Melanoma cells M1626 and M727 were maintained in RPMI with 10% FBS, M525 cells were maintained in RPMI with 5% FBS, HepG2 cells were maintained in DMEM with 10% FBS, and M354 cells were maintained in 1:1 of RPMI with 10% FBS and dermal cell basal media from ATCC (PCS-200-030). Normal melanocytes were maintained in dermal cell basal media from ATCC (PCS-200-030). All cells were grown in Tissue Culture incubators set at 5% CO2 and 37°C degrees. For NRF1 siRNA mediated knockdown, a total of 1 × 10⁵ M727 cells were seeded in 6 well plates and incubated overnight at 37 ° C and 5% CO₂. Cells were then transfected using 8 μl of siRNA transfection reagent (Santa Cruz, Cat. No. sc-29528) and 120 pmol of NRF-1 siRNA (Cat. No. sc-38105) or control siRNA-A (Santa, Cruz, Cat. No. sc-37007).

Detection and quantification of CD47 cell surface protein levels were performed using FACS and an anti-human CD47 antibody (BioLegend, Cat. No. 323106). Briefly, cells were cultured in an exponential phase and harvested before reaching confluency. To harvest the cells, the spent medium was carefully aspirated, cells were washed with PBS, and detached using Accutase™ solution (STEMCELL, Cat. No.07922) for 3 min at 37°C in an incubator with 5% CO₂. The cells were then collected and washed twice with PBS and resuspended in 100 μL of ice-cold cell staining buffer (CSB) (PBS, 1%BSA) at a concentration of 0.5-1x10⁶cells/ml. Following resuspension, cells were stained with anti-human CD47-FITC antibody (1μg/1x10⁶cells) and incubated on ice protected from light for 20 min. After incubation, the cells were centrifuged at 200g for 5 min, the supernatant was carefully aspirated, and the cells were washed twice with 200 μL of CSB. The cells were transferred to FACS tubes and analyzed in a BD FACSymphony A5 cell analyzer, at least 1x10⁴ cells live cells were recorded per cell type. Dead cells were excluded with a DAPI staining solution (10μg/mL). Once flow cytometry analysis was completed, data was analyzed in FlowJo v10.10 software, live cells were gated, and the geometric mean fluorescence intensity (GMFI) was obtained for each cell type. GMFI values were normalized (same number of live events per cell type) and compared between cell lines as well as their respective histograms.

Cells were grown in-vitro on p100 cell culture dishes until reaching ~70% confluency after which they were lysed using 1ml of Trizole reagent. mRNA was precipitated using isopropanol and washed twice with 70% ethanol. cDNA was synthesized using a high-capacity cDNA reverse transcription kit (4368814) and power SYBR green PCR MasterMix (4367659 Thermo Fisher Scientific) was used to quantify total levels of CD47 or NRF-1 transcripts; 18s gene was used for the normalization of qPCR results. Primer pair sequences are listed in Supplementary Table 1 .

For ATAC-Seq 50,000 cells were collected, washed in 50 μl of cold PBS, and resuspended in 50 μl of cold lysis buffer. Nuclei were isolated by centrifugation at 500 x g for 10 min at 4°C and then incubated with a transposition reaction mix for 30 min at 37°C. Following transposition, DNA was then purified using the MinElute PCR Purification kit (Qiagen, Hilden, Germany). The transposed DNA fragments were barcoded initially amplified for 5 cycles, and further amplified by standardized PCR. The resulting ATAC libraries were sequenced on an Illumina HiSeq 2500 platform in Rapid Mode with the 50-bp paired-end reads output mode. Raw reads were aligned to the human genome reference (1000 Genomes) using BWA-MEM (version 0.7.5a) with default settings. ATAC-Seq peaks were identified using the callpeaks function in MACS2 with a threshold set to -q = 0.01. The resulting peak calls were filtered for sequences that mapped to the mitochondria using shell scripts. BigWig files were normalized and integrated with publicly available genomic and epigenomic data for visualization in the UCSC Genome Browser.

