Epigenetic Silencing of BMP6 by the SIN3A–HDAC1/2 Repressor Complex Drives Melanoma Metastasis via FAM83G/PAWS1

Cutaneous melanoma begins as a localized lesion but rapidly becomes lethal after evolving metastatic capacity. Despite recent advances in therapy, 5-year overall survival (OS) of stage IV metastatic melanoma patients is less than 40% in patients treated with BRAF and MEK inhibitors (1, 2) and around 30% to 40% with immune-checkpoint inhibitors (3, 4). The high risk of melanoma recurrence after surgical removal of primary lesions (5) and the significant relapse-free survival benefit of adjuvant therapies with the clinically approved agents, such as BRAF and MEK inhibitors, suggest that invasive and metastatic traits of the disease develop early in disease progression (6). These clinical observations suggest that there is an important early therapeutic window for adjuvant and/or neoadjuvant therapy to block metastasis (7).

A paucity of metastasis-driving mutations and the presence of subpopulations of cells displaying stable metastatic phenotypes in primary tumors suggests the existence of epigenetically regulated metastasis-promoting transcriptional programs (8, 9). Like DNA cytosine methylation, covalent histone modifications are somatically heritable epigenetic marks that play critical roles in regulating gene transcription (10). Such metastable chromatin alterations are thought to be important in the metastatic phenotypes of cancer cells (11). A critical histone modification is the removal of acetyl groups from lysine residues of N-terminal histone tails, which induces the formation of transcriptionally repressive chromatin (12). In fact, several important transcriptional repressor complexes contain class I histone deacetylase 1 and 2 (HDAC1/2) as a core subunit. These epigenetic “eraser” complexes include the repressor complexes of SIN3 (13, 14), NuRD (15, 16), CoREST (17), and MiDAC (18). Existence of these diverse transcription repressor complexes containing the same HDAC1/2 enzymes in mammalian cells indicates that these complexes have distinct biological roles and specific cellular functions (19). However, their roles in metastatic progression of cancers remain largely undefined.

Bone morphogenetic proteins (BMP), members of the transforming growth factor β (TGFβ) superfamily, are secreted ligands for BMP receptor types I and II (BMPR1A/B and BMPR2), with pleiotropic biological functions (20) including embryonic development and tissue homeostasis. Upon binding to the receptor, BMPs activate a well-defined canonical intracellular signaling cascade mediated by phosphorylation of SMAD (small mothers against decapentaplegics) proteins (SMAD1/5/9 in human and SMAD1/5/8 in mouse). The activated SMADs then heterodimerize with SMAD4 and translocate to the nucleus where they regulate target gene transcription (21). Dysregulation of the BMP–SMAD signaling pathway in cancer cells induces paradoxical effects on tumorigenesis and metastasis (22). This can result in cell-autonomous antiproliferative effects in some types of cancer cells, but with increasingly recognized proinvasion and prometastasis roles in various cancer types, as evidenced by their involvements in tumor angiogenesis, epithelial–mesenchymal transition (EMT), modulation of the tumor microenvironment, and cancer stem cells (23–25). On the other hand, there is evidence that the BMP–SMAD signaling pathway can have antimetastatic roles in some cancer systems (26–28). In general, the dual roles of BMP–SMAD signaling in cancer progression is largely determined by the expression levels of each specific BMP and SMAD signaling mediator, in concert with genetic alterations in the tumor cells (29). Therefore, it is important to identify the precise functional role of each BMP and SMAD in cancers, as well as their upstream transcriptional regulators and downstream effectors. Here, we report a metastasis-promoting role of the SIN3A–HDAC1/2 repressor complex, acting via inhibition of the metastasis suppressing function of the BMP6-activated SMAD5 signaling pathway. We also identify a role for FAM83G/PAWS1 as a critical downstream effector of BMP6/SMAD5 signaling in promoting melanoma metastasis. Our findings nominate the SIN3A/HDAC1/2→BMP6/SMAD5→FAM83G signaling axis as a compelling therapeutic target for inhibiting metastatic progression in human melanoma.

BRAF-mutant melanoma cell lines WM1341D, WM983B, 1205Lu, WM852, WM793, WM35, WM167 (gifted from Dr. Meenhard Herlyn, The Wistar Institute), A375 (purchased from ATCC), and 1205Lu derivatives 1205Lu-M1 and 1205Lu-M2 (generated in this study) were maintained in DMEM with 1% penicillin–streptomycin, 1% L-glutamine, and 10% (v/v) FBS. BRAF-wild-type MeWo and NRAS-mutant SKMEL2 cell lines (both purchased from ATCC) were also maintained in the same culture medium as the BRAF-mutant cells. Cells were grown at 37°C under a humidified atmosphere of 5% CO₂. Cell lines (WM793, 1205Lu, 1205Lu-M1, and 1205Lu-M2) was authenticated by STR profiling (Genetica Cell Line Testing, LabCorp) and confirmed human origin.

In vivo mouse experiments were performed in compliance with an animal protocol approved by the Institutional Animal Care and Use Committee at the Center for Discovery and Innovation, Hackensack Meridian Health. Animal experiments were carried out with 10- to 12-week-old NOD/SCID (NOD.Cg-Prkdc^scid/J) mice purchased from The Jackson Laboratory and maintained under standard pathogen-free conditions.

The study included 50 formalin-fixed paraffin-embedded (FFPE) tumor tissues from patients with cutaneous melanomas: 25 primary melanomas and 25 melanoma metastases. All tumor specimens were collected at the time of surgery and banked at the Department of Pathology, Hackensack University Medical Center (HUMC), from January 2017 to December 2019. The retrieved tissue samples (Supplementary Table S1) were categorized by gender, age at diagnosis, and primary and metastatic tumor localization. Additionally, pathologic parameters of primary tumors were collected including Breslow thickness, ulceration, and absence or presence of microsatellite and regression. Pathologic stages of the primary malignant tumors, according to the 2009 American Joint Committee on Cancer staging system, were also retrieved. The primary and metastatic tumor specimens were manually macrodissected to obtain tumor and adjacent normal cells with greater than 90% purity, immediately followed by total RNA extraction. This retrospective study protocol was approved by the HUMC Internal Review Board.

