BRCA2 abrogation triggers innate immune responses potentiated by treatment with PARP inhibitors

Heterozygous germline mutations in BRCA2 predispose to breast and ovarian cancer. Contrary to non-cancerous cells, where BRCA2 deletion causes cell cycle arrest or cell death, tumors carrying BRCA2 inactivation continue to proliferate. Here we set out to investigate adaptation to loss of BRCA2 focusing on genome-wide transcriptome alterations. Human cells in which BRCA2 expression is inhibited for 4 or 28 days are subjected to RNA-seq analyses revealing a biphasic response to BRCA2 abrogation. The early, acute response consists of downregulation of genes involved in cell cycle progression, DNA replication and repair and is associated with cell cycle arrest in G1. Surprisingly, the late, chronic response consists predominantly of upregulation of interferon-stimulated genes (ISGs). Activation of the cGAS-STING-STAT pathway detected in these cells further substantiates the concept that BRCA2 abrogation triggers cell-intrinsic immune signaling. Importantly, we find that treatment with PARP inhibitors stimulates the interferon response in cells and tumors lacking BRCA2.


3
BRCA2 tumor suppressor plays key roles in cell physiology by promoting DNA replication and DNA double-stranded breaks (DSBs) repair via homologous recombination 1 . The latter is well-characterized biochemically and relies on BRCA2 loading the RAD51 recombinase at sites of DSBs that have been processed by resection. Assembly of the RAD51 nucleoprotein filament facilitates the search and subsequent invasion of homologous DNA, which acts as a template for the repair reaction 2 . Whilst BRCA2 function in DNA repair has been studied in the context of DNA damage inflicted by exposure to genotoxic agents (e.g. ionizing radiation, DNA cross linking agents), the role of BRCA2 in replication is intrinsic to cell physiology. In the absence of external challenges, loss of BRCA2 triggers a significant decrease in replication fork progression and a high frequency of stalled forks 3,4 . A subset of these forks collapse and are converted to DSBs which, due to compromised HR repair, accumulate in BRCA2-deficient cells. Moreover, nucleolytic degradation of the stalled forks can occur upon extensive replicative arrest 5,6 . Conceivably, this acute perturbation of DNA replication triggered by inactivation of BRCA2 could lead to rampant genomic instability, which is lethal or severely obstructs cell proliferation in primary human cells 7 . Likewise, disruption of Brca2 gene in mice is embryonically lethal [8][9][10] . Mechanisms of replication stress and DNA damage tolerance mediate cellular adaptation to chronic loss of BRCA2, enable cells to survive and ultimately underlie their tumorigenic potential. Consistent with this, loss of BRCA2 occurs in tumors and is thought to promote tumorigenesis, whilst BRCA2 heterozygous germline mutations increase susceptibility to breast and ovarian cancer, as well as other cancers [11][12][13] .
Here we investigate the possibility that transcriptional alterations provide modalities of cell adaptation to loss of BRCA2, thus preventing cell death or proliferative arrest. We characterized the transcriptome of BRCA2-deficient cells, using a doxycycline (DOX)-inducible shRNA to inhibit BRCA2 expression in human non-small cell lung carcinoma H1299 cells and invasive ductal breast cancer MDA-MB-231 cells. RNAsequencing (RNA-seq) analyses conducted after 4 and 28 days of DOX-induced BRCA2 depletion enabled us to monitor the dynamics of gene expression and to identify substantial transcriptional alterations from early to late stages of BRCA2 inactivation. In the short term, we observed downregulation of cell cycle DNA replication and repair genes, which correlated with marked accumulation of BRCA2-deficient cells in G1. In the long-term, 4 we found that cell cycle re-entry occurred concomitantly with upregulation of a set of genes involved in the innate immune response and controlled by interferon signaling 14 . A similar set of genes was upregulated in BRCA2-deleted primary ovarian tumors. These results support the concept that inducible BRCA2 inactivation in cultured cells recapitulates cellular changes associated with loss of BRCA2 during tumorigenesis and could provide clues to mechanisms of cellular adaptation in tumors or tumor vulnerabilities. Importantly, treatment with PARP inhibitors (olaparib, talazoparib) stimulated upregulation of the interferon signaling genes, which may contribute, in part, to the toxicity of PARP inhibitors to BRCA2-deficient cells and tumors. 5

BRCA2 inactivation elicits changes in gene expression
Previous studies have implicated BRCA2 in gene transcription via interaction with EMSY 15 and have reported differential gene expression between BRCA2-deleted and wild type cells using expression microarrays 16 . BRCA2-deleted cells are terminally adapted and this precluded analysis of progressive changes associated BRCA2 inactivation. Here we used RNA-seq analyses of gene expression to determine how the transcriptome of BRCA2deficient cells is modulated over time. We cultured H1299 and MDA-MB-231 human cells carrying a DOX-inducible BRCA2 shRNA cassette in the presence or absence of 2 µg/mL DOX for 28 days and collected samples for RNA-seq analysis after 4 and 28 days of treatment (Fig. 1a). BRCA2 expression was effectively suppressed by DOX exposure in the short-term (4 days), as well as in the long-term (28 days), as indicated by immunoblotting of cell extracts prepared at these time points (Fig. 1b).
