Quantitative Histone Mass Spectrometry Identifies Elevated Histone H3 Lysine 27 (Lys27) Trimethylation in Melanoma*

Normal cell growth is characterized by a regulated epigenetic program that drives cellular activities such as gene transcription, DNA replication, and DNA damage repair. Perturbation of this epigenetic program can lead to events such as mis-regulation of gene transcription and diseases such as cancer. To begin to understand the epigenetic program correlated to the development of melanoma, we performed a novel quantitative mass spectrometric analysis of histone post-translational modifications mis-regulated in melanoma cell culture as well as patient tumors. Aggressive melanoma cell lines as well as metastatic melanoma were found to have elevated histone H3 Lys27 trimethylation (H3K27me3) accompanied by overexpressed methyltransferase EZH2 that adds the specific modification. The altered epigenetic program that led to elevated H3K27me3 in melanoma cell culture was found to directly silence transcription of the tumor suppressor genes RUNX3 and E-cadherin. The EZH2-mediated silencing of RUNX3 and E-cadherin transcription was also validated in advanced stage human melanoma tissues. This is the first study focusing on the detailed epigenetic mechanisms leading to EZH2-mediated silencing of RUNX3 and E-cadherin tumor suppressors in melanoma. This study underscores the utility of using high resolution mass spectrometry to identify mis-regulated epigenetic programs in diseases such as cancer, which could ultimately lead to the identification of biological markers for diagnostic and prognostic applications.

Melanoma is a deadly variety of skin cancer, accounting for 75% of skin cancer-related deaths. In 2015, melanoma is expected to be the fifth most common cancer in men and the seventh most common cancer in women. According to the World Health Organization, it is estimated that melanoma will result in the death of around 65,000 people globally and 9940 people in the United States in 2015. The high mortality rate associated with metastatic melanoma suggests a lack of efficient diagnostic and prognostic biomarkers (1). EZH2 1 expression has often been positively correlated to the progression of different types of cancer (2,3). Increased expression of EZH2 has been identified in melanoma tissues (4) as well as prostate, breast, bladder, and liver cancers and has been recognized as a prognostic marker for aggressive prostate and breast cancer (5,6). EZH2 is a histone modifier that functions as the catalytic component of the Polycomb Repressive Complex 2 (PRC2) (7,8). It is a lysine methyltransferase and promotes the addition of the repressive marker histone H3K27me2/me3 to target chromatin, thereby inducing chromatin compaction and transcriptional repression. Chromatin condensation/compaction leads to transcriptional repression by restricting access to transcriptional regulators like RNA polymerase II and other transcription-associated factors. Hence, silencing of tumor suppressor genes by H3K27me3 is implicated in the initiation and advancement of different types of cancer (9 -11). H3K27me3-silenced tumor suppressor genes include RUNX3 (12), E-cadherin (13), SLIT2 (14), DAB2IP (15), and FBXO32 (3). RUNX3 and E-cadherin were points of focus for this study.
The tumor suppressor RUNX3 is one of several prognostic biomarkers proposed for melanoma (16). RUNX3 is coded by the RUNX3 gene, and along with RUNX1 and RUNX2 it constitutes the runt domain family of transcription factors. Members of the RUNX family regulate major developmental pathways besides promoting growth arrest in response to oncogenic RAS (17). RUNX3 has been reported to regulate the cell cycle and induce apoptosis by inhibiting cyclin-dependent kinases (12). Hence, RUNX3 suppression is a key step in carcinogenesis in different types of cancer such as leukemia (18), lung cancer (19), and gastric cancer (20). Repression of the tumor suppressor RUNX3 via EZH2-mediated H3K27 tri-methylation leads to increased cellular proliferation in cancers such as breast cancer (9) and neuroblastoma (21), which is key to tumor formation and maintenance. Transcriptional silencing of RUNX3 has also been reported to occur by DNA hypermethylation of CpG islands (22), hemizygous deletion (23), and by miR-532-5p, a micro-RNA targeting RUNX3 mRNA sequences (24).
Tumor progression from in situ to invasive to metastasis involves loss of cell-cell adhesion and expression of factors that allow tumor cells to degrade, to cross basement membrane and endothelial cell barriers, and to migrate to distant tissues. In many epithelial tumors, this process is associated with the loss of plasma membrane-associated adhesion protein E-cadherin (25). E-cadherin is a member of the cadherin family of transmembrane proteins and mediates cell-cell adhesion. Expression of E-cadherin is decreased during the progression of many types of epithelial tumors and has been linked with the development of metastases in different cancers like gastric cancer (26), non-small cell lung cancer (27), and breast cancer (28).