Human BMP6 cDNA ORF in pcDNA3.1⁺/C-(K)DYK vector (GenScript, #OHu16359), FAM83G cDNA ORF in pcDNA3.1⁺/C-(K)DYK vector (GenScript, #OHu17355) and pcDNA3.1⁺ control vector were used to generate stably expressing melanoma cell lines by transfection with Lipofectamine 2000 reagent (Invitrogen, #11668019). Stably transfected cells were selected by G418 (MilliporeSigma, #345810) treatment, and the target gene-expression levels were determined by RT-qPCR and/or immunoblots. Lentiviral plasmid vectors, pLKO.1 Puro-shRNA/SIN3A-1, pLKO.1 Puro-shRNA/SIN3A-2, pLKO.1 Puro-shRNA/FAM83G-1, pLKO.1-shRNA/FAM83G-2 (Millipore Sigma), and pLKO.1 Puro-Empty vector control (Addgene, #8453) were used for the generation of melanoma cell lines stably depleted SIN3A, FAM83G, and scrambled control cells. Lenti-X 293T cells were used for lentiviral production for the stable cell line generation. All plasmid constructs were propagated in DH5α (New England BioLabs) on LB plates or in liquid LB media with 100 μg/mL ampicillin at 37°C.

Cells were treated with Corin, MS-275, and GSK2879552 at the dose of 5 μmol/L for 24 hours. Corin was synthesized in the Dr. Philip Cole's Laboratory at Brigham and Women's Hospital and Harvard Medical School. MS275 (HY-12163) and GSK2879552 (HY-18632) were purchased from MedChemExpress. For the stimulation of melanoma cells with BMP2 and BMP6 proteins, recombinant human BMP2 (cat. #355-BM) and BMP6 (cat #507-BP) proteins purchased from R&D Systems were added to the cells at the concentrations of 0, 50, and 100 μg/mL for 2 hours.

For the relative transcript level assay from cultured cells, total RNA was isolated by TRIzol and quantified by NanoDrop One Spectrophotometer (Thermo Fisher). cDNA was synthesized from 1.5 μg of the total RNA with the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems/Thermo Fisher). Quantitative real-time PCR was performed with KOD SYBR qPCR Mix (TOYOBO), and all reactions were analyzed with QuantStudio 3 Real-Time PCR Systems (Thermo Fisher).

For the quantitative enumeration of circulating melanoma cells in mouse blood, standard curves from the 250 μL of normal mouse blood samples spiked with known numbers of melanoma cells (0, 10, 50, 250, and 1,250 cells of 1205Lu-M2/Luc ectopically expressing luciferase gene) were generated. Total RNA was extracted from the spiked blood samples using QIAamp Blood RNA Mini Kit 9 (Qiagen), and cDNA was synthesized with the SuperScript III First-Strand Synthesis System (Thermo Fisher). Quantitative real-time PCR was performed on the transcripts of the luciferase and human melanocyte cell marker S100A6 genes, and the linear standard curves were generated by plotting the C_T values against the artificially spiked melanoma cell numbers. Blood samples from the experimental mice were processed and assayed as the same as the standard curve generation without spiking melanoma cells. The circulating melanoma cell numbers were extrapolated from the standard curves.

For the quantification of mRNA transcript level from FFPE tissue samples, extraction and quantification of total FFPE-RNA and FFPE-DNA from the macrodissected tumor tissues were performed as published in our previous paper (30). Target gene transcript levels from the FFPE tissues were determined by the absolute quantification assay with duplex real-time qPCR as described in our previous papers (31). Amplifications were performed using a QuantStudio 3 Real-Time PCR Systems (Applied Biosystems/Thermo Fisher). Primers and probe concentrations for each target gene and reference gene (Melan-A) were preoptimized. The expression ratios of target genes (BMP6, FAM83G, and SIN3A) to the reference gene Melan-A were computed to normalize the target gene expression to the size of melanocytic tumors in the specimen, and the resulting data points were used as representations of the levels of target gene expression in the tissue samples. All primers and probes sequences are shown in Supplementary Table S2.

Cell lysate was run on an SDS-PAGE gel in ice and blotted onto Immun-Blot PVDF membrane (Bio-Rad) according to the standard protocol. Primary antibodies used in this study are: rabbit anti-BMB2 (Abcam, #14933), rabbit anti-BMP6 (Abcam, #155963), rabbit anti-SIN3A (Cell Signaling Technology, #8056), rabbit anti-p-SMAD1 (Cell Signaling Technology, #11971), rabbit anti-p-SMAD1/5 (Cell Signaling Technology, #9516), rabbit anti-p-SMAD1/5/9 (Cell Signaling Technology, #11971), mouse anti-SMAD1 (Santa Cruz Biotechnology, #7965), rabbit anti-H3K9ac (Cell Signaling Technology, #9649), rabbit anti-H3K4me1 (Cell Signaling Technology, #5326), rabbit anti-H3K9me2 (Cell Signaling Technology, #9725), rabbit anti-FAM83G (Bethyl Laboratories, #A304-282A), rabbit anti-FOXM1 (Cell Signaling Technology, #5436), rabbit anti-p21 (Abcam, #109520), rabbit anti-p38γ (Cell Signaling Technology, #2307), rabbit antiphospho-p38 (Cell Signaling Technology, #4511), and rabbit anti-GAPDH (Cell Signaling Technology, #5174).

Cells were fixed with formaldehyde to a final concentration of 1% for 7 minutes at room temperature (RT) with constant shaking. For stopping crosslinking, glycine was added to the final concentration of 0.125 mol/L for 5 minutes at RT. Cells were then recovered by centrifugation at 1,500 × g at RT for 5 minutes and washed twice with cold PBS. For nuclear lysis, the pellet was gently suspended with a 1.4 mL cell lysis buffer (5 mmol/L HEPES, 85 mmol/L KCl, 0.5% NP40 and Protease inhibitor cocktail), and incubated on ice for 15 minutes. The nuclei were recovered by centrifugation and suspended in 250 μL of nuclear lysis buffer (50 mmol/L Tris-HCl, 10 mmol/L EDTA, and 1% SDS) and incubated on ice for 10 minutes. Chromatin was sheared using Covaris M220 focused–ultrasonicator and the size of sheared chromatin was determined on 1.2% agarose gel. Pierce Protein A/G Magnetic Beads (Thermo Fisher Scientific) was washed with PBS plus 0.5% BSA solution, and incubated with each 1 μg of CoREST (Millipore Sigma, #07-455), SIN3A (Cell Signaling Technology, #8056), CHD4 (Cell Signaling Technology, #12011), rabbit IgG (Cell Signaling Technology, #2729) antibodies, H3K9ac (Cell Signaling Technology, #9649), H3K3me3 (Cell Signaling Technology, # 9751), and RNA polymerase II (Abcam, #ab26721) followed by washing with PBS/BSA solution. Sheared chromatin was mixed with the magnetic beads and rocked for overnight at 4°C. Bound chromatin was washed two times with low ionic strength buffer (50 mmol/L HEPES, 140 mmol/L NaCl, 1% Triton X-100, 0.1% sodium deoxycholate, and 1 mmol/L EDTA) for 5 minutes at 4°C, and repeated washing with high ionic strength buffer (50 mmol/L HEPES, 500 mmol/L NaCl, 1% Triton X-100, 0.1% sodium deoxycholate, and 1 mmol/L EDTA), and LiCl buffer (10 mmol/L Tris-HCl, 250 mmol/L LiCl, 0.5% NP-40, 0.5% Sodium deoxycholate and 1 mmol/L EDTA). Finally, chromatin was washed with TE buffer (10 mmol/L Tris-HCl, 1 mmol/L EDTA) and eluted with elution buffer (50 mmol/L Tris-HCl, 10 mmol/L EDTA and 1% SDS) containing proteinase K. DNA was extracted using PCI (phenol chloroform isoamyl alcohol) and CI (Chloroform isoamyl alcohol) solutions, precipitated with 3M NaAc and glycogen in 100% cold ethanol, and dissolved in H₂O. The DNA (2 μL) was used for qPCR with Power SYBR Green PCR Master Mix (Thermo Fisher Scientific). Primer sequences are shown in Supplementary Table S2.