Before proceeding with RNA-seq analyses of BRCA2-deficient cells, we performed control experiments in which we addressed whether DOX alone could induce transcriptional changes under our experimental conditions. Parental H1299 and MDA-MB-231 cells were incubated with 2 µg/mL DOX for 4 and 28 days, before collecting samples for RNA-seq analyses ( Supplementary Fig. 1a,b). Hierarchical clustering of sample-tosample Euclidean distances based on the RNA-seq data showed no clear clustering of samples treated with DOX (n = 3 independent experiments) or untreated samples (n = 3 independent experiments) in either cell line, indicative of insignificant alterations between the -DOX and +DOX samples. Moreover, differential gene expression analyses with false discovery rate (FDR) of 5% revealed that only 2 genes were differentially expressed in DOX-treated versus untreated H1299 cells at day 4 and none at day 28 of DOX treatment ( Supplementary Fig. 1c). The corresponding numbers of deregulated genes in MDA-MB-231 were 3 and 7, respectively ( Supplementary Fig. 1d). These results demonstrate that DOX treatment for 4 or 28 days has a negligible impact on gene expression of parental cells lacking BRCA2 shRNA.
In contrast to this, DOX treatment inflicted substantial changes on the transcriptome of H1299 and MDA-MB-231 cells carrying a DOX-inducible BRCA2 shRNA. Hierarchical clustering of sample-to-sample Euclidean distances showed a clear 6 distinction between +DOX (n = 3 independent experiments) and -DOX samples (n = 3 independent experiments) in both cell lines ( Supplementary Fig. 2a,b), indicative of specific, substantial differences in the transcriptome upon BRCA2 abrogation. Consistent with this, DOX-inducible BRCA2 depletion in H1299 cells triggered significant alterations (FDR < 0.05, no fold-change filter) in the expression of 1,363 genes at day 4 and of 479 genes at day 28 of DOX treatment (Supplementary Fig. 2c and Supplementary Table 1

Distinct pathways are deregulated in the short-term versus long-term BRCA2 inactivation
We proceeded to identify which gene sets are differentially expressed in BRCA2-deficient versus -proficient cells, at 4 and 28 days after shRNA induction. We focused on H1299 cells because DOX treatment had a high impact on gene expression in this cell line. We used stringent conditions (FDR < 0.05; Log2Fold Change > |0.5|) which enabled us to identify deregulated genes with high confidence (Fig. 2a, Gene set enrichment analysis of differentially expressed genes based on functional annotation (Gene Ontology -Biological Process database) showed enrichment in specific pathways (Fig. 2c,d). Genes downregulated in BRCA2-deficient cells at day 4 were mainly implicated in cell cycle, chromosome segregation, DNA repair and DNA replication and defined an early, acute response to BRCA2 inactivation. The genes upregulated at day 28 primarily mediated cytokines and immune responses. Interestingly, induction of these genes did not correlate with changes in BRCA2-deficient cell proliferation from early to late time points (data not shown).
