Chromatin Remodeling in Response to BRCA2-Crisis

SUMMARY

Individuals with a single functional copy of the BRCA2 tumor suppressor have elevated risks for breast, ovarian, and other solid tumor malignancies. The exact mechanisms of carcinogenesis due to BRCA2 haploinsufficiency remain unclear, but one possibility is that at-risk cells are subject to acute periods of decreased BRCA2 availability and function (“BRCA2-crisis”), which may contribute to disease. Here, we establish an in vitro model for BRCA2-crisis that demonstrates chromatin remodeling and activation of an NF-κB survival pathway in response to transient BRCA2 depletion. Mechanistically, we identify BRCA2 chromatin binding, histone acetylation, and associated transcriptional activity as critical determinants of the epigenetic response to BRCA2-crisis. These chromatin alterations are reflected in transcriptional profiles of pre-malignant tissues from BRCA2 carriers and, therefore, may reflect natural steps in human disease. By modeling BRCA2-crisis in vitro, we have derived insights into pre-neoplastic molecular alterations that may enhance the development of preventative therapies.

Graphical Abstract

In Brief

Gruber et al. identify EGF-independent proliferation in mammary cells following transient BRCA2 depletion (“BRCA2-crisis”). In the absence of recurrent genomic mutations, chromatin remodeling mediated by NF-κB signaling drives gene expression changes. This in vitro model shares molecular commonalities observed in BRCA2^+/− carriers.

INTRODUCTION

Patients with germline heterozygous mutations in BRCA2 are predisposed to develop breast and ovarian cancers at a higher rate compared to the general population (Mavaddat et al., 2013). Tumors arising in these carriers often exhibit loss of heterozygosity (LOH) at the second BRCA2 allele, resulting in defective homologous recombination DNA repair pathways (HRRs) (Lord and Ashworth, 2016; Venkitaraman, 2014). BRCA2-mutant-driven breast tumors are enriched for the “luminal B” intrinsic molecular subtype (van der Groep et al., 2011) and frequently exhibit increased genomic instability, leading to high mutational burdens.

It remains unclear whether genomic instability is a direct consequence of BRCA2-mutation-associated defects in HRRs, especially as BRCA2 LOH often arises concomitant with p53 loss. Chronic reliance on the alternative and error-prone non-homologous end joining (NHEJ) pathway is posited to cause an accumulation of somatic mutations as “early events” that ultimately drive tumor initiation (Deng, 2001). Conversely, one functional BRCA2 allele may be sufficient for cellular maintenance, but rare circumstances may cause a transient impairment of BRCA2 function, during which time a high-fidelity DNA-damage response cannot be fully supported. We term this acute incident of BRCA2 deficiency “BRCA2-crisis.” For example, it has recently been demonstrated that exposure to environmental aldehydes can transiently impair BRCA2 function in cells (Tacconi et al., 2017; Tan et al., 2017). Repeated periods of BRCA2-crisis could subsequently lead to pre-malignant changes and, ultimately, tumor initiation, with genomic instability occurring as a by-product of transformation and as a “late event.”

These competing theories underscore the need for a better understanding of the earliest genetic and epigenetic changes that occur at the onset of BRCA2-deficient cancer. Prophylactic mastectomy remains the most effective preventative therapy for breast cancer in women with BRCA2 mutations. Since BRCA2 carriers have up to a 65% risk of developing cancer by the age of 70 (Nielsen et al., 2016), development of additional non-invasive alternatives remain important. Paradoxically, despite the frequent LOH by genetic or epigenetic means in tumors, modeling this event remains challenging, due to the embryonic lethality observed in knockout mice (Evers and Jonkers, 2006). Permanent ablation of BRCA2 expression at either the transcriptional or the post-transcriptional level through CRISPR or short hairpin RNA (shRNA) modalities in non-transformed cell lines leads to growth arrest (Feng and Jasin, 2017), rendering these in vitro models less informative with regard to malignant transformation.

Here, we use transient depletion of BRCA2 expression (“BRCA2-crisis”) in non-transformed human breast epithelial cell lines to study early changes associated with malignant transformation. In contrast to models using heterozygous cell lines, this system incurs a primary BRCA2-crisis event in naive cells and provides a unique ability to separate epigenetic changes from genomic mutations. Specifically in BRCA2-depleted conditions, after allowing for the cells to recover and re-establish BRCA2 expression, we discover the establishment of different epigenetic patterns, primarily in the form of chromatin states associated with histone acetylation, in the absence of recurrent genomic mutations. These chromatin changes are associated with nuclear factor κB (NF-κB) signaling that drives proliferation in non-permissive EGF-free conditions, which may model an early step in carcinogenesis. These findings identify a pathway for potential therapeutic intervention to prevent cancer onset and provide a platform to further investigate molecular changes that occur in a preneoplastic setting.

RESULTS

Transient BRCA2 Depletion Causes EGF-Independent Proliferation

We designed an assay to identify early phenotypic alterations caused by an acute cellular deficiency of BRCA2 that we have termed BRCA2-crisis (Figure 1A). Short-term depletion of BRCA2 was followed by a recovery period to allow BRCA2 protein levels to return to normal. This approach was necessary, in part, because permanent ablation of HR factors is not well tolerated in non-transformed cell lines (Feng and Jasin, 2017). Since an established hallmark of cancer is sustained proliferative signaling in the setting of anti-proliferative stimuli (Hanahan and Weinberg, 2011), after BRCA2 levels returned to baseline, selective growth conditions were used to assay for phenotypic alterations induced by BRCA2-crisis (Figure 1A). Non-transformed mammary cells can be propagated in serum-free media supplemented with hydrocortisone, insulin, epithelial growth factor (EGF), and pituitary extract. Removal of any of these factors from cell culture media impairs cellular proliferation, including the growth of non-transformed breast epithelial MCF10A cells (Figure 1B).

Figure 1. Transient BRCA2-Crisis Leads to a Growth Advantage in Non-permissive EGF-free Culture Conditions.

(A) Experimental schema for establishing BRCA2-crisis in vitro: siRNA transfection to human mammary cells leads to protein knockdown, which is sensed by cells and leads to an epigenetic response. When siRNA effects abate, the cells recover from stress, but some epigenetic changes may persist. These changes are then tested by growth assays.

(B) MCF10A cells grown in media lacking the individual factors hydrocortisone, insulin, EGF, or pituitary extract suffer a growth impairment compared to cells grown in complete media. Cell numbers (mean ± SD; n = 3) were measured after 3 days.

(C) MCF10A cells were transfected with siRNAs directed to BRCA1, BRCA2, and a non-target control and then serially passaged. Immunoblots indicate that protein levels return to baseline by passage 4.

(D) MCF10A cells surviving BRCA-crisis with recovered protein levels (after passage 4) exhibit similar growth kinetics in complete (including EGF) media. Data represent mean ± SD; n = 3.

(E) MCF10A cells surviving BRCA2-crisis with recovered protein levels (after passage 4) exhibit a growth advantage in EGF-free media compared to control cells and BRCA1-crisis cells. Data represent mean ± SD; n = 3.

(F) Immunoblots of MCF10A cells transfected with different siRNAs targeting BRCA2 and a non-targeting control siRNA confirm that BRCA2 levels recover to baseline by passage 4.

(G) BRCA2-crisis driven by different siRNAs targeting distinct regions of BRCA2 confirm that transient depletion of BRCA2, as opposed to off-target siRNA effects, leads to a growth advantage in EGF-free media compared to control. Data represent mean ± SD; n = 3.

