UHRF1/UBE2L6/UBR4-mediated ubiquitination regulates EZH2 abundance and thereby melanocytic differentiation phenotypes in melanoma

Cellular heterogeneity in cancer is linked to disease progression and therapy response, although mechanisms regulating distinct cellular states within tumors are not well understood. We identified melanin pigment content as a major source of cellular heterogeneity in melanoma and compared RNAseq data from high-pigmented (HPCs) and low-pigmented melanoma cells (LPCs), suggesting EZH2 as a master regulator of these states. EZH2 protein was found to be upregulated in LPCs and inversely correlated with melanin deposition in pigmented patient melanomas. Surprisingly, conventional EZH2 methyltransferase inhibitors, GSK126 and EPZ6438, had no effect on LPC survival, clonogenicity and pigmentation, despite fully inhibiting methyltransferase activity. In contrast, EZH2 silencing by siRNA or degradation by DZNep or MS1943 inhibited growth of LPCs and induced HPCs. As the proteasomal inhibitor MG132 induced EZH2 protein in HPCs, we evaluated ubiquitin pathway proteins in HPC vs LPCs. Biochemical assays and animal studies demonstrated that in LPCs, the E2-conjugating enzyme UBE2L6 depletes EZH2 protein in cooperation with UBR4, an E3 ligase, via ubiquitination at EZH2’s K381 residue, and is downregulated in LPCs by UHRF1-mediated CpG methylation. Targeting UHRF1/UBE2L6/UBR4-mediated regulation of EZH2 offers potential for modulating the activity of this oncoprotein in contexts in which conventional EZH2 methyltransferase inhibitors are ineffective.


INTRODUCTION
The ability of cancer cells to adapt, survive and proliferate during tumor growth, metastasis, and treatment is a major impediment to improving patient outcomes. As the same mechanisms that enable cancer cell adaptation also fuel intratumoral heterogeneity (ITH), understanding drivers of ITH is a high priority in cancer research [1].
ITH is facilitated by phenotype switching, in which individual cells reversibly switch via epigenetic mechanisms between different transcriptional programs, functional states and differentiation phenotypes [2][3][4]. In melanoma, switching between invasive and proliferative states was shown to be regulated by enhancer of zeste homolog 2 (EZH2) [3,4], with the proliferative state characterized by high expression of the melanocyte transcription factor MITF and low expression of BRN2 and AXL, and opposite expression patterns found in the invasive state.
Highly specific EZH2 methyltransferase inhibitors, GSK126 and EPZ-6438, are currently in clinical trials [13]. Although these have shown antitumor effects in lymphomas with enzymeactivating mutations of EZH2 [22,23] and in ovarian cancer cells with inactivating mutations of ARID1A [24], some cancers appear resistant to EZH2 methyltransferase inhibition but sensitive to genetic depletion of EZH2, raising the possibility that EZH2 promotes tumorigenesis via methyltransferaseindependent mechanisms. It has also been shown that EZH2 enzymatic inhibitors have better activity against EZH2 mutant melanoma cells than wild type cells [25]. Indeed, as EZH2 can promote cancer independently of its methyltransferase activity [25] or by acting as a transcriptional coactivator [26][27][28], targeting EZH2 abundance may be more effective than enzymatic EZH2 inhibition against cancers that are driven by EZH2 methyltransferase-independent mechanisms. Cellular protein activity and stability are regulated by posttranslational protein modifications [29] such as ubiquitination, which involve reversible addition of ubiquitin proteins to lysine residues of target substrates [30]. This is catalyzed by a series of enzymes: (1) ubiquitin-activating enzymes (E1) that use ATP to form ubiquitin-thioester bonds, (2) ubiquitin-conjugating enzymes (E2) that bind active ubiquitin to cysteine residues, and (3) ubiquitin-ligase enzymes (E3) that interact with E2 enzymes, catalyze formation of covalent bonds between ubiquitin and target substrates, and regulate substrate specificity. EZH2 protein undergoes ubiquitin-dependent degradation in different contexts by several E3 ligases, including β-TrCP, SMURF2, PRAJA1, MDM2 and FBW7, [31][32][33][34][35]. Recently, UBE2L6 was shown to be a tumor-suppressor and a prognostic marker in melanoma [36]. Here, we report UBE2L6 as an E2 ubiquitinconjugating enzyme for EZH2 that regulates EZH2 abundance in melanoma cells.