To complement pathway mapping, we performed network analysis which showed how the 574 downregulated genes interact with each other and clustered into different pathways (Fig. 2e). We retrieved high-confidence, experimentally validated proteinprotein interactions through the NetworkAnalyst platform 17 . The network was generated by mapping the significant genes to the STRING interactome 18 and applying a search algorithm to identify first-order neighbours (proteins that directly interact with a given protein) for each of the mapped genes. We generated a highly-connected first-order  Table 2) at day 28 of DOX treatment. REACTOME biological pathway analyses showed enrichment in processes including -interferon signalling and cytokine signalling in the immune system ( Supplementary Fig. 4b) and were therefore analogous to the processes upregulated in H1299 cells after 28 days of DOX treatment. Moreover, we performed RNA-seq analyses of differential gene expression in BRCA2 -/versus BRCA2 +/+ DLD1 cells ( Supplementary   Fig. 5). Pathway deregulation score analyses 19 of the REACTOME database using the top upregulated genes in BRCA2 -/-DLD1 cells identified interferon signaling, cytokine signaling and immune response as the highest scoring pathways.

Induction of innate immune response genes in BRCA2-deficient human cells
Long-term BRCA2 inactivation in H1299 and MDA-MB-231 human cells led to upregulation of common immune processes. To facilitate experimental validation for these findings, we identified common genes induced in both cell lines after 28 days of DOX treatment (Fig. 3a). We intersected the list of genes significantly upregulated in H1299 and MDA-MB-231 cells at day 28 (FDR < 0.05) and further filtered the high-confidence hits using Log2Fold Change > 0.3. This analysis identified 28 genes upregulated in both cell lines (Table 1), which showed enrichment in Gene Ontology processes including cytokine signaling, type I interferon response and immune response (Fig. 3b).
In BRCA2-deficient cells, replication fork instability and nucleolytic degradation inflict replication-associated DNA damage, which conceivably cause cytosolic DNA accumulation and activate innate immune responses 14 . Phosphorylation of STAT1 at Tyr701 is routinely used as a marker for activation of cytokine signaling 20 including interferon type I 21 in response to cytosolic DNA. Consistent with this, we observed enhanced STAT1 Tyr701 phosphorylation (Fig. 3c) in BRCA2-deficient relative to BRCA2-proficient H1299 cells, after 28 days of DOX treatment (Fig. 3c). IRF3 phosphorylation at Ser386, indicative of its nuclear translocation 22 , was also induced by long-term BRCA2 depletion. A similar response was triggered in MDA-MB-231 by BRCA2 depletion for 28 days (Fig. 3d). Thus, BRCA2 inhibition in human cells activates DNA damage responses and stimulates innate immune gene expression, as previously reported for DNA damaging agents 20,21 and SAMHD1 inactivation 23 .

Validation of innate immune response activation upon chronic BRCA2 inhibition
Induction of interferon response gene expression identified using RNA-seq analyses was validated using quantitative RT-PCR (Fig. 4a). mRNA levels of interferon-stimulated genes (IFIT1, IFIT2, IFIT3, IFI6, OAS1, OAS2, ISG15) and cytokine signaling gene (TNFRSF1B) were significantly upregulated upon BRCA2 inactivation for 28 days, whilst no significant changes were detected after 4 days of DOX treatment. Consistent with the concept that interferon signaling triggers JAK/STAT-dependent gene expression, we found that STAT1 siRNA-mediated depletion decreased the mRNA levels of interferonstimulated genes ( Supplementary Fig. 6a). Moreover, we found that a subset of these genes were also upregulated in BRCA2-deleted ovarian tumors ( Next we evaluated the dynamics of immune response gene expression by monitoring changes in mRNA levels with time ( Fig. 5a). Cells carrying a DOX-inducible shRNA against BRCA2 were treated with DOX for 28 days. Gene expression was measured at 2-day intervals using quantitative RT-PCR. mRNA levels for all genes monitored in this manner increased gradually during chronic BRCA2 inactivation in H1299 cells. Immune response gene upregulation was also detected over time in MDA-MB-231 cells ( Supplementary Fig. 7b), albeit to lower levels than in H1299. Additionally, DOXinduced BRCA2 depletion triggered STAT1 phosphorylation at Tyr701 and IRF3 phosphorylation at Ser386 (Fig. 5b). These preceded upregulation of the innate immune response genes, consistent with IRF3 and STAT1 promoting transcription of the interferonstimulated genes. Interestingly, STING protein levels were also markedly increased upon BRCA2 abrogation, while no significant change was detected in its mRNA levels. This may reflect stabilization of this protein, possibly by inhibition of its proteolytic degradation 24 .
