Dual PARP and RAD51 Inhibitory Drug Conjugates Show Synergistic and Selective Effects on Breast Cancer Cells

The genetic principle of synthetic lethality has most successfully been exploited in therapies engaging Poly-ADP-ribose-polymerase (PARP) inhibitors to treat patients with homologous recombination (HR)-defective tumors. In this work, we went a step further following the idea of a local molecular cooperation and designed hybrid compounds M1–M3. The drug conjugates M1–M3 combine Olaparib, the first PARP inhibitor approved for clinical use, with Cpd 1, an inhibitor of RAD51 that blocks its HR functions and yet permits RAD51 nucleoprotein filament formation on single-stranded DNA. While in M2 and M3, the parental drugs are linked by -CO-(CH2)n-CO-spacers (n = 2 and 4, respectively), they are directly merged omitting the piperazine ring of Olaparib in M1. Monitoring anti-survival effects of M1–M3 in six breast cancer cell lines of different molecular subtypes showed that in each cell line, at least one of the drug conjugates decreased viability by one to two orders of magnitude compared with parental drugs. While triple-negative breast cancer (TNBC) cells with frequent BRCA1 pathway dysfunction were sensitive to spacer-linked hybrid compounds M1 and M2 regardless of their HR capacities, non-TNBC cells were responsive to the merged drug conjugate M1 only, suggesting different spatial requirements for dual inhibition in these two groups of cell lines. These results demonstrate that, depending on chemical linkage, dual PARP1-RAD51 inhibitory drugs can either sensitize non-TNBC and re-sensitize TNBC cells, or discriminate between these groups of cells.


Introduction
Application of drug conjugates (hybrid drugs) is an emerging approach to overcome drawbacks of current anticancer treatment, such as insufficient potency and efficacy, high toxicity, and development of resistance [1]. In anticancer drug conjugates, two drugs (pharmacophores) are combined in one molecule displaying simultaneous action at two different targets. In comparison to a combination of two single-target drugs, a hybrid drug may offer an advantage of pharmacokinetic simplicity and fewer drug-drug interactions [2]. A multitude of dual-acting hybrids with pharmacophores belonging to various pharmacological classes, such as DNA-alkylating agents, organoplatinum complexes, histone deacetylase inhibitors, topoisomerase inhibitors, have been reported as potential anticancer agents [1][2][3][4].
Poly-ADP-ribose-polymerase (PARP) inhibitors encompass an established class of anticancer agents used for the treatment of BRCA-mutated advanced tumors [5]. PARP inhibitor treatment of such homologous recombination (HR) deficient tumor cells causes synthetic lethality, as treated cells can neither remove proliferation-associated oxidative base damage by PARP-dependent base excision repair nor by the backup pathway HR at stalled DNA replication forks [6][7][8][9]. In this way, PARP inhibitors target HR-deficient tumor cells with comparatively few side effects in treated patients. Olaparib, the first-inclass PARP-inhibitor, was approved for the treatment of platinum-sensitive and BRCAmutated types of ovarian cancer in the US and the EU in 2014, followed by numerous trials applying Olaparib or other PARP inhibitors as single agents or in combination therapies [10]. However, over the last years, tumor resistance mechanisms towards PARP inhibitor treatment were reported, such as reversion mutations in BRCA1 and BRCA2, loss of BRCA1 promoter hypermethylation, or overexpression of RAD51 [11]. In an attempt to overcome resistance, two series of hybrid ligands combining Olaparib with the histonedeacetylase (HDAC) inhibitor Vorinostat [12] and with the HSP90 inhibitor C0817 [13] have recently been developed.
Familial breast and ovarian cancer, as well as to a lesser extent, also prostate and pancreatic cancer, have been linked with mutations in HR genes including but not limited to BRCA1 and BRCA2, whose gene products act upstream of RAD51 in HR [14]. In exploiting the synthetic lethality principle beyond tumors with hereditary defects in the HR pathway, RAD51 as the key enzyme in HR, has become a novel therapeutic target in oncology [15] and several RAD51 inhibitors have already been reported [16]. In support of this concept, co-treatment of cells with a RAD51 inhibitor sensitizes breast cancer cells to single Olaparib treatment [17]. A dual PARP and RAD51 inhibitor might as well re-sensitize triple-negative breast cancer (TNBC) cells with hereditary and acquired defects in the BRCA1 pathway [17] after development of PARP inhibitor resistance reconstituting the HR machinery [18]. The molecular subtype of TNBC, accounting for approximately 10-15% of breast cancer with a classification based on absence of estrogen receptor (ER), progesterone receptor, and human epidermal growth factor receptor 2 (HER2) markers, is characterized by a more aggressive course of the disease and limited treatment options as compared to the hormone receptor-positive luminal and HER2-amplified breast cancer types [19].
Here, we describe a novel class of anticancer drug conjugates M1-M3 obtained by molecular hybridization of Olaparib and a recently reported blocker of RAD51-mediated D-loop formation (see Figure 1) [20]. This so-called Compound 1 (Cpd1) has been chosen, since it more specifically inhibits HR as compared with other RAD51 inhibitory agents, which also prevent formation of RAD51-ssDNA nucleoprotein filaments, and thereby inhibits both HR and replication fork stabilization [20]. Accordingly, the dual inhibitors will not interfere with colocalization of RAD51 and PARP1 at DNA damage sites such as stalled replication forks. The cell lines chosen for analyzing anti-survival effects of these drugs represented the clinically relevant subtypes of breast cancer, i.e., non-TNBC with luminal and HER2+ breast cancer cells as well as TNBC without and with BRCA1 mutation.