Transient transfection of small interfering RNA (siRNA) targeted to BRCA1 and BRCA2 in MCF10A cells (which are wild type for BRCA1 and BRCA2) led to a reversible depletion of protein levels, as assessed by immunoblot, with normal levels restored after 4 passages, compared to a non-targeting control siRNA transfection (Figure 1C). MCF10A cells surviving BRCA2-crisis exhibited similar growth kinetics in complete media that contained all four required growth factors compared to control-treated cells (Figure 1D). However, when the same experiment was performed in media lacking EGF, cells surviving BRCA2-crisis had a growth advantage compared to cells transfected with BRCA1-targeting or non-targeting control siRNAs, both of which had limited proliferation (Figure 1E). In order to rule out siRNA off-target effects, this process was repeated with two independent siRNAs, each targeting different BRCA2 transcript regions. These conditions recapitulated the growth advantage in EGF-free media compared to control siRNA transfection (Figures 1F and 1G).

siRNAs directed to BRCA1 or BRCA2 in an independent cell line, telomerase-immortalized human mammary epithelial (hTert-HME1; also BRCA1 and BRCA2 wild type) cells, also recapitulated the effects seen in MCF10A, as BRCA2-crisis led to increased growth in EGF-free media compared to control-treated cells (Figures S1A and S1B). Similar to what we observed in MCF10A cells, hTert-HME1 cells that survived BRCA1-crisis had no proliferative advantage over control siRNA in EGF-free media (Figures S1C and S1D), suggesting that EGF-independent growth is specific to BRCA2-crisis. Thus, we focused on BRCA2-crisis and used several modalities to verify these growth effects in vitro. Modeling BRCA2-crisis by doxycycline (dox)-induced, temporally constrained transcriptional repression of BRCA2 using CRISPR interference in hTert-HME1 cells also led to EGF-independent cell proliferation compared to mock-treated cells (Figures S1E and S1F). Similarly, EGF-independent growth advantages were observed when exposing primary human mammary epithelial cells (HMECs) to BRCA2-crisis using dox-inducible shRNA (Figures S1G and S1H). Thus, transient depletion of BRCA2 at transcriptional and post-transcriptional levels enables proliferative advantages in EGF-free media in both immortalized and primary mammary cells.

BRCA2-Crisis Leads to Changes in Chromatin Accessibility without Recurrent Genomic Mutations

We used a multi-omics strategy to explore the molecular changes associated with BRCA2-crisis and the resulting growth advantage in EGF-free media. Whole-genome sequencing (WGS; mean coverage, 34.45 ± 2.3X) was performed at passages 3 and 7 after transfection (9 and 21 days post-transfection, respectively) on hTert-HME1 cells from 13 independent media and siRNA conditions (Figure S2A) to determine whether genomic mutations occurred as an early event to facilitate the observed growth advantage in non-permissive conditions. These conditions included two independent siRNAs targeting different BRCA2 regions and two non-targeting control siRNAs. There was no significant enrichment for single-nucleotide variants, insertions or deletions (indels), or structural variants between siRNA or media conditions (Figure S2A). Similarly, there was no significant skew in single-nucleotide variation (SNV) transition or transversion ratios between cells treated with siRNAs to BRCA2 versus control (Figures S2B and S2C), indicating that the BRCA2-crisis response is not likely a result of selection by de novo driver mutations.

In the absence of recurrent driver mutations, we sought to determine whether BRCA2-crisis could induce epigenetic changes. Targeted bisulfite sequencing of all promoter-associated CpG islands and ATAC-seq (Assay for Transposase-Accessible Chromatin (Buenrostro et al., 2013) were used to assess for changes in genome-wide DNA methylation and chromatin accessibility, respectively, in cells transfected with control or BRCA2 siRNAs and grown in either complete or EGF-free media. The profiled cells surviving BRCA2-crisis had recovered baseline BRCA2 expression and had been passaged 7 times, accounting for 21 total days post-transfection at the time of collection for analysis (12 days in EGF-free media). DNA methylation did not discriminate between siRNA or media conditions by principal-component analysis (PCA), with only 10% variance explained by PC1 (Figure S2D). In addition, no high-confidence differentially methylated regions were identified between the experimental conditions. In contrast, PCA of chromatin accessibility by ATAC-seq (Figure 2A) robustly separated siRNA and media effects, with 68% of variance explained in PC1. Overall, we identified 15,747 differentially accessible regions (false discovery rate [FDR] < 0.01) between BRCA2-depleted and control cells after two passages in EGF-free medium.

Figure 2. Chromatin Accessibility Changes Are Associated with Different Transcriptome Profiles based on siRNA Treatment and Media Conditions.

(A) ATAC-seq PCA of signal track data for merged peak regions across all datasets clusters cells grown in EGF-free conditions separately from complete media. Within EGF-free conditions, PCA is able to segregate BRCA2-crisis cells from control cells. 68% of the variance is explained in PC1.

(B) RNA-seq PCA analysis of counts per gene also clusters cells grown in EGF-free conditions separately from complete media. BRCA2-crisis cells grown in EGF-free conditions cluster separately from control cells. 86% of the variance is explained in PC1.

These changes in chromatin accessibility were associated with different gene expression profiles in BRCA2-crisis cells grown in EGF-free conditions. PCA of RNA-sequencing (RNA-seq) data robustly separated siRNA and media effects, with 86% of the variance explained in principle component 1 (PC1), and cells grown in complete media clustering together regardless of siRNA treatment (Figure 2B). Overall, we detected 3,435 differentially expressed genes (FDR < 0.01) between BRCA2-crisis cells grown in EGF-free conditions compared to complete medium. Although we did not observe any significant changes in EGFR expression levels across any of these conditions, ERBB2 expression increased similarly between control and BRCA2-crisis cells upon transition to EGF-free media (data not shown). Therefore, the observed growth effects were unlikely to be attributed to alterations in EGFR or ERBB2 levels.

BRCA2 Directly Regulates a Transcriptional Program through Chromatin Association

Given the robust differences in transcriptional profiles between groups, we explored the possibility that BRCA2 could have a direct epigenetic effect. Chromatin immunoprecipitation followed by high-throughput sequencing (ChIP-seq) using BRCA2-specific antibodies was performed on hTert-HME1 cells grown in complete media. This identified 15,158 significant peaks (FDR < 0.05; Figure 3A), indicating that BRCA2 possesses an intrinsic ability to bind DNA in addition to its well-described interactions with PALB2 and RAD51C in the DNA damage response (Roy et al., 2011). At genes bound by BRCA2, we found that binding sites were enriched near transcriptional start sites (TSS; Figure 3B). Gene set enrichment analysis (GSEA) identified a number of gene sets associated with BRCA2 binding, including those associated with well-known BRCA2-related functions including cell cycle and response to DNA-damaging stimuli (Figure 3C). However, a number of transcription factor and signal transduction pathways were also recovered, indicating that there may be a direct gene regulatory role for BRCA2 outside of its canonical DNA repair function. Indeed, we identified 380 differentially expressed transcripts 6 days after siRNA knockdown of BRCA2 in complete media; 46.3% of these genes (n = 176) were also directly bound by BRCA2 (Figures 3D and 3E). These data suggest that BRCA2 may regulate transcriptional activity through chromatin association.

Figure 3. BRCA2 ChIP-Seq Identifies Promoter-Proximal Binding and Enrichment of Signal-Transduction Gene Sets.

(A) ChIP-seq peaks enriched by BRCA2 antibodies were identified based on differential binding compared to input control libraries. Pink indicates FDR < 0.05.

(B) Gene body plots from TSS to transcription end site (TES), with 2,000 bp of flanking sequence for BRCA2-bound genes indicate preferential TSS binding.

(C) Gene sets enriched in BRCA2-bound genes. All gene sets are significant to FDR < 0.01.

(D) RNA-seq was performed 6 days after treatment with control or BRCA2 siRNAs (n = 4 in each group), and differentially expressed transcripts were identified and intersected with BRCA2-bound genes identified by ChIP-seq. Significance testing was performed with the hypergeometric test.

(E) Plot of the intersection of 176 BRCA2-dependent transcripts by RNA-seq with BRCA2-binding enrichment by ChIP-seq.

