Activation of NF-κB and p300/CBP potentiates cancer chemoimmunotherapy through induction of MHC-I antigen presentation

Significance T cells recognize their targets via their T-cell receptors (TCRs), which in the case of CD8+ T cells bind to MHC-I:antigen complexes on the surface of target cells. Many cancer cells evade immune recognition and killing by down-regulating MHC-I AgPPM. Here, we show how the histone acetyl transferases p300/CBP together with NF-κB epigenetically regulate expression of MHC-I molecules, immunoproteasome subunits, and peptide transporter to enable proper MHC-I antigen presentation. Notably, this pathway is frequently disrupted in human cancers. We now show that certain chemotherapeutics can augment MHC-I antigen presentation via NF-κB and p300/CBP activation, thereby enhancing cancer cell recognition and killing by effector CD8+ CTLs.

For in-vitro experiments, adherent cells were detached using accutase or non-enzymatic cell dissociation solution (CellStripper Dissociation Reagent), and stained as described above. 4′,6-diamidino-2phenylindole (DAPI) or Live/Dead fixable dye (FVD-eFluor780, eBioscience) were used to calculate and exclude death cells. Gating strategies of different stainings are provided in Figure S1B and Figure S6J-L.

Cell culture experiments
Mammalian cells (see Key Resource Table) were all grown in a humidified incubator with 5% CO2 at 37°C. Myc-CaP, HEK293T and MIA PaCa-2 cells were grown in Dulbecco's Modified Eagle Media (DMEM, Gibco) supplemented with 10% FBS, 2 mM L-glutamine, and 100 U/mL penicillin and streptomycin. TRAMP-C2 cells were grown in DMEM supplemented with 5% FBS, 0.4 mM L-glutamine, 100 U/mL penicillin and streptomycin (Gibco), 5% NuSerum IV (Corning), 0.005 mg/mL bovine insulin (Sigma-Aldrich), and 10 nM dehydroisoandrosterone (Acros Organics). MC38, PC3, and all YUMM cells (4) were grown in DMEM/F12 (Gibco) supplemented with 10% FBS, 2 mM L-glutamine, and 100 U/mL penicillin and streptomycin. B16 and WM793 cells were grown in RPMI 1640 (Gibco) supplemented with 5-10% FBS, 0.4 mM L-glutamine, and 100 U/mL penicillin and streptomycin. Human cell lines were cultured as indicated by the provider (Key Resource Table). Inducible ovalbumin (Ova)-P2A-red fluorescence protein (RFP)-expressing TRAMP-C2 cell lines were developed as previously described (3). Ovalbumin processing results in presentation of SIINFEKL peptide on MHCI (H-2Kb), or variants SIIGFEKL (G4) and EIINFEKL (E1). The SIINFEKL peptide can be recognized by OT-I CD8 + T cells that express a high affinity TCR specific for ovalbumin. OT-I cells were isolated from OT-I mice (5) with CD8 selection beads and cocultured with TRAMP-C2 cells as previously described (3). Cells were regularly tested using a mycoplasma PCR kit (EZ-PCR Mycoplasma Test Kit, Biological Industries) and visualized for their characterization. Cells were counted by Trypan Blue (Gibco) exclusion and seeded into 12 or 24 well plates (Flow Cytometry and RNA processing) or into 6 well and 10 cm culture plates for Immunoblot analysis. Cells were treated in fresh media with agents as indicated for 3-48 h after seeding and collected from plates using Acutase (STEMCELL Technologies) for flow cytometry analysis, TRIzol Reagent (Ambion) for RNA processing, or Immunoblot (IB) Lysis Buffer for IB analysis at indicated time points.

qRT-PCR analysis
Total RNA was extracted using an AllPrep DNA/RNA mini Kit (Qiagen) or TRIzol Reagent (Ambion). RNA was reverse transcribed using a Superscript VILO cDNA synthesis kit (Invitrogen) or RTScript cDNA Synthesis Kit (Empirical Bioscience). Quantitative real-time PCR (qRT-PCR) was performed using Sso Advanced Universal SYBR Green Supermix (Biorad) on Biorad CFX96 machine. The relative expression levels of target genes were measured and normalized against levels of housekeeping genes GAPDH or ACTB. Fold-difference (as relative mRNA expression) was calculated by the comparative Ct method (2 (Ct(housekeeping gene − gene of interest)) ). Primer sequences were obtained from the following PrimerBank site (https://pga.mgh.harvard.edu/primerbank/) or designed using the following site (https://www.ncbi.nlm.nih.gov/tools/primer-blast/index.cgi) and are listed in Supplementary Table 2.

HAT Activity Assay
HAT activity assays were performed with EpiQuik HAT Activity/Inhibition Assay Kit (Epigentek, Farmingdale, NY, USA). Solutions were prepared according to manufacturer's protocol.