In the course of these experiments we also addressed the possibility that loss of

BRCA-deficient human cells and tumors
Small molecule inhibitors of poly [ADP-ribose] polymerase (PARP) have been recently approved for clinical treatment of BRCA1-and BRCA2-deficient ovarian and breast cancers 25,26 . The molecular mechanisms of selective targeting of BRCA-deficient cells and tumors by PARP inhibitors has been investigated extensively. Trapping of the PARP1 on DNA ends obstructs replication fork progression leading to accumulation of single strand breaks, which are converted into lethal DSB lesions [27][28][29][30] .
Induction of DNA damage by PARP inhibitors prompted us to investigate whether treatment of BRCA2-deficient cells with these drugs could also impact on the interferon response. To address this, we treated H1299 cells with DOX for 4 days to induce BRCA2 depletion. Under these conditions, we did not detect upregulation of interferon-stimulated genes using quantitative RT-PCR (Fig. 4a, Fig. 5a). However, incubation with 1 or 10 µM olaparib for 24 hours induced a dose-dependent increase in the mRNA levels of interferonstimulated genes (Fig. 6a). We concluded that olaparib induced innate immune responses even after 4 days of BRCA2 depletion.
Next, we investigated whether the selective toxicity of olaparib against BRCA2deficient cells could be mediated by the interferon response. Interferon signaling is attenuated when STAT1 expression is inhibited using siRNA ( Supplementary Fig. 6a).
Thus, we treated BRCA2-proficient and -deficient H1299 cells with an siRNA against STAT1, before exposing them to 10 µM olaparib for 72 hours. We observed that cells in which BRCA2 was abrogated for 4 days were sensitive to olaparib and that STAT1 inactivation rescued their sensitivity ( Supplementary Fig. 6b). This indicates that the selective elimination of BRCA2-deficient cells by olaparib may be in part mediated by induction of innate immune responses. In contrast, STAT1 inhibition in BRCA2-proficient cells increased susceptibility to olaparib. Moreover, olaparib sensitivity of cells in which BRCA2 was chronically inactivated (28 days) were not affected by loss of STAT1, consistent with a role for STAT1 during the early stages of BRCA2 inactivation ( Supplementary Fig. 6c). Slowing down these processes may be conducive for activation of alternative, BRCA2independent mechanisms of fork protection and re-start, which, in the long-term, sustain cell survival.
The late, chronic response to BRCA2 inactivation consisted of upregulation of interferon-regulated innate immune response genes. best. Initially identified as a mechanism of defense against pathogens by blocking viral infections and priming adaptive immune responses 32 , the innate immune response can also be enlisted to counteract tumor progression 14 . The response operates on the premise that free genomic DNA evicted into the cytoplasm as a consequence of the genomic instability intrinsic to most tumors is recognized as pathogenic by cGAS and triggers STING activation (Fig. 7). This, in turn, facilitates phosphorylation and nuclear translocation of interferon response factors (e.g. IRF3), which drive transcription of the interferon response genes. Secreted interferon acts similarly to other cytokines to activate signaling pathways, in particular the JAK/STAT1 pathway, within the same cell (autocrine response) or in adjacent cells (paracrine response), resulting in induction of interferon-stimulated genes. Expression of these genes can also be triggered by ionizing radiation 20 and chemotherapeutic drugs (e.g. cisplatin, MMC) 21 and requires passage through mitosis with associated micronuclei formation 20,33 . 13 Our results demonstrate that the mRNA levels of interferon-stimulated genes are also enhanced by chronic BRCA2 inactivation. We previously reported that severe replication stress 5 and persistent DNA damage associated with loss of BRCA2 result in aberrant chromosome segregation 4 . Conceivably, this could lead to cytosolic DNA accumulation and interferon signaling, as suggested by the phosphorylation of STAT1 and IRF3, as well as upregulation of interferon response genes observed upon chronic BRCA2 inactivation (Fig. 5a). Consistent with STAT1-dependent transcriptional activation of these genes ( Supplementary Fig. 6a), our results demonstrate that STAT1 Tyr701 phosphorylation, required for its nuclear translocation and activation, is detectable from the early stages of BRCA2 depletion (Fig. 5b).
