Identifying and overcoming a mechanism of resistance to WEE1 kinase inhibitor AZD1775 in high grade serous ovarian cancer cells

Objective : As a result of TP53 gene mutation high grade serous ovarian cancer (HGSOC) is dependent on the G2 checkpoint for the repair of DNA damage and survival. The key role of WEE1 kinase at this checkpoint makes inhibition of WEE1 kinase in combination with DNA damaging agents an attractive therapeutic strategy for HGSOC. Our aim was to characterise resistance mechanisms to WEE1 inhibitor AZD1775 and identify ways to overcome resistance ready for use in the clinic. Methods : AZD1775-resistant HGSOC cell clones were isolated and western blotting, cell cycle analysis, growth assays, RNA-Seq and gene expression analysis were used to characterise resistance mechanisms and investigate a way to overcome resistance. Results : A resistance mechanism previously reported in small cell lung cancer did not operate in HGSOC. Instead, resistance resulted from different cell cycle control pathway changes that slow AZD1775-induced cell cycle progression and reduce accumulation of replication associated DNA damage. One major change was reduced levels of CDK1, the substrate for WEE1 kinase inhibition; another was increased levels of PKMYT1, which can also inhibit CDK1. Increased expression of TGF β signalling to slow cell cycle progression occurred in resistant clones. A TGF β R1 inhibitor overcame resistance in a clone with the highest TGF β R1 receptor expression. Conclusions : Although overexpression of the membrane glycoprotein MDR1 is a common mechanism of drug resistance, it was not involved in our HGSOC cells. Instead AZD1775 resistance resulted from cell cycle control pathway changes that combine to slow AZD1775-induced cell cycle progression and so reduce accumulation of replication-associated DNA damage.


Introduction
Ovarian cancer is the eighth commonest cause of female cancer mortality worldwide [1]. High grade serous ovarian cancer (HGSOC) is the commonest form and is characterised by TP53 mutation and genetic instability. There are two major molecular types, homologous recombination repair (HRR) proficient and deficient [2]. Most HGSOC cases are treated by a combination of surgery and adjuvant chemotherapy with carboplatin and a taxane [3]. Patients with HRR-deficient cancers respond well initially, but the majority relapse after first line chemotherapy due to drug resistance. Neoadjuvant chemotherapy prior to surgery is an attractive alternative for some patients that can provide an early indication of chemosensitivity, but there is currently no agreement about which patients are most appropriate for this treatment option [4]. For HRR-deficient HGSOC poly ADP-ribose polymerase (PARP) inhibitors have provided a major advance in treatment options [5], but inhibitor resistance can continue to limit survival [6].
HGSOCs depend on the G2 checkpoint for DNA damage repair because the G1 checkpoint is inactivated by TP53 mutation. The phosphorylation status of CDK1 controls the G2 checkpoint. DNA damage leads to inhibitory phosphorylation of CDK1 by WEE1 kinase resulting in G2 arrest and so providing time for DNA repair before mitosis [7]. In the absence of G2 arrest cells with damaged DNA enter mitosis prematurely leading to mitotic catastrophe [8][9][10]. Thus, WEE1 kinase inhibition with AZD1775 together with DNA damaging agents is a particularly attractive therapeutic strategy for HGSOC [11]. Phase I trials have reported target engagement and partial responses in patients with BRCA-deficient cancers [12,13]. A series of Phase II studies, involving patients with ovarian and other gynaecological cancers, have reported better progression free survival, or improved response rate with AZD1775 and a DNA damaging agent than with the damaging agent alone, although higher levels of adverse events have also been seen [14][15][16][17][18][19].
Our aim was to determine AZD1775 resistance mechanisms in HRR-proficient and -deficient HGSOC and find a way to overcome resistance that could be ready to deploy as the inhibitor entered the clinic. CDK1 is phosphorylated on Y15 by WEE1 kinase [20], while a different kinase, PKMYT1 (MYT1), phosphorylates CDK1 at T14 as well as Y15 [20]. We hypothesised that resistance could arise from changes in key DNA damage response and cell cycle control pathways. While our study was in progress resistance to AZD1775 in small cell lung cancer (SCLC) was found to be due to increased expression of AXL and MET receptors activating CHK1 via mTOR and pERK pathways. CHK1 then activated WEE1 kinase and inhibited CDC25 [21]. Resistance could be overcome by targeting AXL or mTOR. We also wished to investigate whether this very different resistance mechanism to the one that we have envisaged operates in HGSOC.