To gain functional insight into histone epigenetic mechanisms playing roles in melanoma progression, we performed quantitative mass spectrometric analysis of histone posttranslational modifications (PTMs) in melanocytes, two melanoma cell lines, one representing late stage, aggressive melanoma (WM266-4) and one representing early stage, less aggressive melanoma (WM115), and three types of patient tumors, benign nevi, primary melanoma, and metastatic melanoma. A variety of mass spectrometric approaches have been reported for performing quantitative analyses of histone PTMs (29). These include top-down as well as bottom-up approaches that can be quantified in various ways such as isotope labeling and label-free methods. In this study, we utilized a label-free precursor ion intensity approach for a bottom-up analysis of histone PTMs (30). To increase sequence coverage for the lysine-and arginine-rich histones that yield very small peptides upon trypsin digestion, we used d 6 -acetic anhydride to chemically label unmodified and monomethylated lysines with an isotopically heavy acetyl group, which blocks trypsin activity and thereby increases sequence coverage following trypsin digestion (31,32). By increasing the sequence coverage, we can both increase detection of potential sites of modification as well as obtain larger peptides containing multiple sites of potential modification thereby gaining insight into combinatorial histone modifications on individual histone molecules. Our precursor ion intensity-based analysis of histone peptides revealed that the aggressive melanoma cell line WM266-4 and metastatic melanoma patient samples had elevated levels of trimethylated histone H3 at Lys 27 which is the enzymatic product of EZH2. The biological significance of this finding was validated both in cell lines and patient tissues at the tumor suppressors RUNX3 and E-cadherin, which are transcriptionally silenced in aggressive melanoma. There has been no study to date directly linking increased EZH2-mediated H3K27 tri-methylation to suppression of tumor suppressors RUNX3 and Ecadherin in advanced melanoma. Our results represent some of the first evidence that an epigenetic reprogramming of H3K27me3 can occur in aggressive melanoma leading to direct silencing of tumor suppressor genes.

EXPERIMENTAL PROCEDURES
Cell Culture-The human melanoma cell lines A2058, A375, WM266-4, and WM115 were purchased from ATCC. The melanocytes HEMa-LP (Thermo Fisher Scientific, catalog no. C-024-5C) were primary human epidermal melanocytes isolated from adult skin. For constructing the stable cell line WM115 EZ, cDNA of EZH2 was cloned from pCMVHA hEZH2 (42) into pIRES2-EGFP-Ofd1-Myc plasmid (43) and used to transfect the WM115 cells. The WM115EZ cells were maintained in DMEM in the presence of G418 sulfate (400 g/ml) (EMD Millipore, Billerica, MA). pCMVHA hEZH2 was a gift from Kristian Helin (Addgene plasmid 24230), and pIRES2-eGFP-Ofd1-Myc was a gift from Jeremy Reiter (Addgene plasmid 24560). For all drug treatments, DMSO was used as negative control. GSK126 was purchased from Active Biochemicals (Maplewood, NJ). Cells were cultured with DMEM supplemented with 10% fetal bovine serum (FBS) and 10% penicillin/streptomycin/L-glutamine and incubated at 37°C in a humidified incubator with 5% CO 2 .
Experimental Design and Statistical Rationale-High resolution mass spectrometric analysis was performed for the cell lines melanocytes, WM115, and WM266-4, as well as FFPE patient tumors in both biological and technical triplicates. According to the power.t.test analysis in R, biological triplicate samples from the cell lines and patient tumors with a 30% standard deviation, large effect size, and a significance level of 0.05 generated a power of 0.69. Therefore, the minimum number of samples required for high quality statistical testing was used in the experiment. We performed the mass spectrometric analysis in technical triplicates to account for instrument variability. Spearman correlation analysis indicated we were able to achieve high reproducibility of both technical (p value Ͻ 0.0001) and biological replicates (p value Ͻ 0.0001) (SAS version 9.4). The two data sets were analyzed separately for cell lines WM115, WM266-4, and melanocytes and FFPE tissues.
