BAP1 deficiency causes loss of melanocytic cell identity in uveal melanoma

Three uveal melanoma cell lines OCM1A, 92.1, and Mel290 (kindly provided by Drs. J. Kan-Mitchell, M. Jager and B. Ksander, respectively) were used in this study. All three cell lines are wildtype for BAP1. 92.1 cells contain a GNAQ^Q209L mutation, OCM1A cells contain a BRAF^V600E mutation, and Mel290 cell lines are wildtype for both GNAQ and BRAF; all of the cell lines are wildtype for GNA11 [19]. These cell lines are well-established tools in the field of uveal melanoma research and their mutational status is representative of the spectrum seen in uveal melanoma. Due to the low frequency of BRAF mutations in uveal melanoma, OCM1A cells may not be representative of the vast majority of primary uveal melanomas. All uveal melanoma cell lines were grown in RPMI-1640 supplemented with 10% FBS, L-glutamine, and antibiotics at 5% CO_2. Primary uveal melanoma samples were collected at the time of enucleation and informed consent was obtained for each patient. All samples were confirmed to be uveal melanomas by pathologic evaluation and melanoma cells were isolated and grown as previously described [20]. Primary uveal melanoma cells were grown on collagen-covered tissue culture plates in 5% CO₂ and 4% O₂ in MDMF medium which consists of HAM’s F12 (Lonza, Walkersville MD, USA) supplemented with 1 mg/ml BSA (Sigma-Aldrich, St Louis MO, USA), 2 mM L-glutamine (Lonza), 1X SITE (Sigma-Aldrich), 1x B27 (Gibco, Carlsbad CA, USA), 20 ng/ml bFGF (PeproTech Inc, Rocky Hill NJ, USA), 50 μg/ml Gentamicin (Sigma-Aldrich) and 2.5 μg/ml AmphotericinB (Sigma-Aldrich). Primary melanocytes were isolated from unaffected choroid, obtained at the time of enucleation. Normal uveal melanocytes were handled in the same manner as primary uveal melanoma cells except they were maintained in OPTI-MEM medium supplemented with 10 ng/ml bFGF (PeproTech Inc, Rocky Hill NJ, USA), 10 ng/ml PMA (Sigma-Aldrich), 0.1 mM IBMX (Sigma-Aldrich), 1 ng/ml Heparin (Sigma-Aldrich), 50 μg/ml Gentamicin (Sigma-Aldrich) and 2.5 μg/ml AmphotericinB (Sigma-Aldrich).

Transient knockdown was carried out using BAP1 or control siRNA (Ambion, Austin TX, USA) in 92.1 and Mel290 uveal melanoma cell lines as previously described [3]. Lentiviral-based short hairpin RNA (shRNA) was used to deplete BAP1 or control gene, GFP from cultured cells for long-term experiments. Lentiviral pLKO.1 shRNA vectors for GFP (clonetechGfp-438s1c1) and BAP1 (NM_004656.2-2658s1c1 and NM_004656.2-321s1c1) developed by the RNAi Consortium (TRC)] were purchased from the Children's Discovery Institute/Genome Sequencing Center at Washington University in St. Louis. Viral production and infections were carried out according to The RNAi Consortium recommendations (Broad Institute). Lentiviruses were packaged in 293FT cells (Invitrogen, Carlsbad CA, USA) after cotransfection of the shRNA plasmids with pCMV-dR8.2 dvpr and pCMV-VSV-G lentiviral plasmids (Addgene plasmids #8454 and #8455, deposited by Bob Weinberg, [21]) using TransIT-LT1 (Mirus, Madison WI, USA). Cells were infected for 24 hours with lentiviral supernatants in the presence of 5 μg/ml protamine sulfate. Puromycin (3 μg/mL) was added to the cells at 24 hours postinfection for selection as previously described [16]. With the exception of primary class 1 tumor cells, which were under selection for one week, all infected cells were selected for at least two weeks before use in experiments and were maintained under selection for up to four weeks (referred to as BAP1-deficient or control stable cells).

