The clinical use of MEK inhibitors in uveal melanoma is limited by the rapid acquisition of resistance. This study has used multiomics approaches and drug screens to identify the pan-HDAC inhibitor panobinostat as an effective strategy to limit MEK inhibitor resistance.
Experimental Design: Mass spectrometry–based proteomics and RNA-Seq were used to identify the signaling pathways involved in the escape of uveal melanoma cells from MEK inhibitor therapy. Mechanistic studies were performed to evaluate the escape pathways identified, and the efficacy of the MEK-HDAC inhibitor combination was demonstrated in multiple in vivo models of uveal melanoma.
We identified a number of putative escape pathways that were upregulated following MEK inhibition, including the PI3K/AKT pathway, ROR1/2, and IGF-1R signaling. MEK inhibition was also associated with increased GPCR expression, particularly the endothelin B receptor, and this contributed to therapeutic escape through ET-3–mediated YAP signaling. A screen of 289 clinical grade compounds identified HDAC inhibitors as potential candidates that suppressed the adaptive YAP and AKT signaling that followed MEK inhibition. In vivo, the MEK-HDAC inhibitor combination outperformed either agent alone, leading to a long-term decrease of tumor growth in both subcutaneous and liver metastasis models and the suppression of adaptive PI3K/AKT and YAP signaling.
Together, our studies have identified GPCR-mediated YAP activation and RTK-driven AKT signaling as key pathways involved in the escape of uveal melanoma cells from MEK inhibition. We further demonstrate that HDAC inhibition is a promising combination partner for MEK inhibitors in advanced uveal melanoma.
At this time, there are no effective therapies for advanced uveal melanoma. One of the most thoroughly explored targeted therapies for uveal melanoma is small-molecule inhibitors of MEK. Despite initial clinical responses to MEK inhibition, levels of progression-free survival are very short and the majority of patients fail within 3 months. Here, we used three unbiased platforms (proteomics, RNA-Seq, drug screens) to define the mechanisms by which uveal melanoma cells escaped MEK inhibitor therapy. Our studies identified a complex adaptive response involving G-protein coupled receptor (GPCR)-driven YAP activation and increased receptor tyrosine kinase (RTK)-driven AKT signaling, both of which were suppressed by the pan-HDAC inhibitor panobinostat. The combination of the MEK and HDAC inhibitor was highly effective at limiting therapeutic escape and led to durable antitumor responses in both subcutaneous xenograft and liver metastasis models of uveal melanoma. Together, our results provide the rationale for the clinical cotargeting MEK and HDACs in advanced uveal melanoma.
Uveal melanoma is a highly aggressive tumor derived from the melanocytes of the eye, with a tendency to metastasize to the liver. Although few patients show signs of disseminated disease at diagnosis (∼4%), up to half will eventually succumb to metastatic disease despite successful treatment of the primary tumor (1). The majority of uveal melanomas harbor activating mutations in the small G-proteins GNAQ and GNA11. These mutations (most commonly at Q209L/P) disable the intrinsic GTPase activity, leading to constitutive activation (2, 3). The major downstream signaling target of GNAQ and GNA11 is phospholipase-C (PLC), which hydrolyzes phosphatidylinositol 4,5-bisphosphate to the second messengers: inositol triphosphate (IP3) and diacyl glycerol. Protein kinase C (PKC) is activated by these second messengers in GNAQ/GNA11–mutant melanomas (4).
Recent work has shown that PKC and the small G-protein RasGRP3 are required for the GNAQ/GNA11–driven activation of the MAPK pathway and that the majority of uveal melanomas have constitutive MAPK signaling that contributes to cell growth (5, 6). As a single agent, MEK inhibition has some activity against uveal melanoma cell lines, and is associated with reduced cell proliferation in vitro (7, 8). In light of this promising data, and the FDA approval of MEK inhibitors for BRAF-mutant cutaneous melanoma, a number of clinical trials were undertaken to evaluate MEK inhibitors in uveal melanoma. In an open-label phase II clinical trial of patients with uveal melanoma with no history of prior dacabarzine treatment, use of the MEK inhibitor selumetinib was associated with an increase in PFS from 7 to 16 weeks (9). These initially promising findings led to the initiation of a phase III double-blind clinical trial of selumetinib plus dacarbazine, which unfortunately failed to show any increase in PFS compared with dacarbazine alone (10).
Despite these disappointing results, current strategies continue to focus upon combination therapies that include MEK inhibition as the backbone. There is promising preclinical data that indicates the combination of a MEK and a PKC inhibitor potently induces apoptosis and suppresses tumor growth in mouse xenograft models (5). Multiple other signal transduction cascades are also activated in uveal melanoma including the PI3K/AKT/mTOR signaling pathway, which has been implicated in survival and cell migration (11, 12) and the Hippo tumor suppressor pathway, which plays key roles in tissue homeostasis and organ size (13). Under normal physiologic conditions, the MST1/2 and LATS1/2 kinases phosphorylate and inactivate YAP and TAZ, two transcriptional coactivators implicated in oncogenic transformation (13, 14). In uveal melanoma, GNAQ stimulates YAP through a Hippo-independent mechanism that is initiated through actin polymerization (15). Silencing of GNAQ/GNA11 in uveal melanoma cells led to decreased nuclear accumulation of YAP, with further studies showing that the YAP inhibitor verteporfin abrogates GNAQ/GNA11–driven tumor growth in an orthotopic uveal melanoma ocular xenograft model (15, 16). At this time, little is known about the systems level signaling adaptations of uveal melanoma cells to MEK inhibition. In this study, we used activity-based protein profiling (ABPP) and RNA-Seq to identify key proteins involved in the adaptation of uveal melanoma cells to MEK inhibition, and identified novel drug combinations to overcome this adaptation.