Determination of Survival
The survival was determined by MTT assay (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide) (Sigma-Aldrich/Merck, St. Louis, MO, USA). First, the cells were seeded in 96-well plates at a density of 2 × 10 4 cells per well. On day two, four, and seven, the cells were treated with Olaparib (TargetMol, Boston, MA, USA), M1, M2, M3, or Cpd1 at 15 different treatment concentrations increasing up to 1024 µM in fresh culture medium each. On day nine, the MTT assay was performed. MTT was first dissolved in 1x PBS (Gibco/ThermoFisher Scientific) to a concentration of 5 mg/mL. Then, the 5 mg/mL MTT solution was diluted in OptiMEM (Gibco/ThermoFisher Scientific) and for HCC-1937 in RPMI without phenol red (Gibco/ThermoFisher Scientific) to a final concentration of 1 mg/mL. An amount of 100 µL of this solution were added to each well. After an incubation time of 2.5 h at 37 • C, the MTT solution was removed and the cells suspended in 200 µL 5% HCl/95% isopropanol (Sigma-Aldrich/Merck). The plates were shaken for 10 min and then the optical density measured at 570 nm with the Tecan Sunrise Photometer (Tecan, Crailsheim, Germany). The MTT experiments were conducted in duplicates and repeated twice each.

Western Blot Analysis
The cells were washed with ice-cold PBS, scraped in ice-cold PBS, and centrifuged. The The SDS-Page with a separating gel containing 10% Acrylamide (Roth, Karlsruhe, Germany), was run at 80 V and the proteins were transferred for 50 min at 100 V onto a PVDF membrane (GE Healthcare, Chicago, IL, USA). Membranes were sequentially incubated overnight at 4 • C with the following primary antibodies diluted in 5% milk: anti-Rad51 rabbit H-92, sc-8349 from Santa Cruz Biotechnology, Dallas, TX, USA; mouse anti-α-Tubulin monoclonal antibody ab7291-100 from Abcam, Cambridge, UK, diluted 1:5000. Then, membranes were washed three times for 10 min with TBS-T [20 mM Tris, pH 7.6, 137 nM NaCl (VWR international, Randor, PA, USA), 0.1% Tween20 (Sigma-Aldrich/Merck, St. Louis, MO, USA)] followed by incubation for 1 h at RT with a secondary antibody anti-mouse or anti-rabbit (Rockland, PA, USA; 1:10.000). After three times of 10 min washing steps with TBS-T, the proteins were detected with Clarity TM Western ECL Substrate (Bio-Rad Laboratories, Hercules, CA, USA). The proteins were visualized by ChemiDoc TM MP Imaging System (Bio-Rad Laboratories, Hercules, CA, USA) and quantified with Image Lab 5.2.1 (Bio-Rad Laboratories, Hercules, CA, USA).