Immunoproteasome activity assay
Immunoproteasome subunit LMP7 activity assays were analyzed directly on cell lysates from Myc-CaP cells using LMP7 specific fluorogenic peptide substrate Ac-ANW-AMC (Boston Biochem) (6). Cells were treated with or without indicated concentration of oxaliplatin (2 µM), cisplatin (2 µM), or carboplatin (4 µM) alone or in combination with IFNg (2 ng/mL) for 48 h. One set of cells treated with IFNg alone or in combination with each of the platinoids was treated with 200 nM ONX-0914 (Selleckchem) (7) for 2 h prior to lysis to determine background non-specific AMC fluorescence. Cells were lysed directly in proteasome assay buffer (50 mM Tris-HCl, pH 7.5, 40 mM KCl, 5 mM MgCl2, 1 mM ATP, 0.1% NP-40, phosphatase inhibitors, 1 mM dithiothreitol). 100 µM substrate peptide Ac-ANW-AMC was added to the lysate and reaction was carried out for 30 min-1 h at 30°C in an orbital shaker. Fluorogenic AMC release into solution was assessed by a Tecan M200 plate reader (excitation: 380 nm, emission: 440-460 nm) which provides a direct measurement of LMP7 activity in the cell lysates. The measured LMP7 activity was normalized against total protein concentration of cell lysates and represented as fold LMP7 activity to control untreated Myc-CaP cell lysates. Data are represented as dot scatter plots using Graphpad Prism.

Chromatin Immunoprecipitation Assays
Cells were crosslinked for 10 min with 1% formaldehyde. The reaction was stopped by 5 min incubation with 0.125 M Glycine. Cells were washed, harvested with PBS supplemented with protease and phosphatase inhibitors, and cytoplasmic membranes lysed with lysis buffer (5 mM PIPES, 85 mM KCl, 0.5% NP40). After centrifugation, nuclei were lysed for 10 min on ice with sonication buffer (1% SDS, 10 mM EDTA, 50 mM TRIS pH 8 supplemented with protease inhibitors) and sonicated to obtain chromatin fragments of about 400-600 nucleotides. The lysates were precleared for 1 h at 4°C with 30 µL of protein A agarose (Upstate), spun at 5000 rpm for 1 min, and then the supernatants were collected and stored at 10% of input. Chromatin diluted with 9 volumes of dilution buffer (0.01% SDS, 1.2 mM EDTA, 16.7 mM Tris HCl pH 8, 1.1% Triton X-100, 167 mM NaCl, protease inhibitors) was incubated overnight with 20 µL of Protein A Dynabeads (Invitrogen) coated with the antibodies as indicated. The day after, the immunocomplexes were washed five times with Buffer A (0.1% SDS, 2 mM EDTA, 20 mM Tris-HCl pH 8, 1% Triton X-100, 150 mM NaCl), four times with Buffer B (0.1% SDS, 2 mM EDTA, 20 mM Tris-HCl pH 8, 1% Triton X-100, 500 mM NaCl), and once with Buffer T.E. (10 mM Tris-HCl pH 8, 1 mM EDTA). After the final wash, the immunocomplexes were eluted twice with 250 µL elution buffer (1% SDS, 100 mM NaHCO3) for 15 min in rotation at RT and, upon addition of 200 mM NaCl, the crosslinking reversed with an overnight incubation at 65°C. After de-crosslinking, the samples were digested with proteinase K (Thermo Fisher Scientific) and RNAse A (Thermo Fisher Scientific) for 2 h at 42°C, and the DNA was purified and precipitated. Eluted DNA was analyzed by real time PCR as previously described (8,9) using the indicated primers in Key Resource Table. CRISPR-Cas9 cloning, packaging, transfection Stable and transient CRISPR/Cas9 plasmids were used for in vitro and in vivo experiments, respectively. For stable CRISPR/Cas9 plasmid production, 2 µg of the expression lentiCRISPR v2 (Addgene) (10) was digested for 1 h at 37°C with 1 µL BsmBI (NEB) and 5 µL NEBuffer 3.1 (NEB) into a total volume of 60uL and was gel purified (Qiagen). For gRNA insertion, a pair of 25 nt oligos containing the appropriate overhangs were annealed using 1 µL of each primer at a 100 µM stock concentration, 0.5 µL T4 DNA Ligase (NEB), and 1 µL T4 Ligase Buffer (NEB) into a total ligation volume of 10 µL. Primers were annealed using the following parameters: 37°C for 30 min then 95°C for 5 min and ramped down to 25°C at 5°C/min. The oligos were ligated into the vector for 10 min at room temperature by mixing 1 µL of the digested plasmid, 1 µL annealed oligos, 1 µL T4 Buffer (NEB), and 0.5 µL T4 DNA Ligase (NEB) into a total ligation volume of 10 µL. Plasmids were transformed into 25 µL Max Efficiency Stbl2 (ThermoFisher Scientific) bacteria. Resulting colonies were minipreped (Qiagen) and verified by sanger sequencing before transfection into mammalian cells. Transient CRISPR/Cas9 plasmids were constructed exactly as mentioned above using the expression vector pSpCas9(BB)-2A-GFP (PX458) (Addgene) (11) and a BbsI (NEB) restriction enzyme. gRNA sequences were found on Chopchop and DNA oligos were synthetized. The gRNA sequences are shown in Key Resource Table. MAX Efficiency Stabl2 bacteria containing proper CRISPR/Cas9 were expanded in LB broth at 37°C overnight. Plasmids were extracted from bacteria using QIAGEN Plasmid Mini Kit or Maxi Kit. For stable CRISPR/Cas9, lentivirus was produced in HEK293T cells transfected at 80-90% confluency using Lipofectamine 3000 (Invitrogen) as recommended by the manufacturer and psPAX2 (Addgene) and pMD2.G (Addgene) packaging vectors. Medium was changed 6-8 h after transfection and supernatant was collected after 48-72 h. Viral media was passed through a pre-wetted 0.45 µm PVDF filter (Millipore) and mixed with 10 µg ml -1 Polybrene (Sigma Aldrich) before being added to recipient cells. Infected cells were treated with puromycin to generate stable populations. Transfection with transient CRISPR/Cas9 in Myc-CaP and TRAMP-C2 cells was carried out using Lipofectamine 3000 (Invitrogen) as recommended by the manufacturer. Cells were transfected at 70-80% confluency using 12 µg of the plasmid of interest. Medium was changed 36 h after transfection and cells were sorted 48 h after transfection. Cell sorting was performed on a BD FACS Jazz. For each transfected cell line, at least 4 million cells were collected using 0.5 % Trypsin (Gibco), counted by Trypan Blue exclusion, and resuspended in PBS. Cells were sorted into 96 well plates and GFP negative populations were sorted as control lines. Resulting colonies were expanded and verified by immunoblot analysis, flow cytometry, and qRT-PCR.

