with loss of BAP1 expression and monosomy 3 tumors with expression of BAP1). In addition, we studied disomy 3/BAP1-positive cases with and without extra copies of chromosome 8q.
Our data show that loss of BAP1 protein expression is predominantly related to T cell infiltration in UM, while early gain of chromosome 8q is associated with macrophage infiltration.
Discussion
We show that the presence of extra copies of chromosome 8q in UM is associated with macrophage infiltration, while loss of BAP1 protein expression, with or without loss of chromosome 3, is associated with T cell infiltration in UM.
The chromosomal evolution of aggressive UM is thought to start with a mutation in GNAQ/GNA11 [
36,
37], followed by gain of chromosome 8q that precedes a potential loss of one copy of chromosome 3 and/or mutation in the
BAP1 gene [
20,
38]. We show that in UM with disomy 3 and expression of BAP1, the presence of additional copies of chromosome 8q is highly associated with the increased expression of macrophage-attracting chemokines and a stronger macrophage infiltration. In this subgroup, no effect was found with respect to the production of chemokines associated with T cell infiltration. This phenomenon could not be assessed in monosomy 3 tumors as more than 90% of monosomy 3 UM carry extra copies of 8q. Thus, a gain in copy number of chromosome 8q is associated with an increase in macrophage infiltration.
One might expect that this influx is initiated by activation of the c-Myc gene, a proto-oncogene located on chromosome 8q24, which is upregulated in many types of cancer and has been studied in UM [
18,
39,
40]. It has previously been suggested that c-Myc may be involved in the activation of inflammatory mediators in the tumor microenvironment [
41]. However, we previously observed the opposite, i.e., an association between a high c-Myc expression and a low inflammatory phenotype, making it unlikely that c-Myc is the relevant factor [
42].
Monosomy 3 and loss of BAP1-protein expression are strongly correlated in UM, but we observed several atypical cases which allowed us to separately assess the contribution of chromosome 3 and BAP1-protein expression on the magnitude and type of immune cell infiltration and to pinpoint that it was the loss of BAP1 which was associated with the higher expression of T cell-attracting chemokines and a stronger T cell infiltration in UM. The previous studies have reported that tumor suppressor proteins can be involved in the processes and pathways of tumor-promoting inflammation by interacting with transcription factors such as nuclear factor-κB (NF-κB) [
43]. NF-κB regulates genes which are involved in inflammation and immune responses. A close family member of BAP1 is UCHL1. Similar to BAP1, UCHL1 functions as a tumor suppressor protein [
44] and was recently shown to suppress the NF-κB pathway, thereby negatively affecting the production of type 1 interferon and pro-inflammatory cytokines and chemokines, including CCL5 [
28]. We, therefore, hypothesize that BAP1 may have a similar function as UCHL1 and that loss of BAP1 alleviates the suppression pathways leading to activation of NF-κB, resulting in the production of cytokines and chemokines that attract tumor-specific T cells into UM.
Obvious correlations between genetic changes and the development of an immune infiltrate are not easy to find. Loss of function of several tumor suppressor genes (
p53, PTEN) due to genetic aberrations is known to be associated with inflammation [
45]. Interestingly, loss of BAP1 in an unusual cutaneous tumor, the atypical Spitz nevus, was associated with a higher presence of T cells [
46].
When looking at UM, one of the chemokines that was higher in BAP1-negative than BAP1-positive tumors was
CCL5; another chemokine that was almost significantly higher in BAP1-negative tumors was CXCL10. CCL5 and CXCL10 play a role in the recruitment of T cells [
47]. An influence of BAP1 on NF-κB and the additional release of pro-inflammatory multifunctional chemokines might explain why both macrophages and T cells are found in BAP1-negative UM, but this requires further research.
Another chemokine involved in T cell recruitment is CXCL12, which is the ligand for chemokine receptor CXCR4. Previously, it has been described that CXCR4 is involved in the migration of UM cells to the liver [
48,
49]. In contrast, another group reported that the expression of CXCR4 was, indeed, correlated with lymphocyte infiltration, but had no prognostic relevance in UM patients [
33]. We observed discrepant results for CXCR4 in our Leiden cohort, with one probe showing a higher expression in BAP1-negative tumors, and another one a lower expression (Table
2a).
Previously, it had been shown that tumor-intrinsic active β-catenin signaling restrains tumor-infiltration by T cells, resulting in the escape of tumors from immune surveillance [
50]. The β-catenin protein is encoded by the gene
CTNNB1, which, like
BAP1, is located on chromosome 3p21. Hence, loss of chromosome 3 may also reduce β-catenin expression. In contrast to
BAP1 [
24,
25], however, there is no evidence that the other allele of
CTNNB1 is frequently mutated in UM and thus that the loss of β-catenin signalling is underlying T cell infiltration in UM. In our cohort, we found no correlation between
CTNNB1 expression and the amount of CD8
+ T cells. Furthermore, the number of β-catenin-positively staining tumor cells in UM is around 10% [
51], making its expression an unlikely explanation for the absence of T cell infiltration in most UM.
Our current findings show that alterations in copy numbers or mutations in certain genes can drive a specific type of immune response. As the most common treatment of UM is irradiation and not enucleation, we wondered whether local treatments might affect immune infiltration in UM. No tumors in either the Leiden cohort or the TCGA cohort had received prior irradiation. A previous study from our group showed that more T cells were present in secondarily enucleated eyes after prior irradiation compared to primarily enucleated eyes [
52]. As irradiation influenced our chromosome testing, we could not always analyze the chromosome status in previously irradiated tumors [
53]. We do not yet know how the type of inflammation or irradiation influences the patient’s response to immunotherapy, which at this moment has not been very successful in UM.
In conclusion, we provide evidence that the magnitude and type of immune cell infiltration observed in the subgroup of inflamed UM co-evolve with the sequential genetic changes occurring in UM. The initial infiltration by macrophages is related to a gain in the copy number of chromosome 8q, while additional T cell infiltration is correlated to a loss of functional BAP1-protein expression.