MATERIALS AND METHODS Mice and tumors
Eight-week-old NOD SCID IL2R-/-mice (NSG) were injected subcutaneously with empty vector-or UBE2LB_pLX307-containing 28:B4:F3 melanoma cells (n = 11 mice per group). Tumors were measured weekly using calipers. Once a tumor reached 20 mm diameter, mice were sacrificed. All animal experiments were performed according to approved protocol E/1792/2018/ M (AMREP Animal Ethics Committee of Monash University).

Human melanoma tumor samples
19 pigmented and 20 non-pigmented human melanoma tissue sections were obtained from Melanoma Research Victoria (MRV).
Cell lines, chemicals, plasmids, and siRNA The cell lines, chemicals, plasmids, siRNAs, and qPCR primers used for treatment in this study are listed in Table S1.
For cell lines, mycoplasma tests were performed in our lab. Short tandem repeat (STR) profiling was done in the Australian Genome Research Facility (AGRF) to authenticate cell lines.

Western blot
Total proteins were extracted and analysed as described [37], using 4-20% Mini-PROTEAN TGX Stain-Free Protein Gels (BioRad, 4568096). PVDF membranes were incubated with respective antibodies outlined in Table S1. Signals were detected using Clarity ECL Western blotting Substrate (BioRad) and quantified by ImageJ.

Quantitative PCR and Methylation-sensitive PCR (MSP)
Total RNA was extracted and underwent qPCR according to manufacturer's instructions and as described [37]. RNA expression changes were determined using a ΔΔCt method [38]. Primers are listed in Table S1. Methylation of UBE2L6 promoters was analyzed as described [39].
RNA-Seq data and gene set enrichment analysis FASTQ files were processed via Laxy and the RNAsik pipeline. The GRCh38 reference genome was used for STAR alignment [43] and gene expression counts were performed using featureCounts [44]. Gene counts were analysed using Degust (https://zenodo.org/record/3501067) for differential expression (DE) analysis (limma voom). R package ClusterProfiler (v.4.0.0) was used for Gene Set Enrichment Analysis (GSEA) and overrepresentation analysis (ORA) [45]. For GESA, DE gene lists were ranked by logFC and the testing method was set to "fgsea". For both GSEA and ORA, the following annotated gene sets were tested for significant enrichment: Gene Ontology (GO), Reactome, Hallmark and Oncogenic signatures. Genesets were downloaded from the Molecular signatures database (MSigDB) [46]. Genetic variants were identified using the GATK RNAseq variant discovery pipeline.

The Cancer Genome Atlas (TCGA) survival analysis
Clinical data and mRNA expression profiles for skin melanoma samples in TCGA PanCancer Atlas database were accessed through the public cBioPortal (http://www.cbioportal.org) [47]. Using the R package 'survminer' 0.4.4 and 'survival' 3.1-11, respectively, overall survival (OS) Kaplan-Meier curves were plotted, and the log-rank test was performed.

Liquid chromatography-tandem mass spectrometry (LC/MS) interactome and post-translational modification (PTM) analysis
For LC/MS interactome studies, a Dionex UltiMate 3000 RSLCnano system equipped with a Dionex UltiMate 3000 RS autosampler, an Acclaim PepMap RSLC analytical column and an Acclaim PepMap 100 trap column (Thermo Scientific) were used.
