Introduction

BRCA1 and BRCA2 tumour suppressors have key roles in maintaining genome integrity (Zhao et al, 2019; Tarsounas & Sung, 2020). Firstly, they promote RAD51 loading onto RPA-coated single-stranded DNA (ssDNA) at sites of DSBs, thereby initiating DSB repair via HR reactions. Secondly, they act during DNA replication to protect stalled replication forks against nucleolytic degradation, thus preventing their conversion into detrimental DNA lesions. Thirdly, BRCA1 and BRCA2 promote the restart of stalled forks, thus facilitating genome replication. The replication roles of BRCA1 and BRCA2 are thought to be mediated by loading and/or stabilisation of RAD51 nucleoprotein filaments at sites of stalled forks (Bhat & Cortez, 2018). Consequently, BRCA1/2 inactivation is associated with accumulation of stalled forks, replication-associated DNA damage and genome instability that can drive tumour formation.

Consistent with this, patients with familial breast, ovarian, prostate, pancreatic and other cancers often harbour heterozygous germline BRCA1/2 mutations (Michl et al, 2016b). Moreover, somatic BRCA1/2 mutations have recently been reported in sporadic cancers (Lal et al, 2019). Accumulation of replication-associated DNA damage is characteristic of BRCA1/2-mutated tumours and increases their vulnerability to targeted therapies, with inhibitors of poly(ADP-ribose) polymerases (PARPs) as a prominent example (Lord & Ashworth, 2016). PARPis are specifically lethal to BRCA1/2-deficient tumours because PARP1/2 enzymes, the targets of PARPi, bind to DNA ends. PARPis act by trapping PARP enzymes on DNA ends, leading to replication-associated DNA damage when BRCA1/2 is abrogated (Murai et al, 2012; Pascal & Ellenberger, 2015). In addition, PARPis suppress the ability of PARP enzymes to PARylate their substrates, which is essential for the repair of ssDNA breaks that arise during DNA replication. More recently, PARPis have been shown to substantially potentiate the type I innate immune responses triggered by loss of BRCA1 or BRCA2 (Ding et al, 2018; Pantelidou et al, 2019; Reisländer et al, 2019). Thus, in addition to inflicting replication-associated DNA damage and suppressing PARP1/2-dependent DNA repair, PARPis act by enhancing the immunogenicity of BRCA1/2-deficient tumours, thereby facilitating their elimination by the immune system.

In spite of tremendous efforts towards their optimisation, PARPis currently used in the clinic remain vulnerable to acquired drug resistance. BRCA1-deleted tumours can acquire PARPi resistance via loss of 53BP1 (Bouwman et al, 2010; Bunting et al, 2010), REV7 (Xu et al, 2015) or the shieldin complex (Dev et al, 2018; Noordermeer et al, 2018; Tomida et al, 2018). In BRCA2-deleted tumours, PARPi resistance can be triggered by loss of the poly(ADP-ribose) glycohydrolase PARG (Gogola et al, 2018). Because PARPi resistance occurs frequently in cancer patients with BRCA1/2 mutations (Gogola et al, 2019), there is a clinical need for new therapies, which can target resistant tumours.

Stalling of replication forks at physical barriers that obstruct their progression leads to replication-associated pathologies, collectively known as replication stress (Zeman & Cimprich, 2014). Examples of replication barriers include DNA secondary structures (e.g. G4s), DNA repetitive elements (e.g. minisatellites, rDNA, telomeres) or sites of transcription-replication conflicts (R-loops; (Hamperl & Cimprich, 2016; Techer et al, 2017)). G4s are thought to form spontaneously on G-rich ssDNA displaced during fork movement (Lipps & Rhodes, 2009). They consist of stacks of two or more G-quartets formed by four guanines via Hoogsteen base pairing stabilised by a monovalent cation (Murat & Balasubramanian, 2014). More than 700,000 sequences with G4-forming potential have been computationally identified in the human genome (Chambers et al, 2015). Given their thermodynamic stability, G4s can cause uncoupling of replisome components leading to fork stalling, which has the potential to trigger genomic instability. During physiological DNA replication, G4 are resolved by cellular helicases, thus enabling genome duplication (Tarsounas & Tijsterman, 2013). However, stabilisation of these DNA secondary structures by G4 ligands leads to persistent stalled replication forks and fork collapse, leading to replication-associated DNA lesions. These accumulate specifically in the cells in which key DNA repair pathways (HR or non-homologous end joining, NHEJ) are compromised (Zimmer et al, 2016; Xu et al, 2017).