Histones were acid-extracted from cells grown in culture as reported previously (33) and resolved (20 g of histones per lane) by SDS-PAGE using 4 -20% Novex Tris-glycine gradient gels (Life Technologies, Inc.) and stained with Thermo Fisher Scientific Pierce GelCode Blue stain reagent. For FFPE tissues, total protein lysate was extracted, as reported previously (34,35), and then resolved by SDS-PAGE as above. The region of each gel lane containing the core histones was excised as one piece, diced into small pieces, destained, treated with d 6 -acetic anhydride to chemically block unmodified lysines and monomethylated lysines with an isotopically heavy acetyl, and digested in-gel with trypsin as reported previously (32). Tryptic peptides were separated by reverse phase Jupiter Proteo resin (Phenomenex) on a 100 ϫ 0.075-mm column using a nano-Acquity UPLC system (Waters).
Peptides derived from the two cell lines, WM115 and WM266-4, were eluted using a 40-min gradient from 97:3 to 35:65 buffer A/B ratio. (Buffer A ϭ 0.1% formic acid, 0.5% acetonitrile; buffer B ϭ 0.1% formic acid, 75% acetonitrile.) Eluted peptides were ionized by electrospray (1.9 kV) followed by MS/MS analysis using collisioninduced dissociation on an LTQ Orbitrap Velos mass spectrometer (Thermo Fisher Scientific). MS data were acquired using the Fourier Transform Mass Spectrometry analyzer in profile mode at a resolution of 60,000 over a range of 375 to 1500 m/z. The melanocytes and FFPE samples were eluted using a 30-min gradient from 97:3 to 65:35 buffer A/B ratio. (Buffer A ϭ 0.1% formic acid, 0.5% acetonitrile; buffer B ϭ 0.1% formic acid, 99.9% acetonitrile.) Eluted peptides were ionized by electrospray (2.35 kV) followed by MS/MS analysis using high energy collisional dissociation on an Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific). MS data were acquired using the FTMS analyzer in profile mode at a resolution of 240,000 over a range of 375 to 1500 m/z. MS/MS data were acquired for the top 15 peaks from each MS scan using the ion trap analyzer in centroid mode and normal mass range with a normalized collision energy of 35.0. The data-dependent acquisition approach was suited for these discovery phase studies; however, it is possible that some peaks/peptides could not have been identified due to the nature of the data-dependent approach.
Proteins were identified by searching the UniProtKB database (2015_06 release; restricted to Homo sapiens; 20,207 entries) using an in-house Mascot server (version 2.4; Matrix Science). Peak lists were generated from raw data files by MSFileReader (version 2.2; Thermo Fisher Scientific) and ExtractMSn (January, 2011 release; Thermo Fisher Scientific). Charge state deconvolution and deisotoping were not performed. No contaminants were excluded from the data set. Mascot search parameters were specified as follows: trypsin digestion with up to two missed cleavages; fixed carbamidomethyl modification of cysteine; variable methyl, dimethyl, trimethyl, acetyl, acetyl/ 2 H 3 , and methylϩacetyl/ 2 H 3 modification of lysine; variable phosphorylation of serine and threonine (separate search); 2.0 ppm precursor ion tolerance; 0.50-Da fragment ion tolerance. A reversesequence decoy search was also performed. Peptide and protein identifications were validated using Scaffold (version 4.4; Proteome Software). Peptide identifications were accepted at Ͼ50.0% probability as determined by the Scaffold local false discovery rate algorithm. Protein identifications were accepted at Ͼ95.0% probability and a minimum of two identified peptides. Protein probabilities were assigned by the Protein Prophet algorithm (36). Proteins with similar peptides that could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. Proteins sharing significant peptide evidence were grouped into clusters (supplemental Tables 1-6).
Histone H2A, H2B, H3, and H4 peptide precursor intensity values (m/z versus the peak height) were manually extracted from the raw files using Qual Browser (Thermo Fisher Scientific Xcalibur 2.2 SP1.48) (supplemental Tables 7-9). The intensity values for all histone peptides were log 2 -normalized, centered based on the median value (15.95), and scaled based on the standard deviation (9.4), and the relative percent of unmodified to modified peptide intensity was calculated. This accounts for the identification of the histone molecule versus the level of modification. The median of each technical triplicate value was used to calculate the Student's t test for the biological replicate samples (SAS version 9.4) (supplemental Tables 7-9). The transformed data met the assumption of normality and equality of variances according to the Kolomogorov-Smirnov (p value ϭ 0.489) and the Levene test (p value ϭ 0.1161), respectfully (Unscrambler X, version 10.3, CAMO software). Mass spectrometric data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD002471 and null (pxsubmission number 46549).