For PTM analysis, raw files were searched with Byonic v3.0.0 (Protein-Metrics) using GlyGly at lysine as a variable modification. Only peptides and proteins falling within a false discovery rate (FDR) of 1% based on a decoy database were further analyzed.

Statistical analysis
Analyses were performed using GraphPad Prism version 7. Analyses were performed using log-rank tests, unpaired two-tailed t-tests, one-way ANOVA, or Tukey's multiple comparison tests as appropriate to the data type. P values less than 0.05 were considered significant.

RESULTS
Low pigmented melanoma cells demonstrate upregulation of EZH2 protein and EZH2-target genes To interrogate functional differences amongst melanoma cells according to their levels of melanin pigment, RNAseq was performed on low pigmented cells (LPCs) and high pigmented cells (HPCs) purified from 28:F3:B4 cells by flow-cytometry [42]. GSEA analysis revealed that the LU-EZH2-target-DN gene set [51] was markedly enriched in LPCs (p = 2 × 10 −4 ) (Fig. 1A, B). We thus also compared by GSEA differentially expressed gene signatures (DEGs) in sorted LPCs and HPCs from B16-F10 cells with those of EZH2 silenced B16-F10 cells (Table S2 and S3). Thirty-one of 153 genes upregulated in LPC were EZH2-activated genes (Fig. 1C, top panel Venn diagram) and 96 of 209 genes downregulated in LPCs were EZH2-repressed genes (Fig. 1D, top panel Venn diagram). While over-representation analysis (ORA) by Gene Ontology biological processes of the 31 EZH2-activated genes showed enrichment for ribosomal biogenesis (Fig. 1C, bottom panel and Table S4), the 96 EZH2-repressed genes were enriched in melanin biosynthesis pathways (Fig. 1D, bottom panel and Tables S5 and S6), including Oca2, which was one of the genes most negatively correlated (p = 0.002, Table S7) with EZH2 in the TCGA SKCM. These data suggested that increased EZH2 activity in LPCs compared to HPCs might upregulate ribosomal biogenesis and/ or suppress melanocytic differentiation and melanin synthesis.
To investigate this, we evaluated EZH2 mRNA and EZH2 protein levels, as well as a marker of EZH2's methyltransferase activity, H3K27me3, in HPCs and LPCs from 28:F3:B4, B16-F10 and C006-M1 pigmented melanoma cells ( Fig. S1A-C). Western blotting and immunofluorescence (IF) staining demonstrated that EZH2 and H3K27me3 were higher in LPCs than HPCs (Fig. 1E-H and Fig. S1D). However, no significant difference was observed between LPCs and HPCs in EZH2 mRNA levels ( Fig. 1I, J, and Fig. S1E). An inverse association between melanoma cell pigmentation and EZH2 protein was observed in vivo in PDX melanomas (Fig. 1K).
EZH2 alters pigmentation and malignant behavior independently of its methyltransferase function To investigate EZH2 in melanoma cell pigmentation, we suppressed EZH2 protein and/or EZH2 activity in pigmented melanoma cells. Because high EZH2 expression is seen in LPCs, correlations between EZH2 and pigmentation were investigated in B16-F10 cells, which display variable pigmentation. Unsorted B16-F10 cells were EZH2-silenced using two siRNAs, which induced pigmented phenotypes ( Fig. 2A) and upregulation of melanocytic differentiation markers MITF and TYR that were proportional to knockdown efficiency ( When B16-F10 cells were treated with DZNep, a general methyltransferase inhibitor that also promotes ubiquitinmediated proteasomal degradation, we observed increased MITF, decreased EZH2 and H3K27me3 (Fig. 2G), and more prominent pigmented cell phenotypes (Fig. S2E, F), including by flow cytometric analysis ( Fig. 2H and Fig. S2G). Extracellular melanin was also induced after EZH2 knockdown or DZNep (Fig. S2H).