shRNA cell lines
Bacteria containing desired shRNA Lentivirus-plasmids were obtained from La Jolla Institute for Immunology (LJI)-Functional Genomics lab. Sequences are shown in Key Resource Table (https://rnaireagent.liai.org/scrm_mm_trc/). Bacteria were expanded in LB broth overnight at 37°C and plasmids were extracted using Mini-prep or Maxi-prep (QIAGEN) according to manufacturer's instruction. shRNA plasmids were delivered using lentivirus as described above. After lentiviral infection, cells were selected using puromycin and knockdown was confirmed using qRT-PCR and IB.

Immunostaining and Histology
Tissues were embedded in Tissue-Tek OCT compound (Sakura Finetek, Torrance, CA, USA) and snapfrozen. Tissue sections were fixed in cold acetone/methanol or 3-4% PFA for 3-10 min and washed with PBS. Cultured cells were cultured on coverslips and fixed in 4% paraformaldehyde for 10 min at room temperature. After washing twice in PBS, cells were incubated in PBS containing 10% FBS or 0.2% BSA and 1% donkey serum or goat serum for surface staining or added 0.2% gelatin/0.2% BSA (from cold water fish skin; Sigma-Aldrich) or 0.1% saponin for intracellular staining for 15-30 min to block nonspecific sites of antibody adsorption. Sections were incubated with primary antibodies for 1 h or overnight at RT or 4°C, respectively. After washing, secondary antibodies were added for 1 h at RT. As negative controls, samples were incubated with isotype-matched control antibodies or secondary antibodies only. After staining with DAPI, sections were covered with Vectashield Mounting Medium (Vector Laboratories, Burlingame, CA, USA). Confocal images were captured in multitracking mode on a SP5 confocal microscope (Leica) with 40 x or 63 × Plan Apochromat 1.3 NA objective. Paraffin-embedded specimens from a total of 118 Prostate cancer patients were integrated into a tissue microarray system (TMA) constructed at the Clinical Institute of Pathology at the Medical University of Vienna (MUV). All of the human prostate specimens used for TMA construction were approved by the MUV Research Ethics Committee (1753/2014), as previously described (1). Human liver biopsies were obtained from the Biobank of the Medical University of Graz. Biopsies were registered in the biobank and kept anonymous. The research project was authorized by the ethical committees of the Medical University of Graz (ref. no. 1. 0 24/11/2008). The study protocol was in accordance with the ethical guidelines of the Helsinki Declaration. All human samples were de-identified prior to use in our study. Paraffin-embedded tissue sections were subjected to de-paraffinization and rehydration, and then were immersed in a pre-heated antigen retrieval water bath with a pH 6.1 citrate buffer or Dako Target Retrieval Solution for 20 min at 95-96°C. Sections were then incubated with antibodies against as indicated, and as previously described (1, 3). All staining was done according to manufacturer's protocols (ImmPRESS, Vector Laboratories). DAB (Vector Laboratories, SK-4100) and ImmPACT Vector Red (Vector Laboratories, SK-5105) were used for detection. Nuclei were lightly counterstained with a freshly made haematoxylin solution then further washed in water. Sections were examined using an Axioplan 200 microscope with AxioVision Release 4.5 software (Zeiss, Jena, Germany) or TCS SPE Leica confocal microscope (Leica, Germany). Sections were imaged under a Hamamatsu 2.0-HT Digital slide scanner (Hamamatsu Photonics, EU, Japan, and USA). For measurement, ImageJ 1.49v was used. Serial cut slides were used to show adjacent tissues. Prostate cancer cells were shown by PSA staining and their HLA abundance was measured according to the PSA positive location. Prostate cancer cells were specifically selected using ImageJ, and mean optical density was obtained and statistically compared among groups. For liver cancer slides, cancer cells were the most abundant cell type and could be identified by their histopathological features.