pGL3-BMP6-Lux promoter-reporter construct was kindly provided by Dr. I. Y. Kim and Dr. G. T. Lee at Rutgers Cancer Institute at New Jersey. Cells were transiently transfected with the promoter-reporter construct using Lipofectamine 2000. The DNA–Lipofectamine mixture was then added to the cell suspension (5 × 10⁵ cells in 1.5 mL growth medium) followed by gentle rotation at 37°C for 40 minutes. Then, the transfected cell suspension was incubated overnight in a tissue culture incubator followed by replacement with fresh growth medium containing the indicated doses of Corin. After 48 hours of additional culture, the cells were washed and luciferase activity was measured using Dual-Luciferase Reporter Assay System (Promega, #E1910).

Confluent monolayer cells in serum-free medium were scratched to create wound, and images were taken at the indicated time points. Migration distance was quantified by calculating the distance between the wound edges at a given time point from those at the 0 hour using ImageJ. To check whether the wound gap changes are due to the cell growth rate changes by the BMP6 expression, cell growth assays were performed in parallel. The melanoma cells (3 × 10⁴ cells per well) were seeded and cultured for 20 hours in 12-well plates with serum-free medium, and the cells were counted using Countess II FL cell counter (Thermo Fisher Scientific).

Invasion assay was performed using a 24-well Fluoroblok Transwell insert system (Corning). Matrigel (BD Bioscience/Corning) diluted in a coating buffer (0.01 M pH8 Tris-HCl, 0.7% NaCl) was poured into the inserts (8-μm pore size) and incubated at 37°C for 4 hours. DMEM growth medium with 10% FBS was put to the bottom well of the Transwell chamber, and cells stably expressing BMP2 and BMP6 (1.0 × 10⁵ cells per well) in 100 μL serum-free medium were added to the upper chamber. Then, the invading cells were fixed with methanol for 15 minutes at RT and stained with 500 μL of crystal violet staining solution for 10 minutes followed by 6× gentle washing with PBS. The invading cell numbers were quantified by measuring the stained area using ImageJ.

Cells (2 × 10³ per well) were suspended in 0.5% Noble agar (BD Bioscience) with DMEM growth medium and overlaid on top of the 1.0% Noble agar layer in 6-well plates. After 21 days of culture, colonies were imaged using a digital camera (AmScope MD500) mounted on an inverted microscope with a 4× objective. Three to five images per well were acquired and colony numbers were manually counted, and colony size were calculated using ImageJ.

Cells growing in 4-well format chamber slides (Thermo Fisher) were fixed with 4% formaldehyde for 15 minutes at RT, followed by 3× rinse with PBS for 5 minutes each. Then, immunofluorescence staining of the cells with Alexa Fluor 488 conjugated phalloidin monoclonal antibody (Cell Signaling Technology, #8878) was processed according to the manufacturer's protocol. VECTASHIELD Antifade Medium with DAPI (VECTOR Laboratories, #H-1200–10) was used for counterstaining and mounting. Fluorescence images were obtained with an inverted fluorescence microscopy (Nikon, Eclipse Ti2). F-Actin quantification was performed using Fiji software and macroprogram (32).

Animal experiments were carried out with 10- to 12-week-old NOD.Cg-Prkdc^scid/J mice. BMP6 stably expressing (1205Lu-M1/Luc-BMP6) and FAM83G knockdown (1205Lu-M1/Luc-FAM83G-KD) cells (1 × 10⁴ cells per mouse) were mixed with matrigel (50:50) and implanted s.c. on the NOD/SCID mice. Primary xenograft tumor growth rates were monitored by measuring in vivo bioluminescence intensity using IVIS (PerkinElmer). When primary tumor size reaches approximately 1 cm³ in volume, the xenograft tumors were surgically removed. Two weeks later, mouse blood was collected through the retro-orbital vein, and lungs were harvested. The whole lung tissues were placed in 6-well tissue culture plates and immediately treated with 100 μL of D-luciferin (15 mg/mL) for 10 minutes at RT. After 2× washes with PBS, the ex vivo bioluminescence intensity was measured to determine metastatic tumor burden using IVIS.

1205Lu-M1/Luc (1 × 10⁴ cells per mouse) mixed with matrigel were implanted s.c. on the mice and were allowed to grow for 14 days at the approximate xenograft tumor size of 1 cm³ in volume. The tumor bearing mice were divided into two groups (treatment and control, n = 6 per group) with equal distribution of tumor size, and the primary tumors were surgically removed. At 2 days after surgery, the mice were treated with Corin for two weeks (15 mg/kg per every 2 days). On the day of sacrifice, blood samples were collected through retro-orbital vein, and lung tissues were harvested for the measurement of metastatic tumor burdens of the lungs.

Mouse lung tissues were dissociated into single-cell suspensions using Lung Dissociation Kit (Miltenyi Biotec, #130-095-027) according to the manufacturer's instructions. The cells were fixed and permeabilized with cold 70% ethanol, washed with PBS twice, and stained with fluorochrome-conjugated antibodies for 1 hour at RT in dark condition. Alexa Fluor 647 anti-firefly luciferase antibody was purchased from Abcam. Alexa Fluor 700 anti-human Ki-67 antibody was purchased from BioLegend, and Alexa Fluor 488 anti-human p27 antibody was purchased from Santa Cruz. Each Ki-67 and p27 fluorescence signal was gated, and the signal intensity was analyzed from the first gating of melanoma cells with high fluorescence intensity of Alexa Fluor 647 luciferase signal. Flow cytometry was performed with BD LSRFortessa Flow Cytometer (BD Biosciences), and data were analyzed with FlowJo (FlowJo Inc.).