Although EZH2 methyltransferase inhibitors GSK126 and EPZ6438 fully inhibited EZH2 methyltransferase activity as measured by H3K27me3 (Fig. 2P), they had no or minimal effects on pigmentation, viability or clonogenicity of the melanoma cells we tested ( Fig. 2Q-R). Thus, specific inhibition of EZH2 methyltransferase did not phenocopy the interventions that reduced total EZH2 protein. This suggests that EZH2 abundance, but not its methyltransferase activity, promotes the LPC state.
EZH2 expression is higher in metastases than primary melanomas If EZH2 positively regulates malignant behaviors in melanoma cells, then its expression might be predicted to increase during disease progression. As increasing EZH2 protein expression was  observed from benign nevi to metastatic melanoma [7,16,17] and as melanoma patient survival was inversely correlated with EZH2 expression [7], we sought to extend these findings. Analysis of TCGA melanoma samples showed no correlation between EZH2 mRNA levels and overall patient survival or disease staging in melanoma (Fig. S3A, B), consistent with our findings (Fig. 1I, J and Fig. S1E). However, in 39 melanoma samples obtained from patients in the Melanoma Research Victoria cohort, incrementally increasing EZH2 protein expression was observed from early stage to metastatic disease (Fig. 3A), confirming published data [7].
UBE2L6 promotes ubiquitin-associated proteosomal degradation of EZH2 in melanoma cells To understand discrepancies between EZH2 mRNA and EZH2 protein levels in melanoma cells, we investigated posttranslational mechanisms that might regulate EZH2. Since ubiquitin-mediated proteasomal degradation plays a pivotal role in protein abundance by affecting protein stability [53], we assessed EZH2 levels pre-and post-treatment with a proteasome inhibitor, MG132, in sorted LPCs and HPCs, and in unsorted B16-F10 cells. MG132 treatment increased EZH2 protein in HPCs, but not in LPCs, in a time-dependent manner (Fig. 3B, C), suggesting increased proteasomal degradation of EZH2 in HPCs compared to LPCs.
To elucidate mechanisms underlying this, we interrogated our RNAseq data to identify candidate ubiquitin pathway enzymes involved in the post-translational regulation of EZH2. Among E1 and E2 enzymes, Uba7 (Fig. S3C), and Ube2l6 and Ube2c (Fig. 3D,  Fig., S3D), respectively, were significantly decreased in LPCs compared to HPCs. To explore ubiquitination in EZH2 regulation, we focused on Ube2l6, as Ube2c overexpression had trivial effects on EZH2 protein (Fig. S3E), and E1 enzymes have a wide range of substrates.

UBE2L6 reduces EZH2 abundance, tumorigenicity and metastasis in melanoma
To evaluate functional consequences of UBE2L6-EZH2 interactions in melanoma, we developed stable UBE2L6-WToverexpressing pigmented 28:B4:F3 and IGR37 cells. Overexpression of UBE2L6-WT decreased EZH2 protein and cell viability, clonogenicity, invasion and pigmentation (Fig. 3M, N  and Fig. S3J, K). Following subcutaneous injection into mice, tumor formation was impeded in UBE2L6-overexpressing 28:B4:F3 recipient mice compared to controls (Fig. 3O), although differences in tumor volume were not statistically significant at 15 weeks (Fig. 3P). While control tumors were partially pigmented, they were all pigmented in UBE2L6-overexpressed tumors (Fig. S4A), which also demonstrated lower EZH2 protein levels (Fig. S4B) and seeded fewer metastases (Fig. 3Q-T) that were smaller and also more pigmented than metastases from control tumors ( Fig. 3U and Fig. S4C-E). Interestingly, all tumor bearing mice developed lymph node metastasis irrespective of UBE2L6 status (data not shown).
