Crystal structure-based discovery of a novel synthesized PARP1 inhibitor (OL-1) with apoptosis-inducing mechanisms in triple-negative breast cancer

Poly (ADP-ribose) polymerase-1 (PARP1) is a highly conserved enzyme focused on the self-repair of cellular DNA damage. Until now, numbers of PARP inhibitors have been reported and used for breast cancer therapy in recent years, especially in TNBC. However, developing a new type PARP inhibitor with distinctive skeleton is alternatively promising strategy for TNBC therapy. In this study, based on co-crystallization studies and pharmacophore-docking-based virtual screening, we discovered a series of dihydrodibenzo[b,e]-oxepin compounds as PARP1 inhibitors. Lead optimization result in the identification of compound OL-1 (2-(11-(3-(dimethylamino)propylidene)-6,11- dihydrodibenzo[b,e]oxepin )-2-yl)acetohydrazide), which has a novel chemical scaffold and unique binding interaction with PARP1 protein. OL-1 demonstrated excellent potency (inhibiting PARP1 enzyme activity with IC50 = 0.079 μM), as well as inhibiting PARP-modulated PARylation and cell proliferation in MDA-MB-436 cells (BRAC1 mutation). In addition, OL-1 also inhibited cell migration that closely related to cancer metastasis and displayed remarkable anti-tumor efficacy in MDA-MB-436 xenograft model without apparent toxicities. These findings highlight a new small-molecule PAPR1 inhibitor (OL-1) that has the potential to impact future TNBC therapy.

Currently, various PARP inhibitors, such as Olaparib, Rucaparib, BMN-673, Niraparib and Iniparib ( Fig. 1), are under development indifferent stages of clinical trial [14][15][16][17][18][19][20] . From a chemical point of view, most chemical scaffolds of PARP inhibitors contain amide structure, more new chemical structures can be found in the future 21,22 ; From a biological point of view, although these PARP inhibitors have high PARP1/2 inhibition and anti-tumor activity; however, long-term drug administration will accompany with drug resistance, leading to tumor recurrence and metastasis 23 . Thus, in addition to explore the in-depth drug resistance mechanism of existing inhibitors, as well as the relationship between PARP-mediated signaling pathways and tumor specificity, developing a new type PARP inhibitor with improved therapeutic efficacy and lower toxicity is alternatively promising strategy for TNBC therapy.
With the rapid development of computational methods and structural biology, many studies successfully identifying epigenetic inhibitors using pharmacophore-docking-based virtual screening and co-crystallization studies have been reported [24][25][26] . In this study, by constructing a pharmacophore of PARP1 inhibitor and screening a new chemical skeleton through co-crystallization studies, we designed and synthesized several series of PARP1 inhibitors, then identified a novel PARP1 inhibitor (OL-1). This inhibitor could significantly induce cell death and inhibit cell migration in BRAC1 mutant MDA-MB-436 cells with potent anti-tumor efficacy in vivo. These findings highlight a new small-molecule PAPR1 inhibitor (OL-1) that has the potential to impact future TNBC therapy.

