Synergistic effect of PARP inhibitor and BRD4 inhibitor in multiple models of ovarian cancer

Abstract Ovarian cancer has the highest facility rate among gynaecological tumours. Current therapies including PARP inhibitors have a defect that ovarian tumour is easy to recurrent and become resistant to therapy. To solve this problem, we found that BRD4 inhibitor AZD5153 and PARP inhibitor olaparib had a widespread synergistic effect in multiple models with different gene backgrounds. AZD5153 sensitizes cells to olaparib and reverses the acquired resistance by down‐regulating PTEN expression levels to destabilize hereditary materials. In this study, we used the following multiple ovarian cancer models PDX, PDO and 3D/2D cell lines to elucidate the co‐effect of AZD5153 and olaparib in vivo and in vitro. The similar results of these models further proved that the mechanism identified was consistent with the biological process occurring in ovarian cancer patients after drug treatment. This consistency between the results of different models suggests the possibility of translating these laboratory research findings into clinical studies towards developing treatments.

the PDO model can be used to examine drug activity in primary tumours, it could be beneficial to translational medicine studies 12,13 and have already been used to explore drug effects in ovarian cancer. 14 The patient-derived xenografts (PDX) model is another in vivo model that is widely used in precision medicine research, [15][16][17] and it retains the microenvironment and heterogeneity of the primary tumour. 18 The joint application of PDO and PDX in drug screening and mechanism exploration can be used in models to provide a closer simulation of the condition of patients. 19 In this study, we choose olaparib and AZD5153, respectively, as representative agents to further investigate the synergistic effects of PARP and BET inhibitors. olaparib has been confirmed to show less toxicity and off-target effects clinically 20,21 and it can be used at a high dose to achieve maximal PARP-inhibiting effect. 22,23 AZD5153 is a more specific small molecular inhibitor of the BET protein bromodomaincontaining protein 4 (BRD4). 8,24 The experimental models we selected were PDO and PDX, which mimic the tumour environment in patients.
We not only used these models to investigate the actions of these drugs, but we also explored the mechanisms underlying these actions using experiments that are usually conducted in cell lines. Cell lines were all passaged less than 30 times and were cultured under 37°C, 5% CO 2 incubator. 3D cell was cultured in the same condition as PDO models.

| Clinical specimens
All primary ovarian cancer tissues are anonymized and obtained from Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China, in accordance with the Declaration of Helsinki. All the operations were approved by the Ethics or Institutional Review Board.

| Establishment of PDX model
PDX (Patient-derived xenografts) models were obtained by subcutaneously transplanting fresh tumour tissue into nude mice.

| Generation of PARP inhibitor resistant cells
A2780, HOC7 and ID8 cells were cultured with an increased concentration of olaparib. After 3-4 months of treatment, these cells can grow rapidly in the presence of 10 uM olaparib. Cells were cultured in the absence of olaparib for 1 month. Before use, IC50 was calculated again to confirm its drug resistance.

| Fluorescence in situ hybridization (FISH) assay
The slides were pretreated with xylene and gradient ethanol to dewaxing and hydration, boiled, digest with 200 μL pepsin solution, and then put into 2×SSC at room temperature for 3 min.
Dehydration with gradient ethanol for 2 times and dried at room temperature. Samples and probes were hybridized in an environment protected from light. Then the slides were washed and counterstained.
The probes (LBP Guangzhou, # F.01005-01) can hybrid with chromosome 10 centromere (green signal) and PTEN gene (red signal). The normal cell contained two red and two green signals. A PTEN amplificated signal mode contains more red signals.

| Metaphase spread assay
Cells were planted in six-well plates, and cultured till sub-culture at 70% confluency. Cells were exposed to colchicine (100 ng/mL) (Sellek, S2284) for 3 h, collected and resuspension in hypotonic solution (0.075 M KCl) for 30 min at 37°C incubator. Cells were then fixed in methanol: acetic acid (3:1) at 4°C for 30 min and repeat for 3 times. Then the fixed cells were dropped on precooled slides and put into a 65°C incubator to air-dried. After cooled down, the slides were stained in 3% Giemsa and coded for blind analysis. A total of 25 metaphases was analysed from each sample to detect the presence of chromosomal fragments.
The correlation between gene expression and drug reaction of human tumour cell lines was obtained by using the web-based tool provided by GSCALite (http://bioin fo.life.hust.edu.cn/web/GSCAL ite/). The gene expression and drug reaction data were collected by GSCALite from CCLE, CTRP and GDSC.

| Western blot assay
Cells were lysed with RIPA buffer (Servicebio, G2002-100) containing protease and phosphatase inhibitor cocktail (Servicebio, G2002-100). The lysates were centrifuged at 12000 rpm 4°C for 20 min to collect supernatants. After determining the protein content, the cell lysates were separated by 10% SDS-PAGE and electrotransferred onto 0.45 um PVDF membranes. The membranes were blocked with 5% BSA-TBST at room temperature and then incubated with primary antibodies at 4°C overnight. Secondary antibody (Antgene) was incubated for 1 h at room temperature. Bands were visualized by using WesternBright ECL Kit (Advansta, 190113-13).
Staining intensity was assigned a score using a semi-quantitative four-category grading system: 0, no staining; 1, weak staining; 2, moderate staining; and 3, strong staining. Every core was assessed individually and the mean of three readings was calculated for every case. The tumour cell staining score was determined separately by two independent experts simultaneously under the same conditions. In rare cases, discordant scores were reevaluated and scored based on consensus opinion.

| In vivo small animal imaging technology
C57BL/6 mice were purchased from Beijing HFK Bioscience and raise in specific pathogen-free conditions. ID8 cells are planted intraperitoneally into mice.
Mice were anaesthetised intraperitoneally with 4% chloral hydrate (g / ml) at a dose of 150 ul/20 g body weight. 5 mg of the fluorescent substrate was injected intraperitoneally for 10 min. The anaesthetised and injected mice were placed in the instrument.
The images were collected by the small animal imaging instrument (SIEMENS Inveon), and the same low and high values of fluorescence signals were set for each group.

| Statistical analysis
GraphPad Prism 8 and IBM SPSS statistic 26.0 were used for statistical analysis. p < 0.05 was considered to indicate a significant difference.

| Olaparib treatment can increase the expression of PTEN which relates to olaparib resistance
Firstly, we analysed the difference between olaparib resistant and sensitive ovarian tumour tissues which were classified by PDO drug screening, we could see that PTEN changed significantly. The  RNA-SEQ data implied that the relative olaparib-resistant group had a higher PTEN expression than the sensitive group ( Figure 1A). At the genetic level, the signal mode detected by FISH assay also exhibited that PTEN had been amplificated after olaparib resistance occurred ( Figure 1B). From the NCI-60 tumour cell line data, which was analysed by the cell miner CDB website, it was found that the sensitivity of cells to olaparib was negatively correlated with PTEN expression ( Figure S1A). The mass spectral data of PTEN also indicated that this protein was highly correlated with DNA replication and chromosomal stability ( Figure 1C). The RFC and SMC protein families, which were overlapped between these two datasets, were related to olaparib resistance and affected by PTEN. Furthermore, pathway enrichment showed that pathways related to olaparib resistance were mainly correlated to DNA replication and chromo- Meanwhile, we found that the sensitivity to olaparib was enhanced after PTEN was downregulated, while after PTEN was restored in shPTEN cell lines, the sensitivity of olaparib decreased ( Figure 1H,I).
From the results above, we found that PTEN was essential in olaparib resistance.

| AZD5153 reverse olaparib resistance by reducing PTEN expression
We also compared the shPTEN and original parent cell lines, and the results showed that PTEN knockout triggered the endogenous DNA double strands break (Figure 2A). This phenomenon was observed using the comet assay and suggests that PTEN had a protective effect on the DNA. The examination of the chromosomes showed that olaparib treatment induced the repair of chromosomes in the original cell line, whereas those of the shPTEN cells exhibited complex aberrations at the same concentration of olaparib ( Figure 2B and Figure S1F). This result indicated that PTEN knockout played an important role in the maintenance of DNA stability and chromosomal structure, which could sensitize ovarian cancer cells to olaparib. This is similar to the effect of AZD5153. 7,25 So, we assume that AZD5153 may affect the sensitivity of olaparib by reducing PTEN.
Subsequently, the chip-seq data in the public database further verified the inhibitory effect of the BRD4 inhibitor on PTEN ( Figure 2C). and 3D environments ( Figure 2G).

| AZD5153 and olaparib showed a widespread synergistic cytotoxicity in multiple ovarian cancer models
In vitro ovarian cancer models, 15 cell lines and 22 PDO models (Table 1) were used to test the drug effects, and the result showed a marked synergistic anti-tumour effect on both models ( Figure 3A and Figure S2A). Among the experimental models with various genetic backgrounds, 86.7% (13/15) and 90.9% (20/22) of the cell lines and PDO models, respectively, were more sensitive to co-treatment with olaparib and AZD5153 than either drug alone.  Figure 3D).
Furthermore, phosphorylated H2A.X variant histone (γH2AX) staining verified that cells in the PDO spheroid were killed by the drug co-treatment ( Figure 3E). In the co-treated group, the dissociation of dead cells and decreased in diameter of the live PDO spheroids were obviously stronger than they were in the groups treated with either drug alone. Meanwhile, the combined therapeutic effect of AZD5153 and palbociclib in 2D cell lines were also synergistic lethality ( Figure S2B,C). The synergistic effect was also examined in the A2780 cell line where we measured the response to AZD5153, olaparib, and their combination under 3D conditions ( Figure S2D,E). While a published paper has confirmed that the repairing level was reduced, 26

| Co-treatment with AZD5153 and olaparib damaged DNA by affecting its replication
To further elucidate the mechanisms of combined lethal effect on AZD5153 and olaparib, we established two DNA fibre assays to examine the long-and short-term drug effects on DNA replication. In the long-term assay, we treated the ovarian cancer models for 2 (cell line) or 4 (PDO model) days before performing 5-chloro-2′-deoxyuridine (CIdU) and 5-Iodo-2′-deoxyuridine (IdU) labelling ( Figure 4A,B). After drug treatment, measurement of labelled DNA fibres revealed a significant decrease in CIdU + IdU tract length in the ovarian cancer models ( Figure 4C,D).
The tract length in the group co-treated with both drugs was significantly shorter than it was in the groups treated with either drug alone. The calculation of replication rate 27 also indicated that AZD5153 treatment slowed the fork progression rate of the DNA fibres, which was even slower in the co-treated group than it was in the groups treated with either agent alone. The ratio of IdU to CIdU indicated a steady rate of DNA fork replication. 28 The IdU/CIdU ratio decreased in the drug-treated groups, especially in the cotreated group, showing that the replication forks became unsteady and easier to degrade ( Figure 4E) after treatment.
In the short-time assay, the ovarian cancer cells were treated with drugs in between the CIdU and IdU labelling ( Figure 4F). The condition of the IdU tracts indicated that the drugs immediately affected DNA replication. Furthermore, the results showed that the co-treated group exhibited a typical signal pattern where the green fluorescence of the IdU tracts was much denser and shorter than those of the other groups ( Figure 4G). The groups treated only with olaparib and only with AZD5153 exhibited a denser green signal and shorter IdU tracts, respectively, than those of the control group. We also identified the following three major patterns of fibre labelling, elongated, stalled and new firing ( Figure 4H). 29,30 Olaparib treatment caused the development of more new firing DNA fibres, 31 whereas AZD5153 produced more stalled fibres.
Furthermore, in the replicating DNA fibres, AZD5153 treatment destabilized the forks and caused them to degrade in the early phase.
The combined effects of both olaparib and AZD5153 caused more DNA fibres to break, and they were damaged further.
The change in the protein levels observed using western blotting ( Figure 4I) verified the changes observed in the DNA fibre assay. Olaparib treatment up-regulated proteins related to DNA replication stress and replication fork stability, whereas they were down-regulated by AZD5153. The protein expression levels in the co-treatment group suggested that the DNA fibres were in an unstable state. The disordered DNA replication resulted in strand breaks and the comet assay confirmed that the PDO model and cell lines showed more DNA double-strand breaks after co-treatment than they did following monotherapy with either drug ( Figure 4J,K).

| AZD5153 and olaparib can also cause greater damage in chromosome and lead to apoptosis
We observed increased levels of micronuclei damage following cotreatment with olaparib and AZD5153, which suggests a more severe chromosomal breakage ( Figure 5A). We also investigated the chromosomal damage using a metaphase spread assay ( Figure 5B) and the results showed that chromosomal fragments, breakage and aberration were higher in the co-treated group than they were in the groups treated with either agent alone. This observation indicates that more damage occurred.
Cells with micronuclei and chromosomal breakage usually undergo apoptosis and we examined the cell status using immunofluorescence staining of the γH2AX foci ( Figure 5C). The result suggested that more cells underwent apoptosis occurred after drug treatment, whereas the western blotting also showed similar changes in protein level ( Figure 5D,E). Flow cytometry was also used to directly examine the apoptosis rate ( Figure 5F). Collectively, these results suggest that co-treatment induced greater cytotoxicity than treatment with either drug alone.  volumes of the subcutaneous PDX models were significantly lower after co-treatment than after treatment with either drug alone ( Figure 6A,B). The drug effect did not cause significant weight loss ( Figure S3A).

| AZD5153 plus olaparib inhibit ovarian cancer growth in vivo
After confirming the in vivo synergistic effect, we further tested whether the drug response was similar between the in vivo and in vitro models. First, a PDX-Organoids model was used to examine the drug reactivity ( Figure 6C) and the result showed that co-treatment with AZD5153 and olaparib inhibited the spheroid growth, which was consistent with the findings in the PDX model. The immunohistochemistry results also showed changes in cell amplification and apoptosis that were similar to those observed in vitro ( Figure 6D).
Then, we treated one of the PDX models with olaparib for 3 months, during which we collected tumour samples for paraffin sectioning ( Figure 6E). The FISH assay was used to amplify the PTEN gene over time and the results were consistent with those obtained in vitro.

| DISCUSS ION
In this study, we demonstrated the synergistic effect of AZD5153  which may have caused selection bias. To avoid the influence of this selection bias, we used multiple models to confirm our conclusions.
During the experiment, we also found that even with the same patient, differences occurred in drug responses. The drug responsiveness led us to infer that tumour cells isolated from ascites had the highest activity and drug resistance, and this was likely because they had an innate ability to survive as spheroids. 38 However, the reason for the higher drug resistance of tumour cells in ascites has not been fully elucidated yet.
In conclusion, we present strong evidence supporting the notion that AZD5153 and olaparib have a widespread synergistic effect.
Olaparib treatment upregulates PTEN, and then DNA and chromosome stability will rise, which made ovarian cancer acquire drug resistance. AZD5153 sensitizes cells to olaparib and reverses the acquired resistance by down-regulating PTEN expression levels to destabilize hereditary materials. In this study, we used the following multiple ovarian cancer models PDX, PDO and 3D/2D cell lines to elucidate the coeffect of AZD5153 and olaparib in vivo and in vitro. The similar results of these models further proved that the mechanism identified was con-

ACK N O WLE D G E M ENTS
Thanks to Huili Hu for providing technical guidance, and thanks to everyone in the research team for their help. This work was supported by the National Natural Science Foundation of China (82073259).

CO N FLI C T O F I NTE R E S T S TATE M E NT
We confirm that this work is original and has not been published elsewhere, nor is it currently under consideration for publication elsewhere. We declare that there are no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data generated or analyzed during this study have been included in the article. Additionally, the raw data that support the findings of this study are available from the corresponding author upon reasonable request.

O RCI D
Junpeng Fan https://orcid.org/0000-0002-9413-408X F I G U R E 6 AZD5153 and olaparib inhibit ovarian cancer growth in vivo. (A) Two cases of PDX models which were divided into four groups were given vehicle, AZD5153, olaparib, and dual drugs, respectively. The dosage was shown in the figure. The curve of tumour volume in 21 days is also shown. (B) The figure showed the volume of tumour tissue which were taken after the mouse was sacrificed after drug treatment for 21 days. (C) After PDX-O culturing and drug treatment, the size, and density of the spheroids in each group were shown in a bright field of microscope and by HE staining. The entire status of viability was calculated by CELLTiter GLO 3D, which was shown in the right. (D) Immunohistochemistry showed that the expression of Ki67, γH2AX, and PTEN in PDX#2 tumour tissue was changed after treatment. (E) The signal pattern of the FISH assay was performed before treatment, after 1 month, and after 3 months of olaparib treatment. (F) The figure showed the in vivo small animal imaging results of C57BL/6 mice planted by ID8 cells intraperitoneally and treated with drugs for 21 days.