Next, we sought to identify a molecular mechanism by which BRCA2 promoter binding could alter gene expression patterns. BRCA2 has been shown to co-immunoprecipitate with acetyl-transferase activities toward histones H3 and H4 and the BubR1 spindle assembly checkpoint molecule (Choi et al., 2012; Siddique et al., 1998). Histone acetylation is a major contributing factor to chromatin states, including accessibility, and could help explain the distinct chromatin accessibility states we observed by ATAC-seq. Thus, we profiled individual histone acetylation marks in primary HMECs and hTert-HME1 cells subject to BRCA2-crisis by immunoblot. In control cells, transition to EGF-free culture conditions resulted in a decrease in acetylation at both lysines 5 and 12 on histone 4 (H4K5Ac and H4K12Ac); however, this effect was abrogated in BRCA2-crisis cells (Figures 4A and S3). We did not observe any differences in acetylation levels across all conditions on a sampling of other acetylation sites, including lysine 8 on H4, lysine 27 on H3, or lysines 12 or 120 on H2B.

Figure 4. Altered Chromatin in BRCA2-Crisis Cells.

(A) Immunoblot of histone acetylation marks associated with open chromatin. H4K5Ac and H4K12Ac levels are reduced in control cells upon transition to EGF-free media, whereas they remain constant in BRCA2-crisis cells. HAT1 was used as a loading control.

(B) H4K12Ac ChIP-seq peaks are enriched at regions proximal to TSSs.

(C) Integration of BRCA2 ChIP-seq, H4K12Ac ChIP-seq, and RNA-seq. ChIP-seq heatmaps are plotted along the gene body from −2,000 bp to TSS to transcription end site (TES) to 2,000 bp downstream for all genes in hg19 genome annotation with clustering by k-means (k = 3) on the ChIP-seq signals. RNA-seq data parsed by ChIP-seq clusters show active transcription in genes bound by both BRCA2 and H4K12Ac throughout the gene body (cluster 2). Cells with growth impairment (control cells in EGF-free media) express a subset of genes in clusters 1 and 2, with reduced expression of cluster 2 genes.

(D) BRCA2-dependent transcripts from Figure 3D were intersected with differential H4K12Ac ChIP-seq peak calls from siControl and siBRCA2-treated cells with significance testing by the hypergeometric test.

ChIP-seq also showed global increase of H4K12Ac in BRCA2-crisis hTert-HME1 cells compared to control. We identified 10,149 H4K12Ac peaks in BRCA2-crisis cells, which encompassed 89.9% of the 2,825 peaks identified in control cells. We found that H4K12Ac preferentially bound near TSS, suggesting an increase in active transcription at these genes located within open chromatin (Figure 4B). Antibodies to BRCA2 and H4K12Ac shared similar binding profiles across the genome (p < 1 × 10⁻¹⁵⁵; hypergeometric test), with three distinct clusters of genes based on binding profile. Genes in cluster 1 are bound specifically near the TSS, cluster 2 genes are bound throughout the gene body, and cluster 3 genes are largely unbound by either BRCA2 or H4K12Ac (Figure 4C). We observed distinct gene expression patterns in coordination with these BRCA2 and H4K12Ac binding patterns in BRCA2-crisis cells. Active transcription in proliferative cells, including those grown in complete media and BRCA2-crisis cells dividing in EGF-free media, was most evident in cluster 2 (BRCA2 and H4K12Ac bound throughout the gene body; Figure 4C).

Accordingly, BRCA2-dependent transcripts were enriched for genes within chromatin marked by H4K12Ac, with 46 BRCA2-dependent transcripts also associated with H4K12Ac (p < 1 × 10⁻⁹; Figure 4D). There was a significant enrichment of H4K12Ac-bound genes with transcripts upregulated in EGF-free media, as well as a significant reduction in H4K12Ac-bound genes at downregulated transcripts: 175 upregulated genes, p < 4 × 10⁻⁵ (enrichment), and 103 downregulated genes, p < 1 × 10⁻¹⁴ (depletion). These findings imply that BRCA2 binding and transcriptional regulation associate with H4K12Ac alterations in this model.

BRCA2-Crisis Activates NF-κB Signaling to Drive Chromatin Remodeling and EGF-Independent Growth

In order to identify functional mechanisms by which BRCA2-crisis cells gain a growth advantage in EGF-free media, we performed additional RNA-seq on earlier time points (passages 4 and 6), immediately prior to and after switching cells from complete to EGF-free media. A linear model combining siRNA and media effects was used to identify differentially expressed genes (Figure S4A). At passage 4, when BRCA2 levels have recovered to normal, there were 24 differentially expressed genes (FDR < 0.05) between cells treated with siRNAs to BRCA2 and control (Figures S4A and S4B). CXCL genes were among several gene families that were upregulated in concert with each other. Notably, in untreated, wild-type cells, we observed specific BRCA2 binding to the CXCL locus on chromosome 4 (Figure S4C), suggesting that differential gene expression during BRCA2-crisis may be dependent on BRCA2 chromatin binding.

Others have identified NF-κB as a transcriptional regulator of 41.7% of the genes we observed to be differentially expressed between cells treated with BRCA2 and control siRNAs at an FDR < 0.05 (Figure S4B) (Acharyya et al., 2012). To determine whether NF-κB activation was a function of BRCA2-crisis, we performed immunoblots across several passages following BRCA2 depletion to assay for phosphorylation of the S536 residue of the p65 NF-κB subunit (relA), a post-translational activating modification. This mark peaked at passage 3, specifically in cells transfected with BRCA2 siRNAs in complete media without an increase in total p65 levels (Figure 5A). We profiled cytokines in the media from MCF10A cells during this activation peak (passage 3) to determine whether NF-κB activation was associated with cytokine secretion. Of the 92 cytokines profiled, 21 were detected in supernatant of MCF10A cells, and 7 were significantly increased in BRCA2-crisis cells compared to control cells (CXCL8, VEGF-A, interleukin (IL)-6, IL-1a, CXCL1, MMP-1, and CXCL5; p < 0.05, Figure S5A). Therefore, BRCA2-crisis is associated with cytokine transcription and secretion, an effect that may be dependent on NF-κB pathway activation. Furthermore, BRCA2-crisis-dependent increase in IL-1 transcript levels required ATR kinase (Figures S5B and S5C), suggesting that NF-κB activation is down-stream of DNA damage after BRCA2-crisis.

Figure 5. NF-κB Activation Promotes EGF-free Growth and Chromatin Remodeling.

(A) Time course of immunoblots of hTert-HME1 cells transfected with control and BRCA2 siRNAs, serially passaged in complete media. NF-κB activation occurs at passage 3 specifically in BRCA2-crisis cells as measured by S536 phosphorylation of the p65 subunit of NF-κB. Total levels of p65 NF-κB remain constant.

(B) hTert-HME1 cells were transduced with vector control or constitutively active FLAG-tagged IKK-β (S177E, S181E; FLAG-IKK-β^EE), a potent NF-κB pathway activator. Expression of the IKK-β^EE transgene was detected by immunoblot.

(C) IKK-β^EE expression and NF-κB activation increase cellular proliferation in EGF-free media compared to empty vector control. Data indicate mean ± SD; n = 6. *p = 0.015, Student’s t test.

(D) Physiological NF-κB activation by TNF-α in hTert-HME1 cells is sufficient to increase cellular proliferation in EGF-free media in the absence of BRCA2-crisis. Cells were transfected with control or BRCA2 siRNAs and then serially passaged in EGF-free media and additionally treated with TNF-α or vehicle 3 days prior to each time point. Data indicate mean ± SD; n = 3. *p ≤ 0.05, and **p = 0.005, 3-way ANOVA with a Sidak’s multiple comparisons test.

(E) NF-κB inactivation by IMD-0354 (3 nM) in EGF-free conditions abrogates H4K5Ac and H4K12Ac levels in BRCA2-crisis cells.