For the ChIP assay cells were prepared using truChIP Chromatin shearing kit (PN520127) from covaries. Briefly, cells were cultured in 2 x p150 dishes and treated with fixing buffer have 1% formaldehyde for 10 mins followed by quenching of crosslinking with a quenching buffer for 5 mins. Cells were lysed per the manufacturer’s protocol and isolated nuclei were transferred into AFA tubes (Cat# 520135 covaris) and chromatin shearing was performed for 15 mins in E220 evolution ultrasonicator from covaries with recommended settings. 50μl of sheared chromatin sample was used to isolate DNA ( Supplementary Figure S2 ) using a DNA purification kit from cell signaling technologies (cat# 14209S) thus the total amount of DNA in sheared chromatin was calculated using nanodrop. For immunoprecipitation, the buffers (cat# 14231) and ChIP-grade magnetic beads (cat# 9006S) were purchased from cell signaling and used as per the manufacturer’s protocol. Briefly, 5-7μg of DNA was used per reaction and 2% input material was set aside for quantification using qPCR. Each reaction was incubated overnight with 2.5-3μg of NRF-1, or isotype control IgG antibody followed by 2 hours of incubation with magnetic beads the following day. Immunoprecipitation was performed using a magnetic rack. After successful elution of pulled-down chromatin, DNA was isolated, and real-time qPCR was performed using power SYBR green PCR master mix (cat# 4367659) from Thermo Fisher Scientific. Primer pair sequences are listed in Supplementary Table 1 .

CD47 promoter reporters containing various numbers of NRF-1 binding sites were created by synthesizing respective DNA regions and cloning them into pGF1-4eCOL2A1_LUC lentiviral vector using BamHI and Spe1 restriction sites. Viral packaging was done in HEK293 cells and collected virions were used to infect target cells for two consecutive days. The CD47 promoter activity of endogenous NRF-1 was assayed by measuring Luciferase bioluminescence.

TCGA TARGET GTEx study was used to analyze CD47 expression levels in human melanoma tissues. ENCFF323SMV, ENCFF862EDR, ENCFF605IET, ENCFF785WIA, ENCFF000XJF, ENCFF644ITQ from encodeproject.org studies were used to analyze NRF-1 ChIP seq data across different types of cancer. Patient-derived melanoma cell lines were analyzed for chromatin accessibility at the CD47 promoter region using the data from the study GSE134432. The ATAC-Seq data for the melanoma patient tissues was fetched from https://gdc.cancer.gov/about-data/publications/ATACseq-AWG. For in-silico H3K4^Me3 analysis of patient-derived melanoma tissue from the database GSE33930, bowtie aligner was used to align the sequence reads, followed by the emoval of mitochondrial reads and duplicates using samtools. Macs2 was used to invoke callpeaks function to obtain peakfiles which were used to generate bigwig datafiles to be visualized in IGV.

We previously established that high levels of CD47 protein on freshly isolated human tumor cells are associated with melanoma progression and immune evasion (2). To understand whether augmented CD47 expression results from an increase in corresponding mRNA production, we used metastatic melanoma patient-derived cells M727, M354, and M1626 that were determined to express various levels of CD47 ( Figures 1A, B ) and compared them to normal melanocytes or hepatocellular carcinoma cells, HEPG2, (known for low levels of CD47 protein) using qRT-PCR. CD47 gene is comprised of 11 exons and as a result of splicing, most notably between exons 8-10, multiple mRNA isoforms are translated ( Supplementary Figure 1A ). Therefore, to prospectively quantify levels of all CD47 transcript isoforms we designed multiple primer sets for a common exon 2, as well as, two primer sets that amplify mRNA regions between exon 2-4 and exon 7-8 ( Supplementary Figure 1B , Supplementary Table 1 ). After performing multiple qRT-PCR assays, our results demonstrate that all metastatic melanomas expressed 10-15 folds (depending on the isoform) elevation in CD47 mRNA as compared to normal melanocytes or HepG2 cells ( Figure 1C ). To further explore clinical significance of our findings, we investigated CD47 mRNA levels among 422 melanoma subjects in comparison to 123 normal skin subjects using TCGA TARGET GTEx dataset. Utilizing an unsupervised dimensional reduction approach, we generated UMAP graphs that visualized patterns of clusters across diagnosed subjects and levels of CD47 mRNA expression. Significantly, two distinct clusters were formed based on the CD47 mRNA levels, high and low, which corresponded to the malignant melanoma patients cluster (high) and the normal skin cluster (low) ( Figure 1D ).