MTS assays were performed using CellTiter 96 AQueous Assay reagent (Promega, Madison WI, USA) according to manufacturer's instructions. Bromodeoxyuridine (BrdU) incorporation assays were performed in 96-well plates and colorimetric changes were measured at 370 nm using a Microplate spectrophotometer (SpectraMAX 190; Molecular Devices) as previously described [16]. Flow cytometry was performed using a standard propidium iodide staining protocol as previously described [22] using a FACScan analyzer (BD Biosciences, San Diego CA, USA). The percentage of cells in each phase was determined using FlowJo software. Assays assessing the growth of cells in stem cell conditions were performed by plating 1000 or 2000 cells/well (OCM1A or 92.1, respectively) in 24-well ultra-low attachment plates containing stem cell medium, MDMF. After 5 or 7 days (OCM1A or 92.1, respectively) images were taken at 40X magnification and colony size was measured using ImageJ software. For clonogenic assays using OCM1A and 92.1 cells, flow cytometry (MoFlo; Cytomation) was used to seed one viable cell per well in ultra-low attachment 96-well plates containing MDMF medium as previously described [23]. Cell morphology of primary uveal melanocytes was assessed by phase contrast microscopy as previously described [16].

For HDAC inhibitor studies 0.5 mM, 1.0 mM or 2.0 mM valproic acid (Sigma-Aldrich) dissolved in water was added to BAP1-deficient or control cells for 72 hrs before RNA was isolated.

Soft agar assays were performed as previously described [23]. Plates were stained with MTT after 2 weeks and images were taken 6.7X using a dissecting scope and colonies were counted using ImageJ software. Scratch assays were carried out by plating 2x10⁵ cells/well in 12 well plates. Before scratching with a P200 tip, cells were treated with 5 μg/ml mitomycin C for 2 hrs at 37°C and washed with PBS. Two 100X images were taken per well and a total of three wells were imaged per condition for each experiment. Images were taken at Day 0, 1 and 2 and closure of the scratch was measured using ImageJ. Time-lapse microscopy was performed by plating cells on collagen coated 8-well chamber slides (Thermo Fisher Scientific, Rochester NY, USA) at a concentration of 1000 cells/well. The cells were allowed to attach overnight at 37°C and then imaged using an inverted Nikon Eclipse Ti at 200X every 15 minutes for 16 hrs. Cells were manually tracked using NIS Elements software (Nikon, Melville NY, USA).

Cell lysates for both westerns and immunoprecipitations (IPs) were prepared by resuspending cell pellets in lysis buffer, which consists of 50 mM Hepes pH7.2, 400 mM NaCl, 0.1% NP-40, 0.5 mM EDTA pH8, 2.5 mM DTT, plus protease and phosphatase inhibitors. Samples were then incubated on ice for 10mins before a 10 sec, low-power sonication. After which, samples were spun down to remove cellular debris and supernatants were then used for either westerns or IPs. For westerns 20 μg of protein was loaded for each sample. IPs were performed using combined lysates from OCM1A, 92.1, and Mel290 uveal melanoma cell lines. After sonication, lysates were pre-cleared with ProteinG Sepharose beads (Sigma-Aldrich) for 1 hr and incubated overnight at 4°C with 5 μg of the indicated antibodies. After incubation for 1 hr with fresh sepharose beads, samples were spun down and beads were washed twice with lysis buffer. Proteins were eluted by boiling the samples with 6X SDS loading buffer for 5 mins. IP supernatants (the supernatant removed from the tube after the initial IP pulldown) were kept for western blot analysis and are referred to as cleared lysates. IP samples and cleared lysates were subjected to SDS-PAGE followed by western blotting for the indicated antibodies. Densitometry was performed on western blots using ImageJ software. Antibodies used for IP and western blot were BAP1 (Santa Cruz, Santa Cruz CA, USA), HCF-1 (Bethyl Laboratories, Montgomery TX, USA), α-tubulin (Sigma-Aldrich), and control antibodies rabbit IgG and mouse IgG (Santa Cruz).