Cell culture assays
The human HGSOC cell lines ES-2 and OVCAR8, validated by short tandem repeat profiling, were sourced from the American Type Culture Collection. Culture conditions used are described in Supplementary Methods. All experiments were performed on cultures within 10 passages of supply. Clones resistant to WEE1 inhibitor AZD1775 (S1525, Selleckchem, Houston, TX) were isolated by plating cells at 1000 cells/well in 96-well plates in medium containing the selective AZD1775 concentration. Resistant colonies appeared after around 14 days of selection. Single colonies arising in individual wells were picked, expanded and maintained under selection. Selective medium was changed every 3-4 days. Sensitivity of cell lines to AZD1775 was determined by Sulphorhodamine B (SRB) assay [22]. 2000 cells per well were plated in 96-well plates containing a dilution series of AZD1775 (8 wells for each dilution). DMSO, the solvent for AZD1775, was maintained at <0.01% and had no effect on the assay. Each 5-day growth assay was repeated twice. Sensitivity of AZD1775-resistant clones to the TGFβ receptor1 inhibitor RepSox (S7223, Selleckchem) was also determined by SRB assay. Dose response curves and IC50 values with 95% confidence intervals were obtained using GraphPad Prism. For cell cycle analysis 50000 cells/well were plated in 6well plates. Twelve hours after plating cells were treated for 24 or 72 hours with AZD1775 before flow cytometry was carried out as described [23].

Western blotting
Protein extraction and western blotting were performed as described [24]. Antibodies used are shown in Supplementary Table 1.

RNA-Seq
The first experiment comprised 16

Isolation of clones resistant to AZD1775 from HRR-proficient and HRR-deficient HGSOC cell lines
Two validated HGSOC cell lines [25] were chosen for this study for their ability to form colonies when seeded thinly. Firstly, OVCAR8, with hypermethylation of the BRCA1 gene promoter [26] and shown to be HRR-deficient by a functional homologous recombination assay [27]. Secondly, ES-2, with an IC50 value in an SRB assay to assess sensitivity to PARP inhibitor olaparib that was >10fold higher than for OVCAR8 [28] and so was considered to be HRR-proficient. Like other HGSOC cell lines, both ES-2 and OVCAR8 are reported to have TP53 gene mutations [29]: ES-2, missense mutation in exon7, S241F; OV-CAR8, pathogenic RNA splice site mutation and this was confirmed by DNA sequencing (data not shown).

Cell cycle profiles of AZD1775-resistant clones are unaffected by AZD1775
Cell cycle profiles for the ES-2 parent and two clones resistant to 1000 nM AZD1775 (ES-2 AZ_1000:7_250:3 and ES-2 AZ_1000:2_250:7), derived from different 250 nM progenitor clones, are shown in Fig. 1C. Cultures were untreated or treated for 24 h with 1000 nM AZD1775. In AZD1775-treated ES-2 cultures the normal sharp distinction between cells in S and G2/M phases was lost. There was an increased frequency of cells in G2/M, reflecting the ability of WEE1 inhibition to drive cells into and then trap them in mitosis leading to a reduced frequency of cells in G1. Profiles of both resistant clones were unaffected by AZD1775. Compared to ES-2 the frequency of S-phase cells in both clones was reduced, likely reflecting the slower growth rate observed for all ES-2 and OVCAR8-derived AZD1775-resistant clones (Supplementary Table 3). The two progenitor clones resistant to 250 nM AZD1775 were next treated with 250 nM AZD1775 for 72 h (Fig. 1D). ES-2 cells treated in the same way again showed a highly dysregulated profile, but now with additional populations with subG1 (presumed apoptotic) and >4n DNA content (indicating continued DNA synthesis without cell division). By contrast AZD1775-treated ES-2 AZ_250:1 and ES-2 AZ_250:3 showed normal profiles. Two OVCAR8 clones resistant to 300 nM AZD1775 were also treated for 72 h with 300 nM AZD1775 (Supplementary Fig. 2). As expected, AZD1775-treated resistant clones showed normal cell cycle profiles, but the AZD1775 effect on S-phase in the OVCAR8 parent was not as noticeable as on ES-2.

The mechanism of AZD1775 resistance reported in SCLC does not operate in HGSOC
In SCLC cell lines resistance to AZD1775 was due to increased expression of AXL and MET receptors, leading to increased activation of CHK1 via increased signalling through mTOR and ERK/p90RSK pathways [21]. Western blotting was first carried out to see if the same mechanism of resistance was operating in our HGSOC cell lines ( Fig. 2A). Although AXL and MET expression in OVCAR8 clones was slightly increased, none of the ES-2 and OV-CAR8 clones showed consistent increased expression of AXL and MET receptors, nor of any markers for activation of their signalling pathways as seen in SCLC (mTOR pS2448, S6 pS240/244, and ERK pT202/Y204). When levels of activated CHK1 pS345 were compared in parents and clones both with and without AZD1775 treatment, it was clear that the observed increase in CHK1 pS345 was as a result of exposure to AZD1775 rather than as part of the resistance mechanism (Fig. 2E). Activated CHK1 levels were lower in AZD1775-treated resistant ES-2 clones than their parents. We conclude that the mechanism of AZD1775 resistance reported in SCLC is not seen in HGSOC.

Altered cell cycle control protein levels in AZD1775-resistant HGSOC clones
Sequencing of all the coding exons of the WEE1 gene from all resistant HGSOC clones found no evidence for mutation, ruling out the most direct resistance mechanism, whereby a WEE1 mutation would prevent inhibition by AZD1775. Levels of WEE1 and PKMYT1 kinases and their phosphorylation target CDK1 were then determined in resistant HGSOC clones (Fig. 2B). Increased expression of WEE1 kinase above the parental level was found in OVCAR8 clones, but ES-2 clones showed no consistent change. One ES-2 clone, ES-2 AZ_250:3, showed an increase, but the WEE1 kinase level in its derivative clone, ES-2 AZ_1000:2_250:3, was very low. Expression of PKMYT1 was very low in both ES-2 parent and resistant clones. PKMYT1 expression was much higher in OV-CAR8 and its resistant clones, with the level in OVCAR8 AZ_300:6 marginally increased over the parent. The reduction in total CDK1 level seen in some ES-2 clones in Fig. 2B was studied further in a larger group of resistant clones. A reduced level of total CDK1, most noticeable in clones resistant to 1000 nM AZD1775, was seen in most of the ES-2 derived clones tested, but in only a single OVCAR8 clone (Fig. 2C). This suggests that reduced CDK1 levels could be involved in resistance in ES-2, but not in OVCAR8 clones.
To confirm continued WEE1 kinase inhibition by AZD1775 in resistant clones, cell lysates were prepared from ES-2 and two resistant clones grown with and without 250 nM AZD1775 for 24 h (Fig. 2D). Both parent and resistant clones treated with AZD1775 showed a major reduction in phosphorylation of CDK1 at Tyr15, so demonstrating continued target engagement of AZD1775 with WEE1 kinase in resistant clones. The OVCAR8 parent and three resistant clones grown with or without AZD1775 treatment for 48 h also demonstrated clear inhibition by AZD1775 of WEE1 kinase activity (Supplementary Fig. 3A).
Overexpression of the membrane glycoprotein MDR1 is involved in resistance to olaparib and some chemotherapeutic agents [30]. Its possible involvement in AZD1775 resistance was investigated and the result is shown in the bottom panel in Fig. 2E. A cisplatin resistant colorectal cancer cell line overexpressing MDR1, HCT-15, was used as a positive control [31]. Although MDR1 levels showed very minor variation between resistant clones, levels were >10fold lower than in HCT-15. Continued AZD1775 target engagement, together with failure to observe MDR1 overexpression, both argue strongly against increased expression  of this drug transporter as a mechanism of AZD1775 resistance in our HGSOC cells.

The DNA damage response is altered in AZD1775-resistant HGSOC clones
In addition to preventing G2 arrest and trapping cells in mitosis, inhibition of WEE1 kinase also disables the S-phase checkpoint resulting in faster passage through Sphase and increased levels of replication-associated DNA damage, such as single-and double-stranded DNA breaks, stalled replication forks, and single-stranded DNA regions [32][33][34]. The DNA damage response (DDR) between untreated and AZD1775-treated parent and resistant clones was compared by western blotting (Fig. 2E and quantified in Supplementary Fig. 3B). As expected, all AZD1775treated lysates showed increased levels of the DNA damage marker γH2AX [35] compared to untreated lysates. However, when treated with AZD1775, all resistant ES-2 clones showed reduced γH2AX levels compared to the ES-2 parent. Clone OVCAR8 AZ_300:1 showed the same reduced γH2AX response to AZD1775 as the ES-2 clones, while three other OVCAR8 clones showed a similar increased response as OVCAR8. There were also major reductions in levels of activated DNA-PK, which is involved in dsDNA break repair [36], and of activated RPA, a marker for ssDNA regions [37], in all AZD1775-treated ES-2 clones compared to the treated parent. AZD1775treated OVCAR8 AZ_300:4 showed the same low level of activated DNA-PK as ES-2 clones, while the remaining three OVCAR8 clones showed the same high level of activated DNA-PK as the OVCAR8 parent. Activated RPA levels were slightly lower in three treated OVCAR8 clones than the OVCAR8 parent and marginally increased in OV-CAR8 AZ_300:3.
The activated ATM response mirrored the DNA-PK and RPA result, with generally reduced activation levels in treated resistant clones relative to treated parent. The altered DDR to AZD1775 in resistant clones relative to the parent suggests that the ability to reduce the level of DNA damage resulting from exposure to AZD1775 could be an important component in the development to resistance. Altered DDR was more pronounced in resistant ES-2 than OVCAR8 clones and we also observed interclonal heterogeneity in the response.
Levels of a marker for chromosomal condensation and mitosis, H3 pS10 [38], were 8-10-fold elevated in both treated parental lines compared to untreated cultures ( Fig. 2E and quantified in Supplementary Fig. 3B), reflecting the ability of AZD1775 to prevent G2 arrest resulting in premature entry into mitosis [11]. Whereas, in resistant clones where levels of H3 pS10 could be detected, the level of elevation of H3 pS10 in AZD1775-treated compared to untreated cultures was much lower, indicating that this could also be important for resistance. Although note that failure to see increased H3 pS10 expression on AZD1775 treatment in OVCAR8 AZ_300:2 was due to unusually high expression in the untreated lysate rather than a smaller increase on treatment.

Cell cycle and related pathway genes have altered expression in AZD1775-resistant clones
To better understand the transcriptional events associated with drug resistance we carried out RNA sequencing across our cell lines. Because we were most interested in identifying changes in gene expression in resistant clones that allowed them to grow in the continuous presence of AZD1775, while parental cells died on prolonged exposure, we made two separate gene expression comparisons. At the centre were parental cells treated with AZD1775. To one side they were compared against resistant clones treated with AZD1775. To the other side they were compared against untreated parental cells. Gene Set Enrichment Analysis (GSEA, see Supplementary Methods) of differentially expressed genes between parental and resistant clones was carried out. The analysis revealed particular enrichment across a number of altered gene networks, including cell cycle regulation, TGFβ signalling which has an important role in cell cycle arrest, and apoptosis. Since our western blotting analysis also indicated that altered cell cycle control protein levels were involved in resistance to AZD1775, we decided to concentrate on a detailed comparison of gene expression differences between parent and resistant clones for these three networks (Fig. 3). For each network a clear change in expression was observed between the ES-2 parent (treated and untreated) and the two treated clones resistant to 1000 nM AZD1775. Many highly expressed genes in the parent showed reduced expression in the clones, and vice versa. This shift was not simply due to global alterations, as evidenced by analysis of a similar sized control set of randomly selected genes. An intermediate expression pattern was seen in the clones resistant to 250 nM AZD1775 from which the 1000 nM clones were derived. A higher concentration of AZD1775 would likely have been required for the difference in expression of cell cycle control genes between untreated and treated ES-2 