Transmission Electron Microscopy (EM) and determining Mitochondria/Cytoplasm Density
Myc-CaP cells were plated on poly-l-lysine treated MatTek dishes and treated with Oxali (2 µM) for 24 or 48 h. After treatment, cells were fixed with 2% glutaraldehyde (18426, Ted Pella Inc.) in 0.1 M sodium cacodylate buffer, pH 7.4 (18851, Ted Pella Inc.) containing 2 mM CaCl2 for 5 minutes at 37 o C and then incubated on ice for 1 h. Cells were washed with 0.1 M sodium cacodylate and posted fixed with 2% osmium tetroxide (19150, Electron Microscopy Sciences) in 1.5% potassium ferrocyanide, 1 mM CaCl2 and 0.1 M sodium cacodylate buffer, pH 7.4 for 30 minutes on ice and washed with double distilled water (ddH2O) at RT. Cells were then treated with 0.5% thiocarbohydrazide solution using a 0.22 µm Millex 33 mm PES sterile filter (SLGSR33RS, Sigma-Aldrich) and incubated for 10 minutes at RT. Cells were then washed with ddH2O, treated with 2% osmium tetroxide in ddH2O for 30 minutes and rinsed with ddH2O. The plate was treated with 1% aqueous uranyl acetate (22400, Electron Microscopy Sciences) and incubated at 4 o C overnight. Cells were then washed with ddH2O and treated with en-bloc Walton's lead aspartate staining for 5 minutes at 60 o C. Afterwards, cells were washed with ddH2O at RT followed by an ice cold graded dehydration ethanol series of 20%, 50%, 70%, 90%, 100% (anhydrous) and then washed with 100% (anhydrous) at RT. Cells were infiltrated with one part Durcupan ACM epoxy resin (44610, Sigma-Aldrich) to one part anhydrous ethanol for 30 minutes then with 100% Durcupan resin, a final change of Durcupan resin and immediately placed in a vacuum oven at 60 o C for 48 h. Cells were identified, cut out by jewel saw and mounted on dummy blocks with Krazy glue. Coverslips were removed and 70-80 nm specimen sections were created with a Leica Ultracut UCT ultramicrotome and Diatome Ultra 45 o 4 mm wet diamond knife.
Sections were picked up with 50 mesh gilder copper grids (G50, Ted Pella, Inc). The sections were imaged by FEI Spirit transmission electron microscope at 80kV. Images and montages were collected by a Tietz TemCam F-224 2k by 2k CCD camera and by Serial EM software version 3.1.1a. Mitochondrial volume to cytoplasm volume ratio was collected at a resolution of 11.5 nm per pixel. Mitochondria and cytoplasm counting was done using Adobe Photoshop CS5 Extended version 12.0X64. Image resolution for setting the grid spacing was 160. Images were then reset to 2048 by 2048 pixels. Counting was done using the number of grid cross-hairs that fell on mitochondria and on the cytoplasm. Nuclei were not counted. Mitochondria/cytoplasm density was calculated by dividing mitochondria counts by cytoplasm counts X100. Mean and standard deviation were used for t-test analysis.

cytosolic mitochondria DNA (mtDNA) quantification
Cells were treated with Oxali (2 µM) for 48 h. Total DNA was isolated using Allprep DNA/RNA Mini Kit (QIAGEN) according to manufacturer's protocol. mtDNA was measured by qRT-PCR using primers specific for the mitochondrial D-loop region or a specific region of mtDNA that is not inserted into nuclear DNA (non-NUMT). Tert and b2m were used for normalization. To quantify the cytosolic released mtDNA, cytosolic fractions were depleted from mitochondria by Mitochondria Isolation Kit for Cultured Cells (Thermo Fisher Scientific) following manufacturer's protocol as described previously (12). The quality of the cytosolic fractions were confirmed by IB analysis using VDAC antibodies ( Figure S4L). The amount of mtDNA in cytosolic fraction were measured using qRT-PCR, as described above.

Organoid cultures from Patient-derived xenograft (PDX) model of bone metastatic PCa
This study was carried out in strict accordance with the recommendations in the Guide for the University of California San Diego (UCSD) Institutional Review Board (IRB). Approval was received from the UCSD institutional review board (IRB) to collect surgical specimens from a patient for research purposes. A surgical PCa bone metastasis specimen was harvested from a patient who had progressed to castrate resistant bone metastatic prostate cancer and labelled as Prostate Cancer San Diego 1 (PCSD1) for the purpose of de-identification of patient data, as described previously (13). PCSD1 cells were injected into the femur endosteal space of Rag2 -/gc -/male mice to establish a patient-derived xenograft (PDX) model representing a preclinical model of bone metastatic PCa as previously shown (13). PCSD1 cells were maintained as intra-femoral tumors in male Rag2 -/gc -/mice prior to establishing 3D organoid cultures. All experiments involving xenograft models were conducted under an IACUC approved protocol at the University of California, San Diego. Tumors were processed according to previously established methods (13) with the additional step of immuno-depletion of mouse cells (Miltenyi Biotec) to enrich for human cells. PCSD1 3D organoid cultures were established as previously described (14)