For primary melanocytes and tumor samples total RNA was extracted with TRIzol (Invitrogen) according to the manufacturer’s protocol and purified by ammonium acetate precipitation. RNA was extracted from cell lines using an RNeasy Kit (Qiagen) according to the manufacturer’s protocol. The RNA was DNase treated and reverse transcribed using iScript cDNA Synthesis Kit (Bio Rad, Hercules CA, USA). Primary melanocyte and tumor sample RNA was preamplified for 14 cycles with pooled primers according to the manufacturer's protocol using TaqMan PreAmp Master Mix (Applied Biosystems, Foster City CA, USA). mRNA levels were measured by qPCR using iTaq SYBR Green Supermix (Bio Rad) as previously described [3]. UBC was used as an endogenous control. Primer sequences are listed in Additional file 1.

Gene expression profiling (GEP) was performed on two independent sets of uveal melanoma cell lines (OCM1A, 92.1 and Mel290), each expressing either GFP or BAP1 shRNA for four weeks (12 samples total). Total RNA was isolated using the RNeasy kit (Qiagen). RNA quality was assessed on the Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA). Samples were subjected to gene expression profiling using the HumanHT-12 v4 Expression BeadChip (Illumina, San Diego, CA, USA). Raw expression data were subjected to cubic spline normalization in GenomeStudio (version 2011.1). ANOVA and hierarchical clustering were performed with Partek Genomics Suite (version 6.6) using a significance of P < 0.01 as a threshold for gene inclusion. Significance Analysis of Microarrays (SAM), Version 4.0 [24] was used to generate a ranked gene list, and a threshold of q < 10% was then used to select the most highly significant genes that were up or down regulated after BAP1 loss. This list was used to determine the most highly represented gene ontology categories and genes from this list were selected for validation by qPCR. A pre-ranked file was generated from the SAM output data and run through Gene Set Enrichment Analysis (GSEA), version 2.0.4 [25, 26] to identify significantly enriched gene sets. Gene expression data have been deposited in the NCBI Gene Expression Omnibus and are accessible through GEO Series accession number GSE48863.

Three uveal melanoma cell lines (OCM1A, 92.1, and Mel290) expressing either GFP or BAP1 shRNA for four weeks were subjected to single nucleotide polymorphism (SNP) arrays using Affymetrix Human Genome-Wide SNP 6.0 array (Affymetrix, Santa Clara, CA, USA). DNA was isolated using a DNeasy kit (Qiagen). Copy number and allele ratios were calculated using Partek Genomics Suite (version 6.6). For paired analyses, cell lines expressing GFP shRNA were used as baselines; for unpaired analyses, the Partek distributed baseline was used as a reference. Hidden Markov Model genomic smoothing was used to identify significant regions of amplification and deletion in samples expressing BAP1 shRNA compared to control samples.

Animal experiments were approved by the Washington University in St. Louis Animal Studies Committee. 5–8 week old NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ JAX [non-obese diabetic severe-combined immunodeficient gamma (NSG)] males (Jackson Laboratory) were injected subcutaneously into the flank with 500 OCM1A or 1000 92.1 cells in 50 μl Cultrex (Trevigen, Gaithersburg MD, USA). Tumor size was monitored once a week and the mice were euthanized after 34 (OCM1A) or 64 (92.1) days at which time tumors were collected and measured. Volume of each tumor was calculated using the ellipsoid volume formula (π xyz/6 mm³). Tumors were collected in TRIzol at time of necropsy for RNA isolation. For different experiments 10,000 92.1 cells or 500,000 OCM1A were injected into the tail vein of 5–8 weeks old NSG males or females. Mice were monitored and euthanized after 29 or 44 days respectively. Organs were collected and fixed in 10% formalin. Fixed liver and lungs were cut into 5 mm thick pieces, dehydrated and embedded in paraffin as a single block. Four micron sections were cut and stained with H&E. Overlapping images of the sections were taken at 20X and merged to one image using AdobePhotoshop CS4. Total liver or lung area, and metastasis area were measured using ImageJ 1.45 s for calculation of % of metastasis.