Our previous work led to the discovery that G4 ligands (e.g. pyridostatin, PhenDC, RHPS4) are specifically toxic to HR-compromised cells, including those that are BRCA1/2-deficient (Zimmer et al, 2016). Importantly, we demonstrated that pyridostatin, the G4 ligand with the highest toxicity against the cells lacking BRCA1 or BRCA2, can also kill BRCA1-deficient cells that acquired PARPi resistance (Zimmer et al, 2016). Subsequent studies led to the characterisation of a second G4 ligand, CX-5461, which is also specifically toxic to BRCA1/2-deficient cells and tumours (Xu et al, 2017). The CX-5461 properties unravelled in this study made it a good candidate for further testing in a phase 1 clinical trial in cancer patients carrying DNA repair defects, including BRCA1/2 mutations (https://clinicaltrials.gov/show/NCT02719977). Given that approximately 95% of the drugs tested in phase 1 trials fail to be certified for clinical use (Wong et al, 2019), we concentrated our efforts on characterisation of novel drugs, with mechanisms of action similar to CX-5461, but with potentially superior pharmacological properties, which could therefore be more effective in the clinic. Moreover, a major cause of treatment failure in patients is development of drug resistance. Thus, we reasoned that a better understanding of the mechanism of action of such novel anti-tumour drugs may enable the design of combination therapies for clinical use that could be more effective than individual drugs.

In this study, we characterise the in vivo potential of pyridostatin, a drug with well-characterised toxicity against BRCA1/2-deficient cells in vitro. We show that xenograft tumours lacking BRCA1 or BRCA2 are hypersensitive to pyridostatin and that the in vivo activity of this compound is similar to that of the PARPi talazoparib. Moreover, we demonstrate that pyridostatin shows anti-tumour efficacy in BRCA1-deficient, PARPi-resistant tumour models, including patient-derived xenografts. A key mechanism of action of pyridostatin is activation of cGAS/STING-dependent innate immune responses, which may underlie its efficacy against BRCA1/2-mutated tumours in vivo. We find that the DNA damage inflicted by pyridostatin can be repaired even in the absence of BRCA1 or BRCA2 by C-NHEJ reactions and, consistent with this, we report synergistic effect between pyridostatin and the DNA-PKcs inhibitor NU-7441. A three-drug combination consisting of pyridostatin, NU-7441 and paclitaxel can effectively suppress growth or even eradicate BRCA1- or BRCA2-deficient tumours and we propose that this triple combination represents an effective therapeutic strategy against BRCA-mutated cancers. Altogether, our results suggest that the G4 ligand pyridostatin is a compound suitable for targeting BRCA1/2-deficient tumours and for overcoming PARPi resistance in vivo, thus highlighting its potential for therapeutic development.

Results Pyridostatin has anti-tumoral activity against BRCA2-deficient xenografts

Compounds that bind and stabilise G4s have been shown to be active against BRCA1/2-deficient xenograft tumours established in mice (RHPS4 and CX-5461). However, these have not yet been demonstrated to benefit patients with BRCA mutations. Moreover, BRCA-mutated tumours are difficult to treat because they rapidly develop resistance to targeted therapies (e.g. PARP inhibitors; PARPi). Therefore, it is imperative to identify new G4 ligands that not only eliminate BRCA-deficient tumours but also counteract resistant disease. Our previously published results (Zimmer et al, 2016) demonstrated that the G4 ligand pyridostatin is specifically toxic to BRCA2-deficient cells in vitro. Here, we evaluated the potential of pyridostatin in eliminating BRCA2-deficient xenograft tumours in vivo. To address this, we generated xenografts in CB17-SCID mice using the isogenic BRCA2+/+ (BRCA2-proficient) and BRCA2−/− (BRCA2-deficient) human colorectal adenocarcinoma DLD1 cells (Fig 1A and B). We extensively optimised conditions for in vivo use of pyridostatin and established that a dose schedule of 7.5 mg/kg/day administered intravenously for five consecutive days, followed by a 2-day break and a second 5-day treatment were well tolerated, as demonstrated by the lack of significant weight loss with no adverse clinical signs (Appendix Table S1). Using these conditions, we found that pyridostatin effectively and specifically inhibited growth of xenograft tumours established from BRCA2-deficient DLD1 cells (Fig 1B). As a control, we used the PARPi talazoparib, known for its ability to eradicate BRCA1/2-deficient tumours in mice (Shen et al, 2013) and recently licensed for use in metastatic breast cancer patients carrying BRCA1/2 germline mutations (Litton et al, 2018). The anti-tumoral effect of pyridostatin against the BRCA2-deficient tumours was similar to talazoparib, and neither drug impaired the growth of BRCA2-proficient tumours (Fig 1A; Appendix Table S1).