Mouse lung tissues were fixed in 4% PFA overnight and cryo-protected with sucrose by increasing the concentration up to 30% in PBS at 4°C, and freeze in OCT at −80°C. All lung samples were frozen sectioned and H&E stained at the Roswell Park Cancer Institute. The stained sections were analyzed using Olympus DP80 microscope and Cellsens software.

Total RNA was extracted using TRIzol reagent from the melanoma cells (WM1341D, WM983B, and 1205Lu) stably expressing BMP6 and control cells. RNA was processed with the Ribo-Zero rRNA Removal Kit (Illumina) and further processed using TruSeq RNA Sample Prep Kit (Illumina). Sequencing libraries were prepared following the manufacturer's recommendations. Samples were barcoded and sequenced using standard Illumina chemistry at GENEWIZ. The sequence reads were trimmed to remove possible adapter sequences and nucleotides with poor quality using Trimmomatic v.0.36. The trimmed reads were mapped to the Homo sapiens GRCh19 reference genome available on ENSEMBL using the STAR aligner v.2.5.2b. BAM files were generated as a result of this step. Extracting gene hit counts, unique gene hit counts were calculated by using FeatureCounts from the Subread package v.1.5.2. The hit counts were summarized and reported using the gene_id feature in the annotation file. Only unique reads that fell within exon regions were counted. Because a strand-specific library preparation was performed, the reads were strand-specifically counted.

The comparison of the differential gene expression of BMPs based on tumor and normal from The Cancer Genome Atlas (TCGA; http://xena.ucsc.edu/) and the GTEx (https://gtexportal.org/home/) databases was generated by GEPIA2 (http://gepia2.cancer-pku.cn/; ref. 33). RNA-seq data of the BMPs and FAM83G and the clinical information including overall and progression-free survivals, vital status, and expression levels based on TPMs were downloaded from PanCanAtlas (https://gdc.cancer.gov/about-data/publications/pancanatlas).

Statistical analysis was performed with GraphPad Prism version 8 (GraphPad Software Inc.) or MS Office Excel spreadsheet. Data are presented as the mean ± SD or SE of replicated experiments, as indicated, and presented as individual values, scatter plots, box plots, bar graphs, and heatmap. Significance was determined using paired/unpaired Student t test, ANOVA test, or log-rank test (Kaplan–Meier curves). Correlation coefficient was calculated by Spearman.

RNA-seq data sets generated during this study have been deposited in NCBI's Gene Expression Omnibus under the accession code GSE164495.

Because mutations in genes regulating the BMP/SMAD signaling pathway are rare in melanomas, we reasoned that epigenetic control might be an underlying mechanism regulating this signaling pathway. We therefore assessed expression of the BMP family genes in human cutaneous melanomas compared with the normal skin tissue using publicly available data sets from TCGA (34) and The Genotype-Tissue Expression (GTEx) project (35). The comparison of melanomas with normal skin (a heterogeneous tissue of which melanocytes are simply one component) is not a perfect one, but expression levels in skin provide a stable baseline for comparisons of expression levels among the BMP family genes. We found that BMP2, BMP4, BMP6, and BMP7 mRNA levels are significantly lower in melanoma tissues than in normal skin (Fig. 1A). We also found that the most highly expressed gene in melanomas is BMP1, suggesting the BMP2, BMP4, BMP6, and BMP7 genes are transcriptionally silenced in this tumor type. Survival analysis between patient groups with high versus low expression of BMP2 and BMP6 showed that patients with high expression of these two BMP family genes have a significantly better OS compared with patients with low expression, but opposite effects were observed when we analyzed OS as a function of BMP1 expression (Supplementary Fig. S1A). Expression-level differences of BMP4, BMP7, BMP8A, and BMP8B showed nonsignificant OS effects (Supplementary Fig. S1B). These data suggest that BMP2 and BMP6 might have a suppressive role in disease progression whereas BMP1 might have a promoting role.

Next, to evaluate a potential interplay between mutant BRAF-induced ERK1/2 signaling and BMP-induced SMAD signaling pathways in the melanoma, we analyzed patient survivals as a function of high versus low expression of BMP1, BMP2, and BMP6 in two patient groups: BRAF-mutant and BRAF wild-type. We found that patients with high BMP6 and BMP2 expression had significantly better progression-free survival (PFS) compared with the patients with low expression. However, there were no PFS benefits between high and low expression of BMP2 and BMP6 in the BRAF-wild-type patient group (Fig. 1B; Supplementary Fig. S2). This trend is consistent in the OS analysis (Supplementary Fig. S3), suggesting a functional counteractive cross-talk between the two signaling pathways. Because the RNA-seq data of cutaneous melanomas in TCGA were obtained mostly from metastatic melanoma tissue samples, these observational data in clinical samples, while simply correlative, suggests that BMP2- and BMP6-activated signaling might have an overall antimetastatic role, whereas BMP1 might has a metastasis-promoting role in BRAF-mutant melanomas.

To test whether any of the HDAC1/2 containing transcriptional repressor complexes were involved the transcriptional silencing of these BMP genes, we treated human melanoma cells with small-molecule inhibitors; MS275: a class I HDAC inhibitor, GSK2879552: a lysine-specific histone demethylase 1 (LSD1/KDM1A) inhibitor, and Corin: a dual-action inhibitor of class I HDACs and LSD1 (Fig. 1C; ref. 36) to inhibit the enzymatic activities of HDAC1/2 in the transcriptional repressor complexes (SIN3, CoREST, and NuRD complexes). Subsequent measurements of BMP family members at mRNA and protein levels by RT-PCR and immunoblot analyses (Fig. 1D and E) showed that BMP2/4/6/7 gene expression was significantly increased by the HDAC inhibitors (MS275 and Corin), whereas expression of BMP1 was not affected. However, the LSD1 inhibitor GSK2879552 had no transcriptional activation effect on any of the tested BMP family genes. Putting all this together, these results from in vivo patient samples and in vitro melanoma cell lines suggest that putative metastasis-suppressors BMP2 and BMP6 are silenced at the transcriptional level in melanoma by HDAC1/2 containing transcriptional repressor complexes.