Materials and Methods
RPMI culture medium was purchased from Corning. FBS was purchased from Sigma Chemical Co.. Trypsin, penicillin/streptomycin antibiotics, and puromycin were purchased from Gibco. Trametinib (MEK inhibitor), panobinostat (pan-HDAC inhibitor), pictilisib (PI3K inhibitor), bosentan hydrate (EDNRB inhibitor), verteporfin (YAP inhibitor), entinostat (HDAC1/2/3 inhibitor), and tubastatin A (HDAC 6 inhibitor) were purchased from Selleckchem. PCI-34051 (HDAC8 inhibitor) was purchased from Cayman Chemical. Endothelin-3 was purchased from Sigma Chemical Co. WNT5A was purchased from R&D Systems. Antibodies for Western blot and immunochemistry were purchased from Cell Signaling Technology, Sigma Chemical Co., Millipore, and Abcam. The phospho-receptor tyrosine kinase and phospho-kinase array were purchased from R&D Systems. OptiMEM medium, Lipofectamine 2000, and live/dead viability stain were purchased from Invitrogen/Life Technologies Corp). siRNA for ROR1/2, IGF-1R, and YAP were purchased from Dharmacon RNA Technologies. Nontargeting siRNA was purchased from Santa Cruz Biotechnology. The Endothelin-3 Assay Kit was purchased from IBL.
Uveal melanoma cell lines
The uveal melanoma cell lines 92.1, Mel270, OMM1, MP41, and MM28 were used as described previously (17). All uveal melanoma cell lines were cultured in RPMI1640 supplemented with 10% FBS, l-glutamine, and antibiotics at 5% CO2. All cells were tested for Mycoplasma contamination every month using the Plasmotest-Mycoplasma Detection Test (Invivogen). Last test date: April 18, 2019. Each cell line was authenticated using the Human short-tandem repeat human cell line authentication service (ATCC) and frozen stocks of cells were discarded after 10 passages.
Cell viability assay (MTT assay)
Uveal melanoma cells were plated in triplicate wells (1 × 103 cells/well) and treated with increasing concentrations of MEK inhibitor (trametinib) for 72 hours. Cell viability was determined using the MTT assay as described previously (18).
Colony formation assay
A total of 1 × 103 cells were plated and allowed to attach overnight. The medium and drug/vehicle was replaced every 2 days for 4 weeks. After the specific treatments for each experiment, colonies were stained with crystal violet dye, as described previously (18).
Flow cytometry for apoptosis analysis
A total of 1 × 105 cells were plated and allowed to attach overnight. After the specific treatments for each experiment, Annexin V staining quantification was performed using FlowJo software as described previously (19).
Activity-based protein profiling
ABPP experiments were carried out as described previously (20). Briefly, 92.1 and Mel270 cells were treated with MEKi (25 nmol/L for 24 hours) and then solubilized with lysis buffer. A total of 1 mg of protein from each sample was prepared for labeling, enrichment, and LC/MS-MS analysis. Protein identification and quantification were performed by Andromeda and MaxQuant (v. 22.214.171.124), the values were log2-transformed and normalized (21). Signaling pathways, protein interaction, and process network analysis were carried out using MetaCore (GeneGO). Data are available in PRIDE (PXD013988).
RNA was extracted from 92.1 and Mel270 uveal melanoma cell lines using Qiagen's Rneasy Mini Kit (Qiagen), and screened for quality on an Agilent BioAnalyzer. The samples were then processed for RNA-sequencing (RNA-seq) using the NuGen Ovation Human FFPE RNA-Seq Multiplex System. Briefly, 100 ng of RNA was used to generate cDNA, and a strand-specific library following the manufacturer's protocol (NuGEN Technologies, Inc.). Quality control steps including analysis on the BioAnalyzer RNA chip and qRT-PCR for library quantification were performed. The libraries were then sequenced on the Illumina NextSeq 500 sequencer with a 2 × 75-base paired-end run to generate 40–50 million read pairs per sample. RNAseq data were preprocessed for quality assessments before aligning to the human genome hs37d5 using Tophat v2.0.13 default setting and quantified using htseq-count based on the RefSeq gene model downloaded from USCS Table Browser. Normalized counts were obtained by using the counts function, and differential expression was analyzed using a Wald statistic test implemented in the DESeq function in the DESeq2 Bioconductor package, which performs serial dispersion estimation and negative binomial generalized linear model fitting procedure. A Benjamini–Hochberg Padj value of less than 0.05 was used as a cutoff to determine significantly differentially expressed genes. Data are available in GEO (GSE127948).
Gene set enrichment analysis
Gene set enrichment analysis (GSEA; ref. 22) was conducted utilizing recommended parameters (http://software.broadinstitute.org/gsea/doc/GSEAUserGuideFrame.Html) with gene sets obtained from the Molecular Signatures Database (23) and custom gene sets obtained from the GSEA database.