Gene Expression Analysis-Total RNA from cultured cells was extracted using RNeasy mini kit (Qiagen) following the manufacturer's specifications. The isolated RNA was treated with RNase-free DNase I Amplification Grade (Invitrogen) to digest any traces of DNA contamination. cDNA was synthesized with oligo(dT) primers using Superscript First-Strand Synthesis System (Invitrogen) following the manufacturer's instructions and treatment with RNase H (Invitrogen). qPCR was performed using a reaction mixture containing 1ϫ SsoAdvanced SYBR Green Supermix (Bio-Rad), primers (0.33 M), and DNA (100 ng). The cyclic conditions used were 95°C for 2 min, followed by 45 cycles at 95°C for 15 s and 50°C for 20 s. The following primer pairs were ordered from Integrated DNA Technologies (Coralville, IA) and used for real time analysis: ␤-actin forward (5Ј-CTGGACTTCG-AGCAAGAG-3Ј) and ␤-actin reverse (5Ј-AAGGAAGGCTGGAAGAGT-3Ј); E-cadherin forward (5Ј-TGCCCAGAAAATGAAAAAGG-3Ј) and Ecadherin reverse (5Ј-GTGTATGTGGCAATGCGTTC-3Ј), and RUNX3 forward (5Ј-CAGAAGCTGGAGGACCAGAC-3Ј) and RUNX3 reverse (5Ј-TCGGAGAATGGGTTCAGTTC-3Ј). qPCR was performed using a MiniOpticon real time PCR detection system (Bio-Rad). All mRNA expression was normalized to that of ␤-actin. Graphs were plotted using the GraphPad Prism 6.0.
Invasion Assay-For coating, Matrigel basement membrane matrix (BD Biosciences) was used with BD BioCoat TM cell culture inserts (8-m PET membrane) (BD Biosciences). Uncoated inserts were used as controls. Staining was performed with hematoxylin for 4 min. Cell counts were performed at ϫ200 magnification, and percent invasion was calculated by dividing the number of invaded cells from the number of cells that migrated through the uncoated inserts multiplied by a factor of 100.
Immunohistochemistry-FFPE tissue samples of benign nevus, primary cutaneous melanoma, and cutaneous metastasis of melanoma were provided by the Pathology Department, University of Arkansas for Medical Sciences, following approval from the Institutional Review Board, University of Arkansas for Medical Sciences. For antigen retrieval, slides were heated to 120°C for 20 s in a Decloaking Chamber TM (Biocare Medical, Concord, CA) using 10 mM sodium citrate buffer, pH 6.0. Staining was performed using Vectastain Elite ABC kit (Vector Laboratories, Burlingame, CA). The following antibodies were used for staining: anti-RUNX3 (1:300; rabbit polyclonal, Abcam, catalog no. ab68938); anti-EZH2 (1:2000; rabbit monoclonal, Cell Signaling, catalog no. 5246); anti-E-cadherin (1:1000; rabbit polyclonal, Abcam, catalog no. ab15148); and anti-H3K27me3 (1:200; rabbit monoclonal, Cell Signaling, catalog no. 9733). Slides were counterstained with Mayer's hematoxylin (Thermo Fisher Scientific) for 1 min. Scoring of FFPE tissue samples was performed in a semiblinded fashion by a board-certified dermatopathologist (without access to the pathological diagnosis).
Statistical Analysis-Following scoring of FFPE tumors, a correlation coefficient (r) was calculated for each data set as well as a weighted Cohen's coefficient (). Spearman's correlation coefficient () was used to correlate average scores for both antibodies. For all other in vitro studies, experiments were performed in biological as well as technical triplicates. A Student's t test was used to test for significant differences between two conditions. For experiments with three or more conditions, a two-way analysis of variance was performed with a Sidak's multiple comparisons test to compare all pairs of columns. All values in the graphs represent mean Ϯ S.E. by Student's t test.