Figure 1. Pyridostatin inhibits growth of BRCA2-deficient DLD1 xenograft tumours and inflicts repairable DNA damage in BRCA2-deficient cells

A, B. CB17-SCID male mice were injected intramuscularly into the hind leg muscles with (A) BRCA2-proficient (BRCA2+/+) or (B) BRCA2-deficient (BRCA2−/−) DLD1 cells. Pyridostatin (PDS) was administered intravenously (i.v.; 7.5 mg/kg/day) and talazoparib was administered orally (p.o.; 0.33 mg/kg/day), over the indicated periods of time. Vertical dashed line indicates end of treatment. Tumour volume was measured at the timepoints shown on the graph and expressed relative to tumour volume at the beginning of treatment. Each experimental group included n = 5 mice. Error bars represent SEM. P values were calculated between treated and untreated tumours at day 17, using an unpaired two-tailed t-test. ***P ≤ 0.001; NS, P > 0.05 C. DNA breaks were measured using alkaline comet assay in BRCA2-proficient (+BRCA2) or -deficient (−BRCA2) human DLD1 cells treated with 2 µM of pyridostatin (PDS) for 16 h and released into fresh medium without pyridostatin. Representative images are shown. Scale bar represents 100 µm. D. Quantification of DNA breaks shown in (C). Graph and error bars represent the mean and SEM of n = 3 independent experiments. A minimum of 50 cells were analysed per condition per experiment. P values were calculated using an unpaired two-tailed t-test. *P ≤ 0.05; NS, P > 0.05. E. Quantification of γH2AX foci visualised using immunofluorescence staining in cells treated as in (C). A minimum of 200 cells were analysed per condition per experiment. Graph and error bars represent the mean and SEM of n = 3 independent experiments. P values were calculated using an unpaired two-tailed t-test. ***P ≤ 0.001; NS, P > 0.05. F. BRCA2-deficient (−BRCA2) human DLD1 cells were treated with 2 µM of pyridostatin (PDS) for 24 h. Next, pyridostatin was removed and cells were released in a medium containing 250 nM NU-7441 (NU). Whole-cell extracts were prepared at the indicated timepoints after release and immunoblotted as shown. SMC1 was used as a loading control. KAP1 phosphorylation site is indicated in red. G. Dose-dependent viability assays of BRCA2-proficient (+BRCA2) or -deficient (−BRCA2) human DLD1 cells treated with pyridostatin (PDS) and NU-7441 at the indicated concentrations for 6 days. Graphs represent average values obtained from of n = 3 independent experiments, each performed in technical triplicates.

Data information: Exact P values for (A-B, D-E) are provided in Appendix Table S8.

Source data are available online for this figure.

Furthermore, we investigated the in vivo response to pyridostatin using a second tumour model, established from isogenic BRCA2+/+ and BRCA2−/− colorectal carcinoma HCT116 cells (Xu et al, 2014). Pyridostatin showed selective toxicity against BRCA2-deficient HCT116 cell-derived tumours (Appendix Fig S1A and B; Appendix Table S2), similarly to its effect in DLD1 cell-derived xenografts.

Our previous work showed that pyridostatin treatment causes DNA damage accumulation in cells with compromised HR repair, including BRCA2-deficient cells (Zimmer et al, 2016). Consistently, immunohistochemical (IHC) analyses revealed that BRCA2-deficient, but not BRCA2-proficient, tumours exhibited increased level of the DNA damage marker γH2AX upon exposure to either pyridostatin or talazoparib (Appendix Fig S1C–F). These results indicated that pyridostatin can specifically suppress not only the growth of the cells (Zimmer et al, 2016), but also of tumours lacking BRCA2 and that it acts in vivo by inflicting DNA damage.

Pyridostatin induces DNA damage that is repaired by canonical non-homologous end-joining

In the course of our in vivo experiments using BRCA2-deficient DLD1 and HCT116 cell-derived xenografts we observed that, in spite of initial inhibition, tumours treated with pyridostatin resumed growth at the end of the treatment (Fig 1B; Appendix Fig S1B). This suggested that the DNA damage inflicted by pyridostatin, which underlies its toxicity against these tumours, can be repaired in the absence of BRCA2. We therefore attempted to gain further insight into the origins of the DNA damage induced by pyridostatin in BRCA2-deficient cells and to identify potential DNA repair pathways that can promote its repair.