Statistics
Graphic presentation of cell viability curves was performed using GraphPadPrism versions 8 and 9 (GraphPad Software, San Diego, CA, USA). Statistical significance was determined using extra sum of-squares F-test, nonlin fit. Statistical significances (p values) of differences between mean IC 50 values for unpaired, nonparametric data were first determined via Kruskal-Wallis test followed by a two-tailed Mann-Whitney-U test in case of statistical significance (p < 0.05). Statistics are shown in Supplementary Tables.

Results
Drug conjugates M1-M3 have been designed by molecular hybridization of the PARPinhibitor Olaparib and of the RAD51-inhibitor Cpd1 [20] (Figure 1). In M1, the parental drugs are directly fused omitting the piperazine ring of Olaparib in a similar way as for the previously reported Olaparib-Vorinostat [12], and Olaparib-HSP90-inhibitor [13] anticancer hybrids. As for M2 and M3, the two pharmacophores are linked by short -CO-(CH 2 )n-CO-spacers, n = 2 and 4, respectively.
The hybrid ligands were screened for their inhibitory effects on the viability of six different cancer cell lines, comprising two luminal (ZR75-1, MCF-7), one HER2+ (MDA-MB-453), and three TNBC lines without (MDA-MB-468) and with (HCC-1937, MDA-MD-436) pathogenic BRCA1 mutations (Table 1) [22,23]. The dose-response survival curves are shown in Figure 2a, visualizing that the hybrid compound M1 with Olaparib and Cpd1, directly connected by an amide linkage, kills five out of six cell lines included in our study with maximum efficiency, i.e., all lines with the exception of MDA-MB-436 cells.
Comparison of the corresponding IC 50 values (Table 2) for the drug conjugate M1 reveals a 28-to 56-fold more pronounced inhibition of viability than achieved by the parental drug Olaparib and a 44-to 110-fold higher potency than parental Cpd1 (p < 0.0001, see Statistics in Supplementary Tables). These data indicated a synergistic action of simultaneous PARP and RAD51 inhibition by M1 in these five cell lines.   When compared to the parental drugs, drug conjugates M2 and M3 with the two pharmacophores linked by short spacers did not sensitize non-TNBC cell lines (ZR75-1, MCF-7, MDA-MB-453) but the TNBC cell lines (MDA-MB-468, HCC-1937, MDA-MB-436) under investigation. As for MDA-MB-468 cells, a gradual increase of IC 50 values was noticeable with increasing distance of the pharmacophores, whereby M1 was 6-fold more potent than M2 and M2 4-fold more potent than M3 (p < 0.0001). As for HCC-1937 and MDA-MB-436, M2 and M3 were equipotent (p ≥ 0.1297) suggesting that the small difference in spacer length (two CH 2 -groups) between M2 and M3 had no major effect on cancer cell viability in these BRCA1-mutated cells.
Cell line group comparison of the average drug responses in Figure 2b highlights the striking potency of M1 as compared to all other drugs in non-TNBC cells with >36fold reduced mean IC 50 compared with the values for parental drugs. This was true regardless of the p53 and HER2 status ( Table 1). For the group of TNBC cells, M2 with a two-carbon and M3 with a four-carbon spacer between the parental drugs, showed >7-and >3-fold reduced mean IC 50 values. These observations indicate that the mode-of-action for dual inhibitory drugs targeting PARP and RAD51 are subject to tight 3D constraints in non-TNBC cell lines, while in TNBC cell lines a larger degree of flexibility exists.