To study the effects of BAP1 loss in uveal melanoma cells, we initially used siRNA to achieve at least an ~ 80% depletion of BAP1 protein levels (Figure 1a). This resulted in a 20-40% reduction in cell cycle progression, measured by BrdU incorporation in two different uveal melanoma cell lines (92.1 and Mel290), which persisted through the four day experiment (Figure 1b-c). This observation is consistent with previous findings in MCF10A breast epithelial cells and HeLa cervical carcinoma cells [11, 13, 27]. To study longer-term effects of BAP1 loss, we used short-hairpin RNA (shRNA) expressed from lentiviral vectors, which consistently achieved 70-90% depletion of BAP1 protein levels in three different uveal melanoma cell lines (OCM1A and 92.1, and Mel290) (Figure 2a). BAP1-depleted cells were then compared to those infected with control lentivirus expressing shRNA directed against GFP. Interestingly, there was no substantial difference in cell viability, BrdU incorporation or cell cycle profile between BAP1-deficient and control cells after stable expression of the shRNA constructs for at least 14 days (Figure 2b-d), indicating that the initial cell cycle inhibition caused by BAP1 depletion was transient.

The uveal melanoma cells stably expressing shRNA against BAP1 and control shRNA against GFP were compared using in vitro and in vivo assays of tumorigenicity. Using scratch assays as a measure of cell motility, BAP1-deficient uveal melanoma cells were less motile than control cells (Figure 3a). Prompted by this unexpected finding, we performed time-lapse microphotography and confirmed that BAP1-deficient cells showed less overall movement than control cells (Figure 3b). Similarly, BAP1-deficient uveal melanoma cells were less capable than control cells of anchorage independent growth in soft agar assays (Figure 3c).

To assess the ability to form tumors in vivo, we created flank tumors in NOD-SCID gamma mice using BAP1-deficient versus control uveal melanoma cells. Surprisingly, the BAP1-deficient tumors were smaller than control tumors (Figure 4a). We confirmed that BAP1 was still depleted in these tumors by isolating RNA at the time of necropsy and performing qPCR (Figure 4b). To assess metastatic capacity, we then performed tail vein injections of BAP1-deficient and control uveal melanoma cells in the same mouse strain, and the BAP1-deficient cells formed fewer metastases in the liver and lungs compared to control cells (Figure 4c-e and Additional file 2).

Given these unexpected findings, we wished to gain insights into the role of BAP1 loss in uveal melanoma progression by analyzing the changes in global gene expression associated with BAP1 depletion. We analyzed the transcriptome of all three uveal melanoma cell lines using Illumina BeadArrays at four weeks after stable shRNA expression (Figure 5a). In order to identify the most significantly altered genes, we used Significance Analysis of Microarrays (SAM) with a false discovery rate cut-off of 10% and found 77 genes that were up-regulated (16 of these were pseudogenes or uncharacterized loci), and 6 genes that were down-regulated by BAP1 depletion (Additional file 3 and Additional file 4). The finding that more genes were up-regulated than down-regulated by depletion of BAP1 is consistent with its known role in transcriptional repression as part of the Polycomb PR-DUB complex [14]. The most common Gene Ontology categories included RNA metabolism (14 genes, including 10 specifically involved in RNA splicing), developmental processes (11 genes), ubiquitin system (8 genes), apoptosis (4 genes), cell cycle (4 genes), and epigenetic regulation (4 genes) (Figure 5b). Among the genes involved in the ubiquitin system, three were involved not with ubiquitin-associated protein degradation, but with substrate deubiquitination.

