BAP1 Loss Promotes Suppressive Tumor Immune Microenvironment via Upregulation of PROS1 in Class 2 Uveal Melanomas

Simple Summary Uveal melanoma is a highly metastatic cancer of the eye which is notoriously resistant to therapy. Elucidating the mechanisms of metastasis in order to devise effective therapies has been a major challenge. The strongest genetic risk factor for metastasis in uveal melanoma is the mutational inactivation of the BAP1 tumor-suppressor gene. However, it remains unknown how BAP1 loss promotes tumor progression. Here, we show that BAP1 loss leads to increased expression of PROS1 in uveal melanocytes and melanoma cells, which in turn leads to phosphorylation and activation of the receptor tyrosine kinase MERTK on adjacent macrophages, driving them into a suppressive M2-polarized state. This mechanism could help explain the suppressive tumor immune microenvironment that is characteristic of BAP1-mutant uveal melanomas, and it suggests that BAP1 loss may lead to metastasis at least in part by facilitating immune escape. These findings provide new insights into the role of BAP1 in uveal melanoma, and they nominate new strategies for increasing the efficacy of immunotherapy in this cancer. Abstract Uveal melanoma (UM) is the most common primary cancer of the eye and is associated with a high rate of metastatic death. UM can be stratified into two main classes based on metastatic risk, with class 1 UM having a low metastatic risk and class 2 UM having a high metastatic risk. Class 2 UM have a distinctive genomic, transcriptomic, histopathologic, and clinical phenotype characterized by biallelic inactivation of the BAP1 tumor-suppressor gene, an immune-suppressive microenvironment enriched for M2-polarized macrophages, and poor response to checkpoint-inhibitor immunotherapy. To identify potential mechanistic links between BAP1 loss and immune suppression in class 2 UM, we performed an integrated analysis of UM samples, as well as genetically engineered UM cell lines and uveal melanocytes (UMC). Using RNA sequencing (RNA-seq), we found that the most highly upregulated gene associated with BAP1 loss across these datasets was PROS1, which encodes a ligand that triggers phosphorylation and activation of the immunosuppressive macrophage receptor MERTK. The inverse association between BAP1 and PROS1 in class 2 UM was confirmed by single-cell RNA-seq, which also revealed that MERTK was upregulated in CD163+ macrophages in class 2 UM. Using ChIP-seq, BAP1 knockdown in UM cells resulted in an accumulation of H3K27ac at the PROS1 locus, suggesting epigenetic regulation of PROS1 by BAP1. Phosphorylation of MERTK in RAW 264.7 monocyte–macrophage cells was increased upon coculture with BAP1−/− UMCs, and this phosphorylation was blocked by depletion of PROS1 in the UMCs. These findings were corroborated by multicolor immunohistochemistry, where class 2/BAP1-mutant UMs demonstrated increased PROS1 expression in tumor cells and increased MERTK phosphorylation in CD163+ macrophages compared with class 1/BAP1-wildtype UMs. Taken together, these findings provide a mechanistic link between BAP1 loss and the suppression of the tumor immune microenvironment in class 2 UMs, and they implicate the PROS1–MERTK pathway as a potential target for immunotherapy in UM.


Introduction
Uveal melanoma (UM) is the most common primary cancer of the eye and is often associated with fatal metastasis [1]. UM can be stratified into two prognostically significant subtypes, with class 1 UM having a low metastatic risk and class 2 UM having a high metastatic risk [2,3]. Biallelic inactivation of the tumor-suppressor BAP1 is the distinctive genomic feature of class 2 UM, which occurs through the mutation of one allele and the whole-chromosome loss of the other allele [4]. While we and others have extensively investigated the downstream effects of BAP1 loss in UM cells [5][6][7], the mechanism by which BAP1 loss leads to metastasis remains unclear.
Despite the increasingly effective management of primary UM, there has been no corresponding improvement in patient survival [8], due to a propensity for early micrometastasis and immune escape [9]. UM is an immunoresistant tumor type that responds poorly to checkpoint-inhibitor immunotherapy [10]. This is due, at least in part, to a suppressive tumor immune microenvironment (TIME), including an enrichment of M2 polarized (M2) macrophages, T cells expressing checkpoint or "exhaustion" markers, and increased HLA expression on tumor cells [11][12][13][14][15]. Importantly, the suppressive TIME in UM is strongly associated with mutational inactivation of BAP1 [11,12,16], suggesting a mechanistic link between genomic aberrations and immune suppression.
While it is now widely recognized that the TIME plays a critical role in cancer progression and therapeutic resistance [17], how the tumor genome shapes the TIME remains poorly understood. Here, we used an integrative approach to explore the mechanistic relationship between BAP1 mutations and the TIME in UM.