Of note, both control drugs Olaparib and Cpd1 similarly reduced cell survival of all six cancer cell lines with IC 50 values ranging from 10-57 µM. Thus, though MDA-MB-436 showed the highest sensitivity to Olaparib (IC 50 = 10.21 µM), as would be predicted for BRCA1-mutated cells [8], the Olaparib response of HCC-1937 cells was indistinguishable from the one in the luminal cell lines ZR75-1 and MCF7. In support, Pierce and colleagues [24] determined similar IC 50 -values for MCF-7 and HCC-1937 cells after Olaparib treatments and found no statistically significant difference between IC 50 values from seven non-TNBC and 7 TNBC cell lines in both MTT as well as colony formation assays.
To provide additional information on the RAD51 status in these cell lines, we produced Western Blotting and immunofluorescence microscopic data of RAD51 foci formation at Olaparib-induced DNA damage sites marked by γH2AX (Figure 3). These results indicate that the RAD51 protein level was similar in all cell lines, even though a trend of reduced RAD51 expression was noticed in the BRCA1-mutated HCC1937 and MDA-MB-436 cells (Figure 3a). Interestingly, however, Olaparib-induced RAD51 foci at γH2AX-marked DNA damage sites in MDA-MB-436 cells were in the range of RAD51 foci numbers in non-TNBC or MDA-MB-468 TNBC cells but close to the detection limit in HCC-1937 cells despite γH2AX foci numbers in the normal range (Figure 3b). These data suggest compromised nucleoprotein filament assembly in HCC-1937 cells but not MDA-MB-436 cells, i.e., adaptation in the latter ones.
Finally, we also compared single versus combined Olaparib and Cpd1 treatments in ZR75-1, MCF-7, MDA-MB-468, and MDA-436 cells (Supplementary Figure S11). Interestingly, in none of these cell lines did combined treatment reduce the IC 50 values to the same extent as was observed for dual inhibitor treatment. Thus, no significantly different IC 50 values (e.g., for Olaparib versus Olaparib plus Cpd1 in MDA-MB-468 and MDA-MB-436) or at most 4.7-fold different IC 50 values (e.g., for Cpd1 versus Olaparib plus Cpd1 in ZR75-1) were observed for combined versus single treatments. For comparison, ≤110-fold differences were achieved with dual inhibitor treatments. From this, we conclude that the dual inhibitor synergistically enhances cell killing, an effect that is not reached by combined Olaparib and Cpd1 treatments, at least not to the same extent.
Altogether these findings indicate that neither PARP nor RAD51 inhibition alone nor combined drug treatment can discriminate between these groups of cell lines with in part differing RAD51 status. However, dual inhibition can break pre-existing PARP inhibitor resistance mechanisms bypassing BRCA1 dysfunction in HR and replication fork protection such as previously described for HCC-1937 [18].
MB-436 cells (Figure 3a). Interestingly, however, Olaparib-induced RAD51 foci at γH2AX-marked DNA damage sites in MDA-MB-436 cells were in the range of RAD51 foci numbers in non-TNBC or MDA-MB-468 TNBC cells but close to the detection limit in HCC-1937 cells despite γH2AX foci numbers in the normal range (Figure 3b). These data suggest compromised nucleoprotein filament assembly in HCC-1937 cells but not MDA-MB-436 cells, i.e. adaptation in the latter ones.

