Oncogene-Expressing Senescent Melanocytes Up-Regulate MHC Class II, a Candidate Melanoma Suppressor Function

On acquisition of an oncogenic mutation, primary human and mouse cells can enter oncogene-induced senescence (OIS). OIS is characterized by a stable proliferation arrest and secretion of proinflammatory cytokines and chemokines, the senescence-associated secretory phenotype. Proliferation arrest and the senescence-associated secretory phenotype collaborate to enact tumor suppression, the former by blocking cell proliferation and the latter by recruiting immune cells to clear damaged cells. However, the interactions of OIS cells with the immune system are still poorly defined. Here, we show that engagement of OIS in primary human melanocytes, specifically by melanoma driver mutations NRASQ61K and BRAFV600E, causes expression of the major histocompatibility class II antigen presentation apparatus, via secreted IL-1ß signaling and expression of CIITA, a master regulator of major histocompatibility class II gene transcription. In vitro, OIS melanocytes activate T-cell proliferation. In vivo, nonproliferating oncogene-expressing melanocytes localize to skin-draining lymph nodes, where they induce T-cell proliferation and an antigen presentation gene expression signature. In patients, expression of major histocompatibility class II in melanoma is linked to favorable disease outcome. We propose that OIS in melanocytes is accompanied by an antigen presentation phenotype, likely to promote tumor suppression via activation of the adaptive immune system.


SUPPLEMENTAL DATA
BRAFV600E and control melanocytes were kept in culture under selection for 1 week before being assayed for senescence and gene expression as described (Pawlikowski et al., 2013), using either Affymetrix array (n=3) or RNA-seq (n=1). The full dataset can be obtained from www.ncbi.nlm.nih.gov/geo (accession no. GSE46818). See also   Table S3 for antigen presentation transcript changes detected by RNAseq. See Table S1 for additional information on the dataset.  Table S2 for antigen presentation transcript changes detected by Affymetrix array. See Table S1 for additional information on the dataset. Significance was calculated by students t-test with BH-fdr correction (n=4). Cytokine quantities in filtered culture supernatant from BRAFV600E and vector transduced melanocytes were determined using the human cytokine array G series 3 (Raybio). Culture supernatant was collected 2 weeks post transduction, and 2 days since the medium was last replaced.    Correlation clustered heatmap of a curated list of known proliferation, inflammation and senescence associated secretory phenotype (SASP) genes. The colour intensity represents column Z-score, with red indicating high and blue low expression. The IMR90 fibroblasts were assayed 1 week after infection with BRAFV600E or control vectors.

Cell culture
Multiple batches of different lots of Lightly pigmented neonatal human epidermal melanocytes (Invitrogen) were cultured in medium 254 with human melanocyte growth supplement (HMGS), 100 U/ml penicillin and 100 µg/ml streptomycin (all from Invitrogen). Multiple batches of different lots of Human neonatal epidermal keratinocytes were cultured in EpiLife medium with human keratinocyte growth supplement, 100 U/ml penicillin and 100 µg/ml streptomycin (all from Invitrogen). Keratinocytes were cultured on collagen (Invitrogen) coated plates. IMR90 fibroblasts were obtained from ATCC and cultured in DMEM, supplemented with 20 % (v/v) fetal bovine serum, 100 U/ml penicillin and 100 µg/ml streptomycin (all from Invitrogen).

Lentivirus vector construction, production and infection
Lentivectors encoding CIITA, HRASG12V, NRASQ61K, BRAFV600E, MEK1Q56P, myrAKT, SV40 T-antigen or short hairpins directed against PTEN, CIITA and IL-1B, under the transcriptional control of the cytomegalovirus initial early promoter and puromycin or neomycin resistance from the simian virus 40 promoter were generated using standard methods; details available upon request. Vesicular stomatitis virus G pseudotyped lentivector stocks were produced as described previously (van Tuyn et al., 2007). Melanocytes, keratinocytes and IMR90 fibroblasts were infected overnight in normal culture medium supplemented with 2 (melanocytes and keratinocytes) and 8 µg/ml polybrene, respectively overnight. Followed by 14-32 days of culture in the presence of 1 µg/ml puromycin or 250 µg/ml G418S (Invitrogen) to select for transduced cells.
In all experiments oncogene and control vector transduced cells were kept in culture under selection for 2 weeks before being assayed for senescence and gene expression as detailed below, unless stated otherwise.

Microarray, RNA-seq and analysis of TCGA data
Microarray and RNAseq analysis of melanocytes transduced with BRAF600E expression or control vectors has been described (Pawlikowski et al., 2013), sequences can be obtained from the Gene Expression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. GSE46818, GSE99397).