To determine which HDAC1/2-containing repressor complex is responsible for the transcriptional silencing of BMP2 and BMP6 in melanoma, we carried out chromatin immunoprecipitation (ChIP) assays spanning ∼1.0 kb from the proximal promoter region to the transcription start site using antibodies to the key components of each repressor complex (SIN3A, NuRD, and CoREST) in BRAF(V600E)-mutant melanoma cell lines WM1341D and 1205Lu. This procedure revealed that, among these three complexes, only the SIN3A–HDAC1/2 complex occupies the promoter region of the BMP2 and BMP6 genes (Fig. 2A; Supplementary Fig. S4A). In contrast, none of the HDAC-containing repressor complexes were localized to the BMP1 gene promoter (Fig. 2A; Supplementary Fig. S4A) as anticipated by the fact that BMP1 is highly expressed in melanoma patient samples (Fig. 1A) and HDAC inhibitor treatment has no effect on BMP1 gene expression (Fig. 1D and F). Because SIN3A and SIN3B are reported to form distinctive HDAC1/2-containing repressor complexes that carry out different biological functions (37), we also assessed possible SIN3B occupancy of the BMP6 promoter region in the 1205Lu melanoma cells. This analysis showed that only SIN3A was bound in the proximal promoter region of this gene (Fig. 2B). Furthermore, the region that was occupied by either HDAC1 or HDAC2 overlapped with that occupied by SIN3A (Supplementary Fig. S4B; Fig. 2A) as expected, suggesting that HDAC1/2 is a core component of the SIN3A repressor complex regulating BMP6 gene expression.

To investigate the chromatin structure and Corin drug response of the proximal regulatory region of the BMP2 and BMP6 genes, we performed additional ChIP analysis with antibodies for transcriptionally active histone marks H3K9ac and H3K4me3. Acetylated H3K9 was undetectable in the proximal promoter region of the BMP2 and BMP6 genes, and the dual-action HDAC inhibitor Corin significantly increased the acetylated H3K9 (Fig. 2C). Recruitment of RNA polymerase II (Pol II) was also evident in the BMP6 gene promoter after Corin treatment of the melanoma cells, compared with the vehicle control (Fig. 2D). It is known that H3K4me3 and H3K27me3, active and repressive histone marks, respectively, are found in bivalent chromatin domains enriched in genes associated with pluripotency and differentiation (38). As shown in Fig. 2E, the active bivalent mark H3K4me3 was increased in the BMP6 gene promoter region by Corin treatment compared with vehicle control, whereas there were no changes in the repressive bivalent mark H3K27me3 (Fig. 2E).

To further verify the transcriptional repressive role of the SIN3A–HDAC1/2 complex at the BMP6 locus, we genetically knocked down SIN3A expression by RNA interference using small hairpin RNA (shRNA) and observed that SIN3A depletion increased BMP6 protein expression in the melanoma cells (1205Lu and WM983B; Fig. 2F). Furthermore, the increase of BMP6 gene promoter-reporter (39) activity in the melanoma cells with stably knocked down SIN3A (Fig. 2G) and dose-dependent elevations of the promoter-reporter activity by Corin treatment (Fig. 2H) demonstrated that the SIN3A–HDAC1/2 repressor complex is a key transcriptional regulator of BMP2 and BMP6 expression in human melanomas.

To evaluate the functional role of the BMP2 and BMP6 in melanoma metastatic phenotype development in vitro, we used two approaches: treatment of melanoma cells with recombinant proteins and stable expression of the BMP2 and BMP6 genes. First, to assess the endogenous levels of SMAD signaling activation in melanoma cells and the effect of BMP2 and BMP6 proteins to the signaling pathway, we treated melanoma cells with each recombinant BMP2 and BMP6 protein and determined the phosphorylation status of the BMP-specific SMAD proteins: SMAD1, SMAD5, and SMAD9, by immunoblotting using an antibody recognizing all three p-SMADs. All the melanoma cells tested (WM1341D. WM983B, MeWo, and 1205Lu) showed that base level of the signaling activity was low without ligand stimulation; however, upon treatment with the ligands, the signaling pathway was highly activated (Fig. 3A and C). The HDAC inhibitors Corin and MS275 also induced elevated BMP–SAMD signaling activation compared with vehicle controls (Fig. 3B). To determine which SAMD is the downstream mediator of the BMP6-induced SMAD signaling activation in melanoma cells, we performed immunoblot analyses using three antibodies that recognize a different set of phosphorylated SMADs. Of note, there are no antibodies currently available that are specific to p-SMAD5 and p-SMAD9. BMP6 treatment (50 ng/mL for 2 hours) in 1205Lu melanoma cells induced a minimally increased p-SMAD1 level assayed with an antibody specific to p-SAMD1, whereas a higher level of p-SAMD1/5 was detected with an antibody recognizing both p-SAMD1 and p-SMAD5, suggesting that BMP6 induces SMAD5 phosphorylation in the cells. Furthermore, there was no increase of the p-SMAD level assayed with an antibody detecting all three p-SMAD1/5/9 compared with that assayed with the antibody recognizing p-SMAD1/5 (Fig. 3C). Based on these immunoblot analyses, we conclude that BMP6 activates the SMAD signaling pathway via SMAD5 phosphorylation in melanoma, which is also supported by the increased expression of the hepcidin coding HAMP gene, a known BMP6 target gene (40), in the cells stably expressing BMP6 and in the Corin-treated cells (Fig. 3D). These results suggest that the BMP–SMAD signaling pathway is genetically intact but epigenetically suppressed by the transcriptional silencing of the BMP6 genes mediated via SIN3A–HDAC1/2 complex in melanoma.

To explore the metastasis-related phenotypic changes caused by long-term BMP-induced SMAD5 signaling activation, we generated a melanoma cell line stably expressing BMP6 (1205Lu-BMP6), and evaluated the impact of BMP6 on cell physiology, including cell spreading, volume, granularity, proliferation, migration, invasion, and colony formation capacities. Expression levels of BMP6 in the 1205Lu-BMP6 cells that were stably transfected with a BMP6-expressing vector (pCDNA3.1-BMP6) were similar to that of the cells treated with Corin (Supplementary Fig. S5A), suggesting that the ectopically expressed BMP6 level is comparable to that of chemically induced BMP6 in this cell line. The BMP6 stably expressing cells appeared to be morphologically distinctive compared with the control cells, with the BMP6 expressing cells being larger in size and having increased granularity as measured by the shifting of forward scatter and side scatter in flow cytometry analysis (Supplementary Fig. S5B). Furthermore, we noticed that the BMP6 expressing cells grow in an aggregated fashion compared with the control cells that grow as scattered single cells. BMP6-induced SMAD5 signaling activation also significantly decreased cell invasion (Fig. 3E) and migration (Fig. 3F). Importantly, these two effects were independent of the effects on cell growth rates, because in the migration assay cells grow in serum-free conditions to limit cell proliferation and no significant proliferation changes were observed in the short-term culture conditions (20 hours; Fig. 3F). Moreover, we verified that SMAD5 is a mediator of the BMP6-induced suppression of metastatic potential using an SMAD5-depleted cell line (1205Lu-SMAD5-KD) in which the known SMAD signaling down targets ID1, ID2, and ID3 were not induced by recombinant BMP6 treatment in contrast to the control cells (1205Lu-EV; Supplementary Fig. S5C). We observed the significantly decreased inhibitory effect on cell migration by BMP6 in the SMAD5-depleted cells compared with that of control cells (Supplementary Fig. S6D). Lastly, BMP6 expression reduced cell colony sizes in soft agar (Fig. 3G), consistent with slowing in cell-cycle kinetics and a G₁ phase arrest (Supplementary Fig. S5E).