Discussion
Since their first approval in 2014, PARP inhibitors entered multiple treatment regimens to combat hereditary and sporadic tumor types including but not limited to BRCA1and BRCA2-mutated malignancies [25,26]. While side-effects of PARP inhibitory drugs are comparatively minor, acquisition of drug resistance has become the major challenge [27]. Given that PARP inhibitor resistance mechanisms can escape synthetic lethality by reconstituting the HR machinery in the tumor cells, several clinical trials combining the PARP inhibitor with another DNA repair inhibitory drug have been launched [25,26]. Here, we chose an alternative strategy by designing dual inhibitory drugs targeting both PARP and the key HR molecule RAD51. Thereby, we aimed at a spatial coupling of the two mechanisms underlying synthetic lethality and an increase of the local concentration of both inhibitors at the site-of-action, potentially reducing the pharmacologically necessary dose. Indeed, we found that in all cell lines, at least one of the dual inhibitory drugs M1, M2, or M3 reduced the IC 50 by one to two orders of magnitude compared to the parental drugs. Most interestingly, the drug conjugate M2 with a -CO-(CH 2 ) 2 -CO-spacer between the parental drug moieties discriminated between non-TNBC and TNBC cell lines with sensitivities differing by >20-fold. M3, with a two-carbons-longer -CO-(CH 2 ) 4 -CO-spacer, also clearly separated these cell line sets but to a lesser extent than M2.
Such a large impact of spacing argues for a role of PARP1 and RAD51 in a highly coordinated process in close proximity, very likely within the same nucleoprotein complex. Of note, recent work revealed that both PARP1 and RAD51 are not only key enzymes in base excision repair and HR, respectively [6,9], but are as well engaged in the controlled reactivation of stalled replication forks. Thus, meta-stable RAD51 nucleoprotein filaments aid fork reversal catalyzed by different translocases. Thereafter, stable RAD51 filaments and PARP1 protect reversed forks from unscheduled nucleolytic processing and remodeling by RECQ1, respectively [28,29]. Cpd1 is the only RAD51 inhibitor that specifically blocks D-loop formation during HR but not ssDNA binding, filament formation, or fork reversal by RAD51 in the concentration range up to the IC 50 values determined for breast cancer cell lines in our study [20]. Therefore, Olaparib-Cpd1 hybrid drugs should permit RAD51 loading and colocalization with PARP1 at stalled replication forks [30]. Of note, Olaparib treatment is known to trap PARP1 by allosteric effects on DNA, which correlates with cell killing and which was suggested to guide the development of improved drugs [31]. Therefore, synergy observed in our study with hybrid drugs could be caused by locking PARP1 on reversed forks adjacent to RAD51 filaments, which will lead to fork collapse, i.e., DNA double-strand breaks (DSBs) which, however, can no longer be repaired by HR.
The HR factors BRCA1 and BRCA2 have also been reported to protect forks, namely from nucleolytic attack through stabilization of RAD51 filaments [28]. Therefore, depletion of BRCA1, as frequently caused in TNBC cell lines by genetic or epigenetic mechanisms [17], will increase the plasticity of the fork. Such a scenario can explain why the BRCA1-positive non-TNBC cells were most sensitive to the spacer-free compound M1, suggesting close neighborhood between PARP1 and RAD51. Compatible with loosened spatial constraints, TNBC cell lines showed broader and more heterogeneous responses, with significant cell killing by M2 and M3 and in part also by M1. While the BRCA1-mutated line HCC-1937 showed a gradual decrease of cell killing with the length of the spacer between the parental drugs, BRCA1-depleted MDA-MB-436 cells exhibited the best response to M2 and M3. In MDA-MB-436 cells, the IC 50 values for M1 and Cpd1 were both even higher than for Olaparib, suggesting that M1 cannot simultaneously target PARP1 and RAD51. At this point, we can only speculate about potential changes in MDA-MB-436 cells impacting on interdomain communication between the C-terminal Olaparib-and N-terminal DNAbinding domains of PARP1 [31] or on the properties of residual RAD51 nucleoprotein filaments abrogating such dual interactions of M1. Ultimately, 3D structural analyses of the hybrid drugs bound to their targets are warranted to answer these mechanistically interesting questions.
Regarding candidate signaling pathways to explain these changes in MDA-MB-436 cells, we notice that PTEN, known to upregulate RAD51 [32], is deleted in MDA-MB-468 and HCC-1937 but not in MDA-MB-436 cells [19,33]. Here, we did not find elevated RAD51 expression in MDA-MB-436 cells. MDA-MB-436 cells were also reported to rapidly acquire BRCA1 reversion mutations partially reconstituting HR [34]. Indeed, in a previous study, we measured ≥10-fold higher HR frequencies in MDA-MB-436 compared with other breast cancer cell lines including MCF7, MDA-MB-453, and MDA-MB-468 [23]. In agreement with HR proficiency, here we observed RAD51 nucleoprotein filament formation at sites of Olaparib-induced DNA damage. Therefore, it is conceivable that MDA-MB-436 cells quickly adapted to BRCA1 loss by rewiring their RAD51 status similar to the pre-existing resistance mechanisms described for HCC-1937 cells [18]. Further resistance mechanisms, such as reduced DNA end protection by 53BP1 seen in all the TNBC cell lines investigated in our study [35], may explain why Olaparib sensitivities were in general similar in the TNBC as compared with non-TNBC cell lines.

Conclusions
In conclusion, Olaparib-RAD51 inhibitor conjugates have the potential to break resistance mechanisms to Olaparib treatment in TNBC cells and sensitize breast cancer cells regardless of BRCAness. While close proximity of the parental drugs in the hybrid is critical for their efficacy in non-TNBC cells, a separation of both pharmacophores by-CO-(CH 2 ) 2 -CO-and CO-(CH 2 ) 4 -CO spacers leads to TNBC cell killing. Specific targeting of PARP1 at sites of DNA repair and replication may be desirable when treating oncological diseases to avoid non-specific effects related to other PARP1 functions, such as a regulator of necrosis and inflammation in a sex-specific manner [6,36].

Data Availability Statement:
The data presented in this study are available on request from the corresponding authors.