Results and Discussion
Co-crystallization screening and structure-based pharmacophore of PARP1/inhibitor complex. Numbers of PARP inhibitors have been reported over the past several years, such as Olaparib, Rucaparib, BMN 673, Niraparib and Iniparib ( Fig. 1) [27][28][29] . These previous work had well described that PARP inhibitors occupy the nicotinamide pocket in the NAD + binding site of PARP1, forming key hydrogen bonds and π-π interactions. Firstly, we used virtual screening of chemical libraries that based upon Drugbank and ZINC databases, searching for novel leading compounds with distinctive skeleton (Fig. 2A). Top500 hits were selected by LibDock protocol in the first step. Subsequently, Top100 (PA-1 ~ PA-100) hits were further determined by CDOCKER protocol and selected for co-crystallization screening. As a result, only one compound from Drugbank database (DB00321) named as PA-10 (3-(10,11-dihydro-5H-dibenzo[a,d] [7]annulen-5-ylidene)-N,N-dimethylpropan-1amine) bound to the nicotinamide pocket of PARP1 (PDB ID code 5HA9) in the co-crystallization screening (Fig. 2B). To explore how to modify the leading compound, we constructed the structure-based pharmacophore including ten reported co-crystal structures of PARP inhibitors. The detected pharmacophore features were shown in Table 1 and Fig. 2C. Among these features, four of them were found as common in these complexes, including A1 (hydrogen bond acceptor), D1 (hydrogen bond donor), AR1 (ring aromatic) and H1 (hydrophobic) (Fig. 2D). Therefore, we can further modify the leading compound by increasing the length of carbon chain; increasing substituent containing hydrogen bond donor and changing of aromatic skeleton. And all newly synthesized compounds were designed according to abovementioned structure-based pharmacophore features.
To explore the impact of structure modifications in mother structures, as well as the framework reconstruction, compound 19 and 23 derived from dibenzo[b,e]oxepin-11(6H)-one (18) and dibenzo[b,e]thiepin-11(6H)-one (22) was also prepared (Fig. 6). Isobenzofuran-1(3H)-one (16) was treated with KOH in the temperature of the xylene reflux, then acidified with HCl to obtain compound 17 in 41% yield. The intermediate was then cyclizing in present of trifluoroacetic anhydride and BF 3 .Et 2 O to obtain compound 18 in 90% yield. A witting reaction similar to previous descriptions occurred in present of compound 10 with n-BuLi in −10 °C provided dibenzo[b,e] oxepin-11(6H)-ylidene derivatives (19a-c) in relatively high yields. Compound 22 was synthesized by a similar cyclization reaction in present of PPA and the final products dibenzo[b,e]thiepin-11(6H)-ylidene derivatives (23a-c) were obtained by a witting reaction similar to previous descriptions with n-BuLi in −10 °C.
Structural modification and structure activity relationship analysis. All synthesized compounds were tested to determine their PARP1 inhibition activities, and all compounds were further evaluated by cell viability assay in MDA-MB-436 cells (BRAC1 mutant breast cancer). The clinical small molecular PARP1 inhibitors Iniparib and Olaparib were used as the reference compound. First, 10,  annulen-5-ylidene core through a different length linker were synthesized to improve the molecular flexibility. Disappointingly, these compounds demonstrated negligible effects on PARP1 inhibition comparing with compound PA-10 (Table 2). Further, switch of the terminal N substituents to phenyl, afforded new derivatives 15a-e, showing less improvement in PARP1 activity (Table 3). Therefore, the structural modification of side chain exhibited when n = 1, R1 = R2 = Me, it had best activity. To further explore the impact of core structure, a series of bioisostere was synthesized, compound 19 and 23 was obtained through ibenzo[b,e]oxepin-11(6H)-one (18) and dibenzo[b,e]thiepin-11(6H)-one (22). Interestingly, both compounds displayed significantly enhanced PARP1 activity and anti-proliferative activity (Table 4), especially compound 19b, showing an IC 50 value of 0.75 μM. However, replacing the core structure to anthracen-9(10H)-ylidene or 9H-xanthen-9-ylidene, led to compounds 26a-c and 28a-c, possessing almost no PARP1 inhibitory activity (Table 5). From further analysis of co-crystallization and pharmacophore, we assumed that 2-substituted groups might be an important functional group interacting with PARP1 protein. Therefore, a series of 2-substituted 3-(dimethylamino)propylidene)-6,1 1-dihydrodibenzo[b,e]oxepin derivatives were obtained from compound 31. Lots of new compounds displayed significantly enhanced PARP1 activity, especially compound 33e (hereafter refer to OL-1), showing an IC 50 value of 0.079 μM against PARP1 and 0.736 μM against MDA-MB-436 cells (Table 6) and being 10-fold more potent than leading compound PA-10, while the one of the positive control Olaparib showing an IC 50 value of 0.005 μM     In addition, we also used the built pharmacophore to estimate the enzymatic activities of all synthesized compounds on PARP1 inhibition. As expected, OL-1 also displayed the best potency on PARP1 inhibition with an estimated IC 50 value of 0.29 μM (see Table S1). Consequently, based on abovementioned results, OL-1 emerged as the best leading compound with both potent PARP1 inhibition activity and good anti-proliferative effect against MDA-MB-436 cells. Moreover, we used molecular docking to examine the binding states between OL-1 and PARP1. As a result, OL-1 showed a good binding affinity with PARP1 with two hydrogen bonds formed in GLY863 (Fig. 9A). Then, we performed the 10-ns molecular dynamics (MD) simulations on OL-1/ PARP1 complex, and obtained the low root-mean-square deviation (RMSD) fluctuations, indicating OL-1 could steadily bind with PARP1 (Fig. 9B). After achieving the most potent compound OL-1, we conducted extensive structure-activity relationship (SAR) studies on part A, B, C (Fig. 10), The activity of the seven membered ring in Part A is superior to six membered ring and when the substituted X is O, the activity is better. In Part B, The more activity is shown when R2 is substituted for different amide groups. In Part C, the carbon chain needs a certain length and the best activity is shown when R1 is substituted by tertiary amine group. This analysis is also consistent with the results of previous molecular docking. Therefore, we selected OL-1 as the parent structure to remain unchanged.