Histone H3K27me3 Is Elevated in Late Stage Melanoma
Compared with Early Stage Primary Melanoma-We performed studies using cell lines as well as patient biopsies. For cell lines, we utilized a pair of matched melanoma cell lines that were isolated from a 58-year-old female in 1981 at the Wistar Institute (38). The cell line WM115 was derived from a primary melanoma tumor located in the skin of the right thigh of the patient. Subsequently, the WM266-4 cell line was isolated from the sentinel lymph node metastasis from the same patient. Human melanocytes were used as controls. Histones were isolated by acid extraction from ϳ5 million cells in biological triplicate for melanocytes, WM115 and WM266-4 cells. The isolated histones were resolved by SDS-PAGE (20 g per lane) and visualized by Coomassie staining (Fig. 1A). Gel slices were de-stained and treated with d 6 -acetic anhydride to block trypsin activity at unmodified and mono-methylated lysines with an isotopically heavy acetylation that is readily detected with mass spectrometric analysis (32). As detailed under "Experimental Procedures," precursor ion intensitybased label-free quantitation was used to measure the amount of unmodified and post-translationally modified histone peptides. All experiments were performed in biological triplicate with technical triplicate data collection for each. For this study, we focused on the following more common histone PTMs: lysine monomethylation; lysine dimethylation; lysine trimethylation; serine phosphorylation; and threonine phosphorylation. We identified 22 uniquely modified histone peptides across the four core histones in the two cell lines, 54 in melanocytes and 20 in FFPE tumors. This is not a comprehensive list of histone PTMs, but rather represents the most abundant bulk level histone PTMs that can be identified with

FIG. 1. Histone H3K27me3 is up-regulated in aggressive late stage WM266-4 melanoma cells compared with less aggressive early stage WM115 melanoma cells.
A, histones were isolated by acid extraction from melanocytes, WM115, and WM266-4 cells in biological triplicates, resolved by SDS-PAGE, and visualized by Coomassie staining. Individual histone bands were excised, treated with d 6 -acetic anhydride, digested with trypsin, and analyzed by high resolution mass spectrometry. B, quantitative analysis of histone peptide intensities revealed that H3K27me3 was elevated in WM266-4 melanoma cells relative to WM115 melanoma cells. Melanocytes (control) were used as point of reference for the study. Standard error was calculated for the specific peptide in the biological replicate samples displayed in the chart. The chart displays the normalized relative intensity values for histone H3K27 and its modifications. Error bars represent the standard error of biological triplicates each analyzed by mass spectrometry in technical triplicates. C, immunoblot analysis of the isolated histones showed an increased level of histone H3K27me3 in WM266-4 cells relative to melanocytes and WM115 cells. Histone H3 was used as the loading control. high confidence. Further separation of histones based on the PTM state (as we have done previously (31,33)) can lead to more comprehensive lists of PTMs, which will be the focus of future work. Histone PTMs having a p value less than 0.1 were considered of interest and included H3K27me3 and H4K20me (supplemental Tables 7-9). Because there is evidence of elevated EZH2 in some cancers (4, 6, 10, 39), we decided to pursue H3K27me3 for biological studies. Using the histone peptide-intensity data for peptides containing unmethylated or methylated states of H3K27, we plotted the relative levels of H3K27 methylation in melanocytes, WM115, and WM266-4 cells (Fig. 1B). The overall trend showed low H3K27 methylation in the melanocytes and less aggressive WM115 cells, whereas the more aggressive WM266-4 cells contained higher levels of methylation and most noticeably elevated H3K27me3 relative to WM115. An immunoblot analysis of the histones isolated from these cell lines confirmed the elevated levels of H3K27me3 in WM266-4 cells (Fig. 1C).
When we performed similar studies using human patient biopsies, higher levels of H3K27 trimethylation were observed in metastatic melanoma tissues compared with both benign nevi and primary melanoma tissues (Fig. 2, A and B). This was in agreement to our previous finding in cell lines and was validated by performing immunoblot using whole tissue lysate extracted from FFPE tissues (Fig. 2C).