The set of differentially expressed genes was further analyzed for functional significance using Gene Set Enrichment Analysis. Genes with altered expression upon BAP1 depletion exhibited significant (p < 0.005) enrichment in gene sets involved in proliferation/cell cycle control (14 gene sets), development and stem cell bio-logy (9 gene sets), RNA splicing (6 gene sets), DNA damage repair (6 gene sets), metastasis (6 gene sets), epigenetic regulation (5 gene sets), amino acid metabolism (5 gene sets), the BRCA1/2 pathway (3 gene sets) and mitochondrial activity (3 gene sets) (Figure 5c and Additional file 5, Additional file 6 and Additional file 7). The metastasis gene sets included genes that were up-regulated in metastasizing melanoma, as well as prostate, lung, and pancreatic cancer.

There were 3 BRCA1/2 pathway gene sets indentified, and among the 6 DNA damage repair gene sets, 3 were related to telomere maintenance. Since BRCA pathway deregulation and telomere dysfunction are both associated with amplifications and deletions in cancer cells [28, 29], we wished to determine whether BAP1 depletion might result in such large-scale chromosomal gains and losses in uveal melanoma cells. However, Affymetrix 6.0 SNP arrays showed no differences in chromosome number between BAP1-deficient versus control cells for any of the three uveal melanoma cell lines after 4 weeks of BAP1 depletion. (Additional file 8).

Prompted by these transcriptomic findings, we wished to explore further the possibility that BAP1 inhibits metastasis of uveal melanoma cells by maintaining their differentiated state and impeding their reversion to a stem-like state. Consistent with this hypothesis, depletion of BAP1 caused a down-regulation of canonical genes of the melanocyte differentiation program (MITF, TRPM1, TYR and DCT) (Figure 6a). Similar changes were seen in cultures of primary uveal melanocyte samples from three independent patients stably expressing shRNA against BAP1 or control shRNA against GFP and also in two short-term cultures from fresh primary class 1 tumors (Figure 6a and data not shown). Further, stable depletion of BAP1 in cultured primary uveal melanocytes resulted in cells with fewer dendritic aborizations and less differentiated spindle morphology, both of which suggest melanocyte dedifferentiation (Figure 6b-c). In addition, we saw consistent up-regulation of the stem cell factor NANOG in BAP1-depleted uveal melanoma cells (Figure 6d). OCT4 expression did not change with BAP1 depletion, but this stem cell factor is tightly maintained within a restricted range to avoid differentiation [30].

To assess the capacity for self-replication, which is a measure of “stemness”, BAP1-deficient and control cells were flow sorted, single cells were seeded into separate wells of low-attachment 96-well plates in serum-free stem cell media, and the presence or absence of colonies from each well was assessed at 5 days. The BAP1-deficient cells exhibited a 50% increased capacity for self-replication compared to control cells (Figure 6e). Further, whereas we showed earlier that BAP1-deficient cells produced colonies in soft agar less efficiently than control cells using our usual serum-containing culture media (Figure 3c), the BAP1-deficient cells grew more efficiently than control cells in the limiting stem cell conditions of serum-free media and low attachment plates (Figure 6f and Additional file 9).

As we showed previously, HDAC inhibition reverts primary class 2 uveal melanoma cells to a differentiated, less aggressive class 1 phenotype [16]. Consistent with those results, treatment of BAP1-deficient uveal melanoma cells with an HDAC inhibitor restored the expression of the melanocyte differentiation markers, which were down-regulated by BAP1 depletion, in a dose-dependent manner (Figure 6g).

A major binding partner of BAP1 protein is the transcriptional co-regulator HCF-1, which was recently shown to play a key role in stem cell maintenance, in part through regulation of RNA splicing [31]. As this interaction has not been addressed within the melanocytic lineage, we examined the interaction between endogenous BAP1 and HCF-1 in BAP1-wildtype uveal melanoma cells. Indeed, HCF-1 and BAP1 were found to co-precipitate using antibodies against either protein for immunoprecitapation, and approximately 75% of total cellular BAP1 was in a complex with HCF-1 (Figure 7).