Aberrant replication is commonly associated with DNA damage accumulation (Zeman & Cimprich, 2014). Our previous work (Zimmer et al, 2016) demonstrated that treatment with pyridostatin slows down replication fork progression in BRCA2-deficient cells, suggesting that G4s stabilised by pyridostatin assemble persistent DNA secondary structures that obstruct replication. BRCA2 has a central role in protecting stalled replication forks against nucleolytic degradation, illustrated by extensive shortening of nascent DNA when replication is arrested with hydroxyurea (HU) and BRCA2 is abrogated (Schlacher et al, 2011). We therefore investigated whether pyridostatin can cause fork stalling and degradation in BRCA2-deficient cells, similarly to HU. To address this, we conducted DNA fibre assays, in which successive pulse-labelling with CldU and IdU was followed by 5-h HU or pyridostatin treatment (Appendix Fig S2A). The relative replication track length, expressed as the ratio of IdU to CldU tracks (Michl et al, 2016a), provided means to quantify fork stability. We found that, whilst neither HU nor pyridostatin had an effect on the relative track length in BRCA2-proficient cells, both compounds led to significant attrition of the newly synthesised DNA in BRCA2-deficient cells (Appendix Fig S2A). Addition of mirin, an MRE11 inhibitor, rescued this phenotype. These results suggested that pyridostatin-stalled forks become substrates for MRE11-dependent degradation in the absence of BRCA2.

We then assessed the consequences of replication fork instability in BRCA2-deficient cells, by visualising the DNA breaks induced by pyridostatin and measuring their repair. Distinctly from previous studies, we investigated here whether the DNA damage inflicted by pyridostatin can be repaired in the cells lacking homologous recombination (i.e. BRCA1/2-deficient) after removal of the drug and, if so, which repair pathways are required. To do this, we monitored either DSBs directly using established techniques (comet assay and mitotic chromosomes spreading) or DNA damage markers (i.e. γH2AX foci, ATM/ATR activation) in BRCA1/2-deficient cells, during a recovery period of 3 days after the end of pyridostatin treatment. First, we used alkaline comet assays (Fig 1C and D) to quantify DNA damage by comparing the percentage of tail DNA relative to the total DNA in individual cells. We observed a significant increase in tail DNA, reflecting accumulation of DNA breaks, in BRCA2-deficient but not BRCA2-proficient DLD1 cells, upon treatment with 2 µM of pyridostatin for 16 h. Importantly, after releasing BRCA2-deficient cells into fresh media for 72 h, the level of DNA breakage was reduced to that of untreated cells. Next, we visualised chromosome breaks and aberrations (e.g. radial chromosomes) using Giemsa-stained spreads of metaphase chromosomes (Appendix Fig S2B and C). Treatment with pyridostatin caused an increase in broken/aberrant chromosomes in the cells lacking BRCA2, which were resolved 72 h after removing the compound from the media. These results indicated that pyridostatin-induced DNA damage can be repaired in BRCA2-independent manner after the treatment ended.

Our previous work demonstrated that pyridostatin inflicts DNA damage leading to ATM/ATR-dependent checkpoint activation and G2/M arrest in the cells with compromised BRCA2 function (Zimmer et al, 2016). We therefore investigated whether pyridostatin-induced ATM kinase activation is also attenuated in BRCA2-deficient cells after removal of the compound. Treatment with pyridostatin triggered phosphorylation of KAP1 at Ser824 and phosphorylation of RPA at Ser4/Ser8 (Appendix Fig S3A), both well-characterised ATM targets (Blackford & Jackson, 2017). Importantly, these modifications were diminished to the level of untreated cells within 72 h of release from pyridostatin treatment. In addition, γH2AX and 53BP1 foci, markers for DNA damage accumulation, showed a robust induction upon pyridostatin treatment specifically in the cells lacking BRCA2 (Fig 1E; Appendix Fig S3B and C), and gradually decreased after removal of the compound. Moreover, the ATM-dependent G2/M arrest elicited by pyridostatin in BRCA2-deficient cells (Zimmer et al, 2016), was also reversed by removal of pyridostatin from the media (Appendix Fig S3D), thus recapitulating the ATM signalling attenuation observed under the same conditions (Appendix Fig S3A).

Our results so far demonstrated that G4 stabilisation by pyridostatin inflicts DNA lesions and activates a potent DNA damage response (DDR) in BRCA2-deficient cells, which, surprisingly, become gradually attenuated after pyridostatin removal from the media. This suggested that DNA repair reactions occur in BRCA2-deficient cells, in spite of their compromised HR activity. We first investigated the involvement of alternative non-homologous end joining (A-NHEJ) in these repair events because tumours lacking BRCA1/2 have been reported to rely on this pathway for their survival (Ceccaldi et al, 2015). POLQ is an error-prone polymerase central to A-NHEJ repair (Yousefzadeh et al, 2014), where it facilitates microhomology annealing of ssDNA overhangs generated by end-resection (Kent et al, 2015). Moreover, POLQ is required in C. elegans to prevent genomic instability stemming from endogenous G4s stabilised by genetic ablation of the DOG-1/FANCJ helicase (Kruisselbrink et al, 2008; Castillo Bosch et al, 2014). We inhibited POLQ expression in human BRCA2+/+ and BRCA2−/− DLD1 cells using siRNA (Appendix Fig S4A) and observed that, surprisingly, POLQ abrogation had no effect on the pyridostatin sensitivity of BRCA2−/− cells in clonogenic survival assays (Appendix Fig S4B). Consistent with this, we found that POLQ depletion did not affect either the levels of pyridostatin-induced DNA damaged in BRCA2−/− cells, visualised using γH2AX or 53BP1 foci (Appendix Fig S4C and D), or their repair kinetics in recovery assays (Appendix Fig S4E and F). Taken together, these results indicate that POLQ is not required for the repair of pyridostatin-induced DNA damage in these cells, during or after treatment with the compound.

We next investigated whether the C-NHEJ pathway is implicated in the repair of pyridostatin-induced DNA lesions in the absence of BRCA2. The DNA-PKcs kinase is a core component of the C-NHEJ pathway, which orchestrates ligation of DNA ends by bringing them in close proximity and recruiting the DNA ligase IV/XRCC4 complex to complete the repair reaction (Calsou et al, 2003). We treated BRCA2−/− cells with pyridostatin for 16 h, then released them in media containing the DNA-PKcs chemical inhibitor NU-7441 for 4 days (Fig 1F). Under these conditions, pyridostatin-induced KAP1 phosphorylation at Ser824, a marker of ATM activation, persisted for 96 h after pyridostatin was replaced with NU-7441 in the media. This suggested that pyridostatin-induced DNA damage is repaired by C-NHEJ in BRCA2−/− cells. As a control for the specificity of NU-7441, we tested its effect of on the viability of HAP1 cells carrying a deletion of the PRKDC gene, which encodes DNA-PKcs (Appendix Fig S5A and B). No significant difference in cell viability was detected between PRKDC wild type (PRKDC WT) and PRKDC-deleted (PRKDC KO) cells upon treatment with NU-7441 (Appendix Fig S5A and B) suggesting lack of significant off-target effects, although the inhibitor showed some toxicity against PRKDC KO cells (Appendix Fig S5B). Human cells lacking DNA-PKcs were reported to be sensitive to pyridostatin (Xu et al, 2017). We recapitulated these results using PRKDC-deleted HAP1 cells (Appendix Fig S5C). Importantly, the combination of pyridostatin and NU-7441 showed no additional toxicity to PRKDC−/− HAP1 cells relative to pyridostatin alone, further validating the specificity of this inhibitor for DNA-PKcs (Appendix Fig S5D).

Given that DNA-PKcs is required for the repair of pyridostatin-induced lesions in the absence of BRCA2, we next tested the effect of pyridostatin/NU-7441 combination on BRCA2+/+ and BRCA2−/− DLD1 cell survival using viability assays (Fig 1G). We observed that addition of NU-7441 potentiated the toxicity of pyridostatin against cells lacking BRCA2, consistent with a synergistic effect of the two compounds in this background. Taken together, these results established that the repair of DNA damage induced by pyridostatin in the absence of BRCA2 is dependent on the C-NHEJ and does not require the A-NHEJ repair pathway.

Pyridostatin treatment triggers cGAS/STING-dependent innate immune responses in BRCA2-deficient cells

We and others have shown that BRCA2-deficient cells and tumours activate innate immune responses orchestrated by the cGAS/STING pathway, as a result of spontaneous DNA damage accumulation (Ding et al, 2018; Chabanon et al, 2019; Pantelidou et al, 2019; Reisländer et al, 2019). These immune responses caused by chronic loss of BRCA1 or BRCA2 function are potentiated by PARPi treatment, consistent with the ability of these drugs to increase endogenous DNA damage levels in the cells lacking BRCA2 (Ding et al, 2018; Pantelidou et al, 2019; Reisländer et al, 2019). Because pyridostatin also inflicts DNA damage in BRCA2-deficient cells (Fig 1C–E; Appendix Fig S2B and C; Appendix Fig S3), we investigated whether this G4 ligand can also trigger immune responses.

To address this, we cultured H1299 human cells carrying a DOX-inducible BRCA2 shRNA cassette in the presence or absence of pyridostatin for 3 days (Fig 2A). Immunoblotting demonstrated not only effective suppression of BRCA2 expression by DOX treatment, but also dose-dependent activation of ATM signalling by pyridostatin, illustrated by increased KAP1 phosphorylation at Ser824. The latter was also detected in BRCA2-proficient cells, possibly as a consequence of DNA lesions caused by the long exposure (3 days) to a relatively high concentration of pyridostatin (10 µM). Importantly, we observed that treatment of BRCA2-deficient cells with pyridostatin induced phosphorylation of IRF3 at Ser386, indicative of its nuclear translocation (Lin et al, 1998) and cGAS/STING pathway activation (Ishikawa et al, 2009; Tanaka & Chen, 2012; Sun et al, 2013). We concluded that pyridostatin triggers innate immune responses in BRCA2-deficient cells, as a consequence of DNA damage accumulation. To strengthen this conclusion, we performed time course experiments in which we monitored activation of ATM signalling and cGAS/STING pathway in time (Fig 2B). Our results indicate a clear increase in the levels of KAP1 Ser824 and IRF3 Ser386 phosphorylation in BRCA2-deficient H1299 cells with duration of pyridostatin treatment.

Figure 2. Pyridostatin activates innate immune responses in BRCA2-deficient cells

H1299+shBRCA2DOX cells were grown for 4 days in the presence (-BRCA2) or absence (+BRCA2) of DOX and subsequently treated with 5 or 10 μM pyridostatin (PDS) for 3 days. Whole-cell extracts were immunoblotted as indicated. KAP1 and IRF3 phosphorylation sites are shown in red. SMC1 and GAPDH were used as loading controls. H1299+shBRCA2DOX cells were grown for 4 days in the presence or absence of DOX and subsequently treated with 10 of μM pyridostatin (PDS) for 1, 2 or 3 days. Whole-cell extracts were immunoblotted as indicated. KAP1 and IRF3 phosphorylation sites are shown in red. SMC1 and GAPDH were used as loading controls. Quantitative RT-PCR of cells grown as in (A) and treated with 10 µM of pyridostatin (PDS) for 3 days was performed using primers specific for the indicated genes. mRNA levels are expressed relative to GAPDH and to untreated cells. Error bars represent the SEM of n = 3 independent experiments, each performed in technical triplicate. P values were calculated using an unpaired two-tailed t-test. *P ≤ 0.05; ***P ≤ 0.001. BRCA2+/+ and BRCA2−/− RPE-1 cells were treated with 10 μM of pyridostatin (PDS) for 2 days. Whole-cell extracts were immunoblotted as indicated. KAP1, IRF3 and STAT1 phosphorylation sites are shown in red. Tubulin and GAPDH were used as loading controls. H1299+shBRCA2DOX cells treated as in (B) were fixed and prepared for immunofluorescence with antibody against cGAS. DNA was counterstained with DAPI. Scale bar represents 20 µm. Quantification of cGAS-positive micronuclei per cells shown in (E). Graph and error bars represent the mean and SEM of n = 3 independent experiments. A minimum of 250 cells were analysed per condition per experiment. P values were calculated using an unpaired two-tailed t-test. *P ≤ 0.05; **P ≤ 0.01; NS, P > 0.05.

Data information: Exact P values for (C, F) are provided in Appendix Table S8.

Source data are available online for this figure.

Activation of the cGAS/STING axis also triggers interferon signalling, detectable as enhanced transcription of interferon stimulated genes (ISGs; Reisländer et al, 2020). Consistent with this, quantitative RT-PCR revealed a substantial increase in the mRNA levels of ISGs (IFIT1, IFIT2, ISG15) specifically in pyridostatin-treated BRCA2-deficient cells (Fig 2C).

Next, we generated BRCA2−/− human RPE-1 cells using CRISPR/Cas9-mediated gene deletion, as an additional cellular model for testing the impact of pyridostatin on the cells lacking BRCA2. Loss of BRCA2 protein expression in two different BRCA2−/− RPE-1 clones was demonstrated using immunoblotting and sensitivity to the PARPi olaparib (Appendix Fig S6A). Moreover, BRCA2+/+, but not BRCA2−/−, RPE-1 cells accumulated RAD51 nuclear foci upon exposure to ionising radiation (Appendix Fig S6B), further confirming abrogation of BRCA2 function. Treatment of BRCA2−/− RPE-1 cells with 10 µM of pyridostatin for 48 h induced robust KAP1 Ser824 and IRF3 Ser386 phosphorylation (Fig 2D), indicative of DNA damage accumulation and innate immune response activation, respectively. In these cells, pyridostatin also induced phosphorylation of STAT1 at Tyr701 (Fig 2D), a marker of activated interferon signalling (Shuai et al, 1993).

Genomic instability triggers micronuclei formation, likely as a result of chromosome mis-segregation during mitosis. Recognition of micronuclei by the cytosolic DNA sensor cGAS activates the cGAS/STING pathway and downstream innate immune responses. We monitored the frequency of cGAS-associated micronuclei in the cells treated with pyridostatin (Fig 2E) and observed a time-dependent increase in BRCA2-deficient cells relative to the wild-type counterparts (Fig 2F). These results demonstrate that innate immunity in BRCA2-deficient cells stems from DNA damage and genomic instability elicited by pyridostatin.

Pyridostatin overcomes PARPi resistance in Brca1-deleted cells and tumour models

The clinical efficacy of PARPi against BRCA1/2-deficient tumours is limited by the rapid development of drug resistance (Gourley et al, 2019). Therefore, concerted efforts are currently focused on the development of therapeutical strategies that eliminate resistant disease. Our previously published work (Zimmer et al, 2016) demonstrated that Brca1−/− mouse cells that acquired PARPi resistance via loss of 53BP1 can be targeted by pyridostatin. Here, we tested the ability of this compound to inhibit growth of Brca1−/−Tp53bp1−/− xenograft tumours established in mice (Fig 3A and B). We found that treatment with pyridostatin effectively suppressed tumour growth, in contrast with the PARPi talazoparib (Fig 3B; Table 1). Neither drug had an effect on the growth of Brca1+/+ tumours (Fig 3A).

Figure 3. Pyridostatin impairs growth of PARPi-resistant BRCA1-deficient tumours and triggers DNA damage and innate immune responses in BRCA1-deficient PARPi-resistant cells

A, B. FVB female mice were injected intramuscularly with (A) BRCA1-proficient KP3.33 (Brca1+/+) or (B) BRCA1/53BP1-deficient KB1PM5 (Brca1−/−, Tp53bp1−/−) mouse mammary tumour cells. Pyridostatin (PDS) was administered intravenously (i.v.; 7.5 mg/kg/day) and talazoparib was administered orally (p.o.; 0.33 mg/kg/day), over the indicated periods of time. Vertical dotted line indicates end of treatment. Tumour volume was measured at the timepoints shown on the graph and expressed relative to tumour volume at the beginning of treatment. Each experimental group included n = 5 mice. Error bars represent SEM. P values were calculated between treated and untreated tumours at day 16, using an unpaired two-tailed t-test. ****P ≤ 0.0001; NS, P > 0.05. C, D. BRCA1-proficient KP3.33 (Brca1+/+), BRCA1-deficient KB1PM5 (Brca1−/−) and BRCA1/53BP1-deficient KB1PM5 (Brca1−/−/Tp53bp1−/−) mouse mammary tumour cells were treated with 2 µM of pyridostatin (PDS) for 24 h and released into fresh medium without pyridostatin. Whole-cell extracts were prepared 0 to 72 h after release and immunoblotted as indicated. SMC1 was used as a loading control. KAP1 phosphorylation site is indicated in red. E. BRCA1+/+, BRCA1−/− and BRCA1−/−/TP53BP1−/− RPE-1 cells were treated with 10 μM of pyridostatin (PDS) or 2 μM of olaparib for 2 days. Whole-cell extracts were immunoblotted as indicated. KAP1, IRF3 and STAT1 phosphorylation sites are shown in red. SMC1 and GAPDH were used as loading controls.

Data information: Exact P values for (A, B) are provided in Appendix Table S8.

Source data are available online for this figure.

Table 1. In vivo anti-tumour efficacy of pyridostatin and talazoparib on Brca1+/+ and Brca1−/−Tp53bp1−/− allografts. Treatment Tumour volume inhibition (%) Tumour growth delay (days) Stable disease Body weight loss (%) Toxic deaths Brca1+/+ Pyridostatin 9 0 0/5 0 0/5 Brca1−/−Tp53bp1−/− Pyridostatin 65 10 1/5 0 0/5 Brca1+/+ Talazoparib 11 0 0/5 0 0/5 Brca1−/−Tp53bp1−/− Talazoparib 22 0 0/5 0 0/5 FVB female mice were injected intramuscularly with 4 × 106 cells per mouse. Tumours were allowed grow to approximately 250 mm3 before initiation of treatment (day 1). Mice were treated with pyridostatin (i.v.; 7.5 mg/kg/day) and talazoparib (p.o.; 0.33 mg/kg/day) for five consecutive days, followed by 2-day break and five more days of treatment. Each experimental group included n = 5 mice. Tumour volume inhibition was calculated at the nadir of the effect using the formula: (1 - [tumour volume in treated mice] / [tumour volume in untreated mice]) ×100 and expressed as average for n = 5 mice in each group. Tumour growth delay was calculated as the median time in days required for untreated and treated tumours to reach 700 mm3. Stable disease was defined as mice in which tumour volume did not change for at least 2 weeks after initiation of treatment. Body weight loss is reported as weight at the end of treatment relative to the first day of treatment (%), as average for n = 5 mice in each group.