To test the effect of BMP6–SMAD5 signaling activation on melanoma cell dissemination and metastatic tumor growth, we developed a xenograft-based mouse metastasis model system, using 1205Lu human melanoma cells, a derivative cell line from spontaneous lung metastases of WM793 cells, which was originally isolated from a vertical growth phase human melanoma tumor (41, 42). The 1205Lu-Luc cells stably expressing luciferase were injected s.c. into NSG mice, and lung metastatic cells were isolated to establish a lung-specific metastatic cell line (1205Lu-M1/Luc; Supplementary Fig. S6A and S6B). Another round of this in vivo selection process resulted in a cell line (1205Lu-M2/Luc) which had high potency to metastasize to the mouse lungs from the primary xenograft tumor. To assess the role of BMP6 in metastasis, a BMP6-overexpressing cell line (1205Lu-M2/Luc-BMP6), which is approximately 2-fold higher than that of control cells (1205Lu-M2/Luc-EV; Fig. 4A) was generated and inoculated s.c. into two groups of NSG mice. To separate tumor growth effects from metastatic capacity, we first monitored the primary xenograft tumor growth. As expected from the in vitro data (Fig. 3G; Supplementary Fig. S5E), stable expression of BMP6 significantly slowed the primary xenograft tumor formation and growth (Fig. 4D). The BMP6-expressing primary tumors of the five mice grew to the size of 1.0 cm³ in volume at 28-day post-inoculation compared with the 17 days in the control group (Fig. 4B and C).

To evaluate the effect of BMP6–SMAD5 signaling activation specifically on metastatic tumor dissemination, the primary tumors were surgically removed at a size of approximately 1.0 cm³ in volume from both groups of mice to equalize the primary tumor burden between the BMP6 overexpressing tumor bearing mice and the control tumor cell–injected mice. Two weeks later, blood samples were collected to measure circulating tumor cells (CTC), and the lungs were harvested to measure lung metastatic tumor burdens (Fig. 4B). Approximately 500 L of blood was collected right before harvesting the lung tissues, and further processed to extract total RNA. Messenger RNA levels of human melanocyte lineage-specific cell marker S100A6 gene and ectopically incorporated luciferase gene in the genome of 1205Lu-M2/Luc-BMP6 cells were determined by RT-qPCR. The CTC levels were significantly lower in the mouse group with the BMP6-overexpressing xenograft tumors (n = 5; CTCs were undetectable in the 3 mice without primary tumors) than those of the control group (Fig. 4E). Lastly, we assessed metastatic tumor burden in the lungs by measuring bioluminescence intensities. Metastatic tumors were detected in only 3 out of 5 mice in the BMP6-overexpressing group (Fig. 4F), whereas all 7 mice in the control group had detectable lung metastases. Moreover, the metastatic tumor burdens in the mouse lungs of the BMP6 group were significantly lower than those of the control group (Fig. 4G and H). Differential expression of BMP6 between melanoma lung metastases harvested from the two mice groups was validated by immunofluorescent staining assay (Supplementary Fig. S8D). Taken together, these data implicate suppressive roles of BMP6–SMAD5 signaling in melanoma cell dissemination and metastatic tumor growth.

To determine whether Corin was a potential treatment for inhibiting melanoma metastasis, we utilized the melanoma lung metastasis model mentioned previously. The 1205Lu-M2/Luc cells were subcutaneously inoculated into NSG mice to form primary xenograft tumors. The primary tumors were then surgically removed at day 14 post-inoculation when the tumors were approximately 1.0 cm³ in volume (Fig. 5A and B). The mice were divided into two groups, vehicle control and Corin treatment, with equal distribution of the primary tumor burden between the two groups. Two days after the surgery, the mice were treated with Corin (15 mg/kg, 3 times per week for two weeks i.p.) or with DMSO as a vehicle control. We observed that only 17% of Corin-treated mice (1/6) developed lung metastases as detected by ex vivo bioluminescence imaging, compared with the 100% (6/6) of the control group (Fig. 5C). The lung metastatic tumor burden was also significantly lower in the Corin-treated group relative to the control group (Fig. 5D).

One of the rate-limiting steps in distant organ metastatic tumor formation is the dormancy of disseminated cancer cells, with escape from the dormancy program being a necessary step for metastatic tumor growth (43, 44). To test whether metastatic tumor growth inhibition by Corin is mediated by the induction and enforcement of cell dormancy, we first evaluated the expression levels of disseminated tumor cell dormancy marker proteins p21^Cip1 and phospho-p38 (45, 46) and a dormancy escape-promoting transcription factor, FOXM1 (47), by Corin treatment in multiple cell lines. Corin treatment increased the levels of dormancy markers p21^Cip1 and phospho-p38 while decreasing FOXM1 levels in melanoma cell lines (Fig. 5E). For the in vivo validation of dormancy induction by Corin, we quantitated metastatic melanoma cells in the lungs of mice treated with Corin versus DMSO vehicle and measured the dormancy marker–positive cell fractions by flow cytometry. Fractions of the metastatic tumor cells (luciferase-positive cells) were significantly lower in the lungs harvested from mice treated with Corin compared with the vehicle-treated group (5.5% vs. 1.6%; Fig. 5F), which is paralleled by the proliferation marker Ki-67–positive cell fractions (30.9% vs. 12.7%) of the lung metastatic melanoma cells. However, percentages of the metastatic melanoma cells positive for the dormant cell marker p27^Kip1 (48) were significantly increased in the Corin-treated group compared with the vehicle control group (16.9% vs. 49.2%; Fig. 5F). These results suggest that Corin, or future small-molecule drugs derived from it, may be effective epigenetically acting agents for adjuvant therapy in patients with melanoma.