Tumor Suppressors RUNX3 and E-cadherin Are Epigenetically Silenced in Aggressive Melanoma by EZH2-mediated
Deposition of H3K27me3-Analysis of a panel of melanoma cell lines by immunoblot revealed higher levels of H3K27me3 and EZH2 protein in the metastatic WM266-4 cells as well as other metastatic cells like A375 and A2058, relative to WM115 cells (Fig. 3A). The EZH2 transcript level of the metastatic cell lines, including the WM266-4 cell line, was also found to be higher than WM115 cells (Fig. 3B). The WM115 cells were transfected with a plasmid expressing EZH2 under the control of the CMV promoter, generating WM115EZ cells that showed characteristics intermediate to WM115 and WM266-4 cells (Fig. 3, C and D, and supplemental Fig. 1). EZH2 expression was increased in WM115EZ cells relative to WM115 cells, which resulted in increased levels of H3K27me3 ( Fig. 3C and supplemental Fig. 1, A and B). Elevated levels of H3K27me3 in aggressive melanoma cell lines WM266-4 and WM115EZ are associated with lower protein and mRNA levels of RUNX3 and E-cadherin when compared with the less aggressive melanoma cell line WM115, suggesting EZH2-mediated epigenetic silencing of the tumor suppressor genes RUNX3 and E-cadherin (Fig. 3, C and D). Because EZH2 is an epigenetic regulator known to modulate the expression of several tumor suppressor genes in different types of cancer, we determined the effect of EZH2 small molecule inhibition and depletion on RUNX3 and E-cadherin expression. Treatment of WM115, WM115EZ, and WM266-4 cells with the EZH2 inhibitor GSK126 resulted in depletion of H3K27me3 in all three cell lines (Fig. 3C) with significant restoration of RUNX3 and Ecadherin expression in the EZH2 overexpressing WM115EZ and WM266-4 cells (Fig. 3, C and D). Furthermore, depletion of EZH2 via siRNA knockdown enhanced RUNX3 and Ecadherin expression in WM266-4 cells further corroborating the silencing effect of EZH2 on the two tumor suppressor genes (Fig. 3E).
Tumor Suppressors RUNX3 and E-cadherin Are Transcriptionally Silenced in Aggressive Melanoma by Accumulation of EZH2-catalyzed H3K27me3 at Promoter Chromatin-To determine whether the increase in RUNX3 transcription was a direct effect of reduced H3K27me3 occupancy at the RUNX3 promoter, we performed ChIP to determine the relative level FIG. 2. Histone H3K27me3 is up-regulated in aggressive late stage FFPE metastatic melanoma tumors relative to benign tumors and less aggressive early stage FFPE primary melanoma tumors. A, whole cell extracts from benign nevi, primary, and metastatic FFPE melanoma tumors were prepared in biological triplicates, resolved by SDS-PAGE, and visualized by Coomassie staining. Individual histone bands were excised, treated with d 6 -acetic anhydride, digested with trypsin, and analyzed by high resolution mass spectrometry. B, quantitative analysis of histone peptide intensities revealed that H3K27me3 was elevated in FFPE metastatic melanoma tumors relative to benign nevi and primary melanoma. Standard error was calculated for the specific peptide in the biological replicate samples displayed in the chart. The chart displays the normalized relative intensity values for histone H3K27 and its modifications. Error bars represent the standard error of biological triplicates each analyzed by mass spectrometry in technical triplicates. C, immunoblot analysis of the whole cell extract showed increased level of histone H3K27me3 in metastatic melanoma relative to benign and primary FFPE tissues. Histone H3 was used as the loading control.
of H3K27me3 at the RUNX3 promoter in melanoma cells (Fig.  4A). ChIP performed in the matched melanoma cell lines WM115 and WM266-4, along with the engineered cell line WM115EZ, showed significantly higher enrichment of H3K27me3 at the RUNX3 promoter in aggressive melanoma cell lines WM115EZ and WM266-4 compared with the less aggressive melanoma cell line WM115 (Fig. 4A). EZH2 inhibition via treatment with GSK126 reduced this H3K27me3 occupancy at the RUNX3 promoter in EZH2 overexpressing cell lines, thereby restoring the RUNX3 expression.
ChIP performed with primers specific to the E-cadherin gene promoter showed that H3K27me3 was enriched higher in WM115EZ and WM266-4 cells relative to WM115 cells (Fig.  4B). To determine the mechanism of GSK126-mediated Ecadherin restoration in EZH2-overexpressing melanoma cells, the effect of the drug was tested on H3K27me3 occupancy at the E-cadherin promoter in all three melanoma cell lines (Fig.  4B). H3K27me3 enrichment at E-cadherin promoter was reduced by GSK126 treatment in WM115EZ and WM266-4 cells. Hence, the E-cadherin restoration observed in these cell lines upon being treated by GSK126 is directly linked to decreased EZH2 occupancy at the promoter of this gene.