OL-1 induces cell death in breast cancer cells.
To determine the molecular mechanism of OL-1, we firstly found that OL-1 demonstrated anti-proliferative effects against various breast cancer cell lines, especially in the BRCA1 mutant MDA-MB-436 cells (IC 50 = 5.14 µM) (Fig. 11A). Then we used Hoechst 33258 staining to confirm that OL-1-induced obvious morphologic alterations of apoptosis in MDA-MB-436 cells (Fig. 11B). In addition, we also measured the OL-1-induced apoptotic cell ratio by Annexin V-FITC/PI double staining, which was obviously increased in a concentration-dependent manner (Fig. 11C). Loss of BRCA1 function leads to genome instability because of defection in DNA repair by homologous recombination [30][31][32] . Consequently, BRCA1 deficient or mutant cancer cells are commonly sensitized to the inhibition of PARP/PAR-dependent DNA repair mechanisms due to the high-level of DNA damage 33,34 . Therefore, we determined the effect of OL-1 to inhibit the activation of PARP1 and downstream substrate proteins such as PAR in treated cells by western blot analysis. And we found that OL-1 treatment significantly inhibited the activity of PARP, accompanying with no cleavage of PARP. In addition, the expression of PAR was also decreased in OL-1 treated cells (Fig. 11D). Subsequently, we investigated the involvements of apoptotic markers in OL-1 induced cell death. We found that OL-1 upregulated Bax expression as well as downregulated Bcl-2 expression. And OL-1 treatment also increased cleavage of caspase-3 (Fig. 11D). Moreover, we found that OL-1 could inhibit cell migration of MDA-MB-436 cells (Fig. 11E), indicating OL-1 may inhibit metastasis. These results demonstrate that OL-1 could induce cell apoptosis by inhibiting PARP1 and inhibit cell migration in BRCA1-mutant MDA-MB-436 cells. Based      than the positive control (Iniparib, 100 mg/kg/d) and low dose groups (12.5 mg/kg/d) (P < 0.05) (Fig. 12A,B). For the toxicity study, compared with the control group and the Iniparib group, high dose of OL-1 (25 mg/kg/d) induced 10.05% loss of body weight during the 14 days of treatment (P < 0.01). In addition, the decrease of body weights in low dose group (12.5 mg/kg/d) was not obvious (Fig. 12C). Meanwhile, liver weights of Iniparib group were significantly decreased (P < 0.01), and spleen weights of mice were also affected by Iniparib (P < 0.05). The liver, spleen and kidney weights were not changed in OL-1 treated groups compared to the Iniparib group (Fig. 12D). In according to the balance between anti-tumor efficacy and toxicity, the low dose (12.5 mg/kg/d) was used as the optimum dose for treatment of tumor growth. To test whether OL-1-induced inhibition of tumor growth in vivo was due to reduced cell proliferation, we detected Ki-67 expression in tumor tissues of vehicleand OL-1-treated mice by immunohistochemical analysis. As a result, OL-1 treatment significantly reduced the positive ratio of Ki-67 compared to the control group (Fig. 12E). For further confirm the mechanism of the therapeutic efficacy of OL-1 in vivo, we examined the expression of PARP, PAR and Caspase-3 by western blot analysis. Interestingly, the expression levels of PARP and PAR were highly in accordance with the in vitro results (Fig. 12F).