To test whether NF-κB pathway activation could cause a proliferation advantage in hTert-HME1 cells, a constitutively activated form of the IκB-kinase β with S177E and S181E mutations (IKK-β^EE) was expressed in the hTert-HME1 cell line (Figure 5B). IKK-β^EE expression in EGF-free conditions resulted in increased cell proliferation compared to vectortransduced cells (Figure 5C), suggesting that NF-κB pathway activation is sufficient to drive EGF-independent growth in mammary cells. Indeed, physiological NF-κB activation by tumor necrosis factor α (TNF-α) also restored cell proliferation of control-treated cells in EGF-free conditions to BRCA2-crisis levels (Figure 5D).

To examine the relationship between the previously observed chromatin changes and NF-κB activation, we treated hTert-HME1 cells with IMD-0354, a potent and selective IKK-β inhibitor that prevents NF-κB nuclear translocation. H4K5Ac and H4K12Ac levels in BRCA2-crisis cells were reduced to levels equivalent or even less than control in EGF-free conditions when treated with IMD-0354 (Figure 5E), indicating that chromatin remodeling occurs downstream of NF-κB activation. Thus, NF-κB pathway activation drives a chromatin remodeling response that is concomitant with the stimulation of cell division in the absence of EGF.

In Vitro BRCA2-Crisis Recapitulates Molecular Changes Found in Human BRCA2 Carriers and Increases Tumorigenic Capacity

Breast cancer risk has been previously linked to a p27-driven gene expression profile in CD44+ mammary progenitor cells that is evident in BRCA2 mutation carriers compared to controls (Choudhury et al., 2013). Our analysis of this published dataset confirmed that normal breast tissue from BRCA2 mutant carriers displayed a different transcriptional profile from that of wild-type controls (Figure 6A). Therefore, we investigated whether gene expression changes were shared between our in vitro model of BRCA2-crisis and histologically normal mammary tissue from BRCA2 mutation carriers. GSEA revealed that differentially upregulated genes in CD44+ mammary progenitors from human BRCA2 mutation carriers were enriched in hTert-HME1 BRCA2-crisis cells (Figure 6B; normalized enrichment score, NES: 1.887, FDR < 0.001). Comparatively, in control cells, differentially expressed genes between EGF-free and complete media conditions were not enriched with the CD44+ BRCA2-carrier signature.

Figure 6. BRCA2-Crisis Recapitulates Molecular Features of Human BRCA2 Mutant Carriers and Increases Tumorigenic Potential.

(A) Bottom: heatmap of 10% most variably expressed genes (n = 1,246) via SAGE-seq of CD44-enriched histologically normal human breast tissue from BRCA2 carriers or controls. Data were reanalyzed from previously published results (Choudhury et al., 2013). Top: correlation matrix of these genes indicating strong positive pairwise correlations within control samples (Pearson’s R = 0.92, 0.91, and 0.89) and modest positive correlations within BRCA2 carriers (Pearson’s R = 0.33, 0.28, 0.27). Sample IDs, age, and parity of the samples are as indicated by Choudhury et al. (2013).

(B) A gene set was constructed from upregulated genes in CD44+ mammary progenitors from BRCA2^+/− tissue compared to wild-type controls. This gene set was enriched in differentially unregulated genes in BRCA2-crisis cells grown in EGF-free conditions compared to complete media (NES: 1.87, FDR < 0.001).

(C) The iPSCs from a BRCA2-mutation carrier were corrected using a scar-less genome engineering process to evaluate the effects of having one or two functional BRCA2 alleles in an isogenic system. Immunoblot of H4K12Ac that is increased in BRCA2-crisis cells indicates that epithelial-differentiated cells with only one functional BRCA2 allele also have increased H4K12Ac compared to cells with two functional alleles. Cells in the pluripotent state, regardless of the number of functional BRCA2 alleles, express similar H4K12Ac levels.

(D) In an orthotopic mammary xenograft model, hTert-HME1 BRCA2-crisis cells exposed to EGF-free media had greater mean final tumor volume (51 mm³) than cells grown in complete media (27 mm³ and 16 mm³ for control- and BRCA2-targeted siRNA cells, respectively). p = 0.027, ANOVA (control versus BRCA2-crisis).

Next, we mapped the histone H4 lysine 12-acetyl-linked genes in the BRCA2-crisis model cells onto the transcriptomic changes found in the BRCA2-carrier cells compared to wild-type CD44+ mammary cells in vivo. This revealed that H4K12Ac-associated genes in BRCA2-crisis cells were significantly enriched in upregulated genes and significantly depleted in downregulated genes in the BRCA2-carrier CD44+ mammary cells compared to the wild-type controls: 190 upregulated genes, p < 0.006 (enrichment); 261 downregulated genes, p < 4 × 10⁻³⁸ (depletion). This suggests that lysine 12 acetylation of histone H4 could play a role in altered gene expression profiles in human BRCA2-carrier mammary cells in vivo. Thus, at the molecular level, in vitro BRCA2-crisis could be reflective of primed, pre-malignant mammary cells found in human BRCA2 mutation carriers.

To determine whether genetic heterozygosity of BRCA2 resulted in similar changes in chromatin to in vitro induction of BRCA2-crisis, we examined patient-derived induced pluripotent stem cells (iPSCs) from a BRCA2-mutation carrier. We used a non-integrating viral approach to generate iPSCs from peripheral blood mononuclear cells (PBMCs) isolated from a patient with a single nucleotide deletion that induced a frameshift mutation at S1848, resulting in a premature termination codon (5543ΔA; Figures S6A–S6C). Using a scar-less CRISPR-based genome engineering strategy (Yusa et al., 2011), we corrected the mutant BRCA2 allele, creating an isogenic system to directly assess the effects of having one or two functional BRCA2 copies (Figures S6A–S6C). We next differentiated these iPSCs toward epithelial precursors (Itoh et al., 2011) (Figures S6D and S6E; see STAR Methods) and then surveyed histone acetylation marks by immunoblot. Akin to BRCA2-crisis cells, we observed greater H4K12Ac levels in BRCA2-heterozygous epithelial progenitor cells compared to BRCA2-restored cells. In contrast, we did not observe any differences in H4K12Ac levels relative to BRCA2 copy number in pluripotent culture (Figure 6C). These data suggest that BRCA2-crisis may occur in BRCA2 heterozygous cells during differentiation or in differentiated states.

Finally, we examined the potential for in vitro BRCA2-crisis to have functional consequences on neoplasia. hTert-HME1 cells were subjected to BRCA2-crisis, and following subsequent recovery in complete media, a subset of cells was transitioned to EGF-free conditions. We observed significantly greater mean tumor volume in BRCA2-crisis cells (51 mm³) after exposure to EGF-free conditions in an orthotopic mammary xenograft model in athymic nude mice (Figure 6D) compared to control cells in complete media treated with control siRNA or BRCA2-targeting siRNA (27 mm³ and 16 mm³, respectively; p < 0.0274). Thus, not only may the observed EGF-independent growth phenotype be indicative of increased tumorigenic potential, but exposure to such conditions in addition to BRCA2-crisis may also be necessary to reinforce these incipient epigenetic changes in disease settings.

DISCUSSION

In this study, we establish an in vitro model of BRCA2-crisis with the aim of identifying early, pre-neoplastic molecular alterations. This transient knockdown approach is unique compared to systems focused on heterozygous cell lines or BRCA2-deficient tumors. The difference in gene expression profiles from normal mammary tissues between BRCA2^+/− and BRCA2^+/+ individuals supports the idea that BRCA2 deficiency leads to divergent transcriptional phenotypes prior to oncogenic transformation and LOH. By leveraging a primary BRCA2-crisis in naive mammary cells, we have uncovered an immediate chromatin remodeling response to an acute BRCA2 deficiency. This may be driven, in part, by the discovery of the innate ability of BRCA2 to bind chromatin directly in normal, naive settings. Decreased BRCA2 levels leads to a chromatin remodeling response coupled to NF-κB activation, whereupon transition to EGF-free media induces downstream regulation of genes similar to that observed in BRCA2-mutation carriers. Additionally, BRCA2 heterozygosity was modeled with patient-derived iPSCs to confirm that epithelial-differentiated cells have similarly altered histone acetylation. Taken together, the data suggest that this response confers a BRCA2-crisis-induced growth advantage that may model incipient events that precede neoplasia onset.

EGF stimulation and binding to its nominal receptor, epidermal growth factor receptor (EGFR), regulates cellular proliferation and survival through activation of mitogen-activated protein kinase (MAPK)/JNK/phosphatidylinositol 3-kinase (PI3K) pathways. Hyperactive EGFR signaling is common in breast cancer, where EGFR or HER-2 overexpression, or somatic acquisition of constitutively active, ligand-independent mutants such as EGFRvIII, enhance tumorigenic potential and is associated with poor outcomes (Gan et al., 2013; Masuda et al., 2012; Tang et al., 2000). As the EGFR signaling axis remains the focus of many targeted therapies involving small molecule inhibitors (Ali and Wendt, 2017; Zhang et al., 2009), the acquisition of EGF-independent growth in a pre-neoplastic setting may have implications for understanding tumor initiation.

It should be noted a combination of BRCA2-crisis and EGF-free conditions were required to confer an advantage in ortho-topic mammary xenografts. In this regard, BRCA2-crisis may not confer bona-fide malignant transformation on its own. Instead, BRCA2-crisis may drive an initial step toward disease, and when combined with additional environmental stress, such as growth factor limitation, these molecular changes may serve to prime abnormal growth patterns. Indeed, the EGF-free conditions described here may also model the in vivo environment of normal breast tissue. For example, it has been shown that circulating EGF is limiting for mammary development and tumorigenesis (Kurachi et al., 1985; Vonderhaar, 1987). Since the mammary gland is subject to many changes that occur during natural events such as involution, pregnancy, and lactation, it is possible that the molecular changes we report here may reflect the enactment of physiologically normal cellular processes. However, because these alterations are spurred specifically by BRCA2-crisis, it suggests that driving normal processes outside of a physiologic context may underlie malignant transformation.

This study identifies two related mechanisms associated with BRCA2-crisis: NF-κB activation and increased lysine acetylation on histone H4. These processes are linked in that impairment of NF-κB activation diminished lysine acetylation after BRCA2-crisis. We also show here that BRCA2 possesses chromatin-binding ability and that BRCA2-crisis directs an NF-κB transcriptional response. This NF-κB activation was a transient, but acute, response after BRCA2-crisis, perhaps related to the fact that the chromatin changes observed also do not persist indefinitely. Since the NF-κB transcriptional profile, along with the associated downstream cell-proliferative phenotype, remains established well beyond this peak in passage 3, the acute relA phosphorylation we report here may be representative of the initiation of a typical phosphorylation/de-phosphorylation signaling cascade.

The NF-κB pathway is also subject to BRCA1 transcriptional co-activation (Gardini et al., 2014) and is deregulated in BRCA1-deficient breast tissue (Sau et al., 2016). Thus, although siRNAs targeted to BRCA1 did not lead to the same EGF-independent growth phenotype, our data do not exclude that a similar BRCA-crisis could underlie the expansion of progenitor cells in BRCA1-carrier breast tissue. Both BRCA1 and BRCA2 are recruited to sites containing double-stranded breaks (Lord and Ashworth, 2016). These may represent periods when lower cellular levels of functional BRCA2 may be insufficient and, thus, represent a BRCA-crisis event. In this setting, to fulfill the urgent need for BRCA2 at double-stranded breaks, BRCA2 may be recruited away from chromatin, ultimately leading to a cascade of epigenetic and transcriptional responses. Although BRCA1-induced breast cancer can be derived from a luminal precursor cell (Molyneux et al., 2010), the relevant cell of origin for BRCA2-induced breast cancer remains uncertain (Venkitaraman, 2014). Our model of in vitro BRCA2-crisis provides a tractable system in which to explore various genomic backgrounds to identify this cell of origin, such as ER+ or p53 mutant cells.

Additionally, work by others (Kim et al., 2012) has shown that the NF-κB p65 subunit associates at promoters together with the TIP60 histone acetyltransferase, which has acetyltransferase activity against the tail of histone H4. As TIP60 acetyltransferase activity is increased by DNA damage, one could speculate that NF-κB and TIP60 co-localize to promoters after BRCA2-crisis, resulting in histone H4 acetylation and increased transcriptional output. Therefore, inhibition of this transcriptional axis could represent a molecular target for abrogation of the BRCA2-crisis response.

A single brief yet acute exposure to BRCA2-crisis is sufficient to drive this phenotype transiently in vitro, yet organisms may experience multiple crises over time with unknown cumulative effects. Our isogenic iPSCs revealed that some of the effects of BRCA2-crisis on chromatin–namely, H4K12Ac–are also apparent in cells heterozygous for BRCA2. As we do not observe differences in histone acetylation in pluripotent cells, these results could indicate that differentiation itself triggers a BRCA2-crisis event, perhaps concomitant with diminished BRCA2 levels during differentiation. Future studies and defined mammary differentiation protocols are required to establish the circumstances that constitute physiological BRCA2-crisis and the impact of prolonged and chronic BRCA2 deficiency due to genetic haploinsufficiency. Further work will also be able to determine the therapeutic value and viability of inhibiting this pathway in terms of preventing cancer onset.

STAR★METHODS

LEAD CONTACT AND MATERIALS AVAILABILITY

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Michael Snyder (mpsnyder@stanford.edu). There are restrictions to the availability of the iPSC lines due to the requirement of a Material Transfer Agreement and University approvals.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Animals

Female 6–8 week-old NU/NU mice were purchased from Charles River (Cat# 088). Mice were housed in groups of up to 5 mice per cage in accordance with NIH and Stanford Administrative Panel on Laboratory Animal Care (APLAC) regulations. Mice were maintained with bedding, and freely available food pellets and water in a temperature and humidity-controlled room with a 12-hour light/ dark cycle. Animals were randomly assigned to experimental groups; the mouse handlers were blinded to the treatment groups.

Human samples

Peripheral blood was obtained from a 78 year-old female BRCA2 carrier (mutation: BRCA2 5543ΔA) under a research protocol approved by Stanford Institutional Review Board for use in generating an iPS cell line. Informed consent was obtained from the subject.

Cell Lines

Female human cell lines MCF10A (ATCC Cat# CRL-10317, RRID:CVCL_0598), HMEC (ATCC Cat# PCS-600–010), and hTert-HME1 (ATCC Cat# CRL-4010, RRID:CVCL_3383) were obtained from ATCC and were maintained in MEGM media (Lonza or Promocell). iPSCs were grown in mTeSR-1 (STEMCELL Technologies) on GelTrex (Life Technologies) coated plates, passaged using ReLeSR (STEMCELL Technologies) and plated into media containing 2 µM of ROCK inhibitor (Thiazovivin, STEMCELL Technologies). All cells were maintained in a humidified incubator at 37°C with 5% CO2. Cell line authentication is performed routinely by ATCC.

METHOD DETAILS

Transfection/transduction, cell assays

siRNA transfection was performed with 30 µl of Lipofectamine RNAiMAX (Invitrogen), 3 mL OptiMEM media, 18 picomoles of siRNA and 5E5 – 1E6 cells in 10 cm plates. SMARTvector lentiviral shRNA constructs targeting human BRCA2 was purchased from Dharmacon (Cat# V3SH11252) with doxycycline-inducible Tet-On 3G promoter driving shRNA expression and hEF1a-TurboGFP marker (shRNA IDs 225202137 (TACACTGAGGGTTCTATTT), 226940214 (ATAAGTACTAATGTGTGGT) and 228413532 (CAGCTATTGAAT TACTTTC)). IKK-2 S177E S181E was a gift from Anjana Rao (Addgene plasmid # 11105) (Mercurio et al., 1997). Cell proliferation was measured with a Biorad TC10 with trypan blue. Population doublings were calculated using the formula: PD = 3.32 (log(counted) – log(plated)), where PD = population doublings, counted = number of cells counted, plated = number of cells plated. TNF-α (eBioscience) was used at 25 ng/mL. IMD-0354 was used at 3 nM. ATR inhibitor #1 is AZ20, used at 2 uM. ATR inhibitor #2 is ATR-45, used at 2 uM.

Orthotopic xenograft mouse experiments

1 million hTert-HME1 cells treated with siRNA and either complete or EGF-free media were mixed with an equal volume of Matrigel and injected into mammary fat pads of female 6–8 week-old NU/NU mice. Tumor measurements were performed in blinded fashion by calipers three times per week. Final tumor sizes were assessed when animals were required to be euthanized due to tumor burden or at the end of 16 weeks. There were not enough siControl-treated cells in the EGF-free media condition for tumor injection studies.

iPSCs - Derivation

Peripheral blood samples were collected into BD Vacutainer® CPT Mononuclear Cell Preparation Tube (BD Biosciences) and PBMCs were isolated according to the manufacturer’s protocol. Isolated mononuclear cells were cultured in PBMC medium (GIBCO StemPro-34 SFM (Thermo Fisher Scientific) medium supplemented with hSCF, hFLT-3, hIL-3, hIL-6 (Peprotech) and hEPO (R&D Systems) for 7 days. Then, the expanded cells were reprogrammed into iPSCs using the CytoTune-iPS 2.0 Sendai Reprogramming Kit (Thermo Fisher Scientific). Briefly, 2×10^5 cells were transduced with Sendai virus with MOI of 5:5:3 (KOS:c-Myc:Klf4). Then, the cells were seeded on Matrigel (Corning) coated 6-well plates with PBMC medium. After 6 days, the medium was switched to GIBCO Essential 6 Medium plus human bFGF (Thermo Fisher Scientific) for further reprogramming. Medium was changed daily and then switched to Essential 8 Medium when small iPSC-like colonies were emerged.

The iPSC colonies were manually transferred onto new Matrigel-coated and maintained in Essential 8 Medium. iPS cells were routinely passaged every 4–6 days with Versene (EDTA) solution (Lonza). Selected iPSC clones were further validated for the expression of pluripotency markers with immunostaining.

iPSCs - Genome Engineering

The BRCA2 mutation was edited using dimeric CRISPR RNA-guided FokI nucleases and a dual positive/negative selection strategy using piggyBac transposase (Yusa, 2013). Briefly, sgRNA targeting sites near the BRCA2 mutation (GCTGCATTTTTATTTTTGCA and ATCTAATAGTAATAATTTTG) were designed using http://zlab.bio/guide-design-resources and cloned into pXFokI-dCas9, a gift from Bruce Conklin (Addgene Plasmid #60901) (Miyaoka et al., 2016). Homology arms were derived from PCR amplification of the wild-type chromosome, spanning 1,330 bp, ranging from chr12:32,913,238–32,914,568. A homology-repair template with the wild-type allele was provided on a plasmid containing a PuroΔTK drug selection cassette flanked by piggyBac ITR sites, coinciding with the TTAA at chr13:32,913,972–32,913,975 (Yusa et al., 2011). Plasmids were transiently transfected with nucleofection using the 4D-Nucleofector (Lonza) and the P3 Primary Cell kit (Lonza). Homology-directed repair was selected for using puromycin (500 ng/ml), and positive colonies were isolated and verified using PCR Sanger sequencing. Positive clones were nucleofected with a plasmid expressing excision-only mutant piggyBac transposase (System Biosciences). Negative selection for the PuroΔTK cassette was performed using FIAU (250 nM), and surviving colonies were isolated and verified using PCR and Sanger sequencing.

Epithelial Progenitor Differentiation

Prior to differentiation, iPSCs were plated as 2–10 cell colonies in ROCK inhibitor and allowed to re-attach overnight. The next day differentiation was initiated by switching media to MEGM (Lonza) containing 1 mM all-trans retinoic acid and 10 ng/mL BMP4. Media was changed daily for 4 days and cells were allowed to become confluent. Thereafter they were passaged by trypsinization and replated onto geltrex-coated plates supplemented with ROCK inhibitor for further expansion in MEGM for 2 additional days.

CRISPRi cell line

pHAGE TRE dCas9-KRAB was a gift from Rene Maehr & Scot Wolfe (Addgene plasmid # 50917) (Kearns et al., 2014). LentiGuide-Puro was a gift from Feng Zhang (Addgene plasmid # 52963) (Sanjana et al., 2014). To clone the guide RNA into LentiGuide-puro, the backbone was digested with BSMBI and treated with alkaline phosphatase in a single reaction for 30 minutes at 37°C, followed by gel purification. The BRCA2 promoter guide RNA oligos were phosphorylated with T4 PNK (NEB) with T4 Ligation buffer as an ATP source at 37°C for 30 minutes, then annealed by denaturation at 95°C for 5 minutes followed by ramping to 25°C at a rate of 5°C per minute. The ligation was performed with 2X Quick Ligase (NEB) and transformed to Stbl3 bacteria. Correct clones were confirmed by Sanger sequencing. Lentivirus was prepared by co-transfection of 293T cells with VSV-G pseudotyping plasmid, psPAX2 packing plasmid and lentiviral expression construct. Lentiviral infection of the hTert-HME1 cell line was performed with polybrene (8 ug/mL). Cells were first infected with the pHAGE-TRE-dCas9-KRAB lentivirus, followed by selection in G-418 (500 ug/mL), then a second lentivirus infection was performed with the guide RNA lentivirus, followed by selection with puromycin (4 ug/mL). Doxycycline was used at a final concentration of 1 ug/ml.

Immunoblots, qRT-PCR

Protein extracts were made in RIPA buffer and quantitated by BCA assay and diluted to equal concentrations. Polyacrylamide gel electrophoresis was performed on NuPAGE Novex gradient gels (Thermo Fisher) followed by wet transfer to nitrocellulose membranes. Blocking was briefly performed with 5% non-fat milk in PBST and primary antibody was incubated overnight at 4°C in 5% milk in PBST, then with HRP-conjugated secondary antibody (Cell Signaling) at room temperature for 1 hour followed by washing, then developed with ECL pico or femto (Thermo Fisher). For gene expression assays total RNA was isolated with All-prep (QIAGEN), DNase-treated, then reverse-transcribed with Superscript III (Invitrogen). qPCR was performed with 2x KAPA SYBR Fast Master Mix on a QuantStudio Flex 6 (Applied Biosystems).

Cytokine profiling

Supernatant from MCF10A cells were harvested at passage 3 (9 days) after siRNA transfection and frozen at −80C for shipment to Olink Proteomics for analysis with the ProSeek Multiplex Inflammation I panel.

Genomic profiling

Whole genome sequencing libraries were constructed with Tru-seq PCR-free kits following manufacturer’s instructions and sequenced by Macrogen, Inc. to 30X on HiSeqX with 150 bp paired-end reads. Methyl-seq libraries were constructed with Agilent SureSelectXT Methyl-seq kits with 1 ug of genomic DNA according to manufacturer’s instructions and sequenced on HiSeq with 100 bp paired end reads. ATAC-seq was performed with 50,000 cells according to published methods (Buenrostro et al., 2013) and sequenced by HiSeq 4000 (101 bp paired end reads). The differences in the number of experiments between ATAC-seq, RNA-seq, and Bisulfite-seq were due to technical limitations: The 3^rd replicate from the RNA-seq experiment from control cells in EGF-free media failed, and the cellular arrest when transitioning control cells to EGF-free media ultimately rendered the genomic yield insufficient to perform both bisulfite sequencing and ATAC-seq.

In detail, the ATAC-seq protocol started with PBS washing of the 50,000 cells, followed by lysis (lysis buffer: 10mM Tris-Cl, pH 7.5, 10 mM NaCl, 3 mM MgCl₂, 0.1% (v/v) Igepal CA-630) with gentle pipetting and collection of nuclei by centrifugation (500 x g) at 4°C. Transposition was performed with 25 uL of TD, 2.5 uL of TDE1 (Nextera, Illumina), and 22.5 uL water with gentle rocking at 37°C for 30 minutes. DNA was recovered with minElute purification (QIAGEN). PCR was performed with barcoded primers with NEBNext High-Fidelity 2x PCR Master Mix (NEB) for 5 cycles, then a qPCR side-reaction was performed with 10% of the PCR reaction to identify the number of additional PCR cycles required to complete the logarithmic phase of amplification. The additional cycles were performed on the original PCR reaction, then DNA was purified by minElute and gel purified to remove primer-dimers and remove sequences longer than 800 bp. Libraries were quantified by QuBit (Invitrogen) and Bioanalyzer (Agilent), then sequenced on HiSeq with 100 bp paired-end reads.

RNA-seq libraries were constructed with Script-seq v2 (Illumina) using 50 ng of poly(A)-enriched mRNA (Dynabeads mRNA Direct, Ambion) according to manufacturer’s instructions and sequenced on HiSeq2500 with 100 bp paired-end reads.

ChIP-seq

Log-phase growth cells were crosslinked in 1% formaldehyde at a density of 5×10⁵ cells per milliliter for 10 minutes at room temperature, then quenched with 125 mM glycine for 5 minutes, washed in PBS and snap frozen. Cells were then thawed and nuclei isolated by triton-X permeabilization followed by washing in a low detergent buffer. Then RIPA was added and sonication was performed by Branson (3 pulses of 30 s each with power output 4 W), followed by 14 cycles of sonication on the Bioruptor Pico. Chromatin extracts were then cleared by centrifugation and immunoprecipitation was performed with antibodies and protein A/G agarose beads overnight. Antibodies: Bethyl A300 anti-BRCA2 antibody; Abcam ab46983 anti-H4k12Ac. The next day the beads were extensively washed with RIPA, then PBS, then resuspended in TE with 1% SDS and crosslinks were reversed overnight at 65°Celcius. The next day RNase A treatment and proteinase K treatment were performed, followed by recovery of DNA with the Qiaquick spin columns. High throughput sequencing libraries were then constructed by A tailing, adaptor ligation and 10–15 cycles of PCR followed by library purification, removal of PCR primer-dimers and high-throughput sequencing by HiSeq 4000.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistics

Details of statistical tests can be found in the figure legends and manuscript text, including definitions of the tests used, definition and quantification of n, and quantification of measurement precision (error). Multiple-hypothesis correction was used for high-throughput sequencing studies and pathway analysis with and FDR cutoff of at least < 0.1 required to declare statistical significance. Otherwise, statistical significance was declared for p values < 0.05.

Whole genome sequence alignment and variant calling

Adapters were trimmed with SeqPrep with default settings. Paired-end reads were aligned to hg19 with bwa mem (Li and Durbin, 2009) using default settings, then sorted with samtools (Li et al., 2009). Duplicates were removed with picard-tools version 1.92 Mark-Duplicates command. Bam files were realigned to improve mapping around indels with GATK (McKenna et al., 2010) version 3.3.0 GenomeAnalysisTK and IndelRealigner tools. SNV calls were made with the DKFZ calling pipeline MB.I (Alioto et al., 2015) using the samtools mpileup command, and indels were called with the Platypus (Rimmer et al., 2014) callVariants command. Structural variant calling was performed with CREST (Wang et al., 2011) in the following manner: first extractSClip.pl was run by chromosome for both sample and reference realigned bam files and per chromosome outputs were concatenated. Then reference events were removed with countDiff.pl. Then the SV detection script CREST.pl was run with default settings. All called SVs were manually interrogated in IGV for further quality control.

Targeted DNA Methylation Sequencing

Adapters and reads were trimmed with Trim Galore (version 0.4.1) using default settings for paired-end mode. Paired-end reads were aligned with bsmap (Xi and Li, 2009) version 2.89 to the hg19 reference assembly. Duplicate reads were removed with the picard-tools version 1.92 MarkDuplicates command. Read duplicate ratios were between 8%–21% per MethylSeq library. The bisulfite conversion ratio was > 0.98 for all samples. Per CpG methylation values (beta values) were calculated with moabs (Sun et al., 2014) (version 1.3.0). The moabs mcall command was run with options to correct for M-bias:–genome hg19–cytosineMinScore 20–skipRandomChrom 1–processPEOverlapSeq 1–requiredFlag 2–excludedFlag 256–minFragSize 110–reportCpX G–qualityScoreBase 0–trimRRBSEndRepairSeq 0–trimWGBSEndRepairPE1Seq 5–trimWGBSEndRepairPE2Seq 5. Low-quality reads, as well as reads with overlapping paired ends were excluded for calculation of CpG methylation values. Only CpGs covered by a minimum of 5 high-quality reads and within the SureSelect target regions were considered for further analysis. DMRs were called with metilene (Jühling et al., 2016) version 0.2–4, requiring a minimum of 3 CpGs and a mean methylation difference of 20% between any two groups of samples, and an adjusted p value < 5%. For PCA analysis, mean CpG methylation signal over target capture regions was computed (methylated = 1, unmethylated = 0).

ChIP-seq and ATAC-seq peak calling and analysis

Paired-end 100 bp reads were trimmed with cutadapt version 1.8.1 (Martin, 2011) with flags –u −50 –U −50 –a CTGTCTCTTATACAC ATCTCCGAGCCCACGAGAC -A CTGTCTCTTATACACATCTGACGCTGCCGACGA -O 5 -m 30 -q 15. Bowtie (Langmead et al., 2009) version 1.1.1 was used to align trimmed reads to hg19 with flags -q–phred33-quals -X 2000 -m 1–fr -p 8 -S–chunkmbs 400, followed by samtools sort command. Duplicates were marked with Picard-tools version 1.92, then samtools view with flags –b –f 1 -F12 –L were used to filter mitochondrial mapping reads with a bed file containing all chromosomes except chrM. For ChIP-seq, SPP/phantom (Kharchenko et al., 2008; Landt et al., 2012) was run to obtain the fragment length with maximum strand cross-correlation. MACS2 (Zhang et al., 2008) broadpeak function was then performed with flags -q 0.01–nomodel –shiftsize = 1/2 fragment length obtained from SPP. For ATAC-seq, MACS parameters were callpeak –q 0.01 –nomodel –shiftsize 75. To create a merged peak list from all samples the peak files were concatenated, sorted by samtools, then merged with bedtools (Quinlan and Hall, 2010) merge command. The program align2rawsignal (https://github.com/akundaje/align2rawsignal) was used to create genome-wide signal coverage tracks with normalization to account for depth of sequencing and read mappability with flags kernel (k) = epanechnikov, fragment length (l) = 150 (ATAC-seq) or (l) = 1/2 fragment length from SPP for ChIP-seq, smoothing window (w) = 150, normFlag(n) = 5, mapFilter (f) = 0. Extractsignal (https://code.google.com/archive/p/extractsignal/) was used to export signal from the merged peak list for each bam file. ATAC-seq principle component analysis was then performed with the prcomp command in R. To define total peak counts in the H4K12Ac ChIP datasets, the findOverlapsOfPeaks command in the ChiPpeakAnno R package was used. Diffbind (Ross-Innes et al., 2012)was then used with input controls to define differential peaks between siControl and siBRCA2 conditions, using summits = 250 parameter.

RNA-seq analysis

Paired-end reads were aligned with tophat (Trapnell et al., 2009) version 2.0.11 using the gencode hg19 GTF file gencode.v19.annotation.gtf. Reads were counted with the sumarizeOverlaps function from the GenomicAlignments (Lawrence et al., 2013) R package with options mode = “Union,” singleEnd = F, ignore.strand = T, fragments = T. The gencode.v19.annotation.gtf was used to make a transcript database with makeTranscriptDbFromGFF command and genes were extracted with exonsBy command to use for read counting. DESeq2 (Love et al., 2014) workflow was then followed. A DESeqDataSet command was performed with the count matrix and design model ~media + replicate + siRNA, followed by rlog transformation. Batch effects were removed with removeBatchEffect command from limma (Ritchie et al., 2015). PCA analysis was performed with plotPCA command from affycoretools package. DESeq command yielded differential expression results and FDR cutoff of 0.1 was applied.