How the innate immune response activation impacts on the long-term survival of BRCA2-deficient cells and tumors remains to be elucidated. In our in vitro system, the interferon response induced upon chronic BRCA2 inactivation is associated with partial restoration of cell cycle progression (Fig. 5c). Upregulation of immune signaling factors known to promote cell proliferation (e.g. TNFRSF1B, also known as TNFR2 or p75 34 ) could account for the ability of BRCA2-deficient cells to survive. However, prolonged cell survival in the absence of BRCA2 may potentiate the innate immune response, as successive rounds of DNA replication and chromosome segregation are likely to augment their genomic instability. In the tumor context, ensuing interferon responses may stimulate cytotoxic T-cell activation, analogous to the STING-dependent defense mechanisms elicited by viral infection 35 , which could be onco-suppressive. Studies using mouse models of tumorigenesis driven by Brca2 abrogation 36 must be conducted in order to evaluate this possibility. Conversely, it will be interesting to determine whether BRCA2-deficient tumors that became resistant to chemotherapy have also acquired the ability to escape immune detection.
A recent study has shown that PARP inhibitors inflict DNA damage and promote cGAS/STING pathway activation in vitro, independently of BRCA status 37 . Our results establish that olaparib accelerates upregulation of innate immune response genes specifically in BRCA-deficient cells and tumors. Conceivably, the intrinsically high levels of DNA damage in BRCA2-deficient cells were further increased by olaparib, which thereby potentiated the innate immune response. Our results also suggest that the impact 14 of PARP inhibitors on the viability of BRCA2-deficient cells may be STAT1-dependent, as STAT1 siRNA-mediated inhibition reduced olaparib toxicity in the short term. Whether STAT1 signaling is also required for the long-term anti-tumoral effect of PARP inhibitors remains to be established.
Our discovery that the chronic response to BRCA2 inactivation is associated with innate immune response upregulation, potentiated by PARP inhibitors, has important therapeutic implications. First, it predicts that drugs that specifically kill BRCA2-deficient RNA-seq data processing. Reads were aligned to the human reference genome (GRCh37) using HISAT2 42 and duplicate reads removed using the Picard 'MarkDuplicates' tool (Broad Institute). Reads mapping uniquely to Ensembl-annotated genes were summarised using featureCounts 43 . The raw gene count matrix was imported into the R/BioConductor environment 44 for further processing and analysis. Genes with low read counts (less than ~10 reads in more than 3 samples) were filtered out, leaving sets of 14,000 to 15,000 genes to test for differential expression between conditions, depending on the samples considered.

Differential expression analysis.
The analyses were carried out using the R package Gene set enrichment analysis. The GO and REACTOME pathway enrichment analyses were performed using the Gene Set Enrichment Analysis (GSEA) software 46 with the Molecular Signatures Database collection. The NetworkAnalyst platform 17 and the Cytoscape 3.5 software 47 were used to visualize the first-order protein-protein interactions networks, based on the high-confidence (confidence score > 0.9 and experimental evidence required) STRING interactome database 18 .
Pathway deregulation scores. The Pathifier algorithm 19 calculates a pathway deregulation score (PDS) based on gene expression matrices for each sample. PDS represents the extent to which the activity of a given pathway differs in a particular test sample from the activity in the matched control. Gene sets were retrieved from the REACTOME database 48  Cell viability assays. Cells were seeded in 96-well plates at a density previously estimated to reach 70 -80% confluency after four days in the absence of treatment. Twenty four hours after seeding, drugs were added to the growth medium and viability was determined using resazurin-based assays three days later. Cells were incubated with 10 μg/mL resazurin (R7017, Sigma-Aldrich) diluted in growth medium at 37˚C for 2 hours.           (a, b) in (a, b). Each dot represents one gene; red, genes with significantly altered expresssion values (FDR < 0.05); gray, genes without significantly altered expresssion values (FDR > 0.05).