Distinct modulation of IFNγ-induced transcription by BET bromodomain and catalytic P300/CBP inhibition in breast cancer

Interferon gamma (IFNγ) is a pro-inflammatory cytokine that directly activates the JAK/STAT pathway. However, the temporal dynamics of chromatin remodeling and transcriptional activation initiated by IFNγ have not been systematically profiled in an unbiased manner. Herein, we integrated transcriptomic and epigenomic profiling to characterize the acute epigenetic changes induced by IFNγ stimulation in a murine breast cancer model. We identified de novo activation of cis-regulatory elements bound by Irf1 that were characterized by increased chromatin accessibility, differential usage of pro-inflammatory enhancers, and downstream recruitment of BET proteins and RNA polymerase II. To functionally validate this hierarchical model of IFNγ-driven transcription, we applied selective antagonists of histone acetyltransferases P300/CBP or acetyl-lysine readers of the BET family. This highlighted that histone acetylation is an antecedent event in IFNγ-driven transcription, whereby targeting of P300/CBP acetyltransferase activity but not BET inhibition could curtail the epigenetic remodeling induced by IFNγ through suppression of Irf1 transactivation. These data highlight the ability for epigenetic therapies to reprogram pro-inflammatory gene expression, which may have therapeutic implications for anti-tumor immunity and inflammatory diseases.


Background
Pro-inflammatory cytokines directly activate signaling cascades, however, epigenetic mechanisms are also critical for coordinated inflammatory gene expression. Interferon gamma (IFNγ) is a pleiotropic cytokine that is a key regulator of anti-tumor immunity [1,2]. In the tumor microenvironment (TME), IFNγ induces the expression of genes essential for antigen processing and presentation, thereby promoting immune-surveillance [3]. Gene expression profiling in melanoma patients treated with nivolumab (anti-PD1) alone or combined with Ipilimumab (anti-CTLA4) revealed that an IFNγ-driven gene signature was the most predictive feature for clinical responses to these immunotherapies [4]. In contrast, genetic aberrations that suppress IFNγ signaling are detected in patients relapsing from immune checkpoint blockade demonstrating that tumor-intrinsic IFNγ signaling is a clinically relevant mechanism of immune evasion [5]. This is also supported by genome-scale immune evasion screens that recurrently identify the IFNγ receptors (IFNGR1/IFNGR2) and obligate JAK/STAT members as the most important mediators of sensitivity to

Graphical Abstract
T-cell killing [6,7]. Finally, in vitro models of acquired resistance to T-cell bispecific antibodies and CAR T-cells also converged on IFNγ signaling as a key mediator of immune evasion [8]. Thus, anti-tumor immunity and therapeutic responses to immune oncology agents are critically determined by tumor cell intrinsic IFNγ signaling.
It is now well established that epigenetic therapies may modulate the immunogenicity of tumor cells and inflammatory gene expression in cancer [9]. For example, we, and others, identified that the pre-clinical activity of prototypical BET Bromodomain inhibitor, JQ1, was dependent on an intact host immune system, which was mechanistically linked to suppression of immune checkpoint ligand PD-L1 on tumor cells [10,11]. Suppression of PD-L1 by BET inhibition was evident in the context of 9p24.1-amplified Hodgkin lymphoma, where PD-L1 is constitutively expressed [12], as well as triple-negative breast cancer (TNBC) as a model of IFNγ-dependent expression. While IFNγ-inducible expression of PD-L1 was BET protein-dependent, the wider role of BRD4 or additional epigenetic regulators in regulating cellular responses to IFNγ remains poorly defined.
Herein, unbiased transcriptional and epigenomic profiling was employed to provide mechanistic insight into the hierarchy of events occurring at the chromatin interface following IFNγ stimulation leading to the coordinated transcription of IFNγ target genes. These epigenetic events underpinning IFNγ-induced gene expression were uncoupled by antagonizing distinct nodes using small molecule inhibitors of epigenetic modulators. Taken together, these data illustrate the acute epigenetic remodeling that drives the IFNγ-induced transcriptional response.

IFNγ promotes rapid transcriptional activation in solid tumors
We utilized the AT3 model of TNBC to profile the transcriptional and epigenetic effects of IFNγ. We first performed RNA-sequencing (RNA-seq) in AT3 cells treated acutely with recombinant IFNγ or vehicle control and subsequent differential gene expression analysis (DGEA) revealed that IFNγ stimulation led to a concerted transcriptional response (Fig. 1A) involving predominant up-regulation of numerous genes, including canonical IFNγ target genes Cd274 (PD-L1), Irf1, Tap1, and Socs1 (Fig. 1B). The transcriptional response in AT3 cells was also highly correlated with canonical IFNγ signaling and inflammatory gene expression by Gene Set Enrichment Analysis (GSEA; Fig. 1C). To identify the conservation of the IFNγ response, we analyzed RNA-seq of additional murine solid tumors treated with IFNγ, including B16F10 (melanoma) and MC38 (colon adenocarcinoma), which revealed concordant transcriptional responses (Additional file 1: Fig. S1A, B) across all cell types (Additional file 1: Fig. S1C). Importantly, the IFNγ signature had a strong prognostic significance in a TCGA cohort of breast cancer patients associated with high IFNγ signature positivity correlating with superior survival (Fig. 1D). These data indicate that the molecular processes that regulate tumor cell intrinsic IFNγ target gene expression are of high clinical relevance and prognostic significance in cancer.
Chromatin immunoprecipitation and sequencing (ChIP-seq) for RNA Polymerase II (RNAPII) was performed to assess the processivity across IFNγ-induced genes. IFNγ-stimulated genes were characterized by rapid de novo recruitment of RNAPII following IFNγ treatment with high levels of RNAPII occupancy across the coding region, concurrent with minimal promoterproximal pausing (Fig. 1E). This indicates that unlike EGF responsive genes, which primarily rely on rapid release of RNAPII from the paused state [13], RNAPII initiation is a key rate-limiting checkpoint for IFNγ response genes. In contrast, non-IFNγ-regulated genes exhibited unaltered RNAPII occupancy (Fig. 1E). To gain insight into transcription factors (TFs) driving IFNγ-induced gene expression, de novo motif analysis was performed on the promoter regions of IFNγ-stimulated genes. This demonstrated significant overrepresentation of Interferon-Regulatory Factors (IRF) and the Interferon-Sensitive Response Element (ISRE) motifs (Fig. 1F, G). Overall, these studies indicate that acute IFNγ stimulation leads to active recruitment of RNAPII to a conserved subset of genes putatively regulated by IRF transcription factors.

IFNγ stimulates acute epigenetic remodeling of IRF1-bound cis-regulatory elements
Based on the prevalence of IRF motifs, we performed ChIP-seq for Irf1 following treatment with IFNγ, which identified the loci where Irf1 was recruited following IFNγ-stimulation ( Fig. 2A). As shown in Additional file 1: Fig. S1D, these elements were predominantly localized at intergenic enhancers. It has been reported that IRF TFs physically associate with lysine acetyltransferase (KAT) P300 to activate gene expression [14]]. Therefore, we performed ChIP-seq for P300, which revealed robust co-recruitment to Irf1-bound cis-regulatory elements following IFNγ-stimulation (Fig. 2B). Assay for Transposase-Accessible Chromatin using sequencing (ATACseq) analyses revealed that under steady-state conditions, Irf1-associated cis-regulatory elements exhibit low chromatin accessibility, which was drastically increased following acute IFNγ stimulation (Fig. 2C). Indeed, direct correlation of changes in Irf1 and P300 binding with  (Fig. 2D). Moreover, enhancer regions exhibited the greater increases in chromatin accessibility relative to promoters regions (Fig. 2E). Finally, to validate these Irf1 peak-centric analyses, de novo motif analysis of ATAC-seq peaks that were gained in IFNγ-stimulated cells revealed that IRF motifs are significantly overrepresented (Additional file 1: Fig. S1E). Overall, these data demonstrate that IFNγ stimulation leads to rapid co-recruitment of Irf1 and P300 at selective loci, de novo chromatin remodeling, increased chromatin accessibility and RNAPII recruitment.

IFNγ promotes differential usage of pro-inflammatory enhancers
To test whether recruitment of P300 would lead to hyper-acetylation of adjacent nucleosomes, ChIPseq for H3K27 acetylation (H3K27ac) was performed. Nucleosomes flanking Irf1-bound sites exhibited greatly increased H3K27ac following IFNγ stimulation (Fig. 2F), directly linking Irf1-P300 co-recruitment with local changes in histone acetylation and chromatin accessibility. To investigate changes in H3K27ac more broadly, and infer differential usage of specific enhancers, we identified super-enhancers (SEs) based on H3K27ac signals. Following IFNγ stimulation, there was acquisition of new SE elements proximal to genes involved in IFNγ signaling (Fig. 2G). These data indicate that IFNγ stimulation leads to acute histone acetylation which correlates with an active transcriptional state at specific enhancer elements adjacent to key pro-inflammatory TFs.
To interrogate differential TF usage, ATAC-seq data were further explored to identify TF motifs overrepresented within putative SEs [15]. These autoregulatory sets of TFs, termed the 'core regulatory circuit' (CRC), are represented by calculating the number of individual TF motifs within an SE element regulating the expression of each CRC TF (inward binding) and the number of TF-associated SEs bound by a CRC TF (outward binding). Prior to IFNγ stimulation, the TFs demonstrating the highest connectivity were dominated by TFs previously identified for regulating facets of breast cancer biology ( Fig. 2H; in black), including SOX9, GATA3, and ETV6 [16][17][18]. However, IFNγ stimulation dynamically altered the CRC TF connectivity and led to the acquisition of several interconnected TF nodes involved with pro-inflammatory gene expression (Fig. 2H), including IRF1, IRF2, STAT1, NFKB1, and BCL6. These findings were also recapitulated by calculating the clique fraction, a metric for the participation of individual TFs within interconnected TF networks as a function of the total number of discrete networks (see methods) (Additional file 1: Fig. S1F-G). Thus, IFNγ acutely alters the central repertoire of interconnected TFs that cooperatively regulate cellular transcription.
The Bromodomain and Extra-Terminal (BET) proteins bind acetylated histones and TFs and recruit pTEF-b, thereby stimulating transcriptional elongation [19,20]. Analysis of BRD4 ChIP-seq data showed robust recruitment of BRD4 following IFNγ stimulation to Irf1-bound loci ( Fig. 2I), which may drive effective pause-release post RNAPII initiation (Fig. 1E). Finally, ChIP-seq analysis of RNAPII and H3K4 tri-methylation (H3K4me3), a histone mark that is constitutively associated with active TSS regions (independent of IFNγ stimulation), revealed that RNAPII recruitment in response to IFNγ occurred at promoter-proximal regions demarcated by H3K4me3 ( Fig. 2J-K). Overall, these data provide a putative epigenetic sequence of events whereby recruitment of IRF1 in association with P300 leads to locus-specific chromatin remodeling and histone acetylation, a mark subsequently recognized by "reader" proteins, such as BRD4, RNAPII recruitment and transcriptional activation of IFNγ target genes, which are largely inactive under baseline conditions ( Fig. 2L-M).

Inhibition of BET proteins selectively disrupts IFNγ target gene expression
To experimentally dissect our epigenetic model of IFNγinduced gene expression, BET proteins were antagonized using JQ1 to assess where in the cascade BET proteins were required. According to our sequential model, abrogating BET protein recruitment through JQ1 should interfere with IFNγ-driven transcription, but leave the epigenetic remodeling upstream of nucleosome acetylation unperturbed. Consistent with this hypothesis, Irf1 affected by BET protein inhibition. Similarly, ATACseq revealed that the de novo chromatin remodeling was not reduced by JQ1 co-treatment (Fig. 3D). In contrast, JQ1 co-treatment with IFNγ impaired the recruitment of BRD4 (Fig. 3E) and led to a modest reduction in RNAPII processivity across IFNγ-induced genes (Fig. 3F). Finally, to evaluate the consequences of BET inhibition on mRNA production, we performed RNA-seq under the same conditions. Here, JQ1 co-treatment was associated with highly selective disruption of IFNγ target gene expression, whereby one subset of IFNγ-induced genes was potently suppressed (Fig. 3G), whereas another subset was unaffected by JQ1 co-treatment and remained highly expressed. For example, certain IFNγ-stimulated genes, such as Stat1 and Tap1, were impervious to JQ1 co-treatment (Fig. 3H), whereas immune-suppressive PD-L1 expression was potently suppressed by JQ1-treatment (Fig. 3G). This transcriptional dichotomy remains poorly understood, but suggests that for a subset of genes, effective pause-release can be mediated independently of BRD4 histone-acetyl binding, which is consistent with the notion that the Super Elongation Complex and P-TEFb (CDK9 and Cyclin T) can be recruited in a BRD4 independent manner [21]. These findings are also consistent with numerous studies demonstrating that sub-classification by SEs or other epigenetic/transcriptional features remains insufficient for accurately predicting sensitivity to BET inhibition [22]. Under baseline conditions, JQ1 treatment globally reduced BRD4 binding to chromatin (Additional file 2: Fig. S2A) without altering global chromatin accessibility (Additional file 2: Fig. S2B), even at those regions exhibiting the most robust loss of BRD4 (Additional file 2: Fig. S2C-D), consistent with a recent report [23]. Overall, BET inhibitors reduce BRD4 binding to chromatin and can suppress gene expression downstream of IRF1-driven de novo enhancer remodeling only at a subset of genes/loci.

Contrasting BET Bromodomain inhibition and P300/ CBP acetyltransferase inhibition upon steady-state and IFNγ-induced gene expression
Next, the effect of BET protein inhibition was compared with the effects of inhibiting the catalytic KAT domain of lysine acetyltransferase paralogues, P300 and CBP.
We assessed the transcriptional consequences on both steady-state and IFNγ-inducible gene expression using two chemically distinct catalytic P300/CBP inhibitors, A-485 [24] and A-241 [25]. RNA-seq under steady-state transcription demonstrated that both A-485 and A-241 preferentially suppressed transcription (Additional file 3: Fig. S3A), although A-241 was more potent (Additional file 3: Fig. S3B, C). Accordingly, A-241 was utilized for all subsequent assays. We performed ChIP-seq with reference exogenous genome (ChIP-Rx) for H3K27ac as a direct biomarker of P300/CBP activity following A-241 treatment. Indeed, H3K27ac signal at active cis-regulatory elements demonstrated drastic global reduction following treatment with A-241 (Additional file 3: Fig.  S3D) that was not clearly associated with modulation in chromatin accessibility by ATAC-seq (Additional file 3: Fig. S3E, F). Finally, it was notable that the transcriptional response to A-241 and JQ1 was highly divergent (Additional file 4: Fig. S4A), supporting observations recently made in multiple myeloma [26]. The effects of inhibiting P300/CBP catalytic KAT activity on IFNγ-induced gene expression were explored next. Recapitulating the effects on steady-state transcription (Additional file 3: Fig. S3B, C), A-241 more potently suppressed IFNγ-induced gene expression than A-485 (Additional file 4: Fig. S4B). It was clear that while BET inhibition selectively disrupted the transcriptional response to IFNγ (Fig. 3G), the effects of inhibiting P300/CBP catalytic KAT activity were more global (Fig. 4A) and included transcripts unaffected by BET inhibition, such as Stat1 and Tap1 (Additional file 4: Fig. S4C). Despite suppressing Stat1 transcriptional upregulation (Additional file 4: Fig. S4B), A-241 had no effect on phosphorylation of Stat1 (Y701) downstream of IFNγ receptor stimulation (Fig. 4B), highlighting that inhibition of P300/CBP catalytic KAT activity acts downstream of initial signal transduction. Moreover, global analysis of IFNγ-stimulated genes demonstrated more robust suppression when compared to BET inhibition (Additional file 4: Fig. S4D). These findings highlight that inhibiting P300/CBP catalytic KAT activity robustly modulates cellular transcription. Importantly, these transcriptional defects are broader and clearly dissimilar to the more selective effects of BET bromodomain inhibition, which we anticipate would be even more pronounced if nascent RNA profiling was employed.

Loss of P300/CBP lysine acetyltransferase activity prevents activation of IRF1-bound cis-regulatory elements
To determine how inhibition of P300/CBP impacted the epigenetic remodeling induced by IFNγ, ChIP-Rx [27] for H3K27ac was utilized to investigate the capacity for A-241 to suppress hyper-acetylation of IRF1-associated cis-regulatory elements. As expected, A-241 co-treatment was able to potently suppress IFNγ-induced histone hyper-acetylation (Fig. 4C, D). The requirement of lysine acetylation by P300/CBP for chromatin accessibility during a de novo remodeling process and transcriptional activation process was unclear. Therefore, ATAC-seq in the presence and absence of IFNγ and A-241 was performed, assessing the changes in chromatin accessibility at IRF1-associated cis-regulatory elements. Catalytic P300/CBP KAT activity was required, at least in part, for chromatin accessibility changes associated with IFNγ stimulation (Fig. 4D-E). ChIP-seq for Irf1 under those conditions was also performed, given that Irf1 binding was a critical prerequisite to epigenetic activation of these loci. Strikingly, we found that inhibiting P300/CBP catalytic KAT activity almost completely inhibited the transactivation by Irf1 (Fig. 4D, F), thereby linking loss of TF binding with a failure to subsequently gain chromatin accessibility. Overall, these data indicate that acetylation by P300/CBP regulates transactivation by IRF1 following IFNγ stimulation. More broadly, these findings highlight that P300/CBP are a critical epigenetic dependency that underpins the transcriptional response to IFNγ.

Discussion
Inhibiting the lysine acetyltransferase activity of P300/ CBP is an emergent therapeutic strategy in cancer. Firstgeneration catalytic P300/CBP inhibitor, A-485, was shown to possess anti-tumor activity in models of prostate cancer [24], multiple myeloma (MM), and chronic lymphocytic leukemia [26]. Using a second-generation P300/CBP inhibitor A-241, we could recapitulate key recent findings that showed acute maintenance of chromatin accessibility occurred largely independently of histone acetylation by P300/CBP [26,28,29]. This may be, at least in part, due to P300 remaining bound to chromatin, while its catalytic KAT domain is inhibited, as was shown in MM [26]. One possibility is that combining P300/CBP catalytic KAT inhibitors with P300/CBP Bromodomain inhibitors (to promote displacement of P300/CBP from chromatin) could more profoundly disrupt P300/CBP protein complexes, leading to more pronounced disruption of epigenetic and transcriptional processes. In agreement with this notion, combined inhibition of P300/CBP KAT and Bromodomain modules in MM was shown to additively promote histone hypoacetylation, though these effects were less pronounced than a dual P300/ CBP-targeting proteolysis-targeting chimera (PROTAC) molecule, dCBP-1, which additionally reduced chromatin accessibility [42]. Indeed, emergent small molecule inhibitors and PROTACs targeting P300 and/or CBP will allow for the systematic characterization of the additional functions of P300/CBP beyond its KAT activity, such as the known scaffolding functionality. In contrast to the effects of inhibiting P300/CBP catalytic KAT activity upon cells under steady-state conditions, chromatin accessibility associated with an acute transcriptional stimulus such as IFNγ was highly dependent on the catalytic KAT activity of P300/CBP. While inhibiting P300/CBP catalytic KAT activity abrogated IFNγ-induced expression of certain antigen processing genes (e.g. Tap1), which may potentially limit CD8 + T-cell immunity, this may concomitantly augment natural killer cell-driven killing. Conversely, it is plausible that catalytic P300/CBP inhibitors may be leveraged in auto-immune disorders and graft-vs-host disease to reduce unwanted or excessive T-cell mediated immune responses. Thus, the overall net effect of therapeutically inhibiting P300/CBP catalytic KAT activity on immune responses remains unclear and requires pre-clinical evaluation using syngeneic in vivo model systems. These in vivo studies will be essential to provide independent validation of the epigenetic mechanisms demonstrated here, as well as investigation of the effects of inhibiting P300/CBP catalytic KAT activity on the host immune system and anti-tumor immunity.
We demonstrated that catalytic P300/CBP KAT activity is required for IRF1 transactivation following IFNγ stimulation. We note this effect could be potentially resultant from (1) P300/CBP suppressing transcription of IRF1 itself, (2) loss of a functional acetylation site on IRF1, (3) P300/CBP inhibitors antagonizing the acetylation-independent allosteric interaction between IRF1 and P300 [30], or (4) a combination of (1)-(3). Thus, systematic biochemical assays will be required to deconvolute these possibilities, as well as the broader conservation of these mechanisms across distinct cell/tumor types.

Conclusions
This study provides an epigenetic hierarchy for IFNγstimulated gene expression. Previous efforts to clarify dependencies of IFNγ target genes have largely been performed using genetic depletion approaches, which are particularly problematic for studying the epigenetic proteins that are typically pan-essential. Moreover, these studies have historically been limited to a single IFNγ responsive locus, which fails to capture the global complexity. In contrast, the work detailed herein utilized integrated and unbiased genomics methodologies to evaluate the hierarchy of events that leads to induction of transcription following IFNγ stimulation, as well as functional studies to evaluate this model by antagonizing various epigenetic regulators. Overall, these findings provide fundamentally important insight into the presumed role for certain epigenetic regulators in driving expression of IFNγ target genes, while also highlighting the importance of P300/CBP KAT activity for IRF1 transactivation. (#610185, BD Biosciences), and anti-α-Tubulin (#05-829, Millipore Sigma). Membranes were incubated with horse radish peroxidase (HRP)-conjugated secondary antibodies at room temperature for 1 h and washed at least three times in TBS supplemented with 0.1% v/v Tween20. Immunoreactive bands were revealed using ECL reagents (Amersham ECL or ECL Prime, GE Healthcare) by film exposure (Fujifilm Super RX, Fujifilm) using an Agfa CP1000 developer (Agfa). For both of the cropped immunoblots presented in Fig. 4B, the corresponding uncropped blots are also shown in Additional File 5: Fig. S5A, B.

TCGA correlation analysis
RNA-Seq by Expectation Maximization (RSEM) [31] scaled expression values for TCGA were downloaded from the GDAC Firehose website [32]. Entrez gene IDs were mapped to HGNC gene symbols using the biomaRt (v2.42) R package [33] and collapsed to unique values per gene symbol by selecting the most variable entrez ID among all samples for each gene symbol. Primary samples from the TCGA BRCA cohort were selected using the TCGAbiolinks (v2.14.0) R package [34] and were matched with progression-free interval end points from the TCGA Pan-Cancer Clinical Data Resource [35]. IFNg signature scores were calculated using the Singscore (v1.6) R package [36] from a set of genes found to be strongly interferon induced across multiple cell lines (Additional file 1: Fig. S1C). Samples were then stratified into 'High' (top 90th percentile) and 'Low' (bottom 10th percentile) signature score groups and log-rank p values were calculated using the Survival (v2.38) R package [37].  [38,39]. Gene set enrichment analyses were performed using Gene Set Enrichment Analysis (GSEA) software (v3.0; https:// www. gsea-msigdb. org/ gsea/ index. jsp) using preranked (ranked by t-statistic) and enrichment plots were re-plotted from GSEA output using replotGSEA function in Rstudio. All additional figure generation for RNAsequencing datasets was performed in in Rstudio (v3.5.1).

ChIP-sequencing
Cells were pre-treated with JQ1 (1 µM), A-241 (250 nM), or A-485 (1 µM) for 1 h prior to the addition of recombinant murine IFNy, or vehicle control, for an additional 2 h (3 h total incubation with small molecules). Chromatin immunoprecipitation coupled with next-generation sequencing (ChIP-seq) was performed with reference exogenous genome (ChIP-Rx) using a modified protocol [27]

ATAC-seq and ChIP-seq analysis
Sequencing files were demultiplexed using Bcl2fastq (v2.17.1.14) to generate Fastq files on which QC was performed using FASTQC (v0.11.5). Sequencing reads were then aligned to custom reference genome consisting the mouse genome (Mm10) and the Drosophila melanogaster genome (Dm3) using Bowtie2 (v2.3.3). The resulting SAM files were converted to BAM files using Samtools (v1.4.1) using the view command, which were subsequently sorted, indexed, and potential PCR duplicates removed using the rmdup function. BAM files were converted into BigWig files using the bamCoverage function (Deeptools, v3.0.0) using the following settings (-normalizeUsing CPM-smoothLength 150-binSize 50-e 200 scaleFactor 1). For experiments with external normalization, the reads mapping to either Mm10 or Dm3 genomes were quantified using FeatureCounts (Subread package, v1.5.0) and the percentage of mapped Dm3 reads as a total of total mapped Dm3 + Hg19 reads was calculated. A scale factor was then calculated as the ratio of Dm3 reads in the control treatment condition and the treatment sample, which was then manually applied as the scaleFactor in the bamCoverage function. BigWig files were imported into Integrative Genomics Viewer (IGV, v2.7.0) for visualization of specific loci. Using Deeptools (v3.0.0), heatmaps were generated by computing read average read density (from BigWig files) across defined genomic intervals using computeMatrix, which we subsequently plot using the plotHeatmap command. Average profile plots were created using matrices generated by computeMatrix using a custom script in Rstudio. Annotation of putative super-enhancer regions from H3K27ac ChIP-seq data was performed using Ranking Ordering of Super-Enhancer (ROSE) using a 12.5 k.b. stitching distance and a 2.5 k.b. TSS exclusion to reduce promoter bias. Peak calling was performed with MACS2 with default parameters. Annotation of ATAC-Seq/ChIP-Seq peaks to proximal genes was performed using annotatePeaks.pl (Homer, v4.8). Rstudio (v1.1.46) and R (v3.5.1) were used for all additional analyses and figure preparation using the following R packages: ggplots2, rcolorbrewer.

Data availability
RNA-sequencing data of B16-F10 cells stimulated with IFNγ or vehicle control were downloaded from NIH's Gene Expression Omnibus (GEO) under the accession number GSE134264. RNA-sequencing of MC38 cells stimulated with IFNγ or control was downloaded from GEO (GSE112252). RNA-sequencing of AT3 cells stimulated with IFNγ ± JQ1, or vehicle control, from our previous study was downloaded from GEO (GSE94057). ChIP-sequencing for RNA polymerase II, BRD4, IRF1, and H3K27ac in AT3 cells stimulated with IFNγ ± JQ1, or vehicle control, from our previous study was downloaded from GEO (GSE94130). Next-generation sequencing data generated in this study have been deposited in the NIH's Gene Expression Omnibus (GEO) under the accession number GSE201883.