Single cell RNA-seq (scRNA-seq) processing and analysis
Tumors from the same treatment groups were pooled (n = 7-8 / group). Tumor single cell suspensions were prepared as decribed in the flow cytometry section. Isolated cells were stained with labeled antibodies for CD45-PE, CD4-FITC, CD8a-PE-Cy7 and CD3-APC in cell staining buffer (Biolegend). Dead cells were excluded with Live/Dead fixable dye (FVD-eFluor780, eBioscience). CD45 + CD3 + CD8 + T cells were sorted in PBS with 0.5% BSA and 2 mM EDTA. Droplet-based 3' end massively parallel single-cell RNA sequencing (scRNAseq) was performed by encapsulating sorted live CD8 + CD3 + tumor infiltrating T cells into droplets and libraries prepared using Chromium Single Cell 30 Reagent Kits V3 according to manufacturer's protocol (10x Genomics). Raw sequence data demultiplexing, barcode processing, alignment and filtering were performed using the Cell Ranger Single-Cell Software Suite (v3.1.0). Subsequent filtering and downstream analyses were performed using Seurat (15). Genes expressed in less than 3 cells and cells that express less than 300 genes were excluded from further analyses. Additional filtering of cells was determined based on the overall distribution of mitochondrial gene expression (< 10%) to eliminate dying cells, respectively. Assessment and removal of multiplets was performed using Scrublet (16). Normalization and variance stabilization of remaining data was applied using regularized negative binomial regression (sctransform (17)), including regression of mitochondrial gene expression prior to principle component analysis (PCA). Optimal dimensionality of the dataset was decided after examination of the JackStraw procedure and Elbow plot. The FindNeighbors function was utilized that implements a graph based nearest neighbor clustering approach, and the FindClusters function was used to identify final cell clusters. UMAP was applied for nonlinear dimensional reduction to obtain a low dimensional representation of cellular states. Differential expression between clusters or between samples was determined using the MAST method via the FindMarkers function, using a minimum expression proportion of 25% and a minimum log fold change of 0.25. Unbiased cell type annotation was performed using SingleR (18). Briefly, this framework allows for the annotation of scRNA-seq data to reference transcriptome data sets (ImmGen) of known origin to infer the cellular state of each input cell. Cell annotations in combination with marker gene expression were used to eliminate non-immune and myeloid lineage cells so that subsequent analyses included only T-cell populations. Normalized expression was used for subsequent analyses including heatmap visualization and pathway analysis. Functional pathway enrichment analysis was performed on cluster specific marker gene expression (enrichR) or at the individual cell level (AUCell), and GSEA (clusterProfiler) was used to interrogate pathway enrichment based on differential expression analyses. Query pathways included all hallmark, canonical and gene ontology (GO: Biological Processes) pathways available through MSigDB.

RNA-seq processing and analysis
For in vitro experiments, RNA-seq data was processed and analyzed as described below. Quality control (QC) of sequencing reads was performed using FastQC (v0.11.9) (Andrews et al., 2010). Sequencing reads were aligned to the mouse genome build GRCm38 using annotations from GENCODE (vm12) with the splice-aware RNA-seq aligner STAR (v2.7.3a) (20,21). Following alignment, the raw counts relative to genes were generated by featureCounts (v2.0.0). The RSeQC (v3.0.1) was used to evaluate the quality of alignments and complete QC report compiled with MutliQC (v1.8). Raw feature counts were normalized and differential expression analysis performed using DESeq2 (22). Significance of differential expression was defined by using an adjusted p-value cut-off of 0.05 after multiple testing correction. Gene set enrichment analysis was performed from rank ordered differential expression using the clusterProfiler (23) package in R. Within sample pathway activity was determined by gene set variation analysis (GSVA) (24). Gene sets queried for functional enrichments included the Hallmark, Canonical pathways, and GO Biological Processes Ontology collections available through the Molecular Signatures Database (MSigDB) (25). Transcriptional regulator analysis on the top 500 determined differentially expressed genes was performed using LISA (26).