OL-1 displays potent anti-tumor activity in vivo.
Altogether, these results demonstrate that OL-1 displays potent anti-tumor activity in vivo by inhibiting PARP1 and its substrate PAR.

Conclusions
In this study, we have described the discovery and identification of a potent and highly effective PARP1 inhibitor OL-1 (compound 33e) with a new chemical skeleton. This compound was designed and synthesized based upon co-crystallization studies of a hit compound PA-10. Further in-depth in vitro assays were performed with OL-1, which has displayed potent anti-proliferative activities in breast cancer cell lines, especially in MDA-MB-436 cells (BRAC1 mutation). And PARP enzymatic inhibition assay revealed that OL-1 potently inhibits PARP1 with an IC 50 value of 0.079 μM. Western blot analysis demonstrated that OL-1 significantly inhibited activities of PARP1 and its downstream substrate PAR. In vivo anti-tumor activity assays showed that OL-1 had more potent anti-tumor efficacy than Iniparib in the MDA-MB-436 xenograft model. By the way, OL-1 was also found to inhibit cell migration by in vitro would-healing assay, indicating OL-1 may have a potential to inhibit metastasis in triple negative breast cancer. And yet, preliminary pharmacokinetic studies and its efficacy of combination use with other anti-tumor drugs need further intense studies.

Methods
Chemistry. All reagents used in this study were purchased from commercial sources without any purification. All 1 H-NMR and 13 C-NMR spectra were tested in CDCl 3 or DMSO-d 6 by a Bruker-ARX-400 spectrometer. Chemical shifts were recorded in ppm. HRMS data were obtained by LC-ESI-TOF-MS instrument. The melting points were recorded in open capillaries and were uncorrected.