Because E-cadherin is known to play a role in the regulation of invasion, we wanted to determine whether EZH2 expression is related to invasiveness of melanoma cells. Indeed, wild-type WM266-4 cells demonstrated higher invasive property compared with WM115 cells (Matrigel assay), and overexpression of EZH2 in WM115 cells (WM115EZ cells) increased their invasive ability (Fig. 4, C and D). Furthermore, treatment with GSK126 reduced invasion in WM115EZ and WM266-4 cells but not in WM115 cells, indicating that EZH2 overexpression is a driver of the invasive phenotype observed in these cells. Furthermore, to confirm that EZH2 regulates the invasive potential of cells via E-cadherin modulation, the direct role of E-cadherin on invasiveness of cells was investigated (Fig. 5). Depletion of E-cadherin protein by siRNA knockdown (Fig. 5A) revealed increased invasion in all melanoma cells, with the most significant effect on primary cell line WM115 (having highest E-cadherin expression) (Fig. 5, B and  C). Hence, we have provided conclusive evidence that EZH2 promotes invasion in metastatic melanoma cells by direct epigenetic silencing of E-cadherin.
To validate the EZH2-mediated epigenetic silencing of RUNX3 and E-cadherin by EZH2 in human melanoma tissues, ChIP was performed using 26 FFPE melanoma patient tumor samples (benign nevi ϭ 10, primary melanoma ϭ 9, metastatic melanoma ϭ 7) (Fig. 6, A and B). As observed in the metastatic melanoma cell line WM266-4, the metastatic melanoma FFPE patient tumors exhibited higher H3K27me3 enrichments at both RUNX3 and E-cadherin promoters compared with benign nevi and primary melanoma FFPE tumors (Fig. 6, A and B). This corroborates the hypothesis that EZH2 mediates epigenetic silencing of RUNX3 and E-cadherin in metastatic melanoma via H3K27 tri-methylation of the corresponding promoter region. To determine whether the enrichment of H3K27me3 at the promoter of the tumor suppressor genes in patient tissues correlated to an increased level of bulk H3K27me3 in metastatic patient melanoma (as was the case for melanoma cell lines), immunohistochemical analysis for H3K27me3 was performed on 17 nevi and 14 melanoma FFPE patient samples (Fig. 6C). Fourteen of 17 nevi showed low staining for H3K27me3, although all 14 of the melanoma tissues showed elevated H3K27me3. The level of H3K27me3 was significantly up-regulated in metastatic melanoma patient tissues (p Ͻ 0.0001) indicating that the elevated H3K27me3 detected in melanoma cell culture is reflected in patient melanoma.
RUNX3 and E-cadherin Are Repressed in Patient Metastatic Melanoma-To determine whether increased H3K27me3 in metastatic melanoma mediated by EZH2 is accompanied by reduced levels of RUNX3 and E-cadherin, large scale immunohistochemistry was performed for EZH2, RUNX3, and Ecadherin in the 26 FFPE melanoma tissues previously used for ChIP experiments (Fig. 6, A and B). As anticipated, the staining intensity of RUNX3 (nuclear) and E-cadherin (cytoplasmic) reduced from strong/moderate in benign nevi (Fig. 7A) to moderate/weak in primary melanoma (Fig. 7B) and to weak/ negative in metastatic melanoma (Fig. 7C). Scoring was per- formed in a semi-blinded fashion (no access to pathological diagnosis) by a dermatopathologist (Table I). The 2 analysis revealed significantly lower RUNX3 staining in metastatic melanoma tissues relative to benign nevi (degrees of freedom (df) 2, 2 ϭ 4.96, p Ͻ 0.05) and primary melanoma (df 2, 2 ϭ 4.15, p Ͻ 0.05) which supports our in vitro results. A significant negative correlation was also observed between EZH2 and RUNX3 staining intensity for all three types of tissues (Spear- FIG. 7. Immunohistochemistry in FFPE patient melanoma shows increased EZH2 and reduced levels of RUNX3 and E-cadherin in aggressive melanoma. A, negative to weak staining for EZH2 in benign nevi is associated with moderate to strong staining for RUNX3 and E-cadherin in benign nevi. B, moderate staining for EZH2, RUNX3, and E-cadherin occurs in primary melanoma. C, moderate to strong staining for EZH2 is accompanied by negative to weak staining for RUNX3 and E-cadherin in metastatic melanoma. All images are shown at 400ϫ. The scale bar at the bottom right of each figure corresponds to 50 m.