To understand molecular mechanisms of augmented CD47 mRNA expression in malignant melanoma, we investigated the dynamics of chromatin re-modeling at CD47 promoter using ATAC-Seq. This approach allows the generation of a cell-specific chromatin accessibility landscape to determine the level of open chromatin and subsequent promoter activation. First, we used in-silico ATAC-Seq analysis of the skin cancer melanoma (SKCM) dataset (https://gdc.cancer.gov/about-data/publications/ATACseq-AWG) containing malignant tissues obtained from 13 melanoma subjects (11 samples were analyzed in duplicates) and a melanoma dataset (GSE134432) containing 9 genetically homogenous melanoma patient-derived cell lines to gain insights into the accessibility of the CD47 promoter region. Our results demonstrate that in both freshly resected melanoma samples, as well as, in melanoma patient-derived cell lines, the DNA region corresponding to the CD47 promoter has a significantly increased chromatin accessibility which could be conjectured as a sign of active mRNA transcription ( Figure 2A ). To validate these changes in the chromatin landscape of the CD47 promoter region during melanomagenesis, we performed ATAC-seq on multiple independent patient-derived melanoma cell lines in comparison to normal adult melanocytes. Importantly, we were able to detect strong peak signals associated with an open chromatin structure around the promoter of CD47 in malignant melanomas, which were missing in normal adult melanocytes indicating the closed chromatin structure of the CD47 DNA promoter region in non-transformed cells ( Figure 2B ).

Open chromatin status is often associated with specific histone modifications such as H3K4^Me3, which is commonly found at the promoter regions of actively transcribed genes (20). To determine whether this mechanism can account for an increased CD47 mRNA production during melanomagenesis, we performed in-silico analysis of H3K4^Me3 histone modifications at the CD47 promoter in melanoma patients and normal adult melanocytes using the GSE33930 dataset. Our results demonstrate that peak signals for H3K4^Me3 DNA modification at the CD47 promoter region are significantly stronger in malignant melanomas in comparison to normal melanocytes ( Figure 2C ).

To identify transcriptional regulators of the CD47 promoter we analyzed its DNA region (± 1000bp relative to the transcription start site (TSS)) for the presence of potential TF binding sites using MotifMap and the Eukaryotic Promoter Databases. Results ranking based on FDR (< 0.038) and p-value (< 0.001) scores revealed a number of DNA binding sites within that region (-63bp, -28bp, -22bp, -14bp, and +32bp), all corresponding to the NRF-1 consensus sequence ( Figure 3A ). To validate predicted interactions of NRF-1 with CD47 promoter we first performed in-silico analysis of ChIP-seq datasets derived from different types of cancer. Importantly, strong peak signals corresponding to NRF-1 binding at the CD47 promoter were identified in the majority of tumors (results presented as fold changes over control), except for HepG2 hepatocarcinoma ( Figure 3B ), which was found to express low CD47 protein and mRNA transcript levels ( Figure 1 ).

Next, to determine NRF-1 regulation of CD47 promoter, we performed Chromatin Immunoprecipitation (ChIP) assays on malignant patient-derived melanoma cells, as well as, adult normal melanocytes and HepG2 cells. Following incubation of the sheared chromatin (500-1000bp) with NRF-1 Ab the pulled-down DNA fragments were quantified by PCR using primer sets designed to amplify only the CD47 promoter regions spanning different putative NRF-1 binding sites ( Figure 4A ). To precisely delineate NRF-1 DNA binding regions within the CD47 promoter, we designed multiple primer pairs to enrich for the following sites relative to TSS: Set 1 (-38/+28) includes putative binding sites at -28, -22, and -14bp; Set 2, (-38/+56) includes putative binding sites -28, -22, -14, and +32; Set 3 (-63/-36) includes putative binding site at position -63bp only ( Figure 4A , Supplementary Table 1 ). An additional set of primer spanning the promoter region of MEF2A gene (a validated NRF-1 binding target) was used as a positive control for all ChIP experiments ( Supplementary Figure 2 ). ChIP analysis on the above indicated cell types identified NRF-1 loading at the predicted DNA region of the CD47 promoter with differential affinity for proximal and distal binding sites ( Figure 4B ). Thus, in malignant melanomas there was a significant increase in binding affinity at the most proximal sites (-28, -22, -14) as compared to normal melanocytes (1.7-fold for M727 and 2.4-fold for M1626). Moreover, when compared to HepG2 cells, which have been characterized by the lowest levels of CD47 mRNA and protein, differences in NRF-1 binding affinity were found to be even greater (1.85-fold for M727 and 3.8-fold for M1626 Figure 4B middle panel). Inclusion of the proximal downstream binding site (+32bp) resulted in similar differences in the NRF-1 binding affinity between malignant melanoma and low CD47 expressing cells (adult melanocytes and HepG2 cells) ( Figure 4B right panel). However, differences in NRF-1 binding affinity were diminished at the more distal binding site (-63bp) within the CD47 promoter region ( Figure 4B left panel). In summary, these findings indicate that CD47 transcriptional regulation by NRF-1 is determined by the proximal promoter region in melanoma cells.