Gene set enrichment analysis was performed with GSEA-Preranked (v. 3.0) (Subramanian et al., 2005). Differentially expressed genes were ranked by the (sign of the fold-change)*(-log(pval)) and compared to upregulated and downregulated gene sets obtained from publicly available differential gene expression data between FACS-sorted normal CD44+ mammary cells obtained from human BRCA2-mutation carriers and wild-type controls (Choudhury et al., 2013). At an FDR < 0.05 and a fold-change > 1, 543 genes were downregulated and 245 genes were upregulated.

The gene expression heatmap and correlation matrix were derived from publicly available SAGE-seq expression values (Choudhury et al., 2013). Genes were ranked based on variance determined by standard deviation across all samples, and the top 10% most variable genes (n = 1,246) were plotted with the pheatmap function. The correlation matrix was also derived from expression values for the top 10% of most variable genes using the cormat function and plotted using the ggplot2 package.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
BRCA2	Bethyl	Cat# A303–434A-M; RRID: AB_10952240
NF-κB p65 (phosphor S536)	Abcam	Cat# ab86299; RRID: AB_1925243
BRCA1	Bethyl	Cat# A300–000A; RRID: AB_67367
Histone H4 acetyl K12	Abcam	Cat# ab46983; RRID: AB_873859
Histone H4 acetyl K5	Millipore	Cat# 07–327; RRID: AB_310523
Actin	Thermo Fisher	Cat# MA5–15452; RRID: AB_11001306
HAT1	Abcam	Cat# ab194296; RRID: AB_2801641
NF-κB p65 total	Cell Signaling	Cat# 8242; RRID: AB_10859369
Flag	Sigma-Aldrich	Cat# F1804; RRID: AB_262044
Bacterial and Virus Strains
NEB Stable Competent E. coli	New England Biolabs	Cat# C3040H
NEB 10-beta Competent E. coli	New England Biolabs	Cat# C3019H
Biological Samples
Peripheral blood from a BRCA2-mutation carrier	This paper	N/A
Chemicals, Peptides, and Recombinant Proteins
mTeSR1	StemCell Technologies, Inc.	Cat# 85850
Thiazovivin	StemCell Technologies, Inc.	Cat# 72254
ReLeSR	StemCell Technologies, Inc.	Cat# 05872
Accutase	Innovative Cell Technologies	Cat# AT-104
Knockout DMEM/F-12	Thermo Fisher Scientific	Cat# 12660012
GelTrex (hESC-qualified)	Thermo Fisher Scientific	Cat# A1413302
GelTrex	Thermo Fisher Scientific	Cat# A1413202
Puromycin	Sigma-Aldrich	Cat# P8833
FIAU	Sigma-Aldrich	Cat# SML0632
Proteinase K	Sigma-Aldrich	Cat# 03115879001
AccuPrime Taq HiFi	Thermo Fisher Scientific	Cat# 12346086
BbsI-HF	New England Biolabs	Cat# R3539
2x SYBR Master Mix	KAPA	KK4618
Lipofectamine RNAiMAX	Thermo-Fisher	13778075
Critical Commercial Assays
P3 Primary Cell 4D Nucleofector Kit	Amaxa	Cat# V4XP-3024
SureSelectXT Methyl-seq	Agilent	Cat# G9651
Nextera DNA Flex Library Prep Kit	Illumina	20018704
Deposited Data
RNA-seq data available at GEO	This paper	GSE107280, GSE107121
ATAC-seq data available at GEO	This paper	GSE107119
BRCA2 ChIP-seq data available at GEO	This paper	GSE133450
Histone acetylation ChIP-seq data available at GEO	This paper	GSE133728
Methyl-seq data available at GEO	This paper	GSE1017103
Experimental Models: Cell Lines
BRCA2-Mut iPSC	This Paper	N/A
BRCA2-Rev iPSC	This Paper	N/A
hTert-HME1	ATCC	Cat# CRL-4010
MCF10A	ATCC	Cat# CRL-10317
primary mammary epithelial cells	ATCC	Cat# PCS-600–010
Experimental Models: Organisms/Strains Oligonucleotides
NU/NU mice	Charles River	strain code: 088
Oligonucleotides
BRCA2 sgRNA1 Top: (CACCGCTGCATTTTTATTTTTGCA)	This Paper	N/A
BRCA2 sgRNA1 Bottom: (AAACTGCAAAAATAAAAATGCAGC)	This Paper	N/A
BRCA2 sgRNA2 Top: (CACCGATCTAATAGTAATAATTTTG)	This Paper	N/A
BRCA2 sgRNA2 Bottom: (AAACCAAAATTATTACTATTAGATC)	This Paper	N/A
Puro-TK-F: (CCCATGCACGTCTTTATCCT)	This Paper	N/A
PGK-R: (GGGGAACTTCCTGACTAGGG)	This Paper	N/A
BRCA2-int-R: (CCAATGCCTCGTAACAACCT)	This Paper	N/A
BRCA2–10B-F: (TTTTTGGAAGTTGCGAAAGC)	This Paper	N/A
BRCA2–10B-R: (tcaaaccatactcccccaaa)	This Paper	N/A
BRCA2–5′HDR-F: (CGCGCCTAGGTTCGAAGTTTCATTGAGAT CACAGCTGCCC)	This Paper	N/A
BRCA2–5′HDR-R: (TTACGCAGACTATCTTTCTAGGGTTAATG GCTGCATTTTTATTTTTGCAG)	This Paper	N/A
BRCA2–3′HDR-F: (CGTCACAATATGATTATCTTTCTAGGGTT AAATTGTCCATATCTAATAGTAATAATTTTG)	This Paper	N/A
BRCA2–3′HDR-R: (CTTACTAGTATTTAAATGCTGAGGCTGAG CTGGTCTGAATGTTCG)	This Paper	N/A
Recombinant DNA
pXFokI-dCas9	Miyaoka et al. (2016)	Addgene Plasmid #60901
Excision-only piggyBac Expression Vector	System Biosciences	Cat# PB220PA-1
PuroΔtk Targeting Vector	Yusa et al. (2011)	PMID: 21993621
pCR2.1-TOPO TA	Thermo Fisher Scientific	Cat# 450641
Human lentiviral shRNA to BRCA2	Dharmacon	Cat# V3SH11252–2252-2137, Lot# V3IHSHEG_5139787
Software and Algorithms
GSEA	Subramanian et al. (2005)	PMID: 16199517
BSMAP	Xi and Li. (2009)	https://code.google.com/archive/p/bsmap/
MOABS	Sun et al. (2014)	https://code.google.com/archive/p/moabs/
metilene	Jühling et al. (2016)	https://www.bioinf.uni-leipzig.de/Software/metilene/
Cutadapt	Martin, 2011.	https://cutadapt.readthedocs.io/en/stable/
Bowtie2	Langmead et al., 2009	http://bowtie-bio.sourceforge.net/bowtie2/index.shtml
Picard-tools	N/A	http://broadinstitute.github.io/picard
samtools	Li et al., 2009	http://www.htslib.org
SPP/phantom	Kharchenko et al., 2008; Landt et al., 2012	https://github.com/kundajelab/phantompeakqualtools
MACS2	Zhang et al., 2008	https://github.com/taoliu/MACS
align2rawsignal	N/A	https://github.com/akundaje/align2rawsignal
DiffBind	Ross-Innes et al., 2012	https://bioconductor.org/packages/release/bioc/html/DiffBind.html

Highlights.

• Transient BRCA2 loss (“BRCA2-crisis”) in mammary cells enables EGF-independent growth

• NF-κB activation promotes H4 acetylation after BRCA2-crisis

• BRCA2-crisis increases tumorigenic potential in vivo

• BRCA2-crisis and BRCA2^+/− carriers share common gene expression and chromatin states