Chlorambucil targets BRCA1/2‐deficient tumours and counteracts PARP inhibitor resistance

Abstract Due to compromised homologous recombination (HR) repair, BRCA1‐ and BRCA2‐mutated tumours accumulate DNA damage and genomic rearrangements conducive of tumour progression. To identify drugs that target specifically BRCA2‐deficient cells, we screened a chemical library containing compounds in clinical use. The top hit was chlorambucil, a bifunctional alkylating agent used for the treatment of chronic lymphocytic leukaemia (CLL). We establish that chlorambucil is specifically toxic to BRCA1/2‐deficient cells, including olaparib‐resistant and cisplatin‐resistant ones, suggesting the potential clinical use of chlorambucil against disease which has become resistant to these drugs. Additionally, chlorambucil eradicates BRCA2‐deficient xenografts and inhibits growth of olaparib‐resistant patient‐derived tumour xenografts (PDTXs). We demonstrate that chlorambucil inflicts replication‐associated DNA double‐strand breaks (DSBs), similarly to cisplatin, and we identify ATR, FANCD2 and the SNM1A nuclease as determinants of sensitivity to both drugs. Importantly, chlorambucil is substantially less toxic to normal cells and tissues in vitro and in vivo relative to cisplatin. Because chlorambucil and cisplatin are equally effective inhibitors of BRCA2‐compromised tumours, our results indicate that chlorambucil has a higher therapeutic index than cisplatin in targeting BRCA‐deficient tumours.


Introduction
BRCA1 and BRCA2 germline mutations have been associated with approximately 25% of the familial cases of breast and ovarian cancer Miki et al, 1994;Wooster et al, 1995); therefore, BRCA1 and BRCA2 represent classical tumour suppressor genes (Lord & Ashworth, 2016). In addition, somatic BRCA1 and BRCA2 mutations, as well as their epigenetic inactivation, have been unravelled in a significant proportion of the sporadic cancers, by recent comprehensive genome sequencing studies (Cancer Genome Atlas Research Network, 2011;Cancer Genome Atlas Network, 2012;Curtis et al, 2012;Ali et al, 2014;Pereira et al, 2016). Thus, the subset of patients affected by BRCA1/2 mutations appears to be greater than initially anticipated. BRCA1 and BRCA2 play essential roles in DNA replication and DSB repair (Michl et al, 2016). Both factors promote HR, a DNA repair pathway active during S/G2 phases of the cell cycle, which also provides a mechanism for the re-start of stalled replication forks. Consequently, BRCA1 or BRCA2 abrogation confers exquisite sensitivity to DNA damage-inducing drugs, in particular those inflicting cytotoxic DNA crosslinks (i.e. platinum drugs and DNA alkylators), which interfere with DNA replication.
Small molecule inhibitors of poly(ADP-ribose) polymerase (PARP) are currently at the forefront of clinical research for the treatment of BRCA-compromised breast, ovarian and prostate tumours (Mateo et al, 2015;Mirza et al, 2016;Robson et al, 2017;Litton et al, 2018). PARP inhibitors induce DNA damage indirectly (Lord & Ashworth, 2017) by immobilising PARP enzymes to DNA ends and suppressing their ability to PARylate various substrates (Murai et al, 2012;Pascal & Ellenberger, 2015).
In spite of the fact that platinum drugs and PARP inhibitors show initially good responses in the clinic, most patients acquire resistance to these drugs (Rottenberg et al, 2007;Sakai et al, 2008;Shafee et al, 2008;Tutt et al, 2010;Norquist et al, 2011;Ter Brugge et al, 2016). Thus, there is a clear necessity for identifying new drugs or drug combinations that can target BRCA1/2-deficient cells and tumours. Here, we report the screen of a chemical library containing 1,280 drugs approved for clinical use by the US Food and Drug Administration (FDA). The highest scoring hit in our screen was chlorambucil, a bifunctional alkylator routinely used in chemotherapeutic regimens against CLL (Goede et al, 2014;Jain & O'Brien, 2015). We demonstrate that chlorambucil has high selective toxicity against human cells and xenograft tumours with compromised BRCA1/2 function. Mechanistically, chlorambucil acts by inducing replication stress and DSBs in actively replicating cells. Although similar to cisplatin in targeting BRCA-deficient tumours, chlorambucil shows substantially lower toxicity to normal cells and tissues. Our results suggest that the clinical use of chlorambucil in the BRCA1/2-deficient subset of cancer patients should be re-evaluated.