Induction of melanocytes with conditioned medium
Cell culture supernatant was collected from BRAFV600E, vector or mock transduced melanocytes cultured in parallel at 2 weeks post transduction, and 2 days since the medium was last replaced (n=4 each). Vector transduced cultures were split several times to keep cultures at approximately the same number of cells. Culture supernatant was cleared by centrifugation at 3000 g for 10 minutes, followed by filtration through a 0.22 µm nitrocellulose membrane filter (Elkay), and frozen at -20°C until used in subsequent coculture experiments. For the coculture experiments fresh melanocytes were cultured in parallel in a mixture of half normal culture medium as described above and half conditioned medium from either BRAFV600E, vector or mock transduced melanocytes (n=4 each). Medium was refreshed every 2 days, for a total of 2 weeks, at which time cells were harvested for RNA and RT-PCR analysis.

Genetically Modified Mouse strains
Animals were kept in conventional animal facilities and monitored frequently. All experiments were carried out in compliance with UK Home Office guidelines at the Beatson Institute for Cancer Research mouse facility (Home Office PCD 60/2607) under project license 60/4079. Mice were genotyped by PCR analysis. Mice carrying a tyrosinase promoter driven NrasQ61K gene (Tyr-NrasQ61K) have been described (Ackermann et al., 2005). Mice conditionally expressing the mutant BrafV600E gene under control of tyrosinase driven CRE-ER (Delmas et al., 2003) (Tyr-CRE-ER : LSL-BrafV600E) have also been described (Dhomen et al., 2009). Albino mice carrying the Tyr-NrasQ61K allele were generated by crossbreeding with the albino FVB/NJ (Taketo et al., 1991) strain. Control wild type mice were littermate albino mice lacking the Tyr-NRasQ61K transgene.

Reverse transcription-quantitative polymerase chain reaction analysis
Total RNA was isolated using the RNeasy mini spin kit with DNAse treatment (QIAGEN). Total cDNA was generated using SuperScript III (Invitrogen) from 0.1 -5 µg RNA using random hexamers (Invitrogen), according to manufacturer instructions. Real-time qPCR was performed on 1/50 th of the cDNA, using the primers and FAM-labeled probes (IDT technologies) described in Supplemental Table S6. Reactions were performed on the Chromo4 PCR machine (Biorad), using platinum Taq, and dNTPs from Invitrogen. Transcript levels were quantified using standard curves of known quantities of plasmid DNA and normalized against the geometric mean of GAPDH and ß-actin (ACTB) gene transcripts.

Immunofluorescence microscopy
Cells were plated on glass coverslips and cultured at least 24 hours prior to fixation with 4% neutral buffered formaldehyde for 15 minutes at room temperature. Samples were washed 3 times with PBS, permeabilized with 0.1% triton in PBS for 5 minutes, followed by a further 3 washes with PBS. Cells were blocked for 30 minutes at ambient temperature with 4% bovine serum albumin and 0.02% sodium azide in PBS (blocking solution), followed by 1 to 24 hour labeling with anti HLA-DR (L243; Abcam) at 1 µg/ml in blocking solution at 4 °C. After 3 washes with PBS, cells were labeled with appropriate Alexa568 conjugated secondary antibodies (Invitrogen) in blocking solution for 1 hour at ambient temperature. Finally samples were washed 4 times with PBS and mounted in prolong gold with DAPI (Invitrogen). Images were acquired on the Nikon eclipse 80i fluorescent microscope, and the Olympus Fluoview 1000 IX81 confocal microscope.

Cytokine array
Culture supernatant from BRAFV600E and vector control transduced melanocytes was collected 2 weeks post transduction, and 2 days since the medium was last replaced. Culture supernatant was cleared by centrifugation at 3000 g for 10 minutes, followed by filtration through a 0.22 µm nitrocellulose membrane filter (Elkay). Cytokine quantities were determined using the human cytokine array G series 3 (Raybio), according to manufacturer recommendations. Images were acquired on the Scanarray Express (Perkin Elmer).

ELISA
The human IL-1ß ESILA Kit from Thermo Scientific (EH2IL1ß) was used to measure the levels of IL1ß in the culture supernatants of Melanocytes transduced with BRAF600E or control vector following manufacturer instructions.
SA ß-gal and EdU assay SA ß-gal staining was performed as previously described (Pawlikowski et al., 2013). Staining for EdU incorporation was performed using the Click-iT EdU Alexa Fluor 594 Imaging kit (Invitrogen) according to manufacturer instructions, after a 72 hour pulse with EdU. Note that for all EdU experiments, a somewhat long 72 hour EdU pulse was used to truly be able to show a lack of proliferation in senescent cells, and to allow the generally slow proliferating primary cells (in the absence of BRAFV600E) used in this study to reach a significant percentage of EdU positive cells.