Further to examine UBE2L6 in human melanoma, we performed immunohistochemical (IHC) staining on 19 pigmented and 20 non-pigmented human melanomas. Pigmented regions in tumors exhibited high UBE2L6 but low EZH2 expression, whereas non-pigmented regions displayed the inverse phenotype ( Fig. 4A and Fig. S5A). To complement these observations, we also analyzed UBE2L6 expression in five random fields of tumor sections and plotted it against EZH2 expression scores, revealing a negative correlation between UBE2L6 and EZH2 (linear regression R 2 = 0.62, p < 0.0001; Fig. 4B). Interestingly, a mild decline (p = 0.0575) was noted in cytosolic UBE2L6 from early stage to metastatic melanoma, inversely correlated with EZH2 ( Fig. S5B and Fig. 3A). We also analyzed TCGA melanoma data, finding a correlation between low UBE2L6 mRNA expression and poor melanoma survival (p < 0.0001, Fig. S5C). Interestingly, UBE2L6 mRNA was negatively correlated with its promoter CpG methylation levels (Spearman correlation r = −0.63, p = 2.2 × 10 −16 , Fig. S5D). These data implicate UBE2L6 as a tumor suppressor in melanoma. Fig. 3 Ezh2 is proteasomally degraded via Ube2l6. A EZH2 expression according to melanoma tumor stage in the Melanoma Research Victoria patient cohort. Stage I (n = 12 patients), Stage II (n = 10 patients), Stage III (n = 9 patients) and Stage IV (n = 8 patients). Data are presented as mean ± SD and analyzed by one-way ANOVA plus Tukey's multiple comparison test. *p < 0.05, **p < 0.01. B Ezh2 (red) IF in HPCs and LPCs sorted from B16-F10 cells and treated with 10 µM MG132 or DMSO control for 16 h. DAPI (blue): nuclei. Scale bar: 10 µm. C Ezh2 protein levels determined by western blot in B16-F10 cells treated with 10 µM MG132 or DMSO control for 4 h or 8 h. D E2 ligases downregulated in RNAseq data from LPCs vs HPCs. E Ube2l6 IF in LPCs and HPCs from B16-F10 cells. Scale bar: 10 µm. F Ezh2 and H3K27me3 levels were determined by western blot in B16-F10 cells transfected with Flag-tagged Ube2l6-WT, Flag-tagged Ube2l6-C87A (enzyme-dead) or empty vector. (G) As in (F), with or without 10 µM MG132 treatment for 16 h. H Stability of endogenous Ezh2 protein determined by western blot in B16-F10 cells transfected with Flag-tagged Ube2l6-WT or Flag-tagged Ube2l6-C87A, followed by 50 µg/mL CHX treatment for the indicated times. I Densitometry for western blots shown in (H). Data from three independent experiments are presented as mean ± SD and were analyzed by one-way ANOVA plus Tukey's multiple comparison test. *p < 0.05, **p < 0.01, ***p < 0.001. J Flag-tagged Ube2l6-WT was overexpressed in B16-F10 cells maintained in 10 µM MG132 for 16 h. Interactions between endogenous Ezh2 and Ube2l6 were determined by immunoprecipitation with anti-Ezh2 antibody followed by western blotting with anti-Flag antibody. K HEK293 cells were co-transfected with Flag-tagged EZH2 and Flag-tagged Ube2l6-WT or Flag-tagged Ube2l6-C87A, together with HA-tagged ubiquitin, or (L) HA-tagged ISG15 in the presence of 10 µM MG132. The ubiquitination (K) and ISGylation (L) of EZH2 were determined by anti-HA IP followed by western blot with anti-EZH2 antibody. M 28:B4:F3 and IGR37 cells were infected with V5-tagged empty vector or V5-tagged UBE2L6 lentiviral particles. Positive clones were selected by incubation in 2 µg/µL puromycin for 2 weeks. EZH2 was detected by western blot in stably transfected cells.  (log-rank test). I-K A375 human melanoma cells were transfected with one of two siRNAs against Uhrf1, or scrambled controls, followed by: (I) MSP analysis of UBE2L6 promoter as in (C), (J) EZH2 and UBE2L6 protein estimation by western blot, and (K) cell number by Trypan blue cell counting. L Immunohistochemical staining of UHRF1 and UBE2L6 in representative Schmorl's stained pigmented human melanomas. Scale bar, 50 μm. M UBE2L6 protein scores (x-axis) in melanoma samples negatively correlated with UHRF1 scores (y-axis) in individual patients. Pink dots correspond to non-pigmented and purple dots to pigmented patient samples. P-value calculated from linear regression analysis. R = correlation coefficient. Protein score = percentage of immunopositive cells × immunostaining intensity. Data for D, E and K from three independent experiments are presented as means ± SD, analyzed by one-way ANOVA plus Tukey's multiple comparison test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

The E3 ligase UBR4 interacts with UBE2L6 and EZH2
The activity of E2 conjugating enzymes relies on their cooperation with E3 ligases through which occur direct interactions with substrates. To identify EZH2-bound E3 ligases in melanoma, EZH2-co-immunoprecipitated lysates obtained from a variety of melanoma cell lines (A375, 28:B4:F3, B16-F10, LM-MEL45 and IGR37) were subjected to Liquid Chromatography-Mass Spectrometry (LC-MS). Besides PRC1 and PRC2 complex proteins, we identified novel EZH2 binding proteins (Table S8). We focused on UBR4, because it was the only common EZH2-interacting E3 ligase identified in the cell lines tested (Fig. 5A). No differences in UBR4 expression were seen between HPCs and LPCs (Fig. S5H) and pigmented patient samples (Fig. S5I), in contrast to UBE2L6 (Fig. 3E).
To complement these data, we performed sequence alignment, revealing that the K381 residue has been conserved in evolution, suggesting its critical for the K381 in EZH2 regulation (Fig. 6C). To test EZH2 K381 ubiquitination functionally, EZH2 K381R mutant proteins were expressed in HEK293 cells, inducing EZH2 stability (Fig. 6D, E) and resistance to UBE2L6-mediated EZH2 degradation (Fig. 6F). Accordingly, melanoma cell viability and invasiveness were promoted by overexpression of EZH2-K381R in UBE2L6overexpressing cells (Fig. 6G-I). These data implicate UBE2L6 as a tumor suppressor in melanoma that couples with UBR4 to direct ubiquitination of EZH2 K381 residues and ubiquitin-mediated degradation of EZH2 (Fig. 6J).

DISCUSSION
We identified EZH2 as a master regulator of melanoma ITH and promoter of malignant behavior in melanoma cells. Genetic or pharmacological depletion of EZH2, but not specific inhibition of its methyltransferase activity, inhibited melanoma cell tumorigenicity and invasion, indicating a critical methyltransferaseindependent function for EZH2 in melanoma. This effect was modulated by UHRF1/UBE2L6/UBR4-mediated inhibition of EZH2 through proteosomal degradation, implicating UHRF1 and UBE2L6 as regulators of ITH and disease propagation in melanoma. Targeting these enzymes might be exploitable in anti-melanoma therapy.
Previous studies implicated genes involved in melanin biosynthesis and melanocytic differentiation as key targets of EZH2directed histone methylation [57,58], leading to altered pigmentation and differentiation phenotypes in melanoma cells [59]. Indeed, we found that EZH2 protein was inversely associated with pigmentation in both melanoma cells and patient melanomas. However, unlike other studies, we found that selective EZH2 methyltransferase inhibition using GSK126 or EPZ6438 did not substantially affect melanocytic gene expression [58] or cell pigmentation. This discrepancy may be explained by off-target effects of high inhibitor doses used in other studies.