FAM83G, a gene encoding PAWS1 (Protein Associated With SMAD1) protein, is a substrate of the type I bone morphogenetic protein receptor (BMPRI) kinase and regulates cytoskeletal dynamics (49). However, its role in metastatic tumor progression is undefined. RNA-seq analysis of melanoma cells stably expressing BMP6 suggested that FAM83G is a potential BMP6/SMAD5 downstream target as the two of three cell lines showed significantly decreased FAM83G/PAWS1 expressions after the gains of BMP6 expression (GEO accession code: GSE164495). This was validated by an independent RT-qPCR assay (Supplementary Fig. S7A), Immunoblot analysis also confirmed the decreased FAM83G/PAWS1 protein level by BMP6 whereas the response is heterogeneous as two cell lines (WM1341D and WM164) out of eight melanoma cell lines tested in this study were unresponsive to the BMP6 overexpression (Fig. 6A; Supplementary Fig. S7B). Binding of phosphorylated SMAD5 on the proximal promoter of the FAM83G gene was validated by ChIP-PCR analysis (Fig. 6B) and the decreased FAM83G transcript and protein levels in the SMAD5 knocked-down melanoma cells (1205Lu-SMAD5-KD; Fig. 6C) suggests that FAM83G/PAWS1 is a downstream target silenced by the BMP6/SMAD5 signaling pathway, which may play a role in cancer cell migration and invasion.

To test whether BMP6–SMAD5 signaling exerts metastasis-suppressive effects via downregulation of FAM83G/PAWS1, we performed BMP6-induced phenotype rescue assays, specifically scoring the migration and invasion capacities of melanoma cells. Stable expression of FAM83G/PAWS1 in metastatic melanoma cells (1205Lu and WM983B) rescued the phenotypes of decreased cell migration and invasion caused by the BMP6 overexpression in in vitro assays (Fig. 6D; Supplementary Fig. S7C). Furthermore, significantly decreased F-Actin stress fiber assembly and distribution by BMP6 overexpression in melanoma cells (Fig. 6E), and a prior report of FAM83G/PAWS1 regulating cytoskeletal dynamics and cell migration (49) indicates that BMP6–SMAD5 signaling activation suppresses the key metastatic properties of melanoma cells at least in part via a transcriptional downregulation of FAM83G/PAWS1.

To validate the critical roles of FAM83G/PAWS1 in metastatic melanoma development in vivo, we generated a FAM83G/PAWS1-depleted cell line (1205Lu-M2/Luc/FAM83G-KD) from our lung-specific metastatic cell line (Fig. 7A) and evaluated the lung metastatic potency using the melanoma metastasis model (Fig. 4; Supplementary Fig. S8A and S8B). The FAM83G/PAWS1-depleted cells showed slightly decreased primary tumor growth rate, but this was statistically insignificant compared with that of control cells (1205Lu-M2/Luc-EV; Supplementary Fig. S8C). However, the depleted cells showed significantly decreased lung metastasis formation compared with that of 1205Lu-M2/Luc-EV control cells (Fig. 7B). We also measured blood CTC levels and found lower levels of the melanoma CTCs in the mice with the 1205Lu-M2/Luc/FAM83G-KD (Fig. 7C), suggesting that FAM83G/PAWS1 plays an important role in melanoma cell dissemination. To determine whether the elevated BMP6 level remained in the metastatic tumors of the mouse lungs after enduring multiple steps of the metastatic process, and to validate the inverse correlation between BMP6 and FAM83G/PAWS1 expression that was observed in the in vitro study (Fig. 6A–C; Supplementary Fig. S7 and S7B), we performed immunofluorescence staining of the lung tissues harvested from the mice bearing lung metastases induced by the BMP6-expressing cells (1205Lu-M2/Luc-BMP6) and corresponding control cells (1205Lu-M2/Luc-EV; Fig. 4A and F). These analyses confirmed the higher expression of BMP6 compared with that of the control group (Supplementary Fig. S8D, top), whereas there was an inverse correlation between BMP6 and FAM83G/PAWS1 (Supplementary Fig. S8D, bottom).

To correlate the metastasis-promoting roles of BMP6 and FAM83G/PAWS1 in human metastatic melanoma development, we measured BMP6 and FAM83G gene transcript levels in patient tissue specimens. Fifty archival cutaneous melanoma patient tissues (25 each primary and metastatic melanoma tissue samples) were analyzed for mRNA expression levels of SIN3A, BMP6, and FAM83G genes using quantitative real-time PCR assays. As predicted, BMP6 gene transcript levels were significantly lower in the metastatic melanoma tissues than in primary tumors, whereas FAM38G expressions in the metastatic tissue samples were significantly higher than in primary tissue samples (Fig. 7D). We also evaluated expression correlations between the FAM83G and BMP6 genes in the melanoma and normal skin tissues using RNA-seq data sets from TCGA and GTEx. There was an inverse correlation between the expression levels of the two genes in the tumor and normal skin tissues; however, this correlation significantly diminished in tumor samples compared with normal skin, which may be due to transcriptional silencing of the BMP6 gene in the tumors (Fig. 7E). Together, these data indicate that FAM83G/PAWS1 is a BMP6/SMAD5 target and may significantly contribute to metastasis progression by promoting melanoma cell dissemination.

Epigenetic variation has been implicated in cancer metastasis by experimental and clinical data. First, primary tumors with the same genetic mutations may develop phenotypically distinct metastatic tumors, with activation of specific transcriptional programs (50, 51). Second, there is a linkage between mutations in genes encoding chromatin regulators, including members of the SWI/SNF chromatin remodeling complexes (52), and altered transcriptional programs in metastatic tumor cells. However, investigating and understanding how such epigenetic changes can activate specific metastatic transcriptional programs has been challenging. This study demonstrates that a chromatin remodeling transcriptional repressor complex, SIN3A–HDAC1/2, epigenetically represses BMP6 expression, thereby dysregulating BMP6–SMAD5-controlled gene transcription programs that promote metastatic progression of melanoma cells, including those regulating cytoskeletal dynamics mediated by FAM83G/PAWS1. These altered transcription programs may work cooperatively with oncogenic programs activated by genetic mutations, such as gain-of-function mutations of BRAF in melanoma, resulting in the enhanced metastatic phenotype. Analysis of clinical data also supported this notion, as the association of OS and PFS with higher expression of BMP2 and BMP6 was only observed in melanoma patients with BRAF-mutant melanomas.

SIN3 is a transcriptional coregulator controlling diverse sets of target genes in a context-dependent manner. SIN3 forms transcription repressor complexes by recruiting other cofactors including HDAC1/2. The histone modification and chromatin remodeling activities of the SIN3 complex are exerted via HDAC1/2 and ATPase-dependent chromatin remodeling factors such as BRG1 in the multiprotein complex (53). Core components of the SIN3 complex have no known DNA binding activity, and it is unknown so far which transcription factors (TF) are responsible for the recruitment of the SIN3 complex to the target gene promoters, in particular the BMP6 and BMP2 genes. Identification of such TFs may provide a mechanistic insight on how the SIN3 complex regulates gene transcription in a tissue-specific and context-dependent manner. Hence, further research should be conducted.