Initiation of transcription from eukaryotic promoters often requires binding of the TFs at numerous locations within the promoter region. Identification and validation of multiple NRF-1 TF binding sites within the proximal region to the CD47 TSS (200 bp window) prompted us to investigate the minimum adequate length of DNA required for the CD47 promoter activation. To achieve this goal, we designed and cloned a series of bioluminescent reporter plasmids expressing Luciferase gene under the control of the proximal CD47 promoter region containing various numbers of NRF-1 TF binding sites ( Figure 5A ). Specifically, construct 1 contained all five NRF-1 proximal binding sites and covered the DNA region from -120 to +50bp relative to the TSS, construct 2 contained four out of five NRF-1 binding sites and covered the region from -70 to +30 bp, and lastly construct 3 contained only three out of five NRF-1 binding sites and covered the region from -62 to +30bp of CD47 promoter relative to the TSS. Lastly, we also used a complete promoter deletion variant of Luciferase expressing plasmid as a negative control (construct 4 Figure 5A ). All of the above variants of the CD47 promoter region were synthesized and cloned into the lentiviral plasmid pGF1-4eCOL2A1_Luc (21) using BamHI and Spe1 restriction sites. In order to evaluate endogenous NRF-1 activity at the CD47 promoter, we generated melanoma cell lines to stably express all versions of Luc reporters using previously published lentiviral transduction protocols (22, 23). To obtain statistically significant results, raw Luc bioluminescence values were first normalized to the viral copy numbers (VCN) for each reporter plasmid. Briefly, serial dilutions of the known concentration of each Luc reporter plasmid were premixed with 5ng of carrier genomic DNA (isolated from each of the above-indicated cell lines), and the real-time PCR was performed using Luciferase primers. Obtained CT values were used to generate a standard curve which allowed to calculate the VCN values for the known concentration of plasmids according to the given formula:

Thus obtained linear regression curves allowed us to calculate viral copy numbers for an equal amount of genomic DNA for each individual cell line ( Figure 5B ). Moreover, raw bioluminescence values were also normalized to the amount of total protein present in each lysate. Normalized data demonstrated that initiation of CD47 transcription requires the presence of all five NRF-1 DNA binding sites ( Figure 5C ). Thus, Luc activity was induced 5-8 fold in melanoma cells containing full-length proximal CD47 promoter (Construct 1) as compared to low CD47 expressing HepG2 cells ( Figure 5C ), while CD47 promoter region containing a reduced number of NRF-1 binding elements (Constructs 2 and 3) had marginal or no effect on reporter activity in the same cells ( Figures 5C–E ). In conclusion, our data demonstrates that the promoter region required for efficient CD47 transcriptional activation in melanoma lies between -120 to +50bp relative to the TSS.

To demonstrate functional significance of NRF1 TF in the regulation of CD47 we tranduced melanoma cells expressing CD47_Luc promoter reporter (Construct 1 described above) with siRNAs (a pool of 3) targeting NRF1 mRNA to achieve its downregulation ( Figure 6A ). Subsequent quantification of Luc signal revealed a significant reduction in CD47 promoter activity (by more than 60% (p=0.0149)) in cells transduced with NRF1 siRNA as compared to matching controls transduced with SC siRNA ( Figure 6B ). Next, we profiled these cells for an endogenous CD47 mRNA levels using qRT-PCR and multiple sets of primers corresponding to its various regions. Our data demonstrates that NRF1 downregulation causes at least 50% decrease in CD47 mRNA expression in target cells (p=0.0245 - 0.00361) ( Figure 6C ). Lastly, we used FACS and fluorescent conjugated antibodies to quantify CD47 protein on the cell membrane, critical aspect of its immune modulatory function during tumor progression. Our analysis demonstrates that NRF1 downregulation leads to the significantly reduced proportion of melanoma cells expressing high levels of this innate immunity checkpoint when compared to the original cell population trasduced with control SC-siRNA: 7.04% CD47 ^high//91.8% CD47 ^low for NRF1-siRNA and 34.3% CD47 ^hig//64.1% CD47^low for SC-RNAi transduced cells ( Figure 6D ). In summary, the above described results provide strong experimental evidence to indicate that NRF1 is directly involved in regulation of CD47 expression in melanoma.