Molecular docking and molecular dynamics (MD) simulations.
Virtual screening of candidate PARP1 inhibitor was processed by the LibDock and CDOCKER modules of Accelrys Discovery Studio (version 3.5) 35,36 . All of the compounds contained in the screening library were downloaded from Drugbank (http://www. drugbank.ca/) and sub-library of ZINC build by NIBS (National Institute of Biological Sciences, Beijing), which contains 33,632 drug-like compounds. Energy minimization of the inhibitors was performed by the CHARMm force field 37 . All residues of PARP1 within 10 Å from the binding site of ligand were defined as the binding sphere. Additionally, Smart Minimizer and CAESAR (Conformer Algorithm based on Energy Screening and Recursive build-up) were applied for in situ ligand minimization and generating ligand conformations, respectively. Moreover, in order to detect the binding affinity and complex stability between PARP1 and OL-1, 10 ns MD simulations were processed by GROMACS (version 4.5.5) according to our previous study 38  X-ray diffraction data collection. Nylon loops was used to harvest the soaked PARP1 crystals and then immersed the crystals in mother liquor supplemented with 15% glycerol for 1 min. The synchrotron data were captured on an ADSC Q315 CCD detector (Shanghai Synchrotron Radiation Facility, Shanghai, China). HKL2000 was used to do the data processing.
Structure solution and refinement. PHASER program was used to do the Molecular replacement a probe PARP1 (Protein Data Bank (PDB) ID 4PJT). Then, REFMAC5 program was used to refine rigid-body by using maximum likelihood. The generated model was manually restructured by COOT program prior to refinement again by REFMAC5 program. PARP1 structure was analyzed by PYMOL program. Refinement statistics details were showed in Table S2.
Structure-based pharmacophore models construction. Ten co-crystal structure data of PARP1/ ligand complex were downloaded from the Protein Data Bank (PDB) 39 . The structure-based pharmacophore models were constructed according to our previous study 40 . In brief, all PARP1/ligand co-crystal structures were turned into a generic reference frame set by using "Multiple Structure Alignment (Modeller)" module in Discovery Studio 3.5. Subsequently, ten individual pharmacophore models based on PARP1/ligand complex were constructed by pharmacophore generation protocol of Discovery Studio 3.5. The identified pharmacophore features were filtered based upon the interaction patterns with PARP1 and showed in  Cell viability assay. 5 × 10 3 cells were seeded into each well in 96-well microplates and cultured for 24 h.
Then the cells were exposed to different concentrations of OL-1 for 24 h. After drug treatment, the cell viabilities were detected by MTT assay. Western blot analysis. Western blot analysis was carried out briefly as previous description 41 .
MDA-MB-436 cells were exposed to OL-1 for indicated time. Both floating cells and adherent were collected. The cell pellets were resuspended with RIPA lysis buffer and PMSF (1 mM) (Beyotime, Haimen, Jiangsu, China) and lysed at 4 °C for 1 h. After 12,000 rpm centrifugation for 10 min, the supernatant was collected to determine the protein content by the BCA Protein Assay Kit (CWBIO, Beijing, China). 30 μg cell lysates in each lane were separated by 8-12% SDS-PAGE and transferred onto PVDF membranes. After pre-blocking in TBST with 5% non-fat milk or BSA for 1 h, the membranes were incubated with primary antibodies overnight at 4 °C, and subsequently incubated with HRP-conjugated secondary antibody at room temperature for 1-2 h. Positive signals were detected by using ECL as the HRP substrate after washing with TBST solution.

Mouse experiments and in vivo xenograft tumor model. All experiments protocols used in this study
were carried out in accordance with guidelines of the animal ethics committee (Sichuan University). Thirty-two 6-8 weeks-old female BALB/c nude mice (18-20 g) were subcutaneously injected with MDA-MB-436 cells (1 × 10 7 cells/mouse). Until the tumor volumes reached 100 mm 3 (calculated as V = L × W 2 /2), the mice were randomly divided into four groups. Two groups were treated with different doses of OL-1 by i.p. (intraperitoneal) injection for 14 days (low dose group, 12.5 mg/kg/d; high dose group, 25 mg/kg/d), whereas the control group was treated with equal amount of normal saline (NS), and the positive drug group was treated with Iniparib, 100 mg/kg/d. Body weight and the tumor size were determined every day until the end of the study. All mice were sacrificed at the end of drug treatment. The organs of mice such as spleen, liver and kidney were harvested and weighed. Tumor tissues were detached and fixed in 4% paraformaldehyde for immunohistochemistry or lysed for western blotting.
Immunohistochemical analysis. Immunohistochemical analysis was carried out by the method of our previous study 42 . Samples were dehydrated using gradient ethanol, and subsequently paraffin embedded. The paraffin embedded samples were sliced into 5 μm thickness sections. The obtained sections were incubated with primary antibodies against KI-67 and PAR for 15 min followed by biotinylated secondary antibodies and detected with DAB. Nuclei were counterstained with hematoxylin. The numbers of positive cells were counted in at least 6 fields for each section and statistical analyzed.

Statistical analysis.
All the experiments were independently performed by at least three times. The data were statistical analyzed by One-way ANOVA or Student's t-test of SPSS 17.0 software. All tests with P < 0.05 were considered statistically significant.