It is thought that phenotype switching produces melanoma cell populations of varied proliferative and invasive activity.  5 UBR4 is the E3 ligase of EZH2. A Venn diagram depicting overlap of proteins co-immunoprecipitating with EZH2 from C006-M1, 28:B4:F3, IGR37, A375 and LM-MEL-45 cells (data derived from n = 3 biological replicates). B HA-tagged EZH2 and Flag-HA-tagged UBR4-LD (ligase domain) were co-expressed in HEK293 cells maintained ±MG132. Interactions between EZH2 and UBR4-LD were determined by immunoprecipitation with anti-EZH2 antibody and western blotting with anti-Flag antibody. C EZH2 protein was determined in HEK293 cells overexpressing either UBR4-LD or UBR4-FL (full length). UBR4 overexpression was measured by qPCR (lower panel: quantification relative to vector control, error bars: + SD). D UBR4 in 28:B4:F3, C006-M1 and HEK293 cells was knocked down by siRNA. EZH2 levels determined by western blot (top panels). Knockdown efficiency of UBR4 measured by qPCR as in (C). E Stability of EZH2 according to UBR4 levels (unaltered vs knockdown) was determined by western blot in A375 cells treated with cyclohexamide (CHX). Representative blot shown at the top, timecourse plot at the bottom. F Ubiquitination of EZH2 was determined by anti-HA IP followed by western blot in A375 cells. G HEK293 cells with or without siUBR4 were transfected with HA-tagged EZH2 or V5-tagged UBR4-FL followed by 50 µM MG132 treatment for 4 h. Ubiquitination of EZH2 was determined by western blot with anti-ubiquitin (Ub) antibody. H B16-F10 cells ±siUBR4 were transfected with either Flag-tagged Ube2l6-WT ("wt") or Flag-tagged Ube2l6-C87A ("mut") and EZH2 levels determined by western blot. I-K To evaluate EZH2/UBR4/UBE2L6 interactions, B16-F10 cells ±siUBR4 were transfected with either Flag-tagged Ube2l6-WT ("wt") or Flag-tagged Ube2l6-C87A ("mut") followed by 50 µM MG132 treatment for 4 h and then: (I) EZH2, UBR4 and UBE2L6 protein estimation by western blot, (J) determination of ubiquitination of EZH2 by western blot with anti-Ub antibody, and (K) co-immunoprecipitation of UBR4 and UBE2L6 with EZH2 antibody. Data represent n = 3 biological replicates. Data for D and E are from three independent experiments and presented as means ± SD, analyzed by one-way ANOVA plus Tukey's multiple comparison test. **p < 0.01.
Invasive melanoma cells have lower pigment levels compared with non-invasive cells [4,56,59,60] and Ezh2 promotes the low-pigment phenotype in vivo by suppressing Oca2, among other targets [59]. Although Pinner et al., [60] found that nonpigmented invasive melanoma cells metastasize to lung and lymph nodes and reverted to pigmented phenotypes after colonization, we did not observe this in our study. Rather, we found that melanoma cells with high EZH2 expression generally formed partially pigmented tumors following subcutaneous implantation, with non-pigmented metastatic foci in the liver, and that UBE2L6 over-expressing melanoma cells with low EZH2 formed highly pigmented tumors with either no or only small and pigmented metastatic foci.
It was also suggested that EZH2 is required for efficient lung and lymph node colonization by metastatic melanoma [58,61]. Consistent with this, we found that metastases to the lung and liver were reduced from tumors formed by UBE2L6-overexpressing cells with suppressed EZH2. Interestingly, all tumor-bearing mice developed lymph node metastasis regardless of UBE2L6 status. This suggests that EZH2 may have a greater role in facilitating hematogenous metastasis rather than early dissemination into lymphatics. Alternatively, other UBE2L6 targets may regulate melanoma metastasis.