Having established that pyridostatin triggers DNA damage signalling in BRCA2-deficient cells, which is silenced after compound removal from the media (Appendix Fig S3A), we now tested whether Brca1−/− and Brca1−/−Tp53bp1−/− show a similar repair capacity. To address this, we monitored KAP1 phosphorylation at Ser824, as a marker for DNA damage-induced ATM activation, in the cells treated with pyridostatin and released in media without the compound. Pyridostatin did not trigger ATM activation in Brca1 wild-type cells, but inflicted DNA damage in Brca1−/− cells, which was repaired upon release from treatment (Fig 3C). In contrast, in Brca1−/−Tp53bp1−/− cells pyridostatin-induced ATM activation persisted after compound removal (Fig 3D), indicating that pyridostatin inflicts unrepairable DNA damage in these cells.

To gain an understanding of the mechanism that prevents the repair of pyridostatin-induced DNA damage in the absence of BRCA1 and 53BP1, we conducted cell fractionation experiments. The cells were treated with pyridostatin for 24 h, allowed to recover for 72 h after removal of the compound, followed by cell fractionation (Appendix Fig S7A). Tubulin was used as a marker for the soluble fraction, and SMC1 as a marker for the chromatin-bound fraction. We observed that XRCC4, a central component of the C-NHEJ pathway, was not detectable in Brca1−/−Tp53bp1−/− cells, in contrast to Brca1+/+ or Brca1−/− cells, either in the chromatin fraction or the whole cell extract (Appendix Fig S7A and B). These results suggest that the inability of Brca1−/−Tp53bp1−/− to repair the DNA damage induced by pyridostatin is due to loss of a key C-NHEJ activity (i.e. XRCC4).

Next, we addressed whether pyridostatin can trigger innate immune responses in PARPi-resistant cells. In these experiments, we used human RPE-1 cells carrying BRCA1 gene deletion (BRCA1−/−; Zimmermann et al, 2018). Treatment of these cells with 10 µM of pyridostatin induced DSBs signalling, as shown by KAP1 Ser824 phosphorylation (Fig 3E). Similar results were obtained upon treatment with 2 µM of olaparib, used as a control. Both IRF3 Ser386 and STAT1 Tyr701 phosphorylation were induced by pyridostatin, and to a lesser extent by olaparib, in the BRCA1−/− cells. These results suggested that pyridostatin triggers innate immune responses associated with DNA damage accumulation in the cells lacking BRCA1 (Fig 3E), similarly to BRCA2-deficient cells (Fig 2A, B and D). We next tested the effect of the two drugs in BRCA1/53BP1-deficient RPE-1 cells (BRCA1−/−TP53BP1−/−), which are resistant to PARPi (Dev et al, 2018). Strikingly, phosphorylation of IRF3 Ser386 and STAT1 Tyr701 were increased solely after pyridostatin treatment in these cells (Fig 3E). Consistently, pyridostatin caused higher levels of DNA damage compared with olaparib, as indicated by KAP1 Ser824 phosphorylation. Taken together, these results demonstrate that pyridostatin triggers cGAS/STING-dependent immune responses in BRCA1/2-deficient cells, including those that have acquired PARPi resistance, and that these responses correlate with the pyridostatin ability to inflict ATM-activating DNA damage.

PARPi-resistant BRCA1-deficient patient-derived xenograft tumours are targeted by pyridostatin

To further confirm the sensitivity of PARPi-resistant tumours to pyridostatin, we used ex vivo cultures of patient-derived tumour xenograft cells (PDTCs; Fig 4A), known to recapitulate tumour vulnerability to specific drugs (Bruna et al, 2016). BRCA1-proficient PDTCs (AB521) showed no growth defects when cultured in the presence of pyridostatin. In contrast, VHIO179, a tumour carrying BRCA1 germline truncation and previously shown to be resistant to treatment with PARPi due to an inactivating mutation in the MAD2L2 (REV7) gene (Bruna et al, 2016; Cruz et al, 2018), was hypersensitive to pyridostatin. To investigate whether this response was recapitulated in the tumour context, we grafted the VHIO179 patient-derived tumour xenografts (PDTX) into CB17-SCID female mice and treated them with pyridostatin. This treatment effectively inhibited tumour growth relative to untreated tumours (Fig 4B; Table 2). Overall, these results demonstrated that in vivo pyridostatin has an inhibitory effect against BRCA1-mutated patient xenograft tumours that developed PARPi resistance.

Figure 4. Pyridostatin inhibits growth of BRCA1-deficient human PDTCs and PDTXs