TABLE 1 Scoring of EZH2, RUNX3 and E-cadherin staining for formalin-fixed-paraffin-embedded (FFPE) patient tumors
Scores of twenty six immunohistochemically stained FFPE slides are tabulated where negative, weak, moderate, and strong staining intensities of 10 human benign nevi, 9 primary melanoma and 7 metastatic melanoma patient tumors are represented by 0, 1, 2 and 3 respectively. Positive staining for EZH2 and RUNX3 was localized to the nucleus, while positive staining for E-cadherin was localized to the plasma membrane and cytoplasm. man's correlation coefficient ϭ Ϫ0.42, p Ͻ 0.05). Immunohistochemical staining of the FFPE patient samples using anti-E-cadherin antibody also showed a significant negative correlation between EZH2 and E-cadherin staining intensity for all three tissue types (Spearman's correlation coefficient ϭ Ϫ0.433, p Ͻ 0.05). EZH2 staining (nuclear) was increased in metastatic melanoma tissues relative to benign nevi and primary melanoma as revealed by a 2 (df 3, 2 ϭ 6.54, p Ͻ 0.05), which is consistent with previously reported findings (McHugh et al. (zref4)). DISCUSSION In this study, we used a quantitative mass spectrometry approach to profile bulk levels of histone modifications in melanoma cell culture as well as FFPE melanoma patient tumors. One of the distinct advantages of using a mass spectrometry approach, in addition to the unbiased nature of the analysis, is avoiding the use of antibodies directed to histone PTMs, which have been shown to have issues of specificity (40). Furthermore, by combining biochemical fraction with high resolution mass spectrometry, one can uncover novel and low level histone modifications (29,31,33,41). Additionally, mass spectrometry is the primary approach for uncovering combinatorial histone modifications; those occurring in particular combinations on single histone molecules, as antibodies for combinatorial PTMs on histones, are extremely limited.
Using our bottom-up mass spectrometric approach, we identified for the first time histone H3K27me3 as elevated in aggressive melanoma cells relative to less aggressive melanoma cells in cell culture and in patient tissues ( Figs. 1 and 2). In support of the elevated H3K27me3, we identified elevated levels of the EZH2 methyltransferase that catalyze the methylation of H3K27 (Fig. 3, A and B). This mis-regulated H3K27me3/EZH2 epigenetic program was found to directly silence two tumor suppressor genes RUNX3 and E-cadherin ( Fig. 3) that have been observed to be down-regulated in other cancers via an EZH2-dependent mechanism. Furthermore, we directly linked EZH2 to increased invasive potential of metastatic melanoma cell lines (Fig. 5). We also showed that inhibition of EZH2 using small molecule inhibitors decreased invasion significantly, whereas selective knockdown of E-cadherin increased invasive potential of cells suggesting that EZH2 mediates invasive properties in metastatic melanoma cells by suppression of E-cadherin (Fig. 6). In addition to observing the H3K27me3 mis-regulation in melanoma cell culture, we confirmed the up-regulation of H3K27me3 and subsequent down-regulation of RUNX3 and E-cadherin tumor suppressors in human melanoma tissues (Figs. 6 and 7). Taken together, our results demonstrate the power of using mass spectrometry to survey histone epigenetic patterns as the particular marks mis-regulated in disease states have the potential to serve as diagnostic and prognostic markers. Furthermore, defining altered patterns of histone PTMs in disease provides for the identification of biochemical pathways for targeted drug development. In future studies, we will combine the use of biochemical fractionation (e.g. polyCAT chromatography or acid-urea gels) of histones based on the PTM state to begin probing lower level and combinatorial histone modifications mis-regulated in melanoma.
Acknowledgments-We acknowledge the University of Arkansas for Medical Sciences Proteomics Facility for mass spectrometric support.
* This work was supported by National Institutes of Health Grants R01GM106024, R33CA173264, UL1TR000039, P20GM103625, S10OD018445, and P20GM103429. The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
□ S This article contains supplemental Fig. S1 and Tables S1 to S9. § Both authors contributed equally to this work. ** Co-corresponding authors. To whom correspondence should be addressed: