Chronic interferon-stimulated gene transcription promotes oncogene-induced breast cancer

In this study, Wang et al. investigated which MRE11-dependent responses are tumor-suppressive using a mammary organoid system. They report that MRE11 controls oncogene-induced chromatin accessibility and interferon-stimulated transcriptional programs through the nuclear innate immune sensor IFI205, revealing a link between innate immune transcription and tumor development in the mammary gland.

Abstract

The MRE11 complex (comprising MRE11, RAD50, and NBS1) is integral to the maintenance of genome stability. We previously showed that a hypomorphic Mre11 mutant mouse strain (Mre11^ATLD1/ATLD1) was highly susceptible to oncogene-induced breast cancer. Here we used a mammary organoid system to examine which MRE11-dependent responses are tumor-suppressive. We found that Mre11^ATLD1/ATLD1 organoids exhibited an elevated interferon-stimulated gene (ISG) signature and sustained changes in chromatin accessibility. This Mre11^ATLD1/ATLD1 phenotype depended on DNA binding of a nuclear innate immune sensor, IFI205. Ablation of Ifi205 in Mre11^ATLD1/ATLD1 organoids restored baseline and oncogene-induced chromatin accessibility patterns to those observed in WT. Implantation of Mre11^ATLD1/ATLD1 organoids and activation of the oncogene led to aggressive metastatic breast cancer. This outcome was reversed in implanted Ifi205^−/− Mre11^ATLD1/ATLD1 organoids. These data reveal a connection between innate immune signaling and tumor development in the mammary epithelium. Given the abundance of aberrant DNA structures that arise in the context of genome instability syndromes, the data further suggest that cancer predisposition in those contexts may be partially attributable to chronic innate immune transcriptional programs.

The MRE11 complex controls the DNA damage response (DDR) by governing double-strand break (DSB) repair as well as the activation of the ataxia-telangiectasia mutated (ATM) transducing kinase. ATM activation promotes cell cycle checkpoint induction, influences DNA repair, and induces apoptosis or senescence in particular cellular contexts (Stracker and Petrini 2011). Null mutations of Mre11, Rad50, and Nbs1 are lethal at the cellular and organismal level (Lee and Lu 1999; Zhu et al. 2001; Buis et al. 2008). Hence, conditional or hypomorphic alleles of MRE11 complex components have been used to study its role in governing the DDR and in tumor suppression.

We previously queried an array of mice harboring mutations affecting various facets of the DDR network to understand its role in the response to oncogene activation. In WT mice, several indices of DDR activation were detected in mammary hyperplasias arising 3 weeks following induction of the neuT oncogene, including γH2AX and 53BP1 foci (Gupta et al. 2013). In Mre11^ATLD1/ATLD1 mice, which harbor a hypomorphic Mre11 mutation inherited in the human ataxia-telangiectasia-like disorder (A-TLD) (Stewart et al. 1999), these DDR outcomes were abolished. This phenotype correlated with much more extensive hyperplasias 3 weeks after oncogene activation. Whereas progression to tumors in WT mice was rare (5%), the florid hyperplasias in Mre11^ATLD1/ATLD1 mice presaged the frequent (40%) onset of highly aggressive basal-like tumors. The extent of hyperplasia and formation of DDR foci were unaffected by mutations that disrupted apoptosis (Trp53^515C/515C and Chk2^−/−) and DNA repair (53BP1^−/−), and tumor latency was indistinguishable from WT in those genetic contexts (Gupta et al. 2013).

The previous study also revealed an unrecognized role for the MRE11 complex in oncogene-induced chromatin modifications. neuT-expressing WT mammary tissues exhibited deposition of heterochromatic markers such as macroH2A and phospho-histone H3 (S10) accompanied by an arrest in the G2 phase of the cell cycle. Unlike the formation of DDR foci, these events required both the MRE11 complex and p53 (Gupta et al. 2013). It is not clear which of the downstream functions of MRE11 and p53 are tumor-suppressive in this context.

Interferon-inducible gene 205 (IFI205) belongs to the HIN-200 family and is a regulator of inflammasome activation and type I interferon (IFN) production (Brunette et al. 2012; Cridland et al. 2012; Nakaya et al. 2017). HIN-200 family members are characterized by at least one HIN-200 domain (hematopoietic interferon-inducible nuclear proteins with a 200 amino acid repeat), which binds double-stranded DNA via an oligonucleotide/oligosaccharide binding (OB) fold (Jin et al. 2012; Liu et al. 2014). HIN-200 family members also contain an N-terminal pyrin domain, which mediates assembly into innate immune effector complexes (Cridland et al. 2012; Chen et al. 2015). Most HIN-200 proteins possess a nuclear localization signal and localize to the nucleus (Ghosh et al. 2017). Although the HIN-200 family is believed to be involved in the modulation of type I IFN and inflammasome pathways, the precise roles of most of the individual genes have not been elucidated (Brunette et al. 2012; Cridland et al. 2012).

The cancer susceptibility in Mre11^ATLD1/ATLD1 mice could be attributable to an MRE11 complex-dependent pathway that is activated upon oncogenic stress. An alternative possibility is that MRE11 complex hypomorphism somehow creates a permissive state for oncogene-induced carcinogenesis. We used an organoid system comprising primary mammary epithelial cells (MECs) that harbor an inducible oncogene, neuT, to examine these possibilities. The organoid system enables ex vivo manipulation and reimplantation to query the effects of ex vivo genetic manipulations on tumor development in vivo.

At baseline (i.e., prior to oncogene activation), Mre11^ATLD1/ATLD1 organoids exhibited a chronic interferon-stimulated gene (ISG) signature. This coincided with changes in the accessibility of numerous chromatin loci. The ISG transcriptional signature and the changes in chromatin accessibility at baseline were dependent on IFI205 DNA binding. As observed in the autochthonous mammary analyses noted above (Gupta et al. 2013), oncogene induction in implanted Mre11^ATLD1/ATLD1 organoids led to aggressive basal-like breast tumors. Ablation of Ifi205 in Mre11^ATLD1/ATLD1 organoids reverted chromatin accessibility to that of WT and restored the oncogene-induced heterochromatic changes observed in WT upon oncogene induction. This correlated with a sharp reduction in tumor frequency and increased latency in implanted Ifi205^−/− Mre11^ATLD1/ATLD1 organoids relative to Mre11^ATLD1/ATLD1 organoids. This study is consistent with the interpretation that oncogene-induced chromatin changes previously observed in WT MECs are tumor-suppressive and suggests a previously unrecognized link between ISG transcription and tumor suppression.

Results

We generated WT and Mre11^ATLD1/ATLD1 mice that contain an inducible neuT gene. CAG-rtTA transgenic mice (Premsrirut et al. 2011) in which a reverse tetracycline-controlled trans-activator (rtTA) is driven by a tissue-nonspecific CAG promoter were crossed with TetO-neuT transgenic mice (Yeh et al. 2011), allowing for doxycycline-inducible expression of the oncogene neuT and a linked luciferase reporter (Fig. 1A). Primary mammary epithelial cells (MECs) were harvested for the establishment of primary mammary organoids adapting previously described methods for intestinal and prostate epithelia (Fig. 1B; Clevers 2016; Drost et al. 2016). This system allowed us to control oncogene activation in the organoids ex vivo and achieve mammary tissue-specific oncogene activation in vivo once the organoids were implanted into recipient mice.

Figure 1.

Mammary epithelia derived from WT and Mre11^ATLD1/ATLD1 mice develop into normal organoids. (A) Schematic representation of doxycycline-inducible neuT and the luciferase reporter transgene. Image created with BioRender.com. (B) Experimental strategy for primary mammary epithelial cell isolation and mammary organoid culture. Image created with BioRender.com. (C) Bright-field images of WT and Mre11^ATLD1/ATLD1 mammary organoids. Representative images from three repeats. (D) Immunofluorescence (IF) staining of lineage markers (KRT18 [green], KRT5 [pink], and DAPI [blue]) in WT and Mre11^ATLD1/ATLD1 mammary organoids. Representative images from three repeats. (E) Strategy of mammary fat pad clearance and organoid implantation. Image created with BioRender.com. (F) Hematoxylin and eosin (H&E) staining of mammary tissues 6 weeks after WT and Mre11^ATLD1/ATLD1 organoid implantation. Representative images from three repeats. (G) Uniform manifold approximation and projection (UMAP) plot color-coded by genotype from single-cell RNA-seq in WT and Mre11^ATLD1/ATLD1 mammary organoids (each in three replicates). (H) UMAP plot color-coded by cell lineage signature score from single-cell RNA-seq in WT and Mre11^ATLD1/ATLD1 mammary organoids (each in three replicates). (I) UMAP plot color-coded by expression level of cell lineage markers as indicated (Krt14, Krt5, Trp63, and Itga6 for basal cells and Epcam, Krt18, Krt8, and Elf5 for luminal cells) from single-cell RNA-seq in WT and Mre11^ATLD1/ATLD1 mammary organoids (each in three replicates).

Organoids derived from WT and Mre11^ATLD1/ATLD1 mice develop normal mammary glands

The conditions established for mammary organoids were compatible with normal mammary gland development. Organoids from both WT and Mre11^ATLD1/ATLD1 MECs formed hollow spheres, reminiscent of mammary gland architecture in vivo (Fig. 1C). Immunofluorescence (IF) staining showed that both WT and Mre11^ATLD1/ATLD1 organoids contained KRT5-positive basal cells and KRT18-positive luminal cells (Fig. 1D). WT and Mre11^ATLD1/ATLD1 organoids formed histologically normal mammary glands in vivo after implantation into cleared mammary fat pads of NOD-SCID mice (Fig. 1E,F).

The composition of the mammary organoids was further assessed by single-cell RNA-seq. Transcriptomes from 7393 cells (3850 WT and 3543 Mre11^ATLD1/ATLD1) were analyzed and visualized by uniform manifold approximation and projection (UMAP). The gene expression profiles revealed that luminal clusters and basal clusters were evident in WT and Mre11^ATLD1/ATLD1 organoids (Fig. 1G). Those clusters were verified using common lineage markers (Krt14, Krt5, Trp63, and Itga6 for basal cells and Epcam, Krt18, Krt8, and Elf5 for luminal cells) or a lineage signature transcriptional score generated from normal mammary tissues (Fig. 1H,I; Nguyen et al. 2018). These data demonstrate that WT and Mre11^ATLD1/ATLD1 organoids are indistinguishable in the competency for mammary gland development.

WT and Mre11^ATLD1/ATLD1 organoids show different oncogenic responses after neuT activation

Having verified that WT and Mre11^ATLD1/ATLD1 organoids exhibited the same cellular composition of mammary glands in vivo, we assessed their responses to oncogene activation. Doxycycline was added to organoid cultures to induce neuT expression. Western blot confirmed that neuT was induced to similar levels in WT and Mre11^ATLD1/ATLD1 organoids (Supplemental Fig. S1). As observed in vivo (Gupta et al. 2013), oncogene activation induced DDR and heterochromatic markers in WT but not Mre11^ATLD1/ATLD1 organoids. Western blotting was performed at days 0, 14, and 28 after adding doxycycline. γH2AX was induced at day 28 after neuT activation in WT organoids, and heterochromatic markers HP1-γ and macroH2A2 also increased after oncogene activation (Fig. 2A). In contrast, Mre11^ATLD1/ATLD1 organoids did not show induction of γH2AX after oncogene activation, and the induction of HP1-γ and macroH2A2 was also greatly diminished, consistent with prior findings that induction of DDR and chromatin markers is dependent on the MRE11 complex (Fig. 2A). Hence, the mammary organoid system adapted here recapitulates previously observed responses to oncogenes in vivo (Gupta et al. 2013).

Figure 2.

WT and Mre11^ATLD1/ATLD1 organoids show different oncogenic responses after neuT activation. (A) Western blot of whole-cell extracts taken from WT and Mre11^ATLD1/ATLD1 organoids at the indicated time points after adding 1 μg/mL doxycycline. Representative images from three repeats. (B) Heat map of 5644 genes exhibiting different patterns of changes after oncogene activation between WT and Mre11^ATLD1/ATLD1 organoids from bulk RNA-seq analysis (Padj ≤ 0.01, |FC| ≥ 1.5; four replicates from WT and three replicates from Mre11^ATLD1/ATLD1). Labels at the left summarize the corresponding cluster based on pathway and biological function analysis. (Cluster 1) Stem cell and DNA damage response (DDR), (cluster 3) chromatin and DDR, (cluster 8) immune, (cluster 10) extracellular matrix (ECM) and lipid. (C) Luminal progenitor signature score, mature luminal signature score, and basal signature score calculated from bulk RNA-seq in WT and Mre11^ATLD1/ATLD1 organoids at the indicated time points after adding 1 μg/mL doxycycline (four replicates from WT and three replicates from Mre11^ATLD1/ATLD1). (D) Heat map of chromatin regions (rows) showing different accessibility at the indicated time points after adding 1 μg/mL doxycycline in WT and Mre11^ATLD1/ATLD1 organoids from ATAC-seq analysis (each in three replicates). (Top left) Chromatin regions showing different accessibility after oncogene activation in WT organoids. (Right) The status of the same regions at the same time points in Mre11^ATLD1/ATLD1 organoids is plotted as comparison (88 peaks; Padj ≤ 0.05, |FC| ≥ 1.5). (Bottom right) Chromatin regions showing different accessibility after oncogene activation in Mre11^ATLD1/ATLD1 organoids. (Left) The status of the same regions at the same time points in WT organoids is plotted as comparison (395 peaks; Padj ≤ 0.05, |FC| ≥ 1.5). (E) Venn diagram of chromatin regions showing oncogene-induced changes in accessibility in WT and Mre11^ATLD1/ATLD1 organoids from D.

To examine the potential mechanistic bases for MRE11 complex-dependent tumor suppression in the mammary epithelium, the responses of WT and Mre11^ATLD1/ATLD1 organoids to oncogene activation were assessed. RNA was prepared from WT and Mre11^ATLD1/ATLD1 organoids at days 0, 14, and 28 after doxycycline induction, followed by bulk RNA-seq. Five-thousand-six-hundred-forty-four genes exhibited different changes in expression after oncogene activation between WT and Mre11^ATLD1/ATLD1 (Padj ≤ 0.01, |FC| ≥ 1.5). Among them, ∼30% (1539 out of 5644) were already differentially expressed prior to oncogene activation. Clustering and pathway analysis of genes with different expression patterns revealed signatures related to stem cells, DDR, chromatin, immune response, extracellular matrix (ECM), and lipid metabolism (Fig. 2B). Although WT and Mre11^ATLD1/ATLD1 organoids displayed similar lineage signature scores at baseline, both luminal progenitor signature score and mature luminal signature score decreased significantly in response to oncogene activation in WT organoids, and a reduction of basal signature score was observed in Mre11^ATLD1/ATLD1 organoids, indicating that different oncogene-induced trajectories may exist in WT and Mre11^ATLD1/ATLD1 organoids (Fig. 2C). This is consistent with our previous finding that neuT-induced Mre11^ATLD1/ATLD1 tumors are basal-like breast cancer (Gupta et al. 2013), which has been reported to originate from the luminal progenitors (Debacq-Chainiaux et al. 2009; Molyneux et al. 2010).

Given the difference in oncogene-induced chromatin modifications between WT and Mre11^ATLD1/ATLD1 organoids, the effect of oncogene activation on chromatin accessibility was assessed. ATAC-seq was performed using WT and Mre11^ATLD1/ATLD1 organoids at the same time points as above. Chromatin accessibility analysis revealed substantial differences in response to oncogene activation between these two genotypes. In WT organoids, 0.2% of the total chromatin regions (64 out of 22,950) became more open, and 0.1% of regions (24 out of 22,950) became more closed after oncogene activation (Fig. 2D, top). In contrast, Mre11^ATLD1/ATLD1 organoids exhibited greater changes in chromatin accessibility, with 0.8% of chromatin regions (187 out of 22,950) becoming more accessible and 0.9% of regions (208 out of 22,950) becoming less accessible after oncogene activation (Fig. 2D, bottom). However, only 10 regions showed similar changes in accessibility in both WT and Mre11^ATLD1/ATLD1 organoids, underscoring divergent oncogene-associated chromatin responses in those genotypes (Fig. 2E).

Mre11^ATLD1/ATLD1 organoids exhibit a chronic ISG transcriptional signature and altered chromatin status at baseline

The differential responses and increased tumorigenesis in neuT-expressing Mre11^ATLD1/ATLD1 MECs can be explained by two nonexclusive possibilities. First, the Mre11 complex may control tumor-suppressive pathways that are induced by oncogenic stress. Alternatively, MRE11 complex hypomorphism may create a state of permissiveness for oncogene-driven carcinogenesis. We reasoned that comparison of WT and Mre11^ATLD1/ATLD1 organoids at baseline would address the latter possibility.

Bulk RNA-seq was performed in WT and Mre11^ATLD1/ATLD1 organoids, followed by canonical pathway analysis and gene set enrichment analysis (GSEA). Prior to oncogene activation, an interferon-stimulated gene (ISG) transcriptional program was the most significant difference between WT and Mre11^ATLD1/ATLD1 organoids (Fig. 3A,B). The ISG pathway observed in Mre11^ATLD1/ATLD1 organoids reflected upregulation of multiple ISGs [e.g., Nos2, Ccl20, Ifi205, Ccl5, Ifit1(Isg56), Isg15, Irf7, and Stat1] (Fig. 3C), leading to a stronger ISG transcriptional signature. Single-cell RNA-seq also showed a uniform increase in the ISG signature score in Mre11^ATLD1/ATLD1 organoids, indicating that the enhanced signature seen in the bulk RNA-seq does not originate from a subpopulation of cells (Fig. 3D; Supplemental Fig. S2).

Figure 3.

Mre11^ATLD1/ATLD1 organoids exhibit chronic ISG transcriptional signature and altered chromatin status at baseline. (A) Canonical pathway analysis performed by ingenuity pathway analysis (IPA) using differentially expressed genes (DEGs; Padj ≤ 0.01, |FC| ≥ 1.5; four replicates from WT and three replicates from Mre11^ATLD1/ATLD1) from a comparison between Mre11^ATLD1/ATLD1 and WT organoids at baseline in RNA-seq (Padj < 0.05). Z-score was calculated by comparing Mre11^ATLD1/ATLD1 versus WT. The positive z-score predicts the activation state of the pathway. An absolute z-score of ≥2 is considered significant. (B) Gene set enrichment analysis (GSEA) comparison between Mre11^ATLD1/ATLD1 and WT organoids at baseline in RNA-seq (four replicates from WT and three replicates from Mre11^ATLD1/ATLD1). (C) Volcano plot from a comparison between Mre11^ATLD1/ATLD1 and WT organoids at baseline in RNA-seq (four replicates from WT and three replicates from Mre11^ATLD1/ATLD1). (Green) Significant (Padj ≤ 0.01, |FC| ≥ 1.5), (gray) not significant, (red) significant ISGs. (D) Violin plot of ISG signature score from single-cell RNA-seq in WT and Mre11^ATLD1/ATLD1 organoids (each in three replicates). (E) Western blot of whole-cell extracts taken from WT and Mre11^ATLD1/ATLD1 organoids. Representative image from three repeats. (F) Expression levels of ISGs measured by qPCR in WT and Mre11^ATLD1/ATLD1 organoids. Data are presented as mean of three repeats with standard deviation. (G) Heat map of differentially expressed chromatin modifiers and remodelers from RNA-seq in WT and Mre11^ATLD1/ATLD1 organoids (Padj ≤ 0.01, |FC| ≥ 1.5; four replicates from WT and three replicates from Mre11^ATLD1/ATLD1). Genes marked with an asterisk are known to be influenced by the IFN signaling pathway (relevant references are listed in Supplemental Table S1). (H) Heat map of differential chromatin accessibility regions (rows) between WT and Mre11^ATLD1/ATLD1 organoids from ATAC-seq (Padj ≤ 0.05, |FC| ≥ 1.5; 118 regions more accessible and 338 regions less accessible in Mre11^ATLD1/ATLD1; each in three replicates). (I) Potential of heat diffusion for affinity-based trajectory embedding (PHATE) 3D plot using chromatin modifiers and remodelers color-coded by genotype (left) and ISG signature score (right) from single-cell RNA-seq in WT and Mre11^ATLD1/ATLD1 organoids (each in three replicates).

Orthogonal validation of the RNA-seq analysis was obtained from Western blotting and qPCR. Western blot analysis showed increased protein levels of STAT1 and STING in Mre11^ATLD1/ATLD1 organoids compared with WT (Fig. 3E). qPCR also revealed upregulation of common downstream ISGs (e.g., Cxcl10, Ccl5, Irf7, Isg56, and Isg15) in Mre11^ATLD1/ATLD1 organoids (Fig. 3F). These data confirm the presence of a stronger ISG signature in Mre11^ATLD1/ATLD1 organoids compared with WT organoids at baseline prior to oncogene activation.

Chronic ISG transcription has been shown to trigger widespread changes in chromatin accessibility (Benci et al. 2016). We therefore investigated the potential effect of ISG transcriptional program on chromatin status in the organoids. Bulk RNA-seq analysis revealed that 36 chromatin modifiers and remodelers (e.g., Jade2, Kat2b, Foxa1, and Hmgn5) were differentially expressed in Mre11^ATLD1/ATLD1 compared with WT organoids (Padj ≤ 0.01, |FC| ≥ 1.5) (Fig. 3G). Notably, 18 of these chromatin regulators are known to be influenced by the interferon (IFN) signaling pathway (asterisk in Fig. 3G; Supplemental Table S1; Kitani et al. 1983; Gianni et al. 1996; Grotzinger et al. 1996; Uddin et al. 1999; Ji et al. 2003; Fontana et al. 2004; Lu et al. 2004; Rajsbaum et al. 2008; Sweeney et al. 2011; Sichien et al. 2016; Liu et al. 2017; Subramanian et al. 2018; Hubel et al. 2019; Alexander et al. 2020; Cornut et al. 2020; Fang et al. 2021; Russo et al. 2021; Bhat et al. 2022) and likely impart chromatin changes due to the chronic ISG transcriptional program in Mre11^ATLD1/ATLD1 organoids. ATAC-seq analysis was performed in WT and Mre11^ATLD1/ATLD1 organoids at baseline to assess the difference in chromatin accessibility. The accessibility of 2.0% of chromatin regions (456 out of 22,950) was different between the two genotypes, with 0.5% of the regions (118 out of 22,950) more open and 1.5% of the regions (338 out of 22,950) more closed in Mre11^ATLD1/ATLD1 organoids (Padj ≤ 0.05, |FC| ≥ 1.5) (Fig. 3H).

Single-cell RNA-seq suggested that the ISG signature observed correlated with chromatin status in Mre11^ATLD1/ATLD1 organoids. To estimate the chromatin status and evaluate ISG signature simultaneously at the single-cell level, expression levels of 113 chromatin modifiers and remodelers examined in a previous study (Heide et al. 2022) were used as a proxy of chromatin status. Potential of heat diffusion for affinity-based trajectory embedding (PHATE) analysis was performed and then color-coded by ISG signature score. The resulting 3D PHATE plot revealed a distinct group of cells with differential expression of chromatin modifiers and remodelers (Fig. 3I, left) and higher ISG signature scores (Fig. 3I, right), suggesting a correlation between chromatin status and ISG signatures. These findings demonstrate that the ISG transcriptional program found in Mre11^ATLD1/ATLD1 organoids at baseline correlates with changes in chromatin status.

IFI205 induces ISG signature and mediates chromatin changes in Mre11^ATLD1/ATLD1 organoids

The ISG signature observed in Mre11^ATLD1/ATLD1 organoids was not attributable to the cGAS–STING pathway. As previously observed in Mre11^ATLD1/ATLD1 MEFs (Theunissen et al. 2003), spontaneous chromosome aberrations were elevated in Mre11^ATLD1/ATLD1 organoids compared with WT (P = 0.0485; aberrations: WT zero out of 21, and Mre11^ATLD1/ATLD1 five out of 22) (Fig. 4A,B). However, no significant increase of cytoplasmic dsDNA or cGAMP was detected in Mre11^ATLD1/ATLD1 organoids compared with WT (Supplemental Fig. S3), indicating that cGAS–STING cytosolic DNA sensing is not the major pathway responsible for the ISG signature in Mre11^ATLD1/ATLD1 organoids.

Figure 4.

IFI205 induces ISG signature and mediates chromatin changes in Mre11^ATLD1/ATLD1 organoids at baseline. (A) Example images of metaphase spreads in WT (n = 21) and Mre11^ATLD1/ATLD1 (n = 22) organoids. (B) Quantification of metaphase spreads in A (Fisher's exact test, P = 0.0485; aberrations: WT zero out of 21, Mre11^ATLD1/ATLD1 five out of 22). (C) Schematic representation of IFI205 protein. (D) Expression levels of ISGs measured by qPCR in WT, Mre11^ATLD1/ATLD1, Ifi205^−/−, and Ifi205^−/− Mre11^ATLD1/ATLD1 organoids. Fold changes were calculated as Mre11^ATLD1/ATLD1 versus WT, Ifi205^−/− versus WT, and Mre11^ATLD1/ATLD1 Ifi205^−/− versus WT. Data are presented as mean of three repeats with standard deviation. (E) Western blot of whole-cell extracts taken from WT, Mre11^ATLD1/ATLD1, Ifi205^−/−, and Ifi205^−/− Mre11^ATLD1/ATLD1 organoids. Representative image from three repeats. (F) Expression levels of 44 ISGs (full gene list in Supplemental Table S2) measured by qPCR in WT, Mre11^ATLD1/ATLD1, Ifi205^−/−, Ifi205^−/− Mre11^ATLD1/ATLD1, cGas^−/−, and cGas^−/− Mre11^ATLD1/ATLD1 organoids. Log₂ fold changes were calculated as Mre11^ATLD1/ATLD1 versus WT, Mre11^ATLD1/ATLD1 Ifi205^−/− versus Ifi205^−/−, and cGas^−/− Mre11^ATLD1/ATLD1 versus cGas^−/−. Each dot represents the mean log₂ fold change from three repeats for a specific ISG target. Data are also presented as median with first and third quartile for all the ISG targets. (G) Expression levels of 44 ISGs (full gene list in Supplemental Table S2) measured by qPCR in WT, Mre11^ATLD1/ATLD1, Ifi205^−/−, Ifi205^−/− Mre11^ATLD1/ATLD1, cGas^−/−, and cGas^−/− Mre11^ATLD1/ATLD1 organoids at days 0 and 28 after adding 1 μg/mL doxycycline. Log₂ fold changes were calculated as day 28 versus day 0. Each dot represents the mean log₂ fold change from three repeats for a specific ISG target. Data are also presented as median with first and third quartile for all the ISG targets. (H) Western blot of whole-cell extracts taken from WT, Mre11^ATLD1/ATLD1, Ifi205^−/− Mre11^ATLD1/ATLD1, cGas^−/− Mre11^ATLD1/ATLD1, Stat1^−/− Mre11^ATLD1/ATLD1, and Sting^−/− Mre11^ATLD1/ATLD1 organoids. Representative image from three repeats. (I) Quantification of Mre11^ATLD1/ATLD1 Ifi205^−/− and WT organoid comparison from ATAC-seq using differential chromatin accessibility regions between Mre11^ATLD1/ATLD1 and WT organoids (open, Padj ≤ 0.05 and FC ≥ 1.5; closed, Padj ≤ 0.05 and FC ≤ −1.5; three replicates from WT, four replicates from Mre11^ATLD1/ATLD1, and three replicates from Ifi205^−/− Mre11^ATLD1/ATLD1). (J) Principal component analysis (PCA) plot from ATAC-seq in WT, Mre11^ATLD1/ATLD1, and Mre11^ATLD1/ATLD1 Ifi205^−/− organoids (open, Padj ≤ 0.05 and FC ≥ 1.5; closed, Padj ≤ 0.05 and FC ≤ −1.5; three replicates from WT, four replicates from Mre11^ATLD1/ATLD1, and three replicates from Ifi205^−/− Mre11^ATLD1/ATLD1).

Given that cytoplasmic DNA does not appear to underlie the observed ISG signature, we hypothesized that a nuclear sensor may engage aberrant DNA structures and induce the observed chronic ISG signature. We found that the transcript encoding the nuclear DNA sensor IFI205 was upregulated in Mre11^ATLD1/ATLD1 organoids at baseline (i.e., without oncogene expression; FC = 7.11, Padj = 3.3676 × 10⁻⁹) (Fig. 3C). This gene product belongs to the HIN-200 family (Brunette et al. 2012) and was also recently found to be enriched at stressed DNA replication forks (Wardlaw and Petrini 2022). IFI16, a member of the human HIN-200 family, was frequently altered in human breast cancer in the METABRIC database (21%, 440 out of 2051) (Cerami et al. 2012; Pereira et al. 2016), with amplifications being the predominant type (339 out of 440), suggesting a potential role for IFI16 in the elevated ISG levels in cells with genome instability.

IFI205 binds dsDNA through its HIN domain and activates IFN signaling (Fig. 4C; Ghosh et al. 2017; Dunphy et al. 2018). To determine whether the ISG signature and subsequent chromatin changes were dependent on IFI205, we used CRISPR–CAS9 to inactivate the Ifi205 gene in WT and Mre11^ATLD1/ATLD1 organoids. qPCR showed that while only Cxcl10 level decreased in Ifi205^−/− organoids compared with WT organoids (FC = 0.27), all the ISGs that were elevated in Mre11^ATLD1/ATLD1 (Cxcl10, Ccl5, Irf7, Isg56, and Isg15) were reduced after knocking out Ifi205 in Mre11^ATLD1/ATLD1 organoids (FC, Ifi205^−/− Mre11^ATLD1/ATLD1 vs. Mre11^ATLD1/ATLD1; Cxcl10 = 2.91, Ccl5 = 3.78, Irf7 = 0.89, Isg56 = 1.59, and Isg15 = 1.66) (Fig. 4D). Western blot showed that both IFN-related proteins (STAT1 and STING) and chromatin markers (HP1-γ and macroH2A2) decreased in Mre11^ATLD1/ATLD1 Ifi205^−/− organoids compared with Mre11^ATLD1/ATLD1 organoids, though they remain similar between Ifi205^−/− and WT organoids (Fig. 4E). The CRISPR–CAS9 method used in these organoids did not cause changes in chromatin markers or downstream ISGs (Supplemental Fig. S4).

Oncogene activation causes DNA damage that is likely linked to DNA replication stress (Halazonetis et al. 2008; Macheret and Halazonetis 2015; Fagan-Solis et al. 2020). In turn, DNA damage can lead to ISG induction in both a cytosolic DNA-dependent and -independent manner (Brzostek-Racine et al. 2011; Dunphy et al. 2018; Cheon et al. 2023). To evaluate the contribution of cGAS and IFI205 to the baseline ISG elevation in Mre11^ATLD1/ATLD1, as well as the oncogene-induced ISGs, we analyzed the expression levels of 44 ISGs (full list in Supplemental Table S2) using qPCR array in WT, Mre11^ATLD1/ATLD1, Ifi205^−/−, Mre11^ATLD1/ATLD1 Ifi205^−/−, cGas^−/−, and Mre11^ATLD1/ATLD1 cGas^−/− organoids before and after doxycycline addition (Supplemental Fig. S5). At baseline, the ISGs were elevated in Mre11^ATLD1/ATLD1 compared with WT organoids (median log₂ fold change, 2.68) (Fig. 4F). This ISG induction caused by Mre11 hypomorphism was markedly decreased in the Ifi205 knockout background (Mre11^ATLD1/ATLD1 Ifi205^−/− vs. Ifi205^−/− median log₂ fold change, 0.256) but only slightly reduced in the cGas knockout background (Mre11^ATLD1/ATLD1 cGas^−/− vs. cGas^−/− mean log₂ fold change, 2.15), indicating that the ISG induction in Mre11^ATLD1/ATLD1 organoids at baseline was mainly mediated by IFI205 rather than by cGAS (Fig. 4F). At day 28 after oncogene activation, compared with WT organoids (median log₂ fold change, 2.90), all the other mutants showed decreased ISG induction (Mre11^ATLD1/ATLD1 median log₂ fold change, 0.462; Ifi205^−/− median log₂ fold change, 0.613; Mre11^ATLD1/ATLD1 Ifi205^−/− median log₂ fold change, 0.157; cGas median log₂ fold change, 0.0458; Mre11^ATLD1/ATLD1 cGas^−/− median log₂ fold change, −0.156), suggesting that both IFI205 and cGAS contribute oncogene-induced ISG expression. Given that the induction was lower in cGas^−/− compared with Ifi205^−/− organoids (median log₂ fold change, 0.0458 vs. 0.613), it is likely that cGAS plays a more important role in this process (Fig. 4G).

IFI205 mediated the chromatin changes observed in Mre11^ATLD1/ATLD1 organoids. To investigate the effects of other IFN regulators on chromatin changes, Mre11^ATLD1/ATLD1 cGas^−/−, Mre11^ATLD1/ATLD1 Stat1^−/−, and Mre11^ATLD1/ATLD1 Sting^−/− organoids were generated. Western blot showed that knocking out cGas, Stat1, or Sting in Mre11^ATLD1/ATLD1 organoids failed to inhibit HP1-γ and macroH2A2 protein levels to that observed in Mre11^ATLD1/ATLD1 Ifi205^−/− organoids. Notably, the chromatin markers in Mre11^ATLD1/ATLD1 Sting^−/− organoids were most similar to Mre11^ATLD1/ATLD1 Ifi205^−/− organoids, indicating that the downstream effects of IFI205 in Mre11^ATLD1/ATLD1 organoids may be largely through STING (Fig. 4H).

Ablation of Ifi205 also largely reverted Mre11^ATLD1/ATLD1 chromatin accessibility to WT. ATAC-seq analysis showed that 91.3% of the previously more open chromatin regions and 76.7% of the previously more closed chromatin regions in Mre11^ATLD1/ATLD1 organoids reverted to WT status in Ifi205^−/− Mre11^ATLD1/ATLD1 organoids (Ifi205^−/− Mre11^ATLD1/ATLD1 vs. WT, Padj > 0.05 or |FC| < 1.5) (Fig. 4I; Supplemental Fig. S6). In addition, among the top 20 enriched transcription factor binding motifs in more closed chromatin regions of Mre11^ATLD1/ATLD1 compared with WT organoids, 13 motifs became enriched in more open chromatin regions of Mre11^ATLD1/ATLD1 Ifi205^−/− compared with Mre11^ATLD1/ATLD1 organoids (ranked by distance calculated using K-S statistics, adjusted P-value < 0.01) (Supplemental Fig. S7A–D). Notably, six of these 13 motifs ranked within the top 20 enriched motifs in more open chromatin regions of Mre11^ATLD1/ATLD1 Ifi205^−/− organoids (PAX3, SEBOX, DDX1, TLX2, NOTO, and CDC5L). Principle component analysis (PCA) of ATAC-seq also showed that, on PC2, Ifi205^−/− Mre11^ATLD1/ATLD1 exhibited a trend of shifting back to WT chromatin (Fig. 4J). PC1 likely represents the MRE11-independent impact of IFI205 on chromatin accessibility (Fig. 4J).

In addition to the normalization of baseline ISG signature and chromatin accessibility, IFI205 deficiency in Mre11^ATLD1/ATLD1 organoids also restored the chromatin response to oncogene activation. Western blot was performed using samples collected at days 0, 14, and 28 after adding doxycycline from WT, Mre11^ATLD1/ATLD1, Ifi205^−/−, and Ifi205^−/− Mre11^ATLD1/ATLD1 organoids. Induction of macroH2A2 and HP1-γ could be detected in Ifi205^−/− Mre11^ATLD1/ATLD1 organoids after oncogene activation similar to that in WT (Fig. 5A). To assess alterations in chromatin accessibility after oncogene activation, ATAC-seq was performed at the same time points in WT, Mre11^ATLD1/ATLD1, and Ifi205^−/− Mre11^ATLD1/ATLD1 organoids. PCA derived from oncogene-induced chromatin regions identified in WT and Mre11^ATLD1/ATLD1 organoids revealed more similar trends between Ifi205^−/− Mre11^ATLD1/ATLD1 and WT organoids than between Ifi205^−/− Mre11^ATLD1/ATLD1 and Mre11^ATLD1/ATLD1 organoids, suggesting that Ifi205^−/− Mre11^ATLD1/ATLD1 largely restores oncogene-induced chromatin changes observed in WT (Fig. 5B). These data demonstrate that IFI205 underlies the ISG signature and subsequent chromatin changes observed in Mre11^ATLD1/ATLD1 organoids. Ablation of Ifi205 in Mre11^ATLD1/ATLD1 organoids not only reverts chromatin accessibility to WT at baseline but also recovers the induction of chromatin changes after oncogene activation.

Figure 5.

Knocking out Ifi205 in Mre11^ATLD1/ATLD1 organoids restores oncogene-induced chromatin responses. (A) Western blot of whole-cell extracts taken from WT, Mre11^ATLD1/ATLD1, Ifi205^−/−, and Mre11^ATLD1/ATLD1 Ifi205^−/− organoids at the indicated time points after adding 1 μg/mL doxycycline. Representative images from three repeats. (B) PCA plot from ATAC-seq in WT, Mre11^ATLD1/ATLD1, and Mre11^ATLD1/ATLD1 Ifi205^−/− organoids at the indicated time points after adding 1 μg/mL doxycycline, using the regions that showed oncogene-induced changes in WT and Mre11^ATLD1/ATLD1 organoids (four replicates from WT, Mre11^ATLD1/ATLD1, and Mre11^ATLD1/ATLD1 Ifi205^−/− organoids at day 0 and three replicates from WT, Mre11^ATLD1/ATLD1, and Mre11^ATLD1/ATLD1 Ifi205^−/− organoids at days 14 and 28).

IFI205 triggers ISG signature and chromatin changes through DNA sensing

Based on the outcomes of Ifi205 deficiency in Mre11^ATLD1/ATLD1 MECs, we hypothesized that IFI205 senses aberrant DNA structures (e.g., extrachromosomal DNA or decondensed chromatin) in Mre11^ATLD1/ATLD1 organoids and promotes both the ISG transcriptional signature and the attendant chromatin changes. This proposal explicitly predicts that DNA binding by IFI205 underlies the effect of IFI205 on the Mre11^ATLD1/ATLD1 phenotype. Therefore, we designed a DNA-binding-deficient Ifi205 mutant to test this hypothesis. A previous study showed that six mutations in the AIM2 HIN domain disrupted DNA binding ability (Sung et al. 2012), among which two sites (K248 and K313) were facing toward DNA in the structure (Fig. 6A; Ru et al. 2013) and were conserved within the IFI205 HIN domain (R299 and K364) (Fig. 6B). Hence, the Ifi205 mutant (Ifi205^K364ER299E) harboring both the K364E and R299E mutations was constructed.

Figure 6.

IFI205 triggers ISG signature and chromatin changes through DNA sensing. (A) Structure of the AIM2 HIN domain with dsDNA (PDB: 4jbm) overlaid by the IFI205 structure predicted by Phyre2. Mutations modeled later in IFI205 are colored red, and corresponding conserved sites in AIM2 are colored yellow. (B) Sequence alignment between AIM2 and IFI205. Mutations modeled later in IFI205 are colored blue. (Line) Fully conserved, (colon) strongly similar, (dot) weakly similar. (C) Binding curves from fluorescence anisotropy using 60mer FAM-labeled dsDNA at 50 nM and increasing amounts of recombinant IFI205 or IFI205^K364ER299E protein. Data are presented as mean of three repeats with standard deviation. (D) Western blot of whole-cell extracts taken from Mre11^ATLD1/ATLD1 Ifi205^−/− organoids ectopically expressing noncomplementary empty vector (NC), Ifi205, or Ifi205^K364ER299E. Representative image from three repeats. (E) Expression levels of ISGs measured by qPCR in Mre11^ATLD1/ATLD1 Ifi205^−/− organoids ectopically expressing noncomplementary empty vector (NC), Ifi205, or Ifi205^K364ER299E. Fold changes were calculated as Ifi205-expressing versus NC-expressing and as Ifi205^K364ER299E-expressing versus NC-expressing. Data are presented as mean of three repeats with standard deviation.

DNA binding was assessed using electrophoretic mobility shift assay (EMSA) and fluorescence anisotropy. Recombinant FLAG-tagged IFI205 and IFI205^K364ER299E were purified from Escherichia coli (Supplemental Fig. S8). We measured the DNA binding affinity of IFI205 and IFI205^K364ER299E by fluorescence anisotropy using 50 nM FAM-labeled 60mer dsDNA (Anderson et al. 2008). The K_D of IFI205^K364ER299E was 190.7 nM ± 86.3 nM, eightfold higher than that of IFI205 (23.7 nM ± 9.44 nM), indicating a significant decrease of DNA binding affinity in IFI205^K364ER299E (Fig. 6C). Similar results were observed by EMSA using 20 nM FAM-labeled 60mer dsDNA (Supplemental Fig. S9).

Empty vector, Ifi205, or Ifi205^K364ER299E (both IFI205 proteins were FLAG-tagged) was ectopically expressed in Ifi205^−/− Mre11^ATLD1/ATLD1 organoids followed by Western blot to assess IFN and chromatin markers. Ectopically expressing Ifi205 resulted in much higher levels of STAT1, macroH2A2, and HP1-γ than that in organoids complemented with Ifi205^K364ER299E (Fig. 6D). Consistently, Ifi205-expressing organoids also showed higher ISG induction compared with Ifi205^K364ER299E-expressing organoids (Fig. 6E), verifying that DNA binding by IFI205 underlies the ISG signature and chromatin changes in Mre11^ATLD1/ATLD1 MECs.

Mammary tumorigenesis in Mre11^ATLD1/ATLD1 is Ifi205-dependent

Ifi205 deficiency reverts the oncogenic response of Mre11^ATLD1/ATLD1 to that of WT MECs, including the induction of heterochromatic marks and changes in chromatin accessibility. We reasoned that if these chromatin responses are tumor-suppressive, then Ifi205 deficiency in Mre11^ATLD1/ATLD1 MECs should reduce tumor incidence. To test this hypothesis, WT, Mre11^ATLD1/ATLD1, and Ifi205^−/− Mre11^ATLD1/ATLD1 organoids were implanted into cleared mammary fat pads of NOD-SCID mice. Six weeks after implantation, 0.2 mg/mL doxycycline water was fed to the mice to induce neuT expression. Successful implantation could be verified by luciferase imaging after 1 week of doxycycline feeding, whereas tumors were visible as large luciferase-positive masses after oncogene activation (Fig. 7A). Macroscopic images and H&E staining confirmed the existence of the primary tumor and lung metastasis from Mre11^ATLD1/ATLD1 organoid implantations (Fig. 7B,C). Fifteen WT, 15 Mre11^ATLD1/ATLD1, and 17 Ifi205^−/− Mre11^ATLD1/ATLD1 organoid-implanted mice were followed up for 35 weeks after doxycycline induction. Tumor-free survival in the Ifi205^−/− Mre11^ATLD1/ATLD1 cohort after neuT activation was significantly longer than that in the Mre11^ATLD1/ATLD1 cohort (Ifi205^−/− Mre11^ATLD1/ATLD1/Mre11^ATLD1/ATLD1, HR = 0.1496, P = 0.0414; WT/Mre11^ATLD1/ATLD1, HR = 0.1146, P = 0.0159; log-rank test) (Fig. 7D). To include early-stage tumors that are not palpable, mammary tissues harvested at the end of the experiment were submitted for pathology analysis. Microscopic primary tumor incidence also decreased in Ifi205^−/− Mre11^ATLD1/ATLD1 implantations compared with Mre11^ATLD1/ATLD1 implantations (Mre11^ATLD1/ATLD1, 85.71%; Ifi205^−/− Mre11^ATLD1/ATLD1, 23.53%; P = 0.0010) (Fig. 7E). These data show that ablation of Ifi205 in Mre11^ATLD1/ATLD1 organoids suppresses oncogene-induced tumorigenesis in Mre11^ATLD1/ATLD1 organoid implantation.

Figure 7.

Knocking out Ifi205 in Mre11^ATLD1/ATLD1 organoids suppresses oncogene-induced mammary tumorigenesis. (A) Luciferase images of a Mre11^ATLD1/ATLD1 organoid-implanted mouse at 1 week (left) and 6.5 months (right) after being fed with 0.2 mg/mL doxycycline water. (B) Macroscopic primary tumor (left) and lung sample (right) from a Mre11^ATLD1/ATLD1 organoid-implanted mouse at 7 months after being fed with 0.2 mg/mL doxycycline water. (C) H&E staining of samples from B (arrows point to lung metastasis lesions). (D) Kaplan–Meier tumor-free survival curves of WT, Mre11^ATLD1/ATLD1, and Mre11^ATLD1/ATLD1 Ifi205^−/− organoid-implanted mice after being fed with 0.2 mg/mL doxycycline water (n = 15 WT, n = 15 Mre11^ATLD1/ATLD1, n = 17 Mre11^ATLD1/ATLD1 Ifi205^−/−; Mre11^ATLD1/ATLD1 Ifi205^−/− vs. Mre11^ATLD1/ATLD1, HR = 0.1496, P = 0.0445; WT vs. Mre11^ATLD1/ATLD1, HR = 0.1146, P = 0.0159; log-rank test). (E) Percentage of microscopic primary tumor incidences at the time of experiment termination (35 weeks after being fed with 0.2 mg/mL doxycycline water) in WT, Mre11^ATLD1/ATLD1, and Mre11^ATLD1/ATLD1 Ifi205^−/− organoid-implanted mice (85.71% Mre11^ATLD1/ATLD1, 23.53% Mre11^ATLD1/ATLD1 Ifi205^−/−, P = 0.0010, Fisher's exact test). (F) Schematic model. DNA sensors such as IFI205 are poised to detect aberrant DNA structures at the DNA replication fork. Under normal conditions, such structures are infrequently encountered during unperturbed DNA replication, leading to low levels of ISG transcription and nonpermissive chromatin states. Tumor-suppressive responses can be readily induced when an oncogene is activated in these normal mammary cells. Under stressed conditions such as Mre11 hypomorphism, constant engagement with aberrant DNA structures results in elevated ISG transcription, which in turn creates permissive chromatin states. As a result, these mammary cells lose the tumor-suppressive oncogenic responses and become more susceptible to oncogene-induced tumorigenesis.

Discussion

In this study, we used primary mammary organoids to examine the mechanisms that underlie the suppression of oncogene-induced breast cancer. Having previously established the importance of the MRE11 complex in suppressing neuT-induced breast cancer via in vivo analyses (Gupta et al. 2013), the organoid system afforded several advantages over the in vivo system. It provides sufficient material for genomic, biochemical, and chromatin analyses, and organoids could be manipulated ex vivo and subsequently reimplanted for in vivo development.

Activation of oncogene in organoids fully recapitulated previous in vivo observations with respect to both the indices of DDR activation and the deposition of heterochromatic marks. These outcomes were not seen in Mre11^ATLD1/ATLD1 organoids. Bulk RNA-seq analyses were carried out prior to oncogene activation and 14 and 28 days after oncogene induction and revealed substantial differences between WT and Mre11^ATLD1/ATLD1 organoids; the trajectories of 5644 genes differed between the two genotypes. Consistent with the fact that WT organoids exhibited indices of DDR activation and heterochromatin formation upon oncogene activation, transcription of DDR genes as well as chromatin remodeling and modifying genes was induced to a greater extent in WT than in Mre11^ATLD1/ATLD1 (Supplemental Fig. S10). In addition, the transcriptional profiles of genes involved in lipid metabolism and extracellular matrix components were sharply downregulated by oncogene induction in WT organoids, whereas those genes were induced in Mre11^ATLD1/ATLD1 (Fig. 2B). We speculate that this difference may partially account for the increased metastatic potential exhibited by Mre11^ATLD1/ATLD1 mammary tumors, but further analysis would be required to substantiate that interpretation.

As stated previously, the increased susceptibility of Mre11^ATLD1/ATLD1 could reflect the existence of a MRE11 complex-dependent response to oncogene activation or a state created by MRE11 complex hypomorphism in which the tumor-suppressive response to oncogene-induced biological changes is attenuated.

Certain aspects of the data presented here support the former explanation. For example, in WT organoids, genes governing the DDR as well as chromatin modifiers are induced. In 53BP1-deficient mice, neuT activation did not lead to extensive hyperplasia or tumor susceptibility (Gupta et al. 2013). This argues that the DNA repair pathway(s) influenced by 53BP1 foci are not strongly tumor-suppressive.

Indeed, the data more strongly support the latter idea—that MRE11 complex hypomorphism indirectly promotes neuT-driven breast cancer, possibly via alteration of the chromatin landscape. The chronic ISG transcriptional signature observed at baseline in Mre11^ATLD1/ATLD1 organoids is associated with global changes in chromatin accessibility. Those changes are dependent on IFI205, the inactivation of which largely restores the altered accessibility peaks in Mre11^ATLD1/ATLD1 to that of WT organoids (Fig. 4I). IFI205 deficiency in Mre11^ATLD1/ATLD1 also reverts the response to neuT activation with respect to the induction of heterochromatic markers and changes in chromatin accessibility (Fig. 5A,B). The accumulation of heterochromatic markers is reminiscent of oncogene-induced senescence (Zhang et al. 2005). However, these mammary organoids did not show strong β-Gal staining or irreversible arrest in response to oncogene activation. Nevertheless, these chromatin changes may represent a more tumor-suppressive cell state, and losing such responses may render the cells more susceptible to oncogene-induced tumorigenesis.

The data suggest that IFI205-dependent chronic ISG transcription in the Mre11^ATLD1/ATLD1 mammary epithelium creates a permissive chromatin environment for oncogene-induced tumorigenesis. Notably, the top 20 most significant enriched motifs in closed chromatin regions comparing Mre11^ATLD1/ATLD1 with WT organoids were associated with stemness and embryonic and early brain development functions. These same functions were evident in the top 20 enriched motifs that became open in Mre11^ATLD1/ATLD1 Ifi205^−/− compared with Mre11^ATLD1/ATLD1 organoids (Supplemental Fig. S7A–D). More specifically, there were six overlaps between these two top motif lists (PAX3, SEBOX, DDX1, TLX2, NOTO, and CDC5L), suggesting that transcription factors relevant to stemness and embryonic and early brain development were less accessible in Mre11^ATLD1/ATLD1 organoids but became more accessible following Ifi205 knockout.

In an attempt to investigate whether these motifs contribute to the permissive chromatin environment in Mre11^ATLD1/ATLD1 organoids, we used a previously reported strategy to identify genes with “unrealized potential” (Krausgruber et al. 2020). These genes are characterized by increased chromatin accessibility (calculated by ATAC-seq) but no corresponding increase in gene expression (calculated by RNA-seq) (Supplemental Fig. S11A). Given that the Mre11^ATLD1/ATLD1 mice do not develop spontaneous tumors in the absence of oncogenes, these genes with “unrealized potential” at baseline may underline the different oncogenic responses and varying susceptibilities to oncogene-induced breast tumorigenesis. We found 31 genes with more accessible chromatin but no increase in gene expression in WT compared with Mre11^ATLD1/ATLD1 organoids at baseline, and these genes were only induced in WT organoids after oncogene activation but not in Mre11^ATLD1/ATLD1 organoids (Supplemental Fig. S11B,C). Gene ontology (GO) and pathway analysis revealed that these genes were associated with stem cell, epithelial–mesenchymal transition (EMT), and Notch signaling (Supplemental Fig. S11D–F). This is consistent with the functions of the motifs that were less accessible in Mre11^ATLD1/ATLD1 organoids but became more accessible in Mre11^ATLD1/ATLD1 Ifi205^−/− organoids. We hypothesize that the oncogene-induced responses observed only in WT organoids may contribute to tumor suppression. In contrast, these chromatin regions become less accessible in Mre11^ATLD1/ATLD1 organoids, rendering these genes unresponsive to oncogene activation and more susceptible to oncogene-induced breast tumorigenesis. It also must be considered that the ISG transcriptional program involves myriad genes that either alone or in combination(s) could change other aspects of epithelial cell biology.

What is certain is that the effect is dependent on IFI205 DNA binding, which likely stems from the increased rate of spontaneous DNA damage ensuing from DNA replication in the context of MRE11 complex hypomorphism. Previous studies clearly indicate that the MRE11 complex is intimately associated with DNA replication and that functional decrement of the complex causes DNA replication stress and chromatid breakage (Maser et al. 2001; Mirzoeva and Petrini 2003; Smith and Savery 2005; Sirbu et al. 2013; Wardlaw and Petrini 2022). Moreover, IFI205 localizes to DNA replication forks in both unperturbed and stressed DNA replication conditions (Wardlaw and Petrini 2022); hence, it is temporally and spatially situated to surveil the replication fork. Consistent with this idea, using The Cancer Genome Atlas (TCGA) data sets, we found that gene amplification of IFI16, a member of the human HIN-200 family, in breast cancer patients with high genome instability was associated with inferior overall survival (Supplemental Fig. S12).

Consistent with our finding that the elevated ISG transcriptional program in Mre11^ATLD1/ATLD1 organoids at baseline is IFI205- rather than cGAS-dependent, a recent study reported that the MRE11 complex is indispensable for cGAS activation (Cho et al. 2024). Therefore, the cGAS–STING pathway, by definition, cannot induce ISG expression in Mre11^ATLD1/ATLD1 organoids. Their study showed that mammary cells with Mre11 hypomorphism failed to induce robust IFN signaling after oncogene Myc overexpression due to deficient cGAS activation (Cho et al. 2024), which can be verified in our mammary organoid system using the oncogene neuT (Fig. 4G). However, there are key differences between these two studies. Limited by the setting of their transgenic mice, Mre11 mutation, Trp53 deletion, and Myc overexpression had to be introduced simultaneously in their study, precluding them from capturing the baseline ISG signature that we observed. In addition, their study was done in a p53-deficient background, which inevitably constrained the focus to p53-independent functions of MRE11 and cGAS. In contrast, our study used nonimmortalized epithelia in 3D culture, where no oncogene (at baseline) and p53 deletion were present, thus allowing us to investigate broader aspects of oncogene-induced tumorigenesis. On the basis of our findings and their work, we believe that the cGAS–STING pathway plays an important role in oncogene-induced ISG transcription, whereas IFI205 becomes more influential when cytosolic dsDNA is low and nuclear DNA damage is predominant (as the baseline condition in the present study).

IFN was first discovered as a major defense against viral infections (Lindenmann et al. 1957). ISGs that encode cytotoxic proteins are induced in the acute phase of signaling in response to genotoxic stress such as that induced by treatment with anthracyclines or ionizing radiation (Sistigu et al. 2014; Widau et al. 2014). In contrast, chronic stimulation with low doses of IFN contributes to resistance to DNA damage and correlated with therapy resistance (Gaston et al. 2016). It is now acknowledged that the effects of IFN signaling on cancers are determined by the strength and duration of stimulation; whereas strong and acute IFN responses are cytotoxic, weak and chronic response promote cell survival (Cheon et al. 2023). In support of this conclusion, we found a stronger IFN signature in human breast cancer tissues compared with matched normal tissues, which is potentially associated with the protumor effects of ISG chronic activation, as observed in Mre11^ATLD1/ATLD1 organoids (Supplemental Fig. S13). It is also worth noting that the mammary organoid system consists of pure epithelia, where IFN-α, IFN-β, and IFN-γ are undetectable. Consistent with this, no change in phosphorylated STAT1 was detected without exogenous stimulation (Supplemental Fig. S14). Given that cancer cell-derived ISGs have been reported to contribute to resistance to immune checkpoint blockade and poor outcome (Benci et al. 2019; Espinet et al. 2021), the organoid system and the novel findings in the present study provide a unique opportunity to further investigate the effects of epithelia-derived ISGs on the tumor microenvironment by manipulating the upstream sensor.

Recent findings have illuminated a role for innate immune signaling in maintaining genomic integrity through the combined activities of nuclear DNA sensing and the induction of ISG15, which we showed is required for replication for stability and for the recruitment of factors that promote replisome function (Boj et al. 2015; Dunphy et al. 2018; Wardlaw and Petrini 2022). On the basis of those observations and the data presented in this study, we propose a model in which innate immune sensors and effectors surveil the DNA replication fork under normal growth conditions. The model posits that DNA sensors such as HIN-200 family members are poised to detect aberrant DNA structures at the DNA replication fork that are likely to exist transiently but frequently during unperturbed DNA replication. Under stressed conditions (DDR mutation or high genome instability), engagement of such structures leads to activation of an ISG transcriptional program that culminates in the modulation of chromatin states and the creation of a “permissive” cell state. As a result, the cells lose the tumor-suppressive oncogenic responses and become more susceptible to oncogene-induced tumorigenesis (Fig. 7F).

In this regard, chronic activation of innate immune pathways may contribute to the cancer predisposition and clinical sequalae of many DDR-related mutations as a result of ISG transcriptional activity stimulated by aberrant DNA structures. Further characterization of chromatin changes associated with chronic ISG transcription may provide information regarding tumor initiation and suggest novel treatment modalities.

Materials and methods

Animal experiments

All animal studies were done in compliance with a protocol approved by the Institutional Animal Care and Use Committee of Memorial Sloan Kettering Cancer Center. CAG-rtTA transgenic mice and TetO-NeuT transgenic mice were kindly provided by Scott Lowe (Memorial Sloan Kettering Cancer Center) and Lewis Chodosh (University of Pennsylvania), respectively (Premsrirut et al. 2011; Yeh et al. 2011). Mre11^ATLD1/ATLD1 mice have been previously described (Theunissen et al. 2003). cGas^−/− mice were kindly provided by Liang Deng (Memorial Sloan Kettering Cancer Center) (Yang et al. 2023). Mice were raised in a pathogen-free facility and genotyped by PCR.

Cleared mammary fat pad organoid implantation was done in 3–4 week old female NOD-SCID mice. Mammary fat pad clearance procedure was performed as previously described (Lawson et al. 2015). Briefly, the bridge between the fourth and fifth mammary glands was cut, and the region of the mammary fat pad between the nipple and just after the proximal lymph node was removed. A 1 million single-cells/Matrigel (1:1) mixture from the desired genotype was injected into the remaining portion of the epithelium-free mammary fat pad. The skin incisions were closed by wound autoclips.

Luciferase imaging was performed 6 weeks after organoid implantation. Organoid-implanted mice were fed with 0.2 mg/mL doxycycline water for 1 week before imaging. The image was taken using an Ivis optical imaging system 10 min after D-luciferin IP injection.

Mouse cohorts after successful implantation were palpated for the development of mammary tumors twice per week. Mice were euthanized using humane experimental endpoints or at the termination of experiments. At necropsy, mammary tissues and lung tissues were harvested and fixed in 4% paraformaldehyde. Samples were processed for paraffin embedding, H&E staining, and pathology analysis (HistoWiz).

Primary mammary organoid culture

The fourth mammary glands were harvested from 6–8 week old female mice from the desired genotype, incubated in collagenase/hyaluronidase medium (Stem Cell Technologies 07912), and shaken for 2 h at 37°C. The resulting digestion was spun down and washed twice with D10F medium (DMEM, 10% FBS). Cells were washed four more times with DMEM. Cells were resuspended in 3 mL of trypsin and incubated for 10 min at 37°C. D10F was added to neutralize the trypsin. Cells were spun down, resuspended in 10 U of dispase (Stem Cell Technologies 07923) and 1000 U of DNase (Worthington), and incubated for 10 min at 37°C. D10F was added and cells were passed through a 40 µm filter. Cells were spun down, embedded in Matrigel (Fisher Scientific CB40230C), and overlaid with growth mammary organoid medium. For maintenance, organoids were cultured in growth mammary organoid medium (basal mammary organoid medium plus 50 ng/mL EGF [Peprotech 315-09]) and passaged on a weekly basis using trituration with a fire-polished glass Pasteur pipette or by trypsinization. Basal-state organoids were cultured in basal mammary organoid medium for 1 day, and oncogene-activated organoids were cultured in doxycycline mammary organoid medium (basal mammary organoid medium plus 1 μg/mL doxycycline [MilliporeSigma D3072]) for the indicated time. A full list of ingredients and concentrations is in Supplemental Table S3.

Immunofluorescence

Organoids were washed twice with PBS and incubated with cell recovery solution (Fisher Scientific 354253) for 1 h on ice. Organoids were fixed with 4% PFA for 1 h at 4°C and then permeabilized using 1% Triton X-100 in PBS for 1 h at room temperature. Staining was performed in 0.3% Triton X-100 and 1% BSA in PBS overnight at 4°C with gentle shaking. The antibodies used are listed in Supplemental Table S4. Stained organoids were imaged using an SP8 confocal microscope.

RNA extraction and quantitative PCR (qPCR)

Organoids were isolated from the Matrigel by multiple washes with ice-cold PBS. Total RNA was extracted from the specified organoid using the RNeasy mini kit (Qiagen 74104) with DNase treatment. cDNA was synthesized using RNA to cDNA EcoDry premix (Takara 639547). qPCR was performed using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad 1725272) on a CFX384 real-time system (Bio-Rad). Target gene expression was normalized to Gapdh, and relative expression was determined with the comparative CT method. The primers used are listed in Supplemental Table S5.

Whole-cell extract Western blot

Organoids were isolated from the Matrigel by multiple washes with ice-cold PBS. Cells were lysed in RIPA buffer containing protease inhibitors (Sigma 11873580001) and phosphatase inhibitors (Thermo Fisher Scientific 78420) on ice and sonicated 10 times for 30 sec at 30 sec intervals using a Bioruptor. Protein concentrations were quantified using a bicinchoninic acid (BCA) assay (Thermo Fisher 23227). Lysates were denatured using 5× protein loading dye (300 mM Tris-HCl at pH 6.8, 10% SDS, 0.5% bromophenol blue, 50% glycerol, 500 mM 2-mercaptoethanol). Extract (20–30 μg) was loaded into gradient SDS-PAGE gels (Bio-Rad) and transferred to nitrocellulose membranes (Amersham Proton 0.2 μM NC 10600006). The membranes were blocked in 5% skimmed milk and PBST and probed with the antibodies listed in Supplemental Table S4. Membranes were developed using either ECL or ECL Prime (Amersham RPN2106/RPN2232).

CRISPR–CAS9 knockout

Gene knockout organoids were generated using CRISPR–CAS9-mediated genome editing as described previously (Ran et al. 2013). Briefly, the Ifi205 guide sequence (TGAAGCCGAAGATGAGACCT), Stat1 guide sequence (GGTCGCAAACGAGACATCAT), or Sting guide sequence (CAGTAGTCCAAGTTCGTGCG) was cloned into PX458 (Addgene 48138) using BbsI restriction sites. The plasmid was nucleofected (Nucleofector II, Amaxa) into organoids and sorted for GFP-positive cells 72 h later. A second round of nucleofection was performed and sorted for GFP-positive cells again. Cells were genotyped via sequencing and ICE analysis (https://www.synthego.com).

Metaphase spreads

Organoids were treated with 100 ng/mL KaryoMAX colcemid (Life Technologies 15212012) for 24 h and harvested. Organoids were washed with PBS to remove Matrigel and made into single-cell suspension by trypsinization. Cells were swelled in 0.075 M KCl for 15 min at 37°C and fixed in ice-cold 3:1 (v/v) methanol:acetic acid overnight at −20°C. Samples were dropped onto slides and stained with 5% giemsa (Sigma) for 5 min at room temperature. Following three washes in dH₂O, slides were dried for several hours at room temperature and mounted with Permount medium (Fisher Scientific). Metaphases were imaged on an Olympus IX50 microscope with an Infinity 3 camera (Lumenera) using a 100× objective.

Recombinant protein expression and purification

Mutagenesis for Ifi205^K364ER299E was performed in pCMV6-Ifi205 (Origene plasmid MR206355) using QuikChange Lightning multisite-directed mutagenesis kit (Agilent 200515). Primers used for mutagenesis are listed in Supplemental Table S5. Sequences encoding Ifi205 ORFs and Ifi205^K364ER299E ORFs were amplified by PCR and then subcloned between the BamHI and EcoRI restriction sites of pGEX-6P-1(Addgene). IFI205 proteins were expressed as GST-IFI205-FLAG fusions in the E. coli strain BL21 (DE). Proteins were purified by glutathione sepharose (GE Healthcare GE17-0756-01) and cleaved by HRV-3C protease (MilliporeSigma SAE0045) to remove the GST tag. The resulting samples were injected into Superdex 200 columns (GE Healthcare) for size exclusion chromatography. IFI205 protein-corresponding fractions were combined and purified by anti-FLAG M2 affinity gel (Sigma A2220). A small portion of the final proteins was run through the Superdex 200 columns (GE Healthcare) again for a quality check.

Binding assay

Oligonucleotides were refolded in 50 mM Tris (pH 7.5), 150 mM NaCl, and 1 mM EDTA with slow cooling starting at 95°C down to room temperature. The change in fluorescence anisotropy of the 60mer 5′ FAM-labeled oligonucleotides was measured to determine the relative binding affinities of IFI205 proteins. The labeled oligonucleotides (50 nM) were mixed in 10 μL of reaction volume with the indicated IFI205 proteins at concentrations ranging from 0 to 750 nM in a reaction buffer containing 20 mM Tris (pH 7.5), 50 mM KCl, 0.5 mM TCEP, 10% glycerol, and 0.1% IGEPAL in a 384 well microplate. The binding data were collected on a SpectraMax M5 (Molecular Devices) using a 495 nm excitation wavelength and a 525 nm emission wavelength. Apparent K_D values were calculated from triplicate experiments using a model for receptor depletion and plotted using Prism 7 GraphPad software. The model used was Y = A_f + (A_b − A_f) × {(L + K_D + X) − sqrt[(sqr(− L − K_D − X)] − 4 × L × X)}/(2 × L), where Y is the anisotropy measured, A_b is the anisotropy at saturation (100%), A_f is the anisotropy of the free oligonucleotide, L is the fixed concentration of the oligonucleotide, and X is the protein concentration.

Electrophoretic mobility shift assay (EMSA)

Oligonucleotides were refolded in 50 mM Tris (pH 7.5), 150 mM NaCl, and 1 mM EDTA with slow cooling starting at 95°C down to room temperature. Binding reactions were performed in buffer containing 20 mM Tris (pH 7.5), 50 mM KCl, 0.5 mM TCEP, 10% glycerol, and 0.1% IGEPAL. IFI205 proteins and 60mer 5′ FAM-labeled oligonucleotide at the indicated concentrations were incubated for 30 min at room temperature and then loaded onto a 4% polyacrylamide nondenaturing gel. The gels were imaged using a Typhoon FLA9500 instrument (GE).

IFI205 protein expression in organoids

Sequences encoding Ifi205 ORFs and Ifi205^K364ER299E ORFs were amplified by PCR and then subcloned between the XbaI and BamHI restriction sites of pHFUW-IRES-EGFP. Lentiviruses expressing pHFUW-IRES-EGFP were produced in HEK293T cells using packaging vectors psPAX2 and pVSV-G. Viral supernatant was concentrated using PEG-it virus precipitation solution (SBI LV810A-1). Spinoculation was done at 600g for 1 h at 32°C. An organoid–virus mixture was incubated for an additional 6 h at 37°C in a culture incubator and refreshed with growth mammary organoid medium. Two weeks after infection, organoids with different IFI205 proteins were sorted for a similar intensity of GFP.

cGAMP quantification

Organoids were washed with PBS to remove Matrigel and made into single-cell suspension by trypsinization. Cells were thoroughly resuspended in 120 μL of lysis buffer (20 mM Tris-HCl at pH 7.7, 100 mM NaCl, 10 mM NaF, 20 mM β-glycerophosphate, 5 mM MgCl₂, 0.1% Triton X-100, 5% glycerol) and lysed with a 28.5 gauge needle. Lysates were incubated for 30 min on ice and then centrifuged at 16,000g for 10 min at 4°C, and cGAMP levels were quantified using the 2′3′cGAMP ELISA kit (Arbor Assays NC1595685) according to the manufacturer's instructions.

Quantification of cytoplasmic dsDNA

Organoids were washed with PBS to remove Matrigel and made into single-cell suspension by trypsinization. Cytoplasmic extracts from live cells were isolated using NE-PER nuclear and cytoplasmic extraction reagents (Thermo Fisher Scientific 78833). dsDNA was quantified in cytoplasmic fraction using the SpectraMax Quant AccuClear Nano dsDNA assay kit (Molecular Devices R8357). The samples were read using a SpectraMax M5 (Molecular Devices) plater reader.

Bulk RNA-seq analysis

Organoids were washed with PBS to remove Matrigel, and pelleted cells were frozen until the process was started. Frozen cells were lysed in 1 mL of TRIzol reagent (Thermo Fisher 15596018), and phase separation was induced with 200 µL of chloroform. RNA was extracted from 350 µL of the aqueous phase using the miRNeasy mini kit (Qiagen 217004) on the QIAcube Connect (Qiagen) according to the manufacturer's protocol. Samples were eluted in 34 µL of RNase-free water. After RiboGreen quantification and quality control by Agilent BioAnalyzer, 500 ng of total RNA with RIN values of 7.2–10 underwent poly(A) selection and TruSeq library preparation according to the instructions provided by Illumina (TruSeq stranded mRNA LT kit RS-122-2102) with eight cycles of PCR. Samples were barcoded and run on a HiSeq 4000 in a PE50 run using the HiSeq 3000/4000 SBS kit (Illumina). An average of 61 million paired reads was generated per sample, and the percent of mRNA bases averaged 79%. Fifty base pair paired-end reads were aligned to the mm10 mouse reference genome using STATaligner. Quantification of genes annotated in Gencode vM2 was performed using FeatureCounts. Batch effects were corrected using RUVSeq if samples were submitted at multiple times. Differential expression analysis was done using the DESeq2 package. Differentially expressed genes were defined if they satisfied an adjusted P-value of <0.01 and if the magnitude of fold change was >1.5. Canonical pathway analysis was performed by Qiagen ingenuity pathway analysis (IPA). GSEA was performed by the fgsea package in R. GO analysis and KEGG pathway analysis were performed by the GOstats package in R. Sequencing data are available at Gene Expression Omnibus (GSE246409).

Bulk ATAC-seq analysis

Organoids were washed with PBS to remove Matrigel and made into single-cell suspension by trypsinization. Fifty-thousand freshly sorted live cells (DAPI-negative) from each sample were used for ATAC-seq as previously described (Buenrostro et al. 2013). Briefly, fresh cells were washed in cold PBS and lysed. The transposition reaction containing TDE1 Tagment DNA enzyme (Illumina 20034198) was incubated for 30 min at 37°C. The DNA was cleaned with the MinElute PCR purification kit (Qiagen 28004), and material was amplified for five cycles using NEBNext high-fidelity 2× PCR Master Mix (New England Biolabs M0541L). After evaluation by real-time PCR, eight to 14 additional PCR cycles were done. The final product was cleaned by AMPure XP beads (Beckman Coulter A63882) at a 1× ratio, and size selection was performed at a 0.5× ratio. Libraries were sequenced on a HiSeq 4000 or NovaSeq 6000 in a PE50 or PE100 run using the HiSeq 3000/4000 SBS kit or NovaSeq 6000 S2 or S4 reagent kit (200 cycles; Illumina). An average of 64 million paired reads was generated per sample. ATAC-seq data processing was performed as previously described (Adams et al. 2019). In brief, paired-end reads were aligned to the mm10 genome using Bowtie2. Mapped fragments were shifted for Tn5 tagmentation and used for peak calling with MACS2. Peaks that were not reproducible (IDR < 0.01) and that overlapped the ENCODE mm10 functional genomics region blacklist were discarded to improve quality. Reproducible peaks from each sample were combined to create a genome-wide atlas of accessible chromatin regions. Reads aligned to the atlas peak regions were counted using Bedtools. Batch effects were corrected using RUVSeq if samples were submitted at multiple times. Differential accessibility of the peaks was assessed by DESeq2 to the count table. Peaks were defined as differentially accessible if they satisfied an adjusted P-value of <0.05 and if the magnitude of fold change was >1.5. The heat maps of differentially accessible peaks were generated using Deeptools. Using the CIS-BP transcription factor binding motif (TFBM) reference, the peak atlas was scanned with FIMO to identify peaks likely containing each TFBM (P < 10⁻⁴). Relative transcription factor accessibility was determined using a two one-sided Kolmogorov–Smirnov (K-S) test (B-H-adjusted P-value of <0.01) comparing the distributions of the log₂ fold changes from differential peak analyses across various conditions. Sequencing data are available at Gene Expression Omnibus (GSE246409).

Single-cell RNA-seq analysis

Organoids were washed with PBS to remove Matrigel and made into single-cell suspension by trypsinization. Cells were incubated with Fc receptor blocker (BD 553142) and multiplexed with TotalSeq-A (BioLegend). Freshly sorted live single cells (DAPI-negative) from each sample were used for single-cell RNA-seq on a Chromium instrument (10X Genomics) following the user guide manual for 3′ v3.1. In brief, FACS-sorted cells were washed once with PBS containing 1% bovine serum albumin (BSA) and resuspended in PBS containing 1% BSA to a final concentration of 700–1300 cells/μL. The viability of cells was >80%, as confirmed with 0.2% (w/v) Trypan Blue staining (Countess II). Cells were captured in droplets. Following reverse transcription and cell barcoding in droplets, emulsions were broken and cDNA was purified using Dynabeads MyOne Silane followed by PCR amplification per the manual's instructions. Between 10,000 and 20,000 cells were targeted for each sample. Samples were multiplexed together on one lane of 10X Chromium (using hash tag oligonucleotides [HTOs]) according to a previously published protocol (Stoeckius et al. 2018). Final libraries were sequenced on an Illumina NovaSeq S4 platform (R1: 28 cycles, i7: eight cycles, and R2: 90 cycles). The cell–gene count matrix was constructed using the sequence quality control (SEQC) package (Stoeckius et al. 2018). Viable cells were identified on the basis of library size and complexity, whereas cells with >20% of transcripts derived from mitochondria were excluded from further analysis. UMAP and PHATE were performed using Scanpy in Python. Lineage signature scores were generated using signature genes previously described (Nguyen et al. 2018). ISG signature scores were generated using genes from GO terms (response to interferon-α, response to interferon-β, and response to interferon-γ) (a full gene list is in Supplemental Table S6). Chromatin modifiers and remodelers used in PHATE are listed in Supplemental Table S7. Sequencing data are available at Gene Expression Omnibus (GSE246409).

TCGA data analysis

Mutation counts, IFI16 amplification status, and survival metrics related to breast invasive carcinoma (TCGA; PanCancer Atlas) were retrieved from cBioPortal (https://www.cbioportal.org). Patients with mutation counts exceeding either 130 or 140 were categorized into subgroups based on the absence and presence of IFI16 amplification. Survival analysis and log-rank tests were carried out with the survival package in the R environment.

Transcriptome profiles and clinical data related to breast-invasive carcinoma were retrieved from TCGA using the TCGAbiolinks package in R. DESeq2 was used to generate the IFN signature score (human gene set: REACTOME_INTERFERON_SIGNALING) in primary tumor samples and matched normal samples. A paired t-test was performed to compare the difference between the two groups.

Competing interest statement

The authors declare no competing interests.