ATAC-seq processing and analysis
Permeabilized nuclei were obtained by resuspending cells in 250 µL Nuclear Permeabilization Buffer [0.2% IGEPAL-CA630 (Sigma-Aldrich), 1 mM DTT (Sigma-Aldrich), Protease inhibitor (Roche), 5% BSA (Sigma-Aldrich) in PBS (Thermo Fisher Scientific)], and incubating for 10 min on a rotator at 4°C. Nuclei were then pelleted by centrifugation for 5 min at 500 xg at 4°C. The pellet was resuspended in 25 µL ice-cold Tagmentation Buffer [33 mM Tris-acetate (pH = 7.8) (Thermo Fisher Scientific), 66 mM K-acetate (Sigma-Aldrich), 11 mM Mg-acetate (Sigma-Aldrich), 16 % DMF (EMD Millipore) in Molecular biology water (Corning)]. An aliquot was then taken and counted by hemocytometer to determine nuclei concentration. Approximately 50,000 nuclei were resuspended in 20 µL ice-cold Tagmentation Buffer and incubated with 1 µL Tagmentation enzyme (Illumina) at 37°C for 30 min with shaking 500 rpm. The tagmentated DNA was purified using MinElute PCR purification kit (Qiagen). The libraries were amplified using NEBNext High-Fidelity 2X PCR Master Mix (New England Biolabs, NEB) with primer extension at 72°C for 5 min, denaturation at 98°C for 30s, followed by 8 cycles of denaturation at 98°C for 10 s, annealing at 63°C for 30 s and extension at 72°C for 60 s. Amplified libraries were then purified using MinElute PCR purification kit (Qiagen), and two size selection steps were performed using SPRIselect bead (Beckman Coulter) at 0.55X and 1.5X bead-to-sample volume rations, respectively. Each library was then sequenced on an Illumina NextSeq500 or HiSeq4000 to a depth of >= 25 million usable reads pairs (i.e. after mapping, filtering, and elimination of PCR duplicates as described in the data processing section). We processed the raw fastq files with the ENCODE ATAC-seq pipeline (https://github.com/ENCODE-DCC/atac-seq-pipeline). In detail, we first auto-detected then applied cutadapt "-m 5 -e 0.10" to trim the adapter sequences from the raw fastq files. Then we aligned the trim sequences to mouse reference genome mm10 using bowtie2 (27) with parameter "-X2000 --mm -k 5". Next, the improper mapped, poorly mapped and unmated reads were filtered from the resultant raw bam files using samtools (27) view with parameter "-F 1804 -q30". Then, duplicates were marked by Picard (27) "MarkDuplicate" and removed by samtools "-F 1804". In the end, the final bam files were acquired after removing mitochondrial reads, sorted, and indexed with samtools. For each condition, we pooled the two replicates then used MACS (v2.1.0) (28) with "--shift -75 --extsize 150 -p 0.01 --nomodel --keep-dup all" to call the accessible regions with enriched ATAC-seq signals. Peaks were further filtered against the ENCODE blacklist regions (29). To find differentially accessible regions (DARs), we first merged the called regions from all conditions, then used htseq-count (30) to generate the read count in each region for each condition and each replicate. The count table was subsequently processed with DESeq2 (22) to call the DARs. DARs were called using an absolute log2 fold change cut-off of 1 and a false discovery rate (FDR) cut-off of 10 -5 . We specifically called the DARs: 1) between IFNg-1 ng/ml vs Control; 2) between Oxaliplatin-2 µM vs Control; 3) between Cisplatin-2 µM vs Control; and 4) between Combo of IFNg-1 ng/ml + Oxaliplatin-2 µM vs Control. Additionally, we tried to identify the 'additive' effect of combination treatment by using a reduced cut-off (log2 fold change of 0.585 and FDR of 0.05) to call DARs between Combo and Oxaliplatin only and between Combo and IFNg only. These DARs were then intersected with those identified between Combo vs Ctrl. In another words, the 'additive' effect only occurs when the combination caused significant chromatin accessibility changes over control or either individual treatments alone. In order to find potential transcription factor binding events within DARs, we utilized GIGGLE (31) to query the complete mouse transcription factor ChIP-seq dataset collection (6,751 datasets across 570 transcription factors) in Cistrome DB (32). For each ChIP-seq dataset in Cistrome DB, we asked how many DARs overlap with the top 1,000 most significant peaks of that dataset, then calculate a 'Giggle score' equal to the product of the -log10 p-value and log2 odds-ratio from a Fisher's Exact two-tailed test. We queried the DARs of IFNg vs Ctrl, Oxaliplatin vs Ctrl, Cisplatin vs Ctrl, Combo vs Ctrl, and additive DARs separately.

EP300 and CBP LOF analysis
TCGA MAF and CNV files for PRAD, LIHC, LUSC, and SKCM were downloaded directly from TCGA Genomic Data Commons (GDC) Data Portal in July, 2019 and May, 2020, respectively. Patients were labelled as LOF if they fit one or both criteria: (1) they had at least one CNV loss event, and/or (2) they had a LOF mutation in either CREBBP (CBP) or EP300. We defined LOF mutations as nonsense, frameshift, inframe insertion, inframe deletion, splice site/ splice region mutations and missense mutations not associated with increased EP300/CBP expression. Specifically, the mean of z-scored CREBBP(CBP) and EP300 expression was calculated for each patient and the distributions were compared between patients with and without any p300/CBP mutations. Missense mutations that resulted in a mean expression level greater than the 75 th percentile of non-mutated patients were considered GOF and excluded from the LOF analysis. In the end, the numbers of LOF patients in PRAD (n = 497), LIHC (n = 369), LUSC (n = 486), and SKCM (n = 448) were 79, 170, 280, and 120, respectively.

Human TCGA expression and survival analysis
To compare tumor and normal gene expression in liver and prostate cancer, raw human RNAseq fastq files were downloaded from The Cancer Genome Atlas (TCGA) for samples with clinical and molecular annotations (33). In total, TCGA data from 369 liver tumors (LIHC) and 50 tissue-matched normal samples, 497 prostate tumors (PRAD) and 52 tissue-matched normal samples, 486 lung tumors (LUSC), and 448 skin tumors (SKCM) were downloaded. Transcript levels were quantified by Sailfish v.7.4 in Transcripts per Million (TPM), and we then constructed a matrix describing gene expression (TPM) for each tumor type (LIHC and PRAD) with patient as the columns, genes as the rows. Expression matrices were subsequently log2 transformed after adding a pseudocount of 1. The log2 transformed expression values were then zscore transformed for the two tumor types separately. The associated heatmap for selected genes was generated with Python package seaborn heatmap ( Figure S3B) and clustermap in Python ( Figure S3E). Pearson correlation coefficients and corresponding p-values of gene expressions were calculated and plotted using Python package seaborn jointplot in Python ( Figure S3A). Pairwise Pearson correlation coefficients were calculated using Python package pandas corr function ( Figure S3A, S3G). The latest clinical annotations from all TCGA patients were downloaded (34). Overall survival (OS) was used to plot the Kaplan-Meier survival curves of TCGA LIHC patients with different CNV status ( Figure 4N) with Python package lifelines. Significance was assessed using the log-rank test from the same Python package.