Results
Pharmacological screen for drugs that selectively eliminate BRCA2-deficient cells In order to identify drugs in clinical use that can target specifically BRCA2-deficient cells, we performed a viability screen using the Prestwick chemical library (http://www.prestwickchemical.com/lib raries-screening-lib-pcl.html) containing 1,280 FDA-approved drugs. Since all drugs are suitable for human testing, any compounds identified in this screen could rapidly be repurposed for the treatment of BRCA1/2-mutated patients. We conducted two independent screens, each in triplicate, at drug concentration of 5 lM (Dataset EV1, Appendix Fig S1) using hamster BRCA2deficient VC8 cells and control BRCA2-complemented cells . We demonstrated previously (Chaikuad et al, 2014;Zimmer et al, 2016) that these BRCA2-deficient cells are hypersensitive to PARP inhibitors, ERK1/2 inhibitors and pyridostatin, when compared to BRCA2-proficient counterparts. A similar chemical library screen aiming to identify drugs that target BRCA2-deficiency was previously performed (Evers et al, 2010) using Brca2 À/À mouse mammary tumour-derived cell lines and the LO-PAC â 1280 Sigma library of pharmacologically active compounds (Dataset EV1). The chemical composition of this library was different from that of the Prestwick library used here, with the two libraries having approximately 25% compounds in common.
Among the top scoring hits in our Prestwick library screens (Dataset EV1, Appendix Fig S1), we identified chlorambucil, a bifunctional alkylating agent used in the past for the treatment of breast and ovarian cancer (Williams et al, 1985;Senn et al, 1997), irinotecan, a topoisomerase I inhibitor in use for the treatment of cancer patients with BRCA1 mutations (Kennedy et al, 2004), and disulfiram, an aldehyde dehydrogenase inhibitor used in the clinic as an alcohol deterrent. Our group has recently characterised disulfiram as an agent specifically toxic to BRCA1/2-deficient cells and tumours, with significant therapeutic potential (Tacconi et al, 2017).
Given that our screens were conducted in hamster cells, we validated chlorambucil and irinotecan in BRCA2-deficient human cells. Human colorectal adenocarcinoma BRCA2 À/À DLD1 cells Fig 1A) were hypersensitive to both drugs, when compared with BRCA2 +/+ DLD1 cells. Olaparib and cisplatin were used as controls for selective targeting of BRCA2-deficient cells. Moreover, spheroid cultures established from BRCA2 À/À DLD1 cells recapitulated the chlorambucil sensitivity observed in 2D cultures ( Fig 1B).
Chlorambucil is toxic to BRCA1-deficient tumour cells, including those that acquired olaparib resistance To address the efficacy of chlorambucil against other HR-deficient cells, we assessed the response to this drug in BRCA1-deficient human cells. RPE1 cells immortalised by hTERT overexpression and TP53 knockout, which carry a BRCA1 CRISPR/Cas9-mediated deletion (Zimmermann et al, 2018), were hypersensitive to chlorambucil, as well as to olaparib and cisplatin used as controls (Fig 2A).
Moreover, we tested chlorambucil in cellular models in which BRCA1 gene inactivation is associated with olaparib resistance. Olaparib sensitivity characteristic of Brca1-deleted mouse mammary tumour-derived cells is abrogated upon loss of 53BP1 (Fig 2B;Bouwman et al, 2010;Tacconi et al, 2017). Nevertheless, these cells remained hypersensitive to cisplatin and chlorambucil. Notably, Brca1 À/À 53bp1 À/À cells were more sensitive to cisplatin than Brca1 À/À cells. A similar trend was previously reported in mouse embryonic fibroblasts (Bunting et al, 2012), suggesting that BRCA1/ 53BP1-deficient, olaparib-resistant tumours may be also more responsive to cisplatin in the clinic. This is indeed the case as demonstrated by a recent clinical trial in which patients with BRCA1/2 mutated, PARP inhibitor-resistant ovarian cancers showed a robust response to platinum-based therapies (Ang et al, 2013).
To generate a second model of olaparib resistance, we inactivated REV7 using two different shRNAs in Brca1-deleted mouse cells, as previously described (Xu et al, 2015). Cells lacking both REV7 and BRCA1 were less sensitive to olaparib than BRCA1deficient; however, they were effectively eliminated by cisplatin and chlorambucil treatments (Fig 2C). Thus, chlorambucil, similarly to cisplatin, can eliminate BRCA1-deficient cells that developed PARP inhibitor resistance via 53BP1 or REV7 inactivation.

Cisplatin-resistant BRCA2-deficient cells derived from human tumours are targeted by chlorambucil
To further investigate the therapeutic potential of chlorambucil, we tested its effect in cell lines established from BRCA2-compromised human tumours. Capan-1 cells derived from a pancreatic adenocarcinoma carry a C-terminal BRCA2 truncation, which impairs RAD51 nuclear localisation (Chen et al, 1998). Capan-1 cells showed significantly higher sensitivity to chlorambucil, as well as to cisplatin and olaparib, when compared to MIA PaCa-2 pancreatic cancer cells with normal BRCA2 expression ( Fig 3A).
As a second tumour-derived model, we used PEO1 cells established from a human ovarian tumour carrying a N-terminal BRCA2 truncation, which abrogates HR repair. C4-2 cells, in which wildtype BRCA2 was restored by treatment with cisplatin (Sakai et al, 2009), were used as a control. Viability assays demonstrated that PEO1 cells were hypersensitive to chlorambucil, in contrast to C4-2 cells ( Fig 3B). Notably, PEO1 cells, as well as other human cell lines lacking BRCA1 or BRCA2 (DLD1 BRCA2 À/À , HCT116 BRCA2 À/À and RPE1 BRCA1 À/À ), showed sensitivity to melphalan, another bifunctional alkylator (Appendix Fig S2). These results support the efficacy of other bifunctional alkylators against BRCA1/2-deficient cells, in agreement with previous studies (Evers et al, 2010).
Loss of the chromatin remodelling factor CHD4 confers resistance to cisplatin in BRCA2-deficient PEO1 cells, through unknown, HRindependent mechanisms that confer DNA damage tolerance (Guillemette et al, 2015). We recapitulated this observation by inhibiting CHD4 expression in PEO1 BRCA2-deficient cells ( Fig 3C). Lentiviral shRNA-mediated CHD4 depletion increased resistance of PEO1 cells to cisplatin, whilst it had no effect on the cisplatin response of BRCA2-proficient C4-2 cells. Importantly, chlorambucil effectively eliminated both cisplatin-sensitive and cisplatin-resistant BRCA2deficient PEO1 cells. These results suggest a potential clinical use for chlorambucil in targeting BRCA2-deficient tumours which acquired cisplatin resistance.

Chlorambucil induces replication stress and DNA damage accumulation in BRCA2-deficient cells
Alkylating agents can inflict DNA lesions in the form of intraand inter-strand DNA crosslinks, with a bias towards the latter (Deans & West, 2011). HR repair is an obligatory step in ICL resolution. In cells with compromised HR repair, ICLs interfere with DNA replication, leading to DSB accumulation and cell death (Michl et al, 2016). We therefore addressed the possibility that chlorambucil toxicity to BRCA2-deficient cells is due to ICLinflicted DNA replication and DSB repair defects. The response of BRCA2-proficient and BRCA2-deficient DLD1 cells to chlorambucil was evaluated using time course experiments and immunoblotting A Dose-dependent viability assays of BRCA2-proficient (+BRCA2) or BRCA2-deficient (ÀBRCA2) human DLD1 cells treated with drugs at the indicated concentrations for 6 days. B Human spheroids established from BRCA2-proficient (+BRCA2) or BRCA2-deficient (ÀBRCA2) DLD1 cells were incubated with 1.25 µM olaparib or 0.5 µM chlorambucil over the indicated period of time.
Olaparib (μM) % viability  Dose-dependent viability assays of BRCA1-proficient (+BRCA1) or BRCA2-deficient (ÀBRCA1) human RPE1-hTERT and TP53-deleted cells treated with drugs at the indicated concentrations for 6 days. B, C Dose-dependent viability assays of Brca1 +/+ and Brca1 À/À mouse mammary tumour-derived cell lines treated with drugs at the indicated concentrations for 6 days.  A Dose-dependent viability assays of BRCA2-deficient (Capan-1) or BRCA2-proficient (MIA PaCa-2) human pancreatic carcinoma-derived cells treated with drugs at the indicated concentrations for 6 days. B Dose-dependent viability assays of BRCA2-deficient (PEO1) or BRCA2-proficient (C4-2) human ovarian tumour-derived cells treated with drugs at the indicated concentrations for 6 days. C BRCA2-deficient (PEO1) or BRCA2-proficient (C4-2) human ovarian tumour-derived cells were infected with lentiviruses expressing control or CHD4 shRNAs, followed by selection with puromycin for 72 h. Dose-dependent viability assays were performed on cells treated with drugs at the indicated concentrations for 6 days. for checkpoint activation markers ( Fig 4A). RPA phosphorylation at Ser33, a marker for ATR activation and replication stress (Zeman & Cimprich, 2014), was induced in BRCA2-proficient cells following exposure to 1 lM chlorambucil for 48 h. In contrast, BRCA2-deficient cells, with intrinsic defects in replication fork progression and stability , showed detectable levels of RPA Ser33 phosphorylation even in the absence of any treatment (0 h), and these were markedly increased upon incubation with 1 lM chlorambucil from 16 h onwards. BRCA2-deficient cells also showed elevated levels of KAP1 Ser824 phosphorylation, a signature of ATM-dependent checkpoint activation and DNA damage accumulation. Phosphorylated KAP1 was detected from 24 h of treatment with chlorambucil in BRCA2-deficient cells. As RPA phosphorylation occurs earlier (16 h), this suggests that replication stress may precede DSB formation in response to chlorambucil. Cleaved PARP, an apoptosis marker, was induced only in cells lacking BRCA2. PARP cleavage was detectable from 16 h onwards, similarly to Ser33 RPA phosphorylation, supporting the concept that replication stress underlies chlorambucil toxicity to BRCA2-deficient cells. ATM activation occurs in response to DSB accumulation and leads to cytotoxicity. Therefore, we next quantified the frequency of DSBs and chromosome aberrations in BRCA2 +/+ and BRCA2 À/À cells upon treatment with 1 lM cisplatin or 1 lM chlorambucil for 72 h ( Fig 4B). This concentration was chosen for both drugs because it has the least toxic effects against BRCA2-proficient cells in viability assays (Fig 1A), whilst it induced apoptosis, measured by PARP cleavage (Fig 4A and Appendix Fig S3A), in BRCA2-deficient cells. Both treatments inflicted a significant level of DSBs and chromosome aberrations in BRCA2-deficient cells ( Fig 4B), with cisplatin inducing more lesions than chlorambucil. This reflects the higher cytotoxicity of cisplatin ( Fig 1A), as only 10% of BRCA2 À/À cells remain viable upon exposure to 1 lM cisplatin, in contrast to 30% of the cells treated with 1 lM chlorambucil. More than 90% of BRCA2 +/+ cells are viable after treatment with 1 lM of either drug. DNA damage accumulation leads to checkpoint activation which alters cell cycle progression. Consistent with this, we observed that a high percentage of cisplatin-treated cells arrested in G2/M ( Fig 4C). In contrast, chlorambucil treatment did not cause a significant G2/M arrest, in spite of inducing DNA damage associated with KAP1 phosphorylation (Fig 4A).

Molecular determinants of sensitivity to chlorambucil and cisplatin
Elucidating which pathways are involved in the repair of chlorambucil-and cisplatin-induced DNA lesions is essential for understanding the mechanism of action of these drugs. This is particularly important for cancer chemotherapy, as resistance is often associated with enhanced DNA repair. On the other hand, defining novel vulnerabilities to chemotherapeutic drugs may provide means to sensitise resistant tumours through novel combination therapies more active in the clinic. We therefore used siRNA depletion in human RPE1 cells to identify DNA damage response factors whose inactivation sensitises cells to chlorambucil and/or cisplatin. Both drugs induce Ser33 RPA phosphorylation, indicative of replication stress and ATR activation ( Fig 4A and Appendix Fig S3A). We thus depleted ATR using siRNA and observed that cells lacking ATR are hypersensitive to both drugs, supporting a key role for this checkpoint kinase in the cellular responses to chlorambucil and cisplatin ( Fig 4D). This result is consistent with a recent study, which showed that depletion of CHK1, an ATR phosphorylation target, sensitises cells to chlorambucil and cisplatin (Bruno et al, 2017).
ATR orchestrates cell responses to replication stress, including activation of the Fanconi anaemia (FA) pathway of ICL recognition and repair (Zhang & Walter, 2014). Given the ability of cisplatin and chlorambucil to induce ICLs, it was perhaps not surprising that abrogating FANCD2, a central FA protein, sensitised human cells to both agents (Fig 4E). ATR regulates FA by promoting FANCD2 mono-ubiquitylation (Andreassen et al, 2004), which, in turn, recruits to the chromatin nucleases required for ICL incision and unhooking. The XPF-ERCC1 nuclease makes the first incision at ICL sites. SNM1A nuclease is then recruited to those structures that resemble stalled forks, to complete ICL unhooking in concert with XPF-ERCC1 (Abdullah et al, 2017). Hypersensitivity of ERCC1 and XPF mutant cells to crosslinking anti-cancer drugs is well documented (McHugh et al, 2001); however, whether SNM1A-deficient cells recapitulate this sensitivity is unknown. We found that SMN1A-depleted human cells were also sensitive to cisplatin and chlorambucil (Fig 4E), indicating an important role for this nuclease in ICL repair.
Whilst XPF-ERCC1 is a known downstream effector of the FA pathway, whether SNM1A functions in FA-dependent manner is unknown. To address a possible cooperation between FA and SNM1A in ICL repair, we co-depleted FANCD2 and SNM1A and evaluated the response to cisplatin and chlorambucil. We found that inactivation of both factors did not further sensitise cells compared to FANCD2 depletion alone ( Fig 4E). Control experiments performed in cells depleted of XPF and/or FANCD2 using siRNAs showed a similar pattern (Appendix Fig S4). These results demonstrate the concerted action of SNM1A, XPF and FANCD2 upon ICL induction and place for the first time the SNM1A nuclease within the FA pathway of ICL repair.
The anti-tumoral effect of chlorambucil against BRCA2-deficient xenografts is similar to cisplatin Chlorambucil inhibited specifically the growth of BRCA2-deficient spheroids established from DLD1 cells (Fig 1B), indicative of its potential use in a tumour setting. Thus, we used BRCA2 +/+ and BRCA2 À/À DLD1 cells to generate xenograft tumours in mice (Fig 5A  and B). Chlorambucil had no effect on the growth of BRCA2-proficient tumours (Fig 5A), but it caused a striking reduction in BRCA2deficient tumour growth (Fig 5B). When the drug was administered intraperitoneally at doses of 3 mg/kg daily for 10 days (with a 2-day break after day 5), we observed tumour eradication in all animals within 21 days from treatment initiation.
Chlorambucil doses approved for CLL patient treatment are 0.2-1.6 mg/kg daily. In our mouse experiments, we used the maximum tolerated dose of 3 mg/kg daily (Weisburger et al, 1975;Grosse et al, 2009), which corresponds to 0.25 mg/kg daily patient dose, calculated using FDA conversion guidelines from mouse to human (http://www.fda.gov/downloads/Drugs/. . ./Guidances/UCM078932. pdf), and it is therefore clinically relevant. We further reduced the dose administered in mice and observed that BRCA2-deficient tumours were also eliminated even at doses of 1 mg/kg daily chlorambucil. However, a 0.3 mg/kg daily dose was ineffective (Appendix Fig S5). Thus, chlorambucil is active against BRCA2-deficient tumours even at doses lower than the equivalent doses used in the clinic.
To further investigate the therapeutic potential of chlorambucil, we used ex vivo cultures of patient-derived tumour xenograft cells (PDTCs; Fig 5C). These recapitulate not only tumour heterogeneity, but also tumour vulnerability to specific drugs (Bruna et al, 2016). Chlorambucil was selectively toxic to PDTCs lacking normal BRCA1 expression (STG201, VHIO179; http://caldaslab.cruk.cam.ac.uk/bca pe/) and had a small effect on BRCA1-proficient ones (AB521). Importantly, VHIO179, a tumour carrying BRCA1 germline truncation, is resistant to treatment with PARP inhibitors due to a MAD2L2  A BRCA2-proficient (+BRCA2) or BRCA2-deficient (ÀBRCA2) human DLD1 cells were incubated with 1 µM chlorambucil (Chl). Whole-cell extracts prepared at the indicated time points during treatment were immunoblotted as shown. GAPDH was used as a loading control. B Quantification of chromosome aberrations and chromatid/chromosome break frequencies in BRCA2-proficient (+BRCA2) or BRCA2-deficient (ÀBRCA2) human DLD1 cells incubated with 1 lM chlorambucil or 1 µM cisplatin for 72 h. Data were obtained from three independent experiments and normalised to untreated controls. A minimum of 60 Giemsa-stained metaphases were analysed for each sample. Cis, cisplatin; Chl, chlorambucil. C Quantification of G2/M cell frequency relative to solvent control, using FACS analyses of cells incubated with 1 lM chlorambucil or 1 µM cisplatin for 48 h. Cis, cisplatin; Chl, chlorambucil. D, E Human H1299 cells were treated with control (CTR) or indicated siRNAs 2 days before drugs were added to the media for dose-dependent viability assays. Cell extracts prepared at the time of drug addition were immunoblotted as indicated. SMC1 was used as a loading control.
Data information: (B) Whiskers indicate 10-90 percentile, and red bars indicate mean frequencies of breaks. P-values were calculated using the Mann-Whitney test. ****P < 0.0001. (C) Error bars represent SEM of three independent experiments. P-values were calculated using an unpaired two-tailed t-test. **P ≤ 0.01. NS, P > 0.5. (D, E) Graphs represent average values obtained from three independent experiments, each performed in triplicate. Error bars represent SEM. Exact P-values are included in Appendix Table S1. Source data are available online for this figure.
ª 2019 The Authors EMBO Molecular Medicine 11: e9982 | 2019 PDTCs derived from this tumour were sensitive to chlorambucil, which supports our results obtained with the Brca1-deleted mouse mammary tumour-derived cells, upon REV7 depletion using shRNAs (Fig 2C). These results strengthened the therapeutic potential of chlorambucil for targeting BRCA-deficient human tumours that acquired resistance to olaparib. We next determined chlorambucil anti-tumour activity using paired BRCA2 +/+ and BRCA2 À/À HCT116 colon carcinoma cell  A, B Nude mice (nu/nu) were injected subcutaneously with 5 × 10 6 human DLD1 cells, BRCA2-proficient (A) or BRCA2-deficient (B). Tumour-bearing mice were treated with 3 mg/kg daily chlorambucil administered intraperitoneally (i.p.) for a total of 10 days. Tumour weight was determined on the indicated days after initiation of the treatment. C PDTCs derived from breast cancer samples as previously described (Bruna et al, 2016) were treated with chlorambucil at the indicated doses. Cell survival is represented relative to DMSO control. AB521, ER-negative tumour, no BRCA1 alteration; STG201, tumour with BRCA1 promoter methylation and loss of BRCA1 expression; VHIO179, tumour with BRCA1 germline mutation and MAD2L2 inactivating mutation (olaparib-resistant); http://caldaslab.cruk.cam.ac.uk/bcape/. D CB17/SCID mice were injected intramuscularly with 5 × 10 6 human BRCA2-deficient HCT116 cells. Tumour-bearing mice were treated on the indicated days with chlorambucil or cisplatin administered intraperitoneally (i.p.), or talazoparib administered orally (o.s.) Tumour volume was measured on the indicated days after treatment initiation and was expressed relative to tumour volume at the beginning of treatment (day 1). Scale bar, 40 µm. lines, generated in one of our laboratories (Xu et al, 2014). BRCA2 À/À HCT116 cells showed hypersensitivity to cisplatin and chlorambucil in vitro (Appendix Fig S6A). Xenograft tumours were subsequently established from BRCA2 +/+ and BRCA2 À/À HCT116 cells and assessed for their response to chlorambucil, cisplatin and talazoparib (Appendix Fig S6B). Chlorambucil had no effect on the BRCA2-proficient tumours, but it inhibited the growth of BRCA2-deficient ones. We further used BRCA2 À/À HCT116-derived tumours to compare the anti-tumoral effects of cisplatin, chlorambucil and the PARP inhibitor talazoparib (Fig 5D). Treatment with each of the three drugs showed tumour growth inhibition at the nadir of the effect of 89, 89 and 85%, respectively (Table 1), with a progression-free survival of 19, 21 and 24 days, suggesting similar anti-tumoral activities against BRCA2defective tumours. Importantly, the three drugs inflicted comparable levels of DNA damage evaluated by immunohistochemical staining with an anti-cH2AX antibody (Fig 5D).

Higher cisplatin toxicity in vitro and in vivo compared to chlorambucil
Our cell viability assays ( Fig 1A) indicated that cisplatin is relatively more toxic to BRCA2 +/+ DLD1 cells than chlorambucil, whilst both drugs induced cleaved PARP expression in BRCA2 À/À DLD1 cells (Fig 4A and Appendix Fig S3A). To further explore the differential toxicity of the two drugs to BRCA2-proficient cells, we treated MRC5 primary-like cells with cisplatin or chlorambucil for up to 72 h (Appendix Fig S3B). Under these conditions, we observed a higher accumulation of the apoptotic marker in response to cisplatin than chlorambucil treatment, supporting the notion that cisplatin is more toxic than chlorambucil to non-tumour cells. MRC5 cells have intact p53 pathway (Carlos et al, 2013), which may account for the exquisite toxicity of cisplatin and chlorambucil to these cells. p53 is a key determinant of cell sensitivity to platinum-based drugs (Kelland, 2007). We did not succeed in abrogating p53 expression in MRC5 cells, and instead, we used RPE1 cells in which TP53 was deleted using the CRISPR/Cas9 system to address the role of p53 in the cellular response to chlorambucil. Functional p53 sensitised cells to cisplatin, as well as to chlorambucil (Appendix Fig S3C), whilst its abrogation promoted resistance. This supports the notion that p53-dependent responses mediate the cytotoxicity of these drugs.
The observation that cisplatin is more toxic than chlorambucil in vitro prompted us to assess the relative toxicities of the two drugs in vivo. We therefore treated wild-type mice with the maximum tolerated doses of each drug and evaluated apoptosis in vivo using single-photon emission computed tomography (SPECT) imaging of the apoptosis imaging marker 99m Tc-Duramycin (Palmieri et al, 2018; Fig 6A). SPECT image quantification demonstrated significant accumulation of 99m Tc-Duramycin in the heart, blood and lungs of cisplatin-treated mice. Apoptosis levels in the organs of chlorambucil-treated mice were similar to the control, solvent-treated group. We furthermore used lung and heart from treated mice for immunohistochemistry staining with cH2AX antibody (Fig 6B). Consistent with the biodistribution of the apoptosis tracer 99m Tc-Duramycin, we observed significantly higher cH2AX levels in the heart and lungs of cisplatin-treated mice, relative to the chlorambucil-treated ones. DNA lesion accumulation, visualised with cH2AX staining, was also detected in the organs from chlorambucil-treated animals, indicating that both drugs inflict DNA damage in vivo. However, cisplatin induced more pronounced and more deleterious lesions than chlorambucil. Given the higher toxicity of cisplatin to normal tissues (Fig 6A and B) and the similar anti-tumour effect of the two drugs (Fig 5C), we propose that chlorambucil represents a potential alternative to cisplatin for the treatment of BRCA-deficient tumours.

Discussion
In this study, we report identification of the bifunctional alkylator chlorambucil in a chemical library screen for drugs with specific toxicity against BRCA2-deficient cells. DNA alkylators, including chlorambucil, melphalan and nimustine, were also isolated in a previous screen and shown to be active in vivo against allografted BRCA2-deleted mouse tumours (Evers et al, 2010). Subsequent studies have substantiated the potential of nimustine in targeting BRCA-deficient tumours (Pajic et al, 2017).
Our work demonstrates the specific toxicity of chlorambucil to BRCA1/2-deficient human cells and xenograft tumours. Importantly, BRCA1/2-deficient cells with acquired resistance to olaparib or cisplatin show sensitivity to chlorambucil, suggesting its therapeutic potential against this difficult to treat tumour subset. Together with chlorambucil being non-toxic to normal cells and tissues (see Tumours were allowed to grow to 250 mm 3 before initiation of treatment. Mice were treated with chlorambucil (i.p.; 3 mg/kg/day) or talazoparib (p.o.; 0.33 mg/ kg/day) for five consecutive days, followed by 2-day break and 5 more days of treatment. Cisplatin (i.p.; 3.3 mg/kg/day) was administered for three consecutive days, followed by 4-day break and 3 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 regression was defined as percentage of mice in which reduction of tumour volume, after the initiation of treatment, was maintained for at least 2 weeks. Tumour relapse was defined as percentage of mice in which tumour regrowth was observed after tumour regression. Median progression-free survival was defined as duration (days) of tumour regression. 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.
ª 2019 The Authors EMBO Molecular Medicine 11: e9982 | 2019 below), these results support chlorambucil re-evaluation in cancer patients with BRCA mutations. Chlorambucil was used in the treatment of breast and ovarian cancer until the late 1970s (Barker & Wiltshaw, 1981;Williams et al, 1985;Senn et al, 1997). These early studies did not show a significant benefit of chlorambucil, possibly because its effect was obscured by the lack of molecular markers (e.g. BRCA1/2 status). In some clinical trials, addition of cisplatin to chlorambucil treatment was beneficial (Barker & Wiltshaw, 1981). Although the response to regimens that included cisplatin was initially superior to chlorambucil alone, the overall patient survival was not improved (Williams et al, 1985). Nevertheless, chlorambucil-based therapies for breast and ovarian cancer were abandoned, as cisplatin was approved for broader clinical use and became the leading anti-cancer drug and first-line treatment for various malignancies.
A major problem associated with cisplatin chemotherapy is the emergence of drug resistance (Norquist et al, 2011). Proposed resistance mechanisms include insufficient drug access to DNA, enhanced DNA repair and apoptotic pathways failure (Siddik, 2003).
Chlorambucil resistance in CLL patients has also been documented (Panasci et al, 2001;Norgaard et al, 2004), although not to a similar extent as cisplatin resistance. For both drugs, clinical evidence for a unique resistance mechanism is lacking due to the multifaceted response to these drugs in patients. Clearly, both drugs induce ICLs and DSBs; however, the mechanism linking it to apoptosis has not been elucidated. TP53 mutations known to increase cell tolerance to DNA damage and limit apoptosis can be reliably correlated with cisplatin resistance (Siddik, 2003). We show here that this is also the case for chlorambucil resistance, as indicated by our viability assays using isogenic p53 wild-type and p53-deleted RPE-1 cells.
An important result reported here is that chlorambucil is toxic to cisplatin-resistant BRCA2-deficient cancer cells. Although the mechanism remains to be further elucidated, this may be explained by the fact that the two drugs inflict distinct DNA lesions (chlorambucil induces primarily inter-and cisplatin primarily intrastrand crosslinks) which activate distinct DNA damage response pathways, especially in BRCA-deficient cells with compromised HR repair. A Wild-type Balb/c mice were injected intraperitoneally with solvent (daily) or 3 mg/kg chlorambucil (daily for 5 days) or 3.3 mg/kg cisplatin (daily for 3 days). Uptake of the apoptosis tracer 99m Tc-Duramycin 2 h after intravenous injection was quantified in selected organs using SPECT imaging in the indicated organs. Representative maximum intensity partial projections showing tracer distribution are shown. B Immunohistochemical analyses of cH2AX staining in organs from mice treated as in (A). Scale bar, 25 µm.
Data information: (A) Each experimental group included n = 5 mice. Error bars represent SEM. P-values were calculated using one-way ANOVA followed by Tukey's multiple comparisons test. ****P < 0.0001; *P < 0.05. (B) Organs from n = 3 mice were analysed for each treatment. Error bars represent SD. P-values were calculated using one-way ANOVA followed by Tukey's multiple comparisons test. ****P < 0.0001; *P < 0.05. Exact P-values are included in Appendix Table S1.
Since cases of cisplatin and chlorambucil resistance (Panasci et al, 2001;Norgaard et al, 2004;Norquist et al, 2011) in tumours are documented, it is important to identify which DNA damage repair defects underlie sensitivity to these drugs. This could lead to novel druggable targets that can sensitise tumours to these agents, with the caveat that combination therapies may increase toxicity. We demonstrate that ATR is a major determinant of sensitivity to chlorambucil and cisplatin. Most likely, DNA crosslinks generated by treatment with the two drugs obstruct DNA replication, and the resulting stalled forks require ATR for their re-start and stabilisation (Zeman & Cimprich, 2014). Furthermore, consistent with the well-established role of FA in ICL repair, we demonstrate that FANCD2 inactivation also sensitises cells to either chlorambucil or cisplatin. As a novel target, we identify SNM1A, a nuclease with a poorly understood function in ICL repair, as a factor required for the proliferation of cells treated with chlorambucil or cisplatin. Chemical inhibitors against this nuclease may re-sensitise resistant tumours to chlorambucil and/or cisplatin. Moreover, our results indicate that SNM1A functions in the context of the FA pathway of ICL repair, which helps elucidate the cellular roles of this nuclease.
A second caveat associated with the clinical use of cisplatin is its well-documented toxicity (Kelland, 2007), mainly in the form of nephropathies and gastrointestinal tract disorders. The problem was partially alleviated by the development of new platinum drugs with lower toxicity, but no benefit for patient survival over cisplatin. Chlorambucil, which has been part of the standard therapy for CLL patients for over 50 years either as single agent or in combination with engineered antibodies (Goede et al, 2013(Goede et al, , 2014, shows overall mild toxicity, occasionally in the form of pancytopenia (Rai et al, 2000).
Consistent with the notion that chlorambucil is less toxic than cisplatin in patients, our study demonstrates that in vitro and in vivo chlorambucil treatment triggers lower levels of apoptosis relative to cisplatin in healthy cells and tissues. We visualised apoptosis in mice using SPECT in vivo imaging of a radiolabelled apoptosis tracer and found that it correlates with cH2AX accumulation. IHC staining showed higher cH2AX levels in the organs of cisplatin-treated mice, consistent with cisplatin inflicting more DNA lesions than chlorambucil. To what extent cisplatin toxicity to healthy tissues can be attributed to its ability to induce irreversible intrastrand crosslinks remains to be determined. Importantly, the two drugs were equally effective in suppressing growth of BRCA2-deficient xenografts. Given the lower toxicity of chlorambucil relative to cisplatin in mice and given that the two drugs show comparable anti-tumoral activity against BRCA-deficient tumours, our results suggest that chlorambucil is a drug with a therapeutic index superior to cisplatin against the BRCA-deficient tumour subset. In addition, chlorambucil is a drug administered orally, like the PARP inhibitor olaparib, in contrast to cisplatin which is an intravenous drug.
One potential caveat associated with administering alkylating agents, including chlorambucil, to CLL patients is the relatively high risk of primary tumours at other sites (Maurer et al, 2016). Whilst concrete evidence for this association was lacking in an initial study (Hisada et al, 2001), a subsequent evaluation established an increased (but not significant) incidence of epithelial cancers and acute myeloid leukaemia in chlorambucil recipients (Grosse et al, 2009). Thus, potential carcinogenic effects of chlorambucil must be taken into consideration when evaluating chlorambucil and other agents known to inflict ICLs for clinical use.
Overall the results reported here suggest that the efficacy of chlorambucil assessed in early clinical trials of breast and ovarian cancer patients was obscured by the lack of molecular markers. Thus, patient stratification based on BRCA1/2 status may help identify those tumours vulnerable to chlorambucil. Recent studies have shown that treatments based on the alkylating agent cyclophosphamide are effective in the subset of BRCA1/2-mutated patients (Vollebergh et al, 2014). Our results corroborate the specificity of alkylating agents in this setting and suggest that the therapeutic potential of chlorambucil in BRCA1/2-mutated patients should be re-evaluated.

Materials and Methods
Cell lines and growth conditions mitosis with 0.2 lg/ml KaryoMAXâ Colcemid (Life Technologies) for 16 h.

Prestwick chemical library screen setup and statistical analysis
Two independent screens were performed, each in triplicate. BRCA2-deficient and BRCA2-reconstituted hamster cells, respectively, V-C8 and V-C8 + BRCA2 , were seeded in 96-well plates. Drugs of the Prestwick Chemical Library (http://www.prestwickchemical.com/prestwick-chemica l-library.html) supplied at 10 mM in DMSO were added 24 h later at 5 lM dilution in 100 ll of culture media. Each plate contained DMSO control wells. Following incubation with the drugs for 3 days (screen 1) or 6 days (screen 2), viability was assessed using resazurin-based assays.
The library consisted of 16 plates (32 plates were used per cell line in each screen); therefore, a median plate normalisation procedure was applied, corresponding to a modified version of the robust per cent of sample (Birmingham et al, 2009). Hits were ranked using strictly standardised mean deviation (Dataset EV1; Zhang, 2011).

Cell viability assays
Cells were plated at densities varying between 100 and 2,000 cells per well in 96-well plates. These densities were determined for each cell line individually, so that they reached 80-90% confluency after 7 days in culture in the absence of any treatment. Drugs were added at the indicated concentrations on the following day. Six days later, cell viability was determined by incubating cells with medium containing 10 lg/ml of resazurin for 2 h. Fluorescence was measured at 590 nm using a plate reader (POLARstar, Omega). Cell viability was expressed relative to cells treated with vehicle control of the same cell line, thus accounting for any differences in viability caused by genetic modifications.

Spheroid cultures
Ten thousand cells of colorectal adenocarcinoma DLD1 cells were plated in a well of a 96 round-bottom well, ultra-low attachment plate (Costar) before being centrifuged at 210 × g for 10 min. Spheroids were cultured for 4 days as in published protocols (Friedrich et al, 2009) before removal of half of the media and addition of the drugs diluted in that same volume. For untreated control spheroids, complete media containing the corresponding solvent was added. For drug-treated spheroids, fresh media containing the drug were added every 72 h. Pictures of the spheroids were acquired using a Nikon TE2000-E microscope, and volumes were analysed with MATLAB R2014a.

Immunoblotting
To prepare whole-cell extracts, cells were washed once in 1× PBS, harvested by trypsinisation, washed in 1× PBS and resuspended in SDS-PAGE loading buffer, supplemented with 0.1 mM DTT. Samples were sonicated using a probe sonicator and heated at 70°C for 10 min. Equal amounts of protein (30-100 lg) were analysed by gel electrophoresis followed by Western blotting. NuPAGE-Novex 10% Bis-Tris and NuPAGE-Novex 3-8% Tris-acetate gels (Life Technologies) were run according to manufacturer's instructions.

Preparation of metaphase chromosome spreads
Cells synchronised in mitosis via overnight incubation with 0.1 lg/ml KaryoMAXâ Colcemid (Life Technologies) were collected by mitotic shake-off and swollen in hypotonic buffer (0.03 M sodium citrate) at 37°C for 25 min. Cells were fixed in freshly prepared 3:1 mix of methanol: glacial acetic acid, and nuclear preparations were dropped onto slides pre-soaked in 45% acetic acid prior to being allowed to dry overnight. The following day, mitotic chromosomes were stained using Giemsa (VWR) and viewed with a Leica DMI6000B inverted microscope equipped with a HCX PL APO 100×/1.4-0.7 oil objective.

FACS analysis
BRCA2 +/+ and BRCA2 À/À DLD1 cells treated for 48 h with 1 lM of cisplatin or 1 lM chlorambucil were incubated with 10 lM EdU for 1 h at 37°C. Cells were harvested by trypsinisation and processed for EdU staining using the Click-iT Plus Alexa Fluor 647 Flow Cytometry Assay Kit (Thermo Fisher Scientific). Cells were incubated with 20 lg/ml propidium iodide and 10 lg/ml RNase A (Sigma) in PBS. At least 10,000 cells were analysed by flow cytometry (FACSCalibur, BD Biosciences). Data were processed using FlowJo (FlowJo, LLC) software.

In vivo xenograft experiments
CB17/SCID male mice (CB17/Icr-Prkdc scid /IcrIcoCrl, 6 weeks old and weighing 26-28 g) were purchased from Charles River Laboratories (Calco, Italy). The mice were maintained in high-efficiency, particulate air HEPA-filtered racks and were fed autoclaved laboratory rodent diet. All animal procedures were in compliance with the national and international directives (D.L. 4 March 2014, no. 26; directive 2010/63/EU of the European Parliament and of the council; Guide for the Care and Use of Laboratory Animals, United States National Research Council, 2011).
To generate xenografts derived from HCT116 BRCA2-proficient or HCT116 BRCA2-deficient cells, CB17/SCID mice were injected intramuscularly into the hind leg muscles with 5 × 10 6 cells per mouse. When a tumour volume of approximately 250 mm 3 was evident in BRCA2-proficient (4 days after cell injection) and BRCA2-deficient (6 days after cell injection) xenografts, mice were randomised in vehicle and treated groups and the treatment was initiated. For xenografts derived from DLD1 BRCA2-proficient or DLD1 BRCA2deficient cells, tumours were grown subcutaneously until they reached a mean volume of approximately 100 mm 3 (BRCA2-deficient) and 150 mm 3 (BRCA2-proficient), at which point the treatment was initiated. Each experimental group included five mice.
Chlorambucil (3, 1 or 0.3 mg/kg/day) was administered intraperitoneally for five consecutive days, followed by 2-day break and 5 more days of treatment. Based on previously established maximum tolerated dose (Leonetti et al, 1996), cisplatin (3.3 mg/kg/day) was administered intraperitoneally for three consecutive days, followed by 4-day break and 3 more days of treatment. Talazoparib (0.33 mg/ kg/day) was administered orally for five consecutive days, followed by 2-day break and 5 more days of treatment. At the time points indicated, tumour volumes were measured in two dimensions using a calliper and tumour weight was estimated from tumour volume (1 mg = 1 mm 3 ). Tumour weight inhibition was calculated using the formula a × b 2 /2, where a and b are the long and short sizes of the tumour, respectively, at the nadir of the effect. The number of mice used in each experiment is described in each figure legend.

Ex vivo drug experiments
The ex vivo drug treatment protocol was performed as previously described (Bruna et al, 2016). Briefly, frozen patient-derived tumour xenografts (PDTXs) were thawed and dissociated on the GentleMACS Dissociator (Miltenyi Biotec, Cat ID 130-093-235) using the Tumour Dissociation Kit, human (Miltenyi Biotec, Cat ID 130-095-929) and preset protocol h_tumour_01. Single cells were plated at~40,000 cells/ml in 96-well plates and dosed 72 h after plating. Cell Titer Glo 3D was added to the cells 6 days after dosing. Plates were read on the Pherastar plate reader using the Luminescence module.

In vivo apoptosis detection
Wild-type Balb/c female mice (6 weeks old and weighing 16-23 g) were purchased from Charles River (UK). Animals were housed in individually ventilated cages in sex-matched groups of up to six per cage in an artificial day-night cycle facility with ad libitum access to food and water. All analyses were performed blinded to experimental group assignment.
Mice were injected intraperitoneally with 0.9% saline (n = 5), 3.3 mg/kg/day cisplatin (n = 5) for 3 days or with 3 mg/kg/day chlorambucil (n = 5) for 5 days. 99m Tc-Duramycin was prepared as previously described (Palmieri et al, 2018). The purity was checked by RP18 HPLC and confirmed to be over 95%. Mice were injected intravenously 2 days after the end of treatment with 1 lg of 99m Tc-Duramycin (2-4 MBq) and were imaged 2 h later, using a VECTor â PET/SPECT/CT scanner (MILabs, Utrecht, the Netherlands). After the final imaging session, mice were euthanised by cervical dislocation and selected organs and tissues were removed. The amount of radioactivity in each organ was measured using a 1470 WIZARD gamma counter (PerkinElmer). Counts per minute were converted into MBq using a calibration curve generated from known standards. These values were decay-corrected to the time of injection, and the percentage of the injected dose per gram (% ID/g) of each organ was calculated. A second cohort of mice (n = 3 per group) treated with the same protocol were sacrificed for immunohistochemical staining of selected organs. These animal procedures were performed in accordance with the UK Animals (Scientific Procedures) Act 1986 and with local ethical committee approval.

Immunohistochemistry
Tissues collected from experimental mice were fixed in 10% formalin pH 7.4 (VWR Chemicals) for 24 h before dehydrating in 70% EtOH for 24 h and embedding in paraffin (ThermoScientific, Histostar and Histoplast paraffin). Formalin-fixed and paraffinembedded lungs and hearts were sectioned onto slides (ThermoScientific Superfrost Ultra-Plus) at 3-5 lm thickness. Immunohistochemistry was performed by de-parafinising (two baths of xylene, for 3 min each), dehydrating (two baths of 100% EtOH, for 3 min each) and rehydrating the slides (in baths of 95, 90, 70 and 50% EtOH for 3 min each). Controlled antigen retrieval was induced with citrate buffer (pH 6.0) for 2 min at 110°C (BioCare Medical, decloaking chamber). Slides were then prepared with Dako EnVision+ Dual link system, HRP kit dual enzyme endogenous blocking buffer for 15 min. Slides were incubated with primary antibody (rabbit antimouse phosphorylated H2AX Ser139, Abcam ab11174) diluted 1:500 overnight at 4°C. The secondary detection system was Dako EnVision HRP polymer-labelled rabbit antibodies and DAB was diluted with the substrate buffer 1:25 and washed of the slides with water after 2 min of incubation counterstained with haematoxylin (Sigma-Aldrich) and mounted with coverslips (Sigma-Aldrich DPX mountant for histology, Menzel-Glaser #1.5 coverslips).

The paper explained
Problem BRCA1 and BRCA2 are major tumour suppressors. Germline mutations in these genes are associated with approximately 25% of the familial cases of breast and ovarian cancer and a significant proportion of sporadic cancers show BRCA1 and BRCA2 gene inactivation. Thus, there is a large number of patients affected by BRCA1/2 deficiency. BRCA1 and BRCA2 proteins play key roles in homologous recombination repair and their loss triggers high sensitivity to DNA damage. Drugs that induce DNA lesions are routinely used in the clinic to treat BRCA1/2-deficient tumours (e.g. cisplatin and PARP inhibitors). However, most tumours acquire resistance to these therapies and novel strategies for their elimination are needed.

Results
Here, we report a screen of the Prestwick chemical library of FDAapproved drugs for compounds that target specifically BRCA2-deficient cells. We identify chlorambucil as the highest scoring hit from this screen, which selectively eliminates BRCA1/2-deficient cells and tumours, including olaparib-resistant and cisplatin-resistant ones. Importantly, chlorambucil is substantially less toxic to normal cells and tissues than cisplatin, a drug routinely used in the clinic for cancer treatment.

Impact
Chlorambucil and cisplatin are equally effective in targeting specifically BRCA1/2-deficient tumours. However, chlorambucil exhibits lower toxicity to normal cells. We propose that chlorambucil may provide an effective alternative to cisplatin for the treatment of this tumour subset, either as single therapy or in combination with other agents.