Low immune signatures in luminal breast cancer cells can be increased by agonists that activate Type I IFN.

a, Gene Set Variation Analysis (GSVA) enrichment scores of IFN stimulated gene (ISG) set expression in breast cancer cell lines grouped by subtype and b, across individual breast cancer cell lines. TNA/B: triple negative A/B, Her2: Her2-positive, LumA/B: luminal A/B. c, GSVA enrichment scores of Antigen Presentation (AP) gene set expression in breast cancer cell lines grouped by subtype. d, Quantification of HLA-A, B, C surface levels in the indicated breast cancer cell lines, measured as geometric mean fluorescence intensity (gMFI) by flow cytometry, n=3 for each cell line. e, Simplified schematic of Type I IFN signaling activation by activation of pattern recognition receptors (PRRs). f, Flow cytometry histograms of HLA-A, B, C surface levels in indicated cell lines following treatment as indicated for 24 hours. Plots are representative of 3 separate experiments. g, FACS quantification of ISRE-GFP fluorescence in indicated cell lines, colored by subtype, following treatment as indicated for 24 hours. Data is represented as fold change over mock-treated control, n=3, data are mean ± sd, p-values are from 2-way ANOVA test using Sidak’s multiple comparisons, comparing each treatment vs mock treatment in all cell lines. h, gMFI of HLA-A, B, C surface levels in HMEC cells, as determined by flow cytometry. Cells were treated with 1000 U/ml IFNβ, 1 μM di-ABZI, or transfected with 300 ng/ml IR-Alu for 24 hours. n=3, data are mean ± sd, p-values are from unpaired t-tests.

Disruption of upstream regulators of PRR signaling activates viral mimicry through distinct mechanisms

a, Representative flow cytometry histograms of ISRE-GFP fluorescence and HLA-A, B, C surface expression (geometric mean fluorescence intensity, gMFI, normalized to mode) in CRISPR disrupted HCC1428 control (sgHPRT1i) or ADAR (sgADAR) cells. b, Representative flow cytometry histograms of ISRE-GFP fluorescence and HLA-A, B, C surface expression (geometric mean fluorescence intensity, gMFI, normalized to mode) in HCC1428 cells treated mock or C48 treated as indicated for 7 days. c, GSVA enrichment scores of ISG and AP gene set expression in HCC1428 cells treated with DMSO for 7 days as a control, or 2.5 μM C48 for 2 or 7 days as indicated. n=3, data are derived from RNAseq analysis, boxplots show min to max, line at median. p-values are from ordinary 1-way ANOVA using Dunnett’s multiple comparisons test. d, Geometric mean fluorescence intensity (gMFI) of ISRE-GFP activity and e, HLA-A, B, C surface levels, as determined by flow cytometry, in HCC1428 cells CRISPR disrupted using an intronic cutting control (sgHPRT1i) or 2 independent sgRNAs against STING (TMEM173) or cGAS (MB21D1) and treated as indicated. Samples were harvested 10 days post-electroporation and 7 days post-treatment. n=3, data are mean ± sem, p-values are from 2-way ANOVA using Sidak’s multiple comparisons test. f, Schematic of MCF7/Jurkat generated co-cultured system. g, Luminescence from NFAT- luciferase reporter (in arbitrary units A.U.) of engineered Jurkat cells co-cultured with engineered MCF7 cells: 4 days post-CRISPR disruption with the indicated sgRNAs, h, 7 days post treatment with mock and the indicated concentrations of C48, i, 10 days post CRISPR disruption with the indicated sgRNAs and treated as indicated 7 days, and j, 7 days post-electroporation with sgRNAs against 2 of 3 KDM5 paralogs. g-j Luciferase activity was measured 5 hours after co-culture, n=3, data are mean ± sd, p-values are ordinary 1-way ANOVA using Dunnett’s multiple comparisons test in g-h, ordinary 1-way ANOVA using Sidak’s multiple comparisons test in i, and 2-way ANOVA using Tukey’s multiple comparisons test in j. k. Quantification of ISRE-GFP activity and HLA-A, B, C surface expression, measured as gMFI, in HCC1428 cells 7 days post-CRISPR disruption using sgRNAs against intronic HPRT1 or ADAR as indicated. Cells are treated with Ruxolitinib (Rux) as indicated throughout the 7 days of the experiment. n=3, data are mean ± sd, p-values are from 2-way ANOVA using Tukey’s multiple comparisons test. l, Percent confluence of HCC1428 cells 12 days post CRISPR disruption using sgRNAs against intronic HPRT1 or ADAR. Cells are treated with DMSO or Rux as indicated throughout the 12 days. n=3, data are mean ± sem, p-values are from 2-way ANOVA using Sidak’s multiple comparisons test. m, Quantification of ISRE-GFP and HLA-A, B, C surface levels, determined by flow cytometry and measured as gMFI, in HCC1428 cells treated as indicated for 7 days. n=3, data are mean ± sd, p-values calculated as in k. n, Luminescence of engineered Jurkat cells co-cultured with MCF7:NY-ESO1 treated as indicated for 7 days, measured 5 hours after co-culture. n=3, data are mean ± sd, p-values are from 2-way ANOVA using Tukey’s multiple comparisons test. o. Percent confluence of HCC1428 cells 14 days post-treatment as indicated. Samples were split at the same ratios on day 7 of treatment. n=3, data are mean ± sd, p-values are from 2-way ANOVA using Tukey’s multiple comparisons test.

KDM5 inhibition induces “viral mimicry” and DNA damage specifically in tumor cells, but not in normal epithelial breast cells.

a, Representative immunoblot of the indicated proteins in HCC1428 tumor cells and normal HMEC cells treated with mock or the indicated concentrations of C48 for 7 days. Histone H3 and tubulin were used as loading controls. b, qRT- PCR analysis of STING mRNA expression levels, measured as 2^(Ct values of indicated gene subtracted from Ct value of β-actin), in normal and tumor cells treated as indicated for 7 days. Data are represented as fold change expression over DMSO-treated (0 μM) control. n=3-5, data are mean ± sem, p-values are from repeated measures 1-way ANOVA using Dunnett’s multiple comparisons test. c, Quantification HLA-A, B, C surface levels, determined by flow cytometry and measured as geometry mean fluorescence intensity (gMFI), in normal and tumor cells treated as indicated for 7 days. Data are represented as fold change gMFI over DMSO-treated (0 μM) control. n=3, data are mean ± sd, p-values are from ordinary 1-way ANOVA using Dunnett’s multiple comparisons test. d, qRT-PCR analysis of ISG signature genes from cells treated as indicated for 7 days. Data is analyzed as in b,. e, Percent confluence of HCC1428 and HMEC cells treated with DMSO or the indicated concentrations of C48 over 19 days. Samples were split at the same ratios on day 7 of treatment. n=3, data are mean ± sd, p-values are from ordinary 1-way ANOVA using Tukey’s multiple comparisons test, comparing the calculated doubling time of the first 7 days derived from each curve to that of DMSO-treated (0 μM) control. f, Representative immunofluorescence (IF) images and quantification of γ-H2AX staining (green) in cells treated with mock or the indicated concentrations of C48 for 7 days. Blue = Hoescht. Scale bar = 50 μM. Violin plots (with lines depicting quartiles) are derived from mean fluorescence intensity (MFI) of γ-H2AX staining, normalized to DMSO-treated (0 μM) control. n=8-13 wells, p-values are from mixed effects analysis with using Dunnett’s multiple comparisons test for HCC1428, repeated measures 1-way ANOVA using Dunnett’s multiple comparisons test for HMEC.

KDM5 inhibition induces R-loop formation in repeat regions that harbor increased H3K4me3.

a, Schematic of R-loops and their potential deleterious consequences. b, Left: representative immunofluorescence images of cells incubated with GFP-RNaseH1 D210N protein (green), which detects RNA:DNA hybrids, in HCC1428 cells treated with mock and C48 as indicated. Blue = Hoescht. Scale bar = 50 μM. Right: Mean fluorescence intensity quantification of GFP positivity. n=8 wells (2 independent experiments where 9 fields /well were counted in triplicate), whiskers are from min to max, p-values are from unpaired t-test. c, Representative immunoblot of XPF protein levels in HCC1428 cells transfected with siRNAs against luciferase (control) or XPF and treated as indicated. Samples were harvested 6 days post-transfection and after 5 days of treatment. Calnexin was used as a loading control. d, FACS quantification of ISRE-GFP fluorescence and HLA-A, B, C surface levels in HCC1428 cells transfected and treated as in c,. Data are represented as fold change of gMFI over control (siLuciferase-transfected cells treated with DMSO, labeled as 0 μM C48). n=3, data are mean ± SEM, p-values are from mixed-effects analysis with multiple comparisons. e-f, qRT-PCR analysis of mRNA expression of XPF and genes representing ISG and AP signatures, measured as 2^(Ct values of indicated gene subtracted from Ct value of β-actin), of the indicated genes in HCC1428 cells transfected and treated as in c,. Data are represented as fold change of mRNA expression levels over control. n=3-4, data are mean ± SEM, p-values are from mixed-effects analysis with multiple comparisons. ns = not significant. g, Genomic annotations of spike-in-normalized H3K4me3 peaks that are significantly increased (log2 fold change > 0, padj < 0.05) in C48-treated compared to DMSO-treated CUT-and-Tag generated data from HCC1428 cells. h, Percent of total aligned H3K4me3 reads that map to subtelomeric regions, as defined in (61)or i, to centromeric regions, as defined in UCSC Genome Browser in H3K4me3 CUT-and-Tag data from cells treated as indicated. n=2, p-values are from unpaired t-tests. j, Percent of total aligned H3K4me3 reads that map to repeat regions in CUT-and-Tag data from DMSO or C48- treated HCC1428 cells. n=2, p-value is from unpaired t-test. k, as in j, but separated into individual repeat classes. Only classes that represent over 3% of total aligned H3K4me3 reads are graphed. p-values are from Sidak’s multiple comparisons test. l, Log2 counts per million (CPM) of H3K4me3 reads that map to the indicated class of repeats in CUT-and-Tag data from HCC1428 cells treated as indicated. Values are z-score scaled. Statistical test is a multiple regression model, and p-values are from the interaction terms of “DMSO” and “C48”. m, Log2CPM of S9.6 CUT-and-Tag generated reads in subsets of repeats that harbor significantly increased H3K4me3 after C48 treatment, separated into the indicated repeat classes. Values are z-score scaled. p-values are calculated as in l.

Model. The KDM5 family of epigenetic regulatory enzymes reduces R-loop formation in genomic repeat regions in luminal breast cancer cells.

Genetic disruption or chemical inhibition of KDM5 causes R-loop-mediated DNA damage in luminal breast cancer cells. R-loop generated RNA:DNA hybrids also result in the activation of the CGAS/STING pathway, increasing ISG and AP signatures as well as cell surface MHCI, which may enhance interactions with the TME. Importantly, KDM5 inhibition does not result in DNA damage or activation of the CGAS/STING pathway in normal breast epithelial cells, suggesting that KDM5 inhibitors present a wider therapeutic window in this setting than STING agonists or type I Interferons. Created using https://BioRender.com/wkqs3p3.

Low immune signatures in luminal breast cancer cells can be increased by agonists that activate type I IFN.

a, qRT-PCR analysis of mRNA expression levels of indicated ISG signature and b, AP signature genes in the indicated cell lines at baseline. Delta Ct values are Ct values of indicated gene subtracted from Ct values of β-actin. Samples were run in triplicate and data are mean ± sd. c-d, Quantification of HLA-A, B, C surface levels, measured as geometric mean fluorescence intensity (gMFI), determined by flow cytometry, in the indicated cell lines following treatment with 1000 U/ml IFNβ, 1 μM di-ABZI, or transfected with 2.5 μg/ml polyIC or 300 ng/ml IR-Alu for 24 hours. Breast lines are colored by subtype as indicated in the figure. Data are represented as fold change gMFI of treatment over that of mock treatment. n=3, data are mean ± sd, p-values are from 1-way ANOVA test using Dunnett’s multiple comparisons, comparing the different treatments vs mock treatment in each cell line. For IR-Alu experiment, p-values are from 2-way ANOVA test using Sidak’s multiple comparisons test, comparing each treatment vs mock treatment in all cell lines. e, Examples of flow cytometry histograms of ISRE- GFP fluorescence in the indicated cell lines after 24h of the indicated treatment. Plots are representative examples from 3 independent experiments.

Perturbation of upstream regulators of PRR signaling activates viral mimicry through distinct mechanisms.

a, Heatmap of log2 fold change dropout of sgRNAs targeting the indicated genes in the indicated cell lines injected into immune-competent (ICB) mice versus immune-incompetent (NSG) mice, summarized as reported (17). b, Representative immunoblot of ADAR protein levels in HCC1428 cells electroporated with sgRNAs against intronic HPRT1 (HPRT1i, control) or 2 independent sgRNAs against ADAR (sgADAR), with actin as loading control. Cells were either treated with DMSO (-) or 1 μM ruxolitinib (+) for duration of experiment and were harvested 4 days post-KO c, Representative immunoblots of the indicated proteins in HCC1428 cells electroporated with 2 independent sgRNAs against STING (TMEM173) or cGAS (MB21D1) and treated with DMSO (-) or 5 μM C48 (+). Samples were harvested 10 days post-CRISPR disruption and 7 days post-C48 treatment. d, Representative flow cytometry histograms of ISRE-GFP levels and HLA-A, B, C surface levels in HCC1428 cells CRISPR disrupted with sgRNAs against STING and e, cGAS, and treated as in c, f. Representative immunoblot of the indicated proteins in HCC1428 electroporated with Cas9 only, sgRNAs targeting intronic regions of KDM5A and KDM5B (Aint+Bint) or 2 independent sets of sgRNAs targeting 2 of the 3 KDM5 paralogs found in female breast cancer cells. Samples were harvested 7 days post CRISPR disruption. g, Representative flow cytometry histograms and quantification of HLA-A, B, C surface levels, measured as gMFI, of samples treated as in f,. n=4-5, data are mean ± sd, p-values are from mixed-effects analysis using Sidak’s multiple comparisons test. h, qRT-PCR of mRNA expression levels of representative ISG signature genes, calculated as 2^(Ct values of the indicated gene subtracted from β-actin Ct value). Data are represented as fold change over Aint + Cint control. n=4-5. Data are mean ± sem. p-values are from mixed effects analysis using Dunnett’s multiple comparisons test. i. Model of ADAR’s role in ISG induction, MHCI surface expression, and tumor cell fitness. Created using https://BioRender.com/d84l254. j, Representative flow cytometry histograms of ISRE-GFP fluorescence and HLA-A, B, C surface expression in HCC1428 cells electroporated with gRNAs against intronic HPRT1 (control) or ADAR, and treated with DMSO (-) or 1 μM ruxolitinib. k. Pearson correlations of ISG gene set expression GSVA scores and DepMap 22Q2 dependency (CERES) scores for all genes across all cell lines common to both data sets. Correlation value for ADAR is indicated. l, Representative flow cytometry histograms of ISRE-GFP fluorescence (left) and HLA-A, B, C surface levels (right) in HCC1428 cells treated as indicated. m, qRT-PCR analysis of mRNA expression levels, calculated as 2^(Ct values of the indicated gene subtracted from β-actin Ct value), in HCC1428 cells treated as indicated. n=3, data are mean ± sd, p-values are from ordinary 1-way ANOVA using Tukey’s multiple comparisons test. n, Number of HCC1428 cells 10 days post-CRISPR disruption using sgRNA against intronic HPRT1, or 2 independent gRNAs against STING or cGAS, and treated as indicated for 7 days. n=1-2, data are mean ± sd.

KDM5 inhibition or disruption induces viral mimicry and loss of cell fitness in luminal breast cancer cell lines.

a, qRT-PCR analysis, measured as 2^(Ct values of indicated gene subtracted from Ct value of β-actin), of cGAS expression in cells treated as indicated for 7 days. Data are represented as fold change expression over DMSO-treated (0 μM) control. n=3- 5, data are mean ± sem, p-values are from repeated measures 1-way ANOVA using Dunnett’s multiple comparisons test. b, Representative flow cytometry histograms of HLA-A, B, C surface levels in the indicated cells treated with mock or the indicated concentrations of C48 for 7 days. c, Representative flow cytometry histograms of HLA-A, B, C surface levels (left) and quantification (right) as geometric mean fluorescence intensity (gMFI) in SKBR3 cells treated with DMSO (0 μM) or 5 μM C48 for 7 days. n=3, data are mean ± sd, p-values are from unpaired t-test. d, qRT-PCR analysis of the indicated IFN and AP genes in normal epithelial or tumor cell lines treated with the indicated concentrations of C48 for 7 days. Data is analyzed as in a,. e, qRT-PCR analysis of the indicated ISG signature genes and cGAS in SKBR3 cells treated with DMSO (0 μM) or 5 μM C48 for 7 days. Data are represented as fold change expression over DMSO-treated (0 μM) control. n=3-4, data are mean ± sem, p-values are from unpaired t-tests. f, Growth inhibition after C48 treatment in the indicated cell lines for 21 days. Data are represented as area under the curve (AUC) of CyQuant quantification after treatment with a titration of C48 concentrations. g, Percent confluence of SKBR3 cells treated with the indicated concentrations of C48 over 19 days. Samples were split at the same ratios on day 7 of treatment. n=3, data are mean ± sd, p-values are from ordinary 2-way ANOVA using Tukey’s multiple comparisons test, comparing the confluence of each sample to DMSO-treated (0 μM) control on the same day. ****p<0.0001 for all concentrations compared to 0 μM/DMSO control. h, Heatmap of CERES scores of the various KDM5 paralogs in the indicated breast cell lines, as reported in DepMap 22Q2. The more negative the score, the higher likelihood the gene of interest is essential. i, Percent confluence of HCC1428 cells electroporated with Cas9 only, sgRNAs targeting intronic regions of KDM5A and KDM5C (Aint+Cint) or 2 independent sets of gRNAs targeting 2 of the 3 KDM5 paralogs. n=4-5, p-values are from mixed-effects analysis using Sidak’s multiple comparisons test comparing the calculated doubling time derived from each curve.

KDM5 inhibition increases R-loop abundance in repeat regions that harbor increased H3K4me3.

a, Representative flow cytometry plots of ISRE-GFP fluorescence and HLA-A, B, C surface levels in HCC1428 cells transfected with siRNAs against Luciferase (siLuc) as a control or XPF (siXPF), and treated as indicated. b, Schematic of CUT-and-Tag experiment using Drosphila spike-ins. DMSO- or C48-treated (2.5 μM) cells and Drosophila spike-in nuclei were bound to Concanavalin A beads. Drosophila nuclei were added at a 1:10 ratio of spike-in nuclei to experimental cells for downstream normalization. H3K4me3, S9.6, or spike-in antibodies were added to recruit pA-Tn5 to H3K4me3, RNA/DNA hybrids, or Drosophila H2Av respectively. Associated DNA were tagmented, and libraries were generated and sequenced. Each condition was run in duplicate (n=2). c, Number of reads from H3K4me3 CUT-and-Tag (CTAG) experiment that were successfully mapped to the Drosophila melanogaster genome. d, Volcano plots of H3K4me3 peaks mapped to the human genome without normalization (left) or after normalization to Drosophila spike-in reads (right). Red dots indicate peaks that are significantly increased (log2 fold change > 0, padj < 0.05) in C48 vs DMSO-treated cells, while blue dots indicate peaks that are significantly decreased (log2 fold change < 0, padj < 0.05). e, Genomic annotations of non-normalized H3K4me3 reads that are significantly increased (log2 fold change > 0, padj < 0.05) in CUT-and-Tag data generated from HCC1428 cells, C48-treated compared to DMSO-treated cells if normalization to drosophila spike-in is not performed. f, Percent of repeat class (repClass) that have increased H3K4me3 reads in CUT-and-Tag data generated from HCC1428 cells treated with C48 vs DMSO-treated cells. The numbers at the right of the bars indicate the actual number of repeat subfamilies in each repeat class that have increased H3K4me3. g. Percent of total H3K4me3 reads mapping to individual SINE, LINE- 1/LINE-2 and LTR subfamilies in HCC1428 CUT-and Tag data where the cells were treated with DMSO or C48 for 7 days. Values are an average of two replicates. Values are represented as normalized alignment (% total reads of TE subfamily divided by TE’s prevalence in genome, in base pairs, x 10-6). Dashed line represents where values are equal in both DMSO and C48- treated cells. h, Percent of total aligned S9.6 reads that map to subtelomeric regions, as definedin (61) (left), or centromeric regions, as defined in UCSC Genome Browser (right) in cells treated as indicated. n=2, p-values are from unpaired t-tests. i, Log2CPM of S9.6 reads in subset of repeats that harbor no increases in H3K4me3 (left), significant increases in H3K4me3 (middle), or significant increases in both H3K4me3 and transcription (right) in HCC1428 data from cells treated with DMSO or C48 treatment. Values are z-score scaled. Statistical test is a multiple regression model, and p-value is from the interaction terms of “C48” and “increased H3K4me3”, or “C48” and “increased H3K4me3 and transcription (trxn)”.