FACS
Fluorescence activated cell sorter (FACS) analysis was performed on a FACSCalibur system (Becton Dickinson), using standard methods. Where stated cells were stained with 5 µM CSFE or 1 µg/ml PI, or labeled with mouse anti CD3 conjugated to allophycocyanin (Biolegend).

Mixed Leukocyte Reaction
White blood cells were isolated according to standard protocols from excess human donor buffy coats using ficoll density gradient centrifugation, and labeled for 5 minutes with 5µM CSFE at room temperature. Unincorporated CSFE was removed by three washes with PBS. WBCs were plated at a density of 1x10 5 cells per well in 96-wells conical wells (not-cell culture treated). Melanocytes previously transduced with BRAFV600E or control vector were added to 5x10 5 cells per well. The co-cultures were maintained for 6 days in RPMI+10% FBS, medium was refreshed daily. FACS analysis was performed to assay WBC activity.

Expression correlation network of TCGA skin cutaneous melanoma data
To generate the expression correlation network (ECN) from the TCGA skin cutaneous melanoma RNA-seq V2 data (Fig 3A), firstly a matrix of expression values by gene (n = 20,531) and patient (n = 375) was generated. Next for each pairwise combination of genes, the Pearson Correlation Coefficient (PCC) of expression values across all patients was calculated. Next, to reduce the number of potentially meaningless connections two filtering steps were applied: Firstly, correlations between two genes below 0.6 were filtered out. Secondly, for each gene, correlations equal to or above 0.6 were ranked (highest first).
Correlations that were not ranked amongst the top ten correlations of both genes were removed. Finally to generate the network, the resulting data was input into an equally weighted Fruchterman-Reingold force-directed algorithm, using a k value of 0.015 and 1000 iterations. Genes were set as nodes and correlations between two genes as edges.

RNA-seq of human Melanoma and Melanocyte cell lines
Paired-end reads were aligned to the human genome (hg19) using a splicing-aware aligner (tophat2) (Kim et al., 2013). Reference splice junctions were provided by a reference transcriptome (Ensembl build 73), and novel splicing junctions determined by detecting reads that spanned exons that were not in the reference annotation. Bigwig files were generated from aligned reads using library size normalization, and uploaded to the UCSC genome browser (Kent et al., 2002).

RNA-seq of mouse WT and NrasQ61K lymph nodes
Inguinal lymph nodes were prepared from 350 day old Tyr-NrasQ61K mice and WT littermates. Total RNA was isolated using the RNeasy mini spin kit with DNAse treatment (QIAGEN). And prepared for RNAseq according to manufacturer instructions (Illumina) and as previously described (Pawlikowski et al., 2013). Paired-end reads were aligned to the mouse genome (mm10) using a splicing-aware aligner (tophat2) (Kim et al., 2013). RNAseq quality and control metrics have been listed in Table  S5. Reference splice junctions were provided by a reference transcriptome (Ensembl build 74), and novel splicing junctions determined by detecting reads that spanned exons that were not in the reference annotation. Aligned reads were processed to assemble transcript isoforms, and abundance was estimated using the maximum likelihood estimate function (cuffdiff) from which differential expression and splicing is derived . Genes of significantly changing expression were defined as FDR corrected p-value <0.05.

RNA-seq Heatmaps
For each gene of biotype coding and status known in the reference transcriptome (Ensembl build 74) the FPKM value was calculated based on aligned reads, using Cufflinks . Z-Scores were generated from FPKMs. Clustering was performed using the R library hclust2 and the Pearson method.

Gene Ontology analysis using David
Genes were uploaded to David (http://david.abcc.ncifcrf.gov), and a functional analysis performed using a background of Ensembl build 74 genes and the molecular functions GO terms.

Kaplan Meier Curve of 10 year skin cutaneous melanoma patient survival
The TCGA skin cutaneous melanoma normalized RNA-seq and patient clinical data was downloaded from the TCGA website (http://cancergenome.nih.gov/). Patients were filtered to include only those with both RNA-seq and clinical data, and a patient follow up date greater than 0. Patients were then grouped into quartiles by normalised CIITA, HLA-DRA and HLA-DRB1 expression level. Kaplan Meier Curves were plotted for the upper and lower quartiles of expression, using the R-library: survival (version 2.38-3). P-values were calculated using the function coxph.