Neoantigen analysis
HLA genotyping and mutation calling was performed for HLA-A, HLA-B, and HLA-C genes, which encode the human MHC-I complex. TCGA samples available for LIHC, PRAD, LUSC, and SKCM were typed with Broad Institute's Polysolver (35). Missense and small insertions and deletions (indel) mutations were taken from the MAF files described above. We used the netMHCpan4.0 tool (36,37) to obtain mutation affinity scores for all TCGA patients' HLA alleles. To determine whether a mutation would be effectively bound as a neoantigen to the MHC-I complex, we calculated Patient Harmonic-mean Best Rank (PHBR) scores (38). For binarizing affinity or PHBR scores, we used score cutoffs of <= 2 and <= 0.5 for binding and strong binding MHC-I neoantigens, respectively (36,38). Mutations with scores > 2 we considered non-binding (36,38). Expression of neoantigens were estimated using the transcriptomic reads of patient-specific TCGA mRNAseq bam files using bam-readcounts (39). Fraction of neoantigens was calculated as (number of neoantigen)/(total number of mutations). P values were calculated using the Wilcoxon ranksum test to determine the significance for each neoantigen attribute (number of neoantigens and fraction of neoantigens) between patients with each tumor type (PRAD, LIHC, LUSC, and SKCM) or between LOF and non-LOF ( Figure 4P, Figure S3H, S3I). The Wilcoxon tests implemented in the scipy.stats Python package were used for these analyses.

Mass spectrometry
MHCI peptide profiling was performed for the H-2Db and H-2Kb ligandome of the murine adenocarcinoma cell line MC-38 treated with four different conditions, (1) an untreated control, (2) IFNg 2 ng/mL, (3) Oxali 4 µM, and (4) a combination of IFNg 2 ng/mL and Oxali 4 µM. Briefly, 6.7 × 10 8 cells were lysed, split into two equal groups and MHCI molecules were immunoprecipitated in parallel using two different antibodies: the H-2Kb-specific antibody derived from the Y3 hybridoma, and the H-2Db-specific antibody from the B22.249 clone, as described previously (40); both were crosslinked to Protein A sepharose resin (Repligen) via DMP chemistry (Thermo Fisher Scientific). After overnight immunoprecipitation at 4°C, MHCI peptides were eluted from the antibody-resin with 0.1 M acetic acid/0.1% trifluoroacetic acid and purified via solid phase extraction (Empore C18) before mass spectrometric analysis. Peptides were loaded onto a trapping column, washed and eluted onto a 75 µm analytical column, both were packed with Luna C18 resin (Phenomenex) and separated by reversed-phase chromatography (nanoAcquity UPLC system, Waters, Milford, MA) using a 120 min gradient. The gradient, composed of solvent A (0.1% formic acid in water) and solvent B (0.1% formic acid in acetonitrile), went from 5% to 25% B in 90 min, 25% to 50% B in 20 min and 50% to 90% B in 10 min. The flow rate was 350 nL/min. The eluted peptides were analyzed by datadependent acquisition (DDA) in EThcD mode on a Fusion Lumos mass spectrometer (Thermo Fisher Scientific). Mass spectral data was acquired using methods comprising of a full scan (survey scan) of high mass accuracy in the Orbitrap at 60,000 FWHM resolution followed by MS/MS scans at 15,000 FWHM resolution. The instrument was run with a 3s cycle from MS and MS/MS. Data were processed through the MaxQuant software v1.5.3.17. Andromeda database search results were filtered at the 1% PSM false discovery rate (FDR) and allowing for one unique peptide per protein.
MS analysis on differentially expressed proteins between Oxaliplatin-treated and control Myc-CaP cells was performed using SWATH strategy. Proteins were extracted from control and Oxaliplatin-treated Myc-CaP cells after 24 h using lysis buffer (50 mM Hepes, 6 M urea, 2 M thiourea and 1× protease inhibitor cocktail), and then digested by trypsin (Promega, sequence grade) using FASP (Filter Aided Sample Preparation). The resulting peptides were analysed with an AB Sciex 5600+TripleTOF mass spectrometer (Concord, Ontario, Canada) interfaced to an EkspertTMNanoLC 425 system (Dublin, CA). Data acquisition parameters and SWATH-MS Data Analysis method were described previously [Scientific RepoRts | 7:45913 | DOI: 10.1038/srep45913], except that proteins were identified by searching mass data against UniProt mouse instead of the human database (containing 50190 sequences). Proteins with significant expression level changes after treatment with adjusted p < 0.05 and FC ≥ 1.2 or FC ≤ 1/1.2 in at least two out of three biological replicates were regarded as differentially expressed proteins and subjected to further pathway enrichment analysis.

Quantification and statistical analysis
Data is presented as either mean ± s.e.m. or median of continuous values and was analyzed by two-sided Students' t-test or Mann-Whitney test for comparison of two groups, respectively D'Agostino & Pearson test and/or Shapiro-Wilk test were used to test the normality of sample distribution. One-way ANOVA or Kruskal-Wallis test was used to compare three or more groups data analyses. Bonferroni's or Dunn's multiple comparison test was applied to compare all pairs of groups. Fisher's exact Chi-square t test was used to calculate statistical significance of categorical values between groups. Two-way ANOVA test was used for tumor growth analysis. Two tailed P values of ≤ 0.05 were considered significant. Linear regression was used to determine the correlation between two different variables. Power calculation was used to confirm (p < 0.05 with a 95% probability, two-tailed) the accurate sample size. Experiments were repeated independently at least two or three times with similar results. GraphPad Prism software was used for statistical analyses.