Phospho-receptor tyrosine kinase and Phospho-kinase array analysis
A Human Phospho-RTK Array Kit (ARY001B) and a Human Phospho-kinase Array Kit (ARY003B) were used to measure the relative level of different RTKs and kinases. 92.1 and Mel270 uveal melanoma cell lines were treated with MEKi (10 nmol/L for 48 hours), the lysed, and 300-μg protein were incubated with the array according to the manufacturer's protocol.
Cells were plated and allowed to attach overnight. After the specific treatments for each experiment, proteins were extracted and blotted as described previously (24). Total and phospho-proteins were analyzed (Supplementary Table S1) and then the membranes were stripped and reprobed for GAPDH/vinculin/β-actin.
tRNA was isolated using Qiagen's Rneasy Mini Kit (Qiagen). TaqMan Gene Expression Assay primer/probes were used as shown in Supplementary Table S2. GAPDH was used to normalize the genes of interest. qRT-PCR reactions were carried out as described previously (18).
Cells from uveal melanoma cell lines were plated and allowed to attach overnight in complete RPMI medium with 10% FBS. After 24 hours, this medium was replaced with Opti-MEM and the cells were transfected with 50 nmol/L siRNA for ROR1 (Dharmacon SMARTpool; L-003171-00-0005), 50 nmol/L siRNA for ROR2 (Thermo Fisher Scientific; #4390824), or 50 nmol/L siRNA for IGF-1R (Dharmacon SMARTpool; L-003012-00-0005) in complex with Lipofectamine 2000 overnight. Nontargeting siRNA was added as a siRNA control (Santa Cruz Biotechnology; sc-37007). After 24 hours of the transfection, medium was replaced by complete RPMI medium with 10% FBS and treated with MEKi (10 nmol/L) for 72 hours.
Cell proliferation assay
A total of 5 × 104 cells were plated and allowed to attach overnight. After the specific treatments for each experiment, cells were counted using Trypan Blue reagent. The percentage of total cells was normalized to the percentage of control cells as described previously (25).
Transfections and luciferase assays
We have used a YAP/TAZ–responsive synthetic promoter driving reporter plasmid, named 8xGTIIC-luciferase (Addgene, #34615) for the assay. Overnight seeded 92.1 and Mel270 cells (120 × 103 cells/well in a 6-well plate) were transfected with the construct using FuGENE (Promega; #31985-070) for 8 hours. Media was changed for another 12 hours prior to the drug(s) treatment with MEKi (10 nmol/L), HDACi (10 nmol/L), or both for 48 hours. Harvested cells were washed once with PBS and luciferase assay was performed according to the manufacturer's protocol and plotted the values normalized against Control, without any inhibitor treatment.
A total of 3 × 103 cells were seeded in 8-Well Lab-Tek II Chamber Slides and allowed to attach overnight. On the next day, cells were treated, then fixed, permeabilized and stained using YAP antibody (#14729). The information about the immunofluorescence antibodies can be found in Supplementary Table S3. Slides were mounted with ProLong Antifade with DAPI in accordance with a previously described protocol (26). Glass slides were observed with a Leica TCS SP8 AOBS laser scanning confocal microscope through a 63X/1.4NA oil immersion Plan Apochromat objective. Laser line at 405 nm was used to excite the DAPI fluorophores. Images were captured at 200 Hz scan speed with photomultiplier detectors using LAS X software version 3.1.5. Images were analyzed using the Definiens Tissue Studio v4.7 (Definiens AG) software suite. Cells were segmented by the nuclear stain DAPI (blue) and phalloidin (red) channel was used as cytoplasm marker for cell simulation. The image was analyzed as an 8-bit image and intensity of each RGB channel was measured from 0 to 255 grayscale fluorescent units. The cells were then quantified for green and red intensity per field for nucleus and for cytoplasm.
Endothelin-3 assay kit
92.1, Mel270, MP41, and OMM1 uveal melanoma cells were seeded in 96-well plates at 1.0 × 104 cells/well. After 24 hours, cells were treated with MEKi (10 nmol/L), HDACi (10 nmol/L), or MEKi + HDACi (10 nmol/L each one) for 24–72 hours. Cell supernatant was collected and incubated according to the manufacturer's protocol.
92.1, Mel270, MP41, and OMM1 uveal melanoma cells were seeded in 384-well plates at 1.0 × 103 cells/well. All compounds of the library were diluted to 0.5 or 2.5 μmol/L, and the experiment was performed in duplicate. A total of 289 compounds from an in-house library were tested. Compounds were aliquoted by a Biotek Precision Pipetting robot. Cell viability was measured by Cell-Titer-Glo (Promega G7572) at 72 hours posttreatment.
Subcutaneous xenograft model.
Eight-week-old female CBySmn.CB17-Prdkc scid/j mice (Stock No: 001803 - Jax) were subcutaneously injected with 1.0 × 106 92.1 or MP41 uveal melanoma cells per mouse. The tumors were allowed to grow for 3 weeks and mice were randomly separated with similar average initial tumor volumes, with a total of 3 mice per group.
Liver metastasis model.
Eight-week-old female NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (Stock No: 005557, Jackson Laboratory) were injected with 2.0 × 105 MP41 uveal melanoma cells per mouse into the tail vein. The tumors were allowed to grow for 4 weeks and mice were randomly separated with similar average initial tumor volumes, with a total of 3 mice per group.
The mice were imaged on 7T Horizontal Magnet (Agilent ASR 310) and (Bruker Biospin, Inc. BioSpec AV3HD), using a 35 mm birdcage coil (m2m imaging corp). Anatomical coronal images were obtained with a TurboRARE sequence with echo time/repetition time (TR/TE) = 1,585/15 ms, 33 slices, slice thickness of 0.6 mm, field of view (FOV) = 30 × 30 mm2, image size 256 × 256. Respiration gating was used to minimize motion artifacts. Liver metastases were manually contoured on all MR slices using ImageJ (https://imagej.nih.gov/ij/index.html). A custom program was written in MATLAB (MATLAB 2018a, The MathWorks, Natick, 2018.) to extract voxels within the manually drawn contours and compute total metastases volume burden and individual metastases volumes, in mm3. In both protocols, mice were treated with MEKi (trametinib, 1 mg/kg gavage, daily), HDACi (panobinostat, 20 mg/kg i.p., three times a week), or the combination of both agents for 30 days (xenograft model) or for 21 days (liver metastasis model). The control group received both vehicles (for trametinib: 0.5% methylcellulose + 0.5% Tween-80 molecular grade sterile water; for panobinostat: 5% dextrose in water). Mouse weight and tumor volumes (½ × L × W2) were measured every 72 hours. All animal experiments were carried out in agreement with ethical regulations and protocols approved by the University of South Florida Institutional Animal Care and by The Institutional Animal Care and Use Committee (IACUC number IS00002983). The IACUC protocol did not permit survival to be an experimental endpoint.
Results are expressed as mean ± SD of a triplicate of at least three independent experiments. One-way ANOVA was used followed by a TUKEY-KRAMER posttest to test for multiple comparisons with a given significance level of P < 0.05. Significant differences between the control and treated groups are indicated by ***, P < 0.001; **, P < 0.01; *, P < 0.05.
Activity-based proteomic profiling identifies signaling pathways implicated in the escape of uveal melanoma cells from MEK inhibitor therapy
We began by characterizing the MEK inhibitor response of a panel of GNAQ/GNA11–mutant uveal melanoma cell lines that were derived from primary and metastatic lesions (92.1, Mel270, MP41: primary, OMM1 and MM28: metastatic). It was found that although the MEK inhibitor trametinib (MEKi) inhibited the growth of all of the uveal melanoma cell lines, these reductions were modest (Fig. 1A; Supplementary Table S4), and associated with regrowth of colonies in all cases (Fig. 1B and C). Levels of MEKi (10–25 nmol/L trametinib)-induced apoptosis were also minor compared with those seen in cutaneous melanoma (Fig. 1D; ref. 27). Little apoptosis induction was observed at 24 hours. To better understand the process of adaptation that occurs when uveal melanoma cells are treated with a MEKi, we treated two GNAQ-mutant uveal melanoma cell lines (92.1 and Mel270) with trametinib (25 nmol/L, 24 hours) and performed activity-based protein profiling (ABPP; Fig. 1E; ref. 20). This method, which uses mass spectrometry to quantify ATP uptake levels of proteins through transfer of a desthiobiotin tag to lysine in the active site of enzymes and kinases, allows signaling activity to be mapped in a comprehensive manner (Fig. 1E; ref. 20). The ABPP studies demonstrated that MEK inhibition increased the ATP uptake of 128 proteins and 98 proteins in the 92.1 and Mel270 cells, respectively (Fig. 1F). Use of STRING analysis allowed us to identify an enrichment for activated proteins implicated in proliferation and survival, ribosome function, metabolism, and the cytoskeleton (Fig. 1G). The top pathways modulated by MEKi treatment included cell metabolism, cytoskeletal remodeling, apoptosis, AKT signaling, IGF signaling, WNT signaling, FGFR signaling, and melanocyte differentiation (Supplementary Fig. S1).
Trametinib induces adaptive AKT signaling in uveal melanoma cells
We next focused on the potential therapeutic escape pathways identified from our ABPP screen and first considered the PI3K/AKT pathway. GSEA analysis of RNA-Seq data generated from uveal melanoma cells treated with trametinib (25 nmol/L, 24 hours) demonstrated an enrichment for genes implicated in PI3K/AKT signaling (Fig. 2A; Supplementary Table S5). This was confirmed in kinome arrays that showed a consistent upregulation of AKT in both cell lines following MEK inhibition (Fig. 2B) and an increase in FAK signaling in the Mel270 cells—confirming the link between MEKi and cytoskeletal rearrangement observed in the ABPP data. The AKT data were confirmed by Western blot, with MEKi found to increase phosphorylation of AKT at T308 (Fig. 2C shows fold-increase by densitometry). The potential role of rebound AKT signaling in the escape of the melanoma cells from MEKi therapy was validated by the ability of the PI3K inhibitor pictilisib (PI3Ki) to significantly increase the apoptotic response to trametinib (MEKi; Fig. 2D). Although there was some evidence that the PI3Ki also suppressed the outgrowth of MEK inhibitor–treated uveal melanoma cells in colony formation assays, the effects were incomplete and tumor cells were still able to evade therapy (Fig. 2E and F).
IGF1R and ROR1/2 activate AKT signaling following MEK inhibition
As adaptive AKT signaling frequently results from increased RTK signaling, we returned to our RNA-Seq data and identified an increase in receptor protein kinase (ES score 0.41) and receptor tyrosine kinase (ES score 0.44) expression (Fig. 3A; Supplementary Tables S6 and S7). These findings were confirmed by RTK arrays, with MEKi being found to increase the phosphorylation of multiple RTKs including IGF-1R (in the 92.1 cells), as well as ROR1 and ROR2 (in both the 92.1 and Mel270 cell lines; Fig. 3B). qRT-PCR and Western blot analyses demonstrated increases in IGF-1R, ROR1, and ROR2 mRNA and protein expression following MEKi (Fig. 3C and D). Although the link between IGF-1R and AKT signaling is well known, less is known about whether ROR1 and ROR2 activate AKT signaling. Treatment of 92.1 uveal melanoma cells with WNT5A, the ligand for ROR1 and ROR2 receptors (28), confirmed a time-dependent increase in AKT phosphorylation (Fig. 3E). Silencing of ROR1/2 (Fig. 3F) in the 92.1 cells inhibited the increases in AKT phosphorylation observed following MEKi treatment (Fig. 3G). Silencing of IGF-1R in combination with the MEKi was found to increase cell death and decrease the numbers of 92.1 cells, but not Mel270 cells, a result consistent with the increased IGF-1R signaling seen only in the 92.1 cell line (Fig. 3F, H, and I). In contrast, silencing of ROR1/2 enhanced the effects of MEKi in terms of decreased cell survival and apoptosis induction in both of the cell lines evaluated (Fig. 3F, H, and I).
MEK inhibition increases YAP activity leading to increased cell survival
As inhibition of PI3K/AKT did not fully abrogate escape from MEK inhibition, we next turned our attention to other possible pathways. One signal transduction cascade known to be critical for uveal melanoma progression, frequently upregulated following cytoskeletal rearrangement is the prooncogenic mediator of Hippo signaling, YAP (15, 29). Analysis of the RNA-seq data by GSEA showed a significant gene enrichment (ES score 0.27) for Hippo pathway targets with a positive correlation among the genes overexpressed after MEKi treatment (10 nmol/L, 24 hours; Fig. 4A; Supplementary Table S8). Although YAP is constitutively activated in uveal melanoma cells, MEKi was noted to further stimulate YAP transcriptional activity in a reporter assay (Fig. 4B) and this was accompanied by an increase in the levels of nuclear YAP accumulation (Fig. 4C and D). The increase in mRNA levels of a number of YAP pathway transcriptional target including YAP, connective tissue growth factor (CTGF), amphiregulin (AREG), and cysteine rich angiogenic inducer 61 (CYR61) also occurred following MEKi treatment (Fig. 4E). Increased expression of two of the main transcriptional targets of YAP/TAZ activity including CTGF and AREG was also seen by Western blot after MEKi treatment (10 nmol/L, 72 hours; Fig. 4F). The role of YAP signaling in therapeutic escape was demonstrated by the ability of the YAP inhibitor verteporfin (YAPi) to decrease colony formation in response to MEK inhibition compared with either drug alone (Fig. 4G and H). In addition, siRNA knockdown of YAP (Fig. 4I) increased the level of MEKi-induced apoptosis in multiple uveal melanoma cell lines (Fig. 4J).
Adaptive GPCR signaling increases YAP activity following MEK inhibition
G-protein–coupled receptors (GPCR) are known to be strong activators of YAP signaling. In line with this, it was found that MEK inhibition led to a strong induction of GPCR expression in our GSEA analysis (Fig. 5A; Supplementary Table S9). Multiple GPCRs showed increased expression including GPR158, GP133, and endothelin-receptor B (EDNRB; Fig. 5B). A potential role for EDNRB signaling in the adaptive YAP signaling was confirmed through studies in which exogenous endothelin-3 (ET-3) was found to increase both YAP reporter activity, nuclear localization of YAP, and induction of YAP-target genes in four uveal melanoma cell lines (Fig. 5C and D; Supplementary Fig. S2A–S2C). In each case, the endothelin receptor B (EDNRB) antagonist bosentan was found to block the ET-3–mediated increases in YAP activation (Fig. 5C and D; Supplementary Fig. S2A–S2C). Mechanistically, it was noted that MEKi treatment led to the release of ET-3 from the uveal melanoma cell lines by ELISA (Fig. 5E) and that the MEKi-mediated increase in YAP reporter activity could be abrogated by the EDNRB antagonist (Fig. 5F). Together, these results suggested that MEK inhibition led to ET-3 release from the uveal melanoma cells and this functioned in an autocrine manner to activate YAP through EDNRB. No increases in YAP activity were found following treatment with ET-1 (Supplementary Fig. S3).
Histone deacetylase inhibitors increase the effects of MEK inhibition in uveal melanoma
Our studies identified both RTK-mediated AKT and GPCR-driven YAP signaling as pathways utilized by uveal melanoma cells to escape MEKi therapy. As there are no known drugs that target both AKT and YAP, we performed an unbiased screen of 289 compounds to identify potential drug combination partners that could limit escape from MEK inhibition (drugs listed in Supplementary Table S10). The drug library used covers all major target classes, including kinases, receptor tyrosine kinases, phosphatases, receptor agonists, proteases/proteasome, PARP1, epigenetic enzymes, Hedgehog, HSP90 and Notch, and reflects the current landscape of targeted agents approved for use or have been considered for clinical development (Fig. 6A). Among these, several drugs were identified with some activity against one or more uveal melanoma cell lines including PI3K inhibitors (GSK2126458, idelalisib), two kinesin inhibitors (ispinesib, SB743921), CDK inhibitors (dinaciclib), H3K27 histone demethylase (GSK-J4), and mTOR (sapanisertib; Fig. 6B). The drug class with the most prominent effect across all four cell lines was the HDAC inhibitors (HDACi; Fig. 6B). To further determine whether other epigenetic inhibitors could also enhance the effects of MEKi (trametinib), we evaluated inhibitors of DOTL1 (EPZ5676), EZH2 (tazemetostat), LSD1 (GSK 2879552), DNMT (decitabine), HAT (anacardic acid), and HDACi (panobinostat) alone and in combination with trametinib. These studies demonstrated that the pan-HDACi panobinostat was the most effective at enhancing the antiproliferative effects of MEKi (Supplementary Fig. S4). We next turned our attention to more specific HDACis, including entinostat (HDAC1/2/3i), tubastatin (HDAC6i), and PCI-34051 (HDAC8i) and noted that panobinostat (HDACi) was the most effective among all of these across all four cell lines in both MTT and colony formation assays (Fig. 6C–E). Further support for the potential role of HDAC activity in the escape of uveal melanoma cells from MEKi therapy was suggested by the increase in global protein deacetylation observed in our uveal melanoma cell lines following MEKi treatment (Supplementary Fig. S5). It was found that cotreatment of multiple uveal melanoma cell lines, including 92.1, MP41, Mel270, and MM28 with the MEKi–HDACi (trametinib–panobinostat) combination was associated with significantly (P < 0.05) higher levels of apoptosis compared with either single agent (Fig. 6F). These effects were specific to uveal melanoma cells, with no apoptosis seen in 3 different primary uveal melanocyte cell lines (Supplementary Fig. S6). The apoptotic response was paralleled by an increased induction of cleaved caspase-7 and cleaved PARP in uveal melanoma cell lines (Fig. 6G).
Trametinib plus panobinostat induces uveal melanoma regression in vivo
We next asked whether the enhanced therapeutic efficacy of MEKi and HDACi resulted from the combined inhibition of YAP and AKT signaling. It was found that cotreatment of the uveal melanoma cells with panobinostat and trametinib effectively suppressed the adaptive AKT and YAP signaling and the release of ET-3 seen following MEK inhibitor treatment (Fig. 7A and B; Supplementary Fig. S7). From a mechanistic standpoint, we identified the PI3K/AKT pathway as being significantly downregulated by GSEA analysis of our RNA-Seq dataset (Supplementary Fig. S8A). A pairwise comparison of the effects of each drug demonstrated the MEKi–HDACi combination to also strongly induce PTEN at the mRNA and protein level (Supplementary Fig. S8B and S8C; Supplementary Table S11). To validate the MEKi–HDACi combination in vivo, we first generated xenografts of two human uveal melanoma models (92.1 and MP41). After formation of a palpable tumor in the subcutaneous model (around 14–21 days; 100–200 mm3), mice were treated with vehicle, MEKi (trametinib, 1 mg/kg, orally, daily), HDACi (panobinostat, 20 mg/kg, i.p. 3× week), or combination of both agents for 30 days. The control group received both vehicles. The combination of MEKi and HDACi led to a significant and durable suppression of uveal melanoma growth compared with either drug alone. (Fig. S7C and S7D; Supplementary Fig. S9A and S9B). Although single-agent MEKi was more effective against MP41 uveal melanoma cells than 92.1 cells, its effects in both models were relatively short-lived and the tumors reinitiated growth. IHC staining confirmed that single-agent MEKi was associated with increased levels of pAKT and YAP/TAZ and that the combination of MEKi and HDACi simultaneously suppressed pAKT, and YAP/TAZ expression (Fig. 7E). Advanced uveal melanoma is typically associated with the development of liver metastases. To explore the effectiveness of the MEKi–HDACi combination in this setting, we injected MP41 uveal melanoma cells into the tail veins of mice and allowed liver metastases to form (around 28 days; 5–10 mm3). Once the presence of liver metastases was confirmed by MRI imaging, treatment with MEKi (trametinib, 1 mg/kg, orally, daily), HDACi (panobinostat, 20 mg/kg, i.p. 3× week), or combination of both agents for 21 days was initiated. It was noted that although the MEKi was associated with some reduction in liver tumor burden, the MEKi–HDACi combination was associated with more profound and durable antitumor responses than either drug alone (Fig. 7F and G; Supplementary Fig. S10). Together these results confirmed our in vitro findings and demonstrated that the addition of a pan-HDACi could inhibit the pathways involved in the escape from MEK inhibitor therapy, limiting uveal melanoma growth at both subcutaneous and at clinically relevant liver metastasis sites.
Although significant progress has been made in the development of systemic therapies for the treatment of advanced cutaneous melanoma, little improvement has been made in the management of metastatic uveal melanoma. Unlike cutaneous melanoma, uveal melanoma has proven extremely resistant to immunotherapy, with anti–CTLA-4 therapy being associated with responses of <7% and no appreciable improvement in overall survival (30, 31). The anti–PD-1 and anti–PD-L1 antibodies have proven similarly ineffective, with the largest clinical trial to date demonstrating response rates of approximately 4% and a limited level of disease control (32). It has been speculated that the low mutational burden, and therefore the low level of neoantigen expression, of uveal melanoma versus cutaneous melanoma may underlie the lack of immunotherapy response.
Another strategy to treat uveal melanoma is targeted therapy, in which kinase inhibitors are used to selectively target the oncogenic drivers responsible for tumor growth and progression. This strategy has been very successful in the treatment of cancers with strong oncogene addiction such as BRAF-mutant melanoma and EGFR and ALK-mutant lung cancers (33–35). Uveal melanomas typically harbor activating mutations in the small G-proteins GNAQ and GNA11, which are not kinases and therefore not easily tractable to drug development (2). Instead, the development of targeted therapies in uveal melanoma has focused upon kinases and pathways downstream of GNAQ/GNA11. The most extensively explored targeted therapy in uveal melanoma to date are the MEK inhibitors. This class of drugs are FDA-approved in the single-agent setting, and in combination with BRAF inhibitors, for the treatment of advanced BRAF-mutant cutaneous melanoma (17, 36). Most MEK inhibitor studies in uveal melanoma to date have focused upon selumetinib (AZD6244). In a phase II open-label clinical trial of advanced uveal melanoma, selumetinib treatment yielded an improved progression-free survival compared with either dacarbazine or temozolomide (9). Despite these initially encouraging results, a subsequent phase III double-blinded trial of selumetinib plus dacarbazine showed no improvement in PFS compared with dacarbazine alone (10). Although disappointing, these findings fit with our growing understanding of how cancer cells respond to MEK inhibition, with multiple studies demonstrating that initial MEK inhibitor responses are followed by adaptive signaling and transcriptional changes that lead to therapeutic escape (37, 38).
The goal of this study was to define the patterns of adaptive signaling in uveal melanoma cells treated with MEK inhibitor therapy and to identify combination partners that limited therapeutic escape. To achieve this, we used a mass spectrometry–based ABPP approach to comprehensively map global protein ATP uptake following MEK inhibitor treatment (20). Changes in multiple pathways were identified, with some of the most prominent being those associated with the organization of the cytoskeleton, PI3K/AKT signaling, and RTK signaling. Inhibition of RAF and MEK is known to trigger a rapid transcriptional reprogramming that is associated with increased RTK expression. This phenomenon was first described for breast cancer, in which chronic MEK inhibitor treatment led to widespread increase in RTK expression that allowed for recovery of signaling through MAPK and other pathways (37, 39). Similar findings have been also reported in many other cancers including BRAF and NRAS-mutant melanoma; where BRAF and MEK inhibition frequently leads to a relief of feedback inhibition and increased signaling through multiple RTKs including IGF-1R, EGFR, ERBB3, EphA2, and c-MET (14, 40–43). There is good evidence that targeting these compensatory pathways improves the response to MAPK-targeted drugs in both in vivo mouse models and in clinical settings (14, 36). To investigate whether this also occurred in GNAQ-mutant uveal melanoma cell lines, we performed RTK arrays and identified increased IGF1-R and ROR1/2 activity following MEK inhibition. In BRAF-mutant melanoma, RAF and MEK inhibition typically leads to recovery of MAPK signaling, and in some cell lines, adaptive AKT signaling (27, 39). Here, we found that MEK inhibition in uveal melanoma cells led to increased AKT and FAK signaling and that was mediated through IGF-1R and ROR receptors. Although the combination of MEK and PI3K-AKT-mTOR inhibition was suggested to be superior to MEK inhibition alone in multiple preclinical uveal melanoma models (11, 12), our results demonstrated that resistance to the MEK-PI3K inhibitor combination still occurred.
YAP is a transcriptional coactivator and tumor promoter, whose nuclear localization and activity is regulated by the Hippo pathway. In GNAQ/GNA11–mutant uveal melanoma cells, YAP is activated by the guanine nucleotide exchange factor Trio leading to YAP activation via Rho and Rac (15, 29). Increased signaling through Rho and Rac leads to increased actin dynamics and the release of YAP from its inhibitory complex with the actin-associated protein angiomotin (15). Once free of this complex, YAP is free to migrate to the nucleus and initiate transcription. Although there is good evidence that YAP is a driver of uveal melanoma progression, this pathway has yet to be implicated in the escape of uveal melanoma cells from MEK inhibitor therapy. Our results herein demonstrate that treatment with MEK inhibitors increased YAP activity further and likely constituted an important therapy escape mechanism. YAP signaling is known to be activated through GPCRs, with our RNA-seq studies identifying a whole series of candidate receptors that were upregulated following MEK inhibition. Among these was EDNRB, a GPCR activated by all three members of the endothelin family. There is good evidence that EDNRB signaling is involved in melanocyte development, with studies showing severe deficits in melanocyte numbers in mice that are null for EDNRB (44, 45). EDNRB signaling is also implicated in melanoma with levels of expression being correlated with melanoma progression and the increased development of melanoma brain metastases in in vivo models (46, 47). There is also evidence from cutaneous melanoma that EDNRB antagonists reduce melanoma growth in vitro and in in vivo xenograft models (48, 49). Other work in BRAF-mutant melanoma demonstrated that BRAF inhibition often leads to increased EDNRB receptor expression and that this confers enhanced sensitivity to the BRAF–endothelin receptor antagonist combination (50). Further evidence suggests that autocrine endothelin-1 might also regulate melanoma heterogeneity following BRAF inhibition and could mediate the switch to an Axl-high/MITF-low (drug resistant) phenotype (51). We here demonstrate that uveal melanoma cells released ET-3 in response to MEK inhibition and that the resulting increase in EDNRB signaling activates YAP signaling, leading to increased cell survival. Although it is likely that EDNRB plays a role in the increased YAP signaling observed following MEK inhibition, it is unlikely to be only GPCR involved, and it is possible that different uveal melanomas may have unique GPCR dependencies. One potential strategy to target multiple G-proteins (and GPCR) simultaneously could be through allosteric inhibitors of GDP/GTP exchange with recent studies demonstrating that GTP exchange inhibitors such as the depsipeptide FR900359 have activity against GNAQ-mutant uveal melanoma cell lines (52). The increased GPCR expression noted following MEK inhibition might be expected to increase the adhesion of uveal melanoma cells to the extracellular matrix, potentially decreasing their metastatic potential (53).
As our goal was to develop novel therapeutic strategies that limited adaptive signaling, we undertook a drug screen to identify potential combination partners for the MEK inhibitors. Our initial analysis identified HDAC inhibitors as a class of drugs with significant single-agent activity. The HDACs constitute a family of enzymes that catalyze the hydrolysis of acetyl groups from acetylated proteins, the best characterized of which being the N-terminal tails of histones (54). Inhibition of multiple HDACs, using the pan-HDAC inhibitor panobinostat was found to be superior to multiple other epigenetic inhibitors including EZH2, DOTL1, HATs, and LSD1 in enhancing the cytotoxic activity of MEK inhibition. There is already good evidence that HDAC inhibitors, including the class III inhibitor tenovin, and a number of pan-HDAC inhibitors (TSA, depsipeptide butyrate) have activity against uveal melanoma cell lines, through affects upon FAS, p21,p27, p53, c-JUN, and β-catenin expression (55–57). In cutaneous melanoma, there is also evidence that HDAC inhibition can restore sensitivity to BRAF inhibition following the onset of resistance (58–60). At the mechanistic level, HDAC inhibition was noted to suppress both AKT and YAP signaling following MEK inhibition, with the effects on AKT mediated, in part, by increased expression of the PI3K/AKT pathway suppressor PTEN. To our knowledge, this is the first demonstration that HDAC inhibitors inhibit YAP signaling. The effectiveness of the panobinostat–trametinib combination was demonstrated in two in vivo uveal melanoma subcutaneous xenograft models, with IHC analysis showing the addition of panobinostat to inhibit AKT and YAP signaling. Of clinical significance, the MEKi–HDACi combination also had good antitumor activity against uveal melanoma liver metastases, the major site of disseminated disease. Panobinostat is an HDAC inhibitor that was FDA-approved in 2015 for the treatment of relapsed multiple myeloma. Our finding that a clinically approved pan-HDAC inhibitor was effective at simultaneously limiting YAP and AKT signaling in uveal melanoma cells suggests this could be a good candidate for future clinical development. At this time, the only agent with proven anti-YAP activity is verteporfin, and while it is FDA-approved for local photodynamic therapy in the treatment of macular degeneration, it is unlikely to have much utility as a systemic therapy for metastatic uveal melanoma. Indeed, even preclinical studies in xenograft models of uveal melanoma have resorted to multiple strategies to improve efficacy, such as mixing liposome-encapsulated verteporfin with uveal melanoma cells prior to xenografting (29). We therefore believe that the combination of trametinib and panobinostat is worthy of future investigation in patients with metastatic uveal melanoma.
Disclosure of Potential Conflicts of Interest
J.D. Licht reports receiving commercial research grants from Celgene. J.W. Harbour holds ownership interest (including patents) in Castle Biosciences and is a consultant/advisory board member for Castle Biosciences, Aura Biosciences, and Immunocore. No potential conflicts of interest were disclosed by the other authors.
Conception and design: F. Faião-Flores, M.F. Emmons, K.S.M. Smalley
Development of methodology: F. Faião-Flores, M.F. Emmons, B. Saha, B. Fang, S.P. Chellappan, U. Rix, K.S.M. Smalley
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Faião-Flores, F. Kinose, B. Saha, B. Fang, S.P. Chellappan, J.W. Harbour
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Faião-Flores, M.F. Emmons, M.A. Durante, B. Fang, J.W. Harbour, K.S.M. Smalley
Writing, review, and/or revision of the manuscript: F. Faião-Flores, M.F. Emmons, M.A. Durante, B. Saha, B. Fang, J.M. Koomen, U. Rix, J.D. Licht, J.W. Harbour, K.S.M. Smalley
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Faião-Flores, J.M. Koomen
Study supervision: F. Faião-Flores, S.S. Maria-Engler, K.S.M. Smalley
The authors would like to thank Dr. Manali Phadke for technical assistance and support. This work is supported by the Bankhead-Coley Program of the State of Florida 7BC05, and the NIH R21 CA216756. It has been also supported, in part, by the SAIL Core Facility, the IRAT Core Facility, the Flow Cytometry Core Facility, the Analytic Microscopy Core Facility, the Proteomics and Metabolomics Core Facility, the Molecular Genomics Core and the Tissue Core Facility at the H. Lee Moffitt Cancer Center & Research Institute (Tampa, FL), an NCI designated Comprehensive Cancer Center (P30-CA076292). This work has also been supported, in part, by Fapesp (grant no. 2013/05172-4 and 2015/10821-7).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.