Cell Lines
UMC026 cells with and without CRISPR-Cas9-mediated deletion of the first exon of BAP1 (BAP1 −/− ) were established and cultured in our laboratory from normal uveal melanocytes in a patient undergoing enucleation, as previously described [18]. RAW 264.7 monocyte-macrophage-like cells (ATCC, Manassas, Virginia) were grown in DMEM media with 10% Tet-Free FBS at 37 • C in 20% oxygen and 5% CO 2 . MP41 and MP46 cells were gifts from Dr. Sergio Roman-Roman [19] and were cultured as previously described [20]. Mel202 and 92.1 were gifts from Drs. B. Ksander and M. Jager, respectively. UMC026, 92.1, Mel202, and MP41 were engineered to allow tetracycline-inducible knockdown of BAP1 as previously described [21]. The BAP1-mutant cell line MP46 UM was engineered to allow for tet-inducible expression of exogenous BAP1 by stable lentiviral integration of pLV-TET-HA-BAP1-WT and pLVX-TET-ON (Clontech) constructs. pLV-TET-HA-BAP1-WT was created by PCR amplification and subsequent recombination of a full-length BAP1 cDNA fragment into pLV-TET-PURO (Addgene #26430). The plasmids were packaged into lentiviral particles by transient co-transfection into HEK293T cells with pMD2G and psPAX2 packaging plasmids using JetPrime reagent (Polyplus), and the virally transduced MP46 cells selected with puromycin (2 µg/mL) and geneticin (500 µg/mL). RNA-seq was performed as previously described [21]. Briefly, RNA was isolated with Direct-zol RNA kit (Zymo), and melanin pigment was removed using OneStep PCR Inhibitor Removal Kit (Zymo), according to the manufacturers' instructions. Libraries were prepared and sequencing was performed by the Sylvester Comprehensive Cancer Center Oncogenomics Shared Resource at the University of Miami. Sequencing quality was assessed using FastQC (v0.11.3, Simon Andrews, Cambridge, UK). Reads were trimmed using Trim Galore (v0.6.5, Felix Krueger, Cambridge, UK) [22], aligned to the human genome builder hg38/GRCH38 using STAR (v2.5, Alexander Dobin, CA, USA) [23] and counts were generated using RSEM (v1.3.3, Bo Li and Colin Dewey, Boston, MA, USA) [24]. Venn diagram was generated by overlap of RNA-seq datasets included genes with >30% upregulation, and FPKM > 10. Significance of PROS1 expression in shBAP1 and BAP1 tumor samples was calculated with a ratio-paired t-test of normalized read counts across all datasets.

ChIP-Seq
Chromatin immunoprecipitation (ChIP) followed by next-generation sequencing (ChIP-seq) was performed using 20 million cells per experiment, which were crosslinked for 7 min with 1% formaldehyde. Chromatin was sonicated to an average fragment size of 200-500 base pairs with a Covaris M220 sonicator. A measure of 10 µg of antibody was used for each ChIP experiment. Libraries were prepared using the NEBNext Ultra 2 kit and sequenced by the University of Miami Oncogenomics Shared Resource with >20 million reads per sample. Reads were quality filtered by Trim Galore! [22] and aligned to the hg38 genome with Bowtie2 [25]. Normalized coverage tracks were generated with MACS2 [26], and plotted with SparK [27].

Single-Cell RNA Sequencing (scRNA-Seq)
A total of 59,915 cells from 11 UM samples were analyzed by scRNA-seq as previously described [11]. Using the first 20 principal components of variably expressed genes, dimensionality reduction was conducted with Seurat (version 4.1.0, Yuhan Hao, New York, NY, USA) using Uniform Manifold Approximation and Projection (UMAP) methodology [28,29]. Cell type annotations were assigned as previously described [11]. Differentially expressed genes were identified using Wilcoxon Rank Sum test of log normalized counts. Dimensional reduction plots, dot plots, and scatter plots were generated using Seurat. Significance of correlation between the expression of genes was conducted in R using Spearman's rank-based measure of association.  Table S1). A total of 91,532 cells were analyzed and the percentage of cells staining positive for BAP1, PROS1, CD163, and Phospho-MERTK was determined. Additionally, the percentage of cells that were co-stained for CD163 and phospho-MERTK was determined.