Fig. S1. RNA-seq and ATAC-seq analysis of cancer cells treated with chemotherapeutics. (A) A schema describing the cell culture experiments. (B) Flow cytometric analysis confirming induction of MHC-I expression in viable cells subjected to the indicated treatments. (C) Distance (Pearson correlation) matrix (left)
and PCA analysis (right) determined from bulk RNA-seq data. Duplicates were used for each condition. (D) Volcano plots depicting DEG analyses that compare each treatment to control. The number of significantly (FDR < 0.05) upregulated (red) and downregulated (blue) genes is shown. (E) Distance (Pearson correlation) matrix (left) and PCA analysis (right) determined from bulk ATAC-seq data. Duplicates were used for each condition. (F) Volcano plots depicting DAR analyses that compare each treatment to control. The number of significantly (FDR < 10 -5 ) opened (red) and closed (blue) regions is shown. Two-sided t-test (means ± s.e.m), and Mann-Whitney test (median) were used to determine significance. *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.             (1) control (5% dextrose), (2) Oxali (6 mg/kg; weekly), (3) anti-PD-L1 (10 mg/kg; weekly), and (4) Oxali + anti-PD-L1 (weekly). After 4 cycles, during which tumor size was measured with a caliper, the mice were euthanized and analyzed. Each dot represents a treatment group average ± s.e.m. Two-sided t-test (means ± s.e.m), and two-way ANOVA were used to determine significance. (B) UMAP representation of total T cell populations profiled by scRNA-seq. Eleven distinct clusters were identified. (C) Heatmap representing the 10 most significant marker genes for each cluster. (D) Proportional contributions of each individual cluster to the overall T cell population in each sample is shown (Fig. 7B)  (1) control (n=6), (2) Oxali + anti-PD-L1 (weekly) (n = 5) and analyzed as described in (A). Each dot represents a treatment group, mean ± s.e.m. Two-sided t-test (means ± s.e.m), and two-way ANOVA were used to determine significance. (B) Mice bearing s.c. tumors generated by control and Ifngr2-ablated (clone 1 and 2) Myc-CaP cells were treated as indicated. Treatment cycles and tumor measurements were performed as described in Fig. 7A. Each dot represents a treatment group mean ± s.e.m. Two-sided t-test (means ± s.e.m), and two-way ANOVA were used to determine significance. (C) Mice bearing s.c. Myc-CaP tumors generated by control and p300silenced cells as described in Fig 7A. Total tumor RNA isolated from indicated treatment groups was analyzed by qRT-PCR for expression of the Ifngr2.   CaP tumor cell suspensions were stained for CD45, CD8, CD44, PD-1 and Tim-3 or CD45,  CD8, TNF, IFNg and CD107a. FVD-eF780 was used to exclude dead cells. The gating scheme for tumor cells (A) and splenocytes (B) is shown. Plots depict overlay comparison the percentages of CD8 + cells expressing TNF and IFNg isolated from Oxali + anti-PD-L1-treated and untreated tumor-bearing mice is shown. The analyses of the corresponding groups are presented in Fig.  7M-7N, S16C-16K. (C-K) Single spleen and tumor cell suspensions were analyzed by flow cytometry for the percentage of TI-CD4 + (C) and TI-CD8 + (D), and effector CD8 + T cell subsets in spleen and tumors as indicated (E-K). *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant. Specific n values are shown in (C-K).

Fig. S17.
Oxali-enhanced immune rejection requires NF-kB signaling and summary. (A) Mice bearing s.c. Myc-CaP tumors generated by control and Rela-silenced cells were allocated into 4 treatment groups as indicated. Total tumor RNA was analyzed by qRT-PCR for expression of Ifngr2, Psmb9, and Tap1 mRNAs. Mann-Whitney test (median) were used to determine significance. (B) Scheme of vaccination experiments (left). Two groups of mice were immunized with lysates of Oxali-killed control (MC wt ) and p300-silenced (MC-p300 D ) Myc-CaP cells. After 7 days, the mice were s.c. inoculated with live MC-p300 D Myc-CaP cells (n = 5 per group). Live MC-p300 D Myc-CaP cells were also implanted into non-immunized mice as a control. Tumor Growth curves are shown (right). Two-sided t-test (means ± s.e.m), and two-way ANOVA were used to determine significance. One-way ANOVA analysis and multiple comparison confirmed the results. *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant. Specific n values are shown in (A, B). (C) A schematic summarizing the different signaling responses triggered by low dose chemotherapy and leading to MHC-I AgPPM induction and increased IFNg responsiveness. :   Table S1: Comparison of CNVs occurrence in MHC-I AgPPM genes between EP300/CBP LOF and non-LOF patients in PRAD and LIHC.

SI Appendix Tables
CNV: copy number variation, gain: gain one or more additional copies of a gene, loss: loss one or both copies of a gene, No CNV: the number of copies for the gene does not vary from expected (two copies), LOF: patients with loss of function in CREBBP (CBP) and/or EP300 caused by gene alteration including deleterious single nucleotide variants and/or CNV deletion (Methods).