BMPs have been shown to exert dual effects on malignant progression in cancer depending on cancer type (29). In melanoma, BMP4 and BMP7 have been shown to play a role in enhanced tumor cell migration and invasion (54). BMP2 and BMP4 contribute to tumor growth by promoting angiogenesis via paracrine signaling (55). Conversely, different studies have reported that BMP7 inhibits melanoma cell proliferation and metastatic progression (56). The functional roles of BMP6 in the metastatic progression have not been well defined. Hu and colleagues reported that BMP6 in breast cancer inhibits metastasis via regulation of MMP-1 expression (57). Although our study supports the metastasis-suppressive role of the BMP6 in melanoma, a recent study from Stieglitz and colleagues showed that increased expression of BMP6 in melanoma cells and tissues, and BMP6 depletion caused significantly delayed tumorigenesis in a mouse model (24). This seemingly contrasting result may be explained considering several factors attributed to the different experimental conditions. First, our study focused on the BMP6 role in metastatic phenotype development using melanoma cell lines and patient tissue specimens, whereas the study by Stieglitz and colleagues focused more on tumorigenesis. Second, we used a xenograft-based metastasis model in an immune-compromised mouse strain with human melanoma cells, whereas Stieglitz and colleagues used transgenic mouse melanoma model in which the host immune cells played a role in delayed onset of tumorigenesis and progression. Thus, although both approaches have advantages and disadvantages, our system measures purely the metastasis phenomenon, independent of the host immune response. Third, our study showed an antimetastatic role of BMP6 only in the context of human tumor with BRAF-mutant melanomas: the cell lines we used were mostly BRAF-mutant cells including 1205Lu cell line used for the metastasis model, which was not considered in the study by Stieglitz and colleagues. Taken together, it is likely that BMP6 plays a dual role in melanoma tumorigenesis by promoting tumor formation, whereas a metastasis-suppressive role by inhibiting cancer cell dissemination in the late stage of metastasis progression particularly in BRAF-mutant melanomas.

FAM83G is one of eight members of FAM83 family proteins that are conserved in vertebrates. The FAM83 proteins share a conserved domain DUF1669 (domain of unknown function 1669; ref. 58). A recent study shows that the FAM83 proteins interact with casein kinase 1 isoforms and regulate WNT signaling (59). Although recent studies implicate that FAM83 family proteins may play in diverse biological functions, precise molecular and pathogenic functions in cancer progression remain poorly characterized. Although some biological functions of FAM83G are known, such as promoting noncanonical BMP transcription target genes and cytoskeletal dynamics (49, 60), we report here a mechanistic aspect of how FAM83G expression is controlled in cancer cells, a repressive transcriptional target by BMP6–SMAD5 signaling, hence forming a potential negative feedback loop in regulation of the BMP–SMAD signaling cascade. Furthermore, our study also demonstrates a critical role of the FAM83G in cancer cell dissemination, which is an important step of metastatic spreading. Nevertheless, exact molecular mechanisms of how cytoskeletal dynamic in cancer cells controlled by FAM83G are unclear. Determination of such mechanisms will provide a potential therapeutic target for adjuvant therapy for cancer patients and warrants further research.

A group of small-molecule inhibitors of HDACs has been approved for the treatment of hematologic malignancies (61). HDAC inhibitors can not only block the catalytic activity of HDACs in the multiprotein complex but also disrupt the interaction dynamics with other core components (62), indicating that the inhibitory activity of a given HDAC inhibitor may differ once the protein resides in a macromolecule complex (18). In fact, a few inhibitors disrupting SIN3 interactions with binding partners have been identified as novel therapeutics against triple-negative breast cancer (63). These results suggest that SIN3A–HDAC1/2 complex is a druggable macromolecular assembly. Although Corin was initially developed for targeting HDAC1/2 and LSD1 in the CoREST complex (36), it also displays a high potency to inhibit HDACs in the SIN3A complex as demonstrated in this study. The functional link uncovered in our study between inhibiting the SIN3A–HDAC1/2 complex, activating BMP6–SMAD5-mediated signaling, and the resulting suppression of the metastatic phenotype may be one of the underlying mechanisms of the antimetastatic effect of Corin, suggesting a new potential therapeutic agent for melanomas.

After surgical excision of early-stage melanomas, subsets of high-risk patients will subsequently relapse with disseminated disease. Thus, adjuvant therapies in melanoma must target residual micrometastatic cancer cells (7). Clinically available standard-of-care agents for targeted therapy and immunotherapy are now moving into the adjuvant setting (64, 65). Given the highly effective role of Corin in inhibiting cancer cell dissemination and colonization in the preclinical melanoma metastasis model described here, it may be an effective epigenetically acting drug for melanoma adjuvant therapy, warranting further evaluation in a clinical setting. Given the distinct mechanism of action of Corin as compared with oncogene-targeting and immunotherapeutic agents, therapeutic combinations of Corin with such drugs may increase the efficacy of the adjuvant therapy in melanomas.

P.A. Cole reports personal fees from Scorpion Therapeutics outside the submitted work; in addition, P.A. Cole has a patent for US Patent App. 16/892,825 pending; and has been a cofounder and equity holder in Acylin Therapeutics and consultant for AbbVie, which have been involved in epigenetic drug discovery. P.A. Cole has also served as a consultant for Constellation and Epizyme. B. Ryu reports grants from NIH during the conduct of the study. No disclosures were reported by the other authors.

D. Min: Data curation, formal analysis, investigation, visualization, and methodology. J. Byun: Data curation, investigation. E.-J. Lee: Data curation and investigation. A.A. Khan: Investigation. C. Liu: Investigation. O. Loudig: Investigation. W. Hu: Investigation. Y. Zhao: Investigation. M. Herlyn: Resources, writing–review and editing. B. Tycko: Formal analysis, investigation, writing–review and editing. P.A. Cole: Resources, writing–review and editing. B. Ryu: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, writing–original draft, and project administration.

We thank Dr. Isaac Yi Kim and Geun Tack Lee at Rutgers Cancer Institute of New Jersey for kindly providing BMP6-luciferase promoter-reporter plasmids. We would like to recognize Dr. Johannes Zakrzewski at the Center for Discovery and Innovation, Hackensack Meridian Health for sharing Lenti-X 293T cells. We thank Dr. Rajit Malliah at the Department of Pathology, Hackensack University Medical Center for pathologic information of patient tissue samples. This work has been supported by an NIH grant R01 CA212639 to B. Ryu and P.A. Cole.

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.