Tumor-suppressive and anti-metastatic roles of UBE2L6 in vivo have not been described, and we found are mediated in melanoma by the ubiquitin-conjugating function of UBE2L6, working cooperatively with the E3 ligase UBR4, to induce EZH2 ubiquitination and protein degradation [53]. We found that UBE2L6 abundance is itself regulated by UHRF1, which induces suppressive UBE2L6 promoter methylation and supports oncogenic EZH2 proteins. UHRF1 acts as an oncogene in various cancers, including melanoma [54,55], and downregulates UBE2L6 in cervical cancer [54]. Our studies extend understanding of this oncoprotein by demonstrating UHRF1/ UBE2L6/UBR4-mediated EZH2 promotion to enhance LPC phenotypes and disease progression in melanoma, providing a mechanism as to how higher levels of EZH2 are associated with late-stage disease.
Most anti-cancer strategies focused on EZH2 targeting have assumed that EZH2 methyltransferase activity is the key oncogenic activity of this protein. However, oncogenic and methyltransferase-independent functions of EZH2 have been described [26][27][28]62], leading to strategies that reduce EZH2 abundance. For example, in breast cancer, DZNep reduces EZH2 protein stability by upregulating the E3 ligase PRAJA1 [35], and MS1943 depletes EZH2 and induces cytotoxicity [52]. Although we found that DZNep reduces EZH2 in melanoma by inducing UBE2L6 expression, its short half-life [63,64] and broad effects on ubiquitination hinder clinical application. Targeted depletion of EZH2 through facilitating UHRF1/UBE2L6/UBR4mediated ubiquitination of its K381 residue offers a specific and appealing alternative.

DATA AVAILABILITY
All relevant data are available from the authors upon request. Fig. 6 The K381 residue on EZH2 is ubiquitinated by UBE2L6/UBR4 in melanoma cells. A Prediction of ubiquitination sites on human EZH2 was made using UbPred, shown as low (green), medium (blue) and high (red) confidence lysine ubiquitination sites (www.ubpred.org) (B) LC-MS analysis of mouse Ezh2 K376 ubiquitination in B16-F10 cells. C Sequence alignment of residues 368-392 of human (Hs) EZH2 protein against Mus musculus (Mm, mouse), Rattus norvegicus (Rn, rat), Bos taurus (Bt, cow), Gallus gallus (Gg, chicken), Danio rerio (Dr, zebrafish), Drosophila melanogaster (Dm, fruit fly), Anopheles gambiae (Ag, mosquito), and Caenorhabditis elegans (Ce, worm) showing conservation of K381 residues. D HEK293 cells with HA-tagged EH2-WT or HA-tagged EZH2-K381A were treated with 50 µg/mL CHX (cycloheximide) for the durations indicated. Stability of HA-tagged EZH2 protein was determined by western blot with anti-HA antibody. E Quantitated time course of EZH2 protein levels from (D). F B16-F10 cells were transfected with either Flag-tagged Ube2l6-WT or Flag-tagged Ube2l6-C87A (mutant) and HAtagged EZH2-WT or HA-tagged EZH2-K381A, and HA-EZH2 levels determined by western blot with anti-HA antibody. G A375 and IGR37 cells transduced with V5-UBE2L6-WT were transfected transiently with either HA-tagged EZH2-WT or HA-tagged EZH2-K381R. Endogenous EZH2 levels were determined 48 h post-transfection with EZH2 antibody, HA-EZH2 levels were determined with anti-HA antibody and V5-UBE2L6 levels with V5 antibody. H Cells treated as in (G) were counted by Trypan blue 48 h post-transfection. I Invasion of cells treated as in (G) was measured 72 h post-transfection using Boyden chamber matrigel invasion assays followed by crystal violet staining. J Proposed model for UHRF1/UBE2L6/UBR4-mediated regulation of EZH2 and thereby melanocytic differentiation phenotypes in melanoma. Data for E and H from three independent experiments are presented as means ± SD, analyzed by one-way ANOVA plus Tukey's multiple comparison test. *p < 0.05, **p < 0.01, ***p < 0.001.