ELISA
Media was collected from cell culture plates following 72 h incubation of UMC026 cells with and without BAP1 knockout. Media samples (1 mL) were centrifuged at 1000× g for 10 min, and 800 µL was removed and frozen at −80 • C. Cell count was determined for each culture plate by trypsinization of cells (2 mL, 0.5% trypsin) and quantification by spectrophotometer. PROS1 protein quantification was performed using a commercially validated, enzyme-linked immunosorbent assay (Catalogue # MBS9427967, MyBioSource, Inc., San Diego, CA, USA) for three different sets of media from UMC026 cells with and without BAP1 knockout. ELISA measurements were performed in triplicate using a VersaDoc spectrophotometer instrument (Bio-Rad; Hercules, CA, USA).

Confocal Immunocytochemistry
UMC026 cells with and without BAP1 knockout were grown in culture on glass coverslips coated with poly-L-lysine. Cells were fixed in 4% paraformaldehyde (v/v in PBS) for 10 min, washed 2× in PBS and incubated in 0.8% glycine (v/v in PBS) for 10 min. Cells were then permeabilized for 30 min in 0.05% Tween-20 (v/v in PBS), incubated in 0.27% ammonium chloride (v/v in PBS) for 10 min, and washed 3× in PBS. Cells were blocked with 5% BSA, 1% NGS and 0.5% Tween-20 (v/v in TBS) for 1 h and washed 3× in TBS. Cells were incubated in anti-PROS1 primary antibody (Proteintech, Rosemont, Illinois, 16910-1-AP) overnight at 4 • C, washed 3× in TBS and then incubated in secondary antibody (Cell Signaling Technology, Danvers, MA, USA, 7074P2) for 1 h at room temperature. Samples were then washed 3× in TBS, mounted with SlowFade Diamond Antifade mounting medium with DAPI (Thermo Fisher Scientific, Waltham, MA, USA), and imaged using an SP8 Leica laser scanning confocal microscope (Leica Microsystems Inc, Buffalo Grove, IL, USA).

Statistical Analysis
A two-tailed t-test was used for continuous data and chi square analysis for categorical data, using GraphPad Prism software version 8.0 for Windows, GraphPad Software (San Diego, CA, USA, www.graphpad.com, accessed on 15 March 2022). For scRNA-seq differential expression analysis, two-sided nonparametric Wilcoxon rank sum test with Bonferroni correction using all genes was used. Pearson correlation was used for correlating expression between genes.

Accession Codes
Submission of the RNA-Seq and ChIP-seq data generated in this study to the Gene Expression Omnibus is in process.

BAP1 Regulates PROS1 by Epigenetic Mechanisms
To identify the genes regulated by BAP1, we analyzed RNA-seq data for differentially expressed genes in three UM cell lines (92.1, Mel202, MP41) and one cell line derived from normal human uveal melanocytes (UMC026) following shRNA knockdown of BAP1, and in 80 human UMs from The Cancer Genome Atlas (TCGA) database that were wildtype versus mutant for BAP1. By plotting genes with 30% increase in FPKM and a minimum of 10 FPKM, we found that only two genes were upregulated across all of these datasets, PROS1 and GDF15 ( Figure 1A), both of which have been linked to immunosuppressive macrophage polarization [31,32]. Upregulation was significant for both PROS1 (p = 0.0027) and GDF15 (p = 0.0194) when comparing BAP1-competent samples with BAP1-deficient samples across all datasets. Here, we focused on PROS1 and its potential role in suppressing the TIME in BAP1-deficient UM. Deletion of BAP1 in UMC026 cells and knockdown of BAP1 in Mel202 (class 1) cells resulted in upregulation of PROS1, whereas ectopic expression of BAP1 in BAP1-deficient MP46 (class 2) cells resulted in downregulation of PROS1 ( Figure 1B and Supplementary Figures S1 and S2). Furthermore, knockdown of BAP1 resulted in increased acetylation of histone H3 at lysine 27 (H3K27ac) around the PROS1 locus ( Figure 1C). These findings suggest that BAP1 represses PROS1 at least in part by epigenetic mechanisms, and that BAP1 loss leads to increased PROS1 expression.

Single-Cell Sequencing Analysis of PROS1 and MERTK in Uveal Melanomas
To further explore the relationship between BAP1 and PROS1, we analyzed single-cell RNA sequencing from 11 human UM samples, as previously described [11] (Figure 2A). There was a strong association between BAP1-mutant class 2 tumors and PROS1 expression (adj. p-value < 10 −300 ) ( Figure 2B). We then performed a further analysis limited to tumor-associated macrophages, which clustered according to GEP class ( Figure 2C). Macrophages derived from class 2 tumors demonstrated significantly increased expression

Single-Cell Sequencing Analysis of PROS1 and MERTK in Uveal Melanomas
To further explore the relationship between BAP1 and PROS1, we analyzed singlecell RNA sequencing from 11 human UM samples, as previously described [11] (Figure  2A). There was a strong association between BAP1-mutant class 2 tumors and PROS1 expression (adj. p-value < 10 −300 ) ( Figure 2B). We then performed a further analysis limited to tumor-associated macrophages, which clustered according to GEP class ( Figure 2C). Macrophages derived from class 2 tumors demonstrated significantly increased expression of the macrophage receptor tyrosine kinase MERTK (adj. p-value = 1.5 × 10 −9 ) ( Figure  2D), which was strongly associated with expression of the M2 polarization marker CD163 (adj. p-value 1.1 × 10 −12 ) ( Figure 2E).

Validation of PROS1 Upregulation and MERTK Activation in Class 2 Uveal Melanomas
Consistent with these findings, multicolor immunohistochemistry in human class1/BAP1wildtype and class2/BAP1-mutant UM samples revealed a significant association between BAP1 loss and PROS1 upregulation in UM cells and MERTK phosphorylation in tumorassociated macrophages (Table 1, Figures 3A and S3-S6). Furthermore, macrophages expressing the M2 polarization marker CD163 were enriched in BAP1-mutant UM samples as previously reported [33]. In BAP1-mutant UM samples, the subset of CD163-positive cells also positive for phospho-MERTK (double-positive cells) was significantly higher than in class 1 UM cells ( Figure 3B), consistent with increased MERTK signaling in immunosuppressive macrophages of BAP1-mutant vs. BAP1-wildtype UM.  ciation between BAP1 loss and PROS1 upregulation in UM cells and MERTK phosphory-lation in tumor-associated macrophages (Table 1, Figure 3A and Supplementary Figures  S3-S6). Furthermore, macrophages expressing the M2 polarization marker CD163 were enriched in BAP1-mutant UM samples as previously reported [33]. In BAP1-mutant UM samples, the subset of CD163-positive cells also positive for phospho-MERTK (doublepositive cells) was significantly higher than in class 1 UM cells ( Figure 3B), consistent with increased MERTK signaling in immunosuppressive macrophages of BAP1-mutant vs. BAP1-wildtype UM.  Table 1.  Table 1.

PROS1 Upregulation following BAP1 Loss Triggers MERTK Phosphorylation in Macrophages
PROS1 can function as a secreted protein, and it can also be localized to the cell membrane and cytoplasm [34][35][36][37]. In UMC026 cells, knockout of BAP1 did not result in an increase in secreted PROS1 in the cell culture media (Figures 4A and S7), whereas we found PROS1 localized to the cell membrane and cytoplasm ( Figures 4B and S8), suggesting that it may function through cell-cell interaction rather than paracrine signaling in this setting. PROS1 is known to promote M2 macrophage polarization by stimulating phosphorylation of the MERTK receptor [31,[37][38][39]. Thus, we performed coculture experiments to study the effect of BAP1 loss in UMC026 cells on MERTK phosphorylation in Cancers 2022, 14, 3678 9 of 14 RAW 264.7 monocyte-macrophage cells [40,41]. Indeed, BAP1 knockout in UMC026 cells resulted in a significant increase in MERTK phosphorylation on cocultured RAW 264.7 cells (Figures 4C,D and S9), which was blocked by knockdown of PROS1 in BAP1-KO UM026 cells ( Figures 4E,F, S9 and S10).

Discussion
Inactivation of BAP1 is the single most consistent mutation associated with metastatic death in UM [4,42], and yet it remains unclear how BAP1 loss promotes metastasis. BAP1 is a deubiquitinating enzyme with many binding partners and substrates, and it has been proposed to affect a wide range of processes such as transcriptional regulation, DNA repair, and metabolism [43]. Interestingly, loss of BAP1 in uveal melanocytes and UM cells results only in a subtle phenotype in vitro [7], suggesting that the dramatic metastatic phenotype associated with BAP1 loss in vivo may be the result of complex multicellular interactions that are not adequately captured by cell culture experiments. BAP1-mutant UMs display a suppressive TIME enriched for M2 polarized macrophages and T cells expressing checkpoint or "exhaustion" markers such as LAG3, TIM3, and TIGIT [11,12,16,44], similar to findings in other cancer types [45]. Consequently, BAP1 loss may promote metastasis at least in part by allowing tumor cells to evade the patient's immune response. However, it remains unknown how BAP1 mutations may mechanistically lead to immune suppression.
As a potential explanation, we found here that BAP1 loss results in upregulation of PROS1 in UM cells through epigenetic mechanisms involving H3K27ac accumulation at the PROS1 locus, consistent with our previous findings [18]. Upregulation of PROS1 has also been associated with increased metastatic risk in other cancer types [46][47][48]. PROS1 is a ligand and agonist of the MERTK receptor [49], which when activated by phosphorylation triggers signaling pathways in macrophages that suppress proinflammatory M1 polarization and promote anti-inflammatory M2 polarization [37,39,50]. M2 polarized macrophages lead to further suppression of the TIME by secreting cytokines that inhibit T cells and other immune cell types [38]. Indeed, M2 macrophages may be a primary driver of the suppressive TIME in UM [13], and they are associated with suppression of T cells related to therapeutic resistance to tebentafusp, a T-cell redirection therapy, and the only FDA-approved medication for metastatic UM [51].
The critical role of cancer genomic aberrations in promoting immune evasion, cancer evolution, and cancer metastasis has become increasingly apparent [52,53]. Our findings reveal a potential mechanistic link between BAP1 mutations and immune evasion in UM via transcriptionally mediated increased PROS1 expression and phospho-activation of tumorassociated macrophage MERTK receptors ( Figure 5), and they propose new possibilities for overcoming resistance to immunotherapy in UM.
We chose to employ a coculture model in this study because UM cell lines are not more metastatic in existing animal models following BAP1 knockdown [7]. We believe that one reason for this lack of adequate animal model is that current models rely largely on immunodeficient mice, where the mechanism of immunosuppression associated with BAP1 loss that we reveal here would not be operative. It is becoming increasingly clear that many tumor suppressors drive cancer progression by their effects on immune evasion [54].
A strength of our study was the use of UMC026 cells derived from normal human uveal melanocytes instead of uveal melanoma cell lines in the coculture experiments. This allowed us to validate that the findings were specific to BAP1 loss and not the result of private genomic aberrations peculiar to a specific UM cell line [55]. Due to interspecies ligand compatibility for MERTK [45], murine RAW 264.7 cells are preferable to human monocyte/macrophage cell lines for coculture experiments because they do not require a differentiation step before use and manifest low levels of phosphorylated MERTK at baseline [40]. Reagents used for human monocyte/macrophage differentiation such as PMA (Phorbol 12-myristate 13-acetate) and LTA (Lipoteichoic acid) can themselves perturb MERTK expression and signaling, which would mask the phenomenon we sought to evaluate in these experiments [40].
T cells and other immune cell types [38]. Indeed, M2 macrophages may be a primary driver of the suppressive TIME in UM [13], and they are associated with suppression of T cells related to therapeutic resistance to tebentafusp, a T-cell redirection therapy, and the only FDA-approved medication for metastatic UM [51].
The critical role of cancer genomic aberrations in promoting immune evasion, cancer evolution, and cancer metastasis has become increasingly apparent [52,53]. Our findings reveal a potential mechanistic link between BAP1 mutations and immune evasion in UM via transcriptionally mediated increased PROS1 expression and phospho-activation of tumor-associated macrophage MERTK receptors ( Figure 5), and they propose new possibilities for overcoming resistance to immunotherapy in UM.  Indeed, the MERTK inhibitor sitravatinib has been shown to circumvent resistance to immune checkpoint blockade in immunoresistant cancer types through its effects on the suppressive TIME [38]. These findings warrant further investigation using in vivo models to explore the potential role for targeting the PROS1-MERTK pathway in UM.

Conclusions
Mutational inactivation of BAP1 in UM may lead to a suppressive TIME at least in part by upregulation of PROS1 in tumor cells and phospho-activation of MERTK in tumor-associated macrophages. These findings nominate MERTK as a potential target for inhibition to increase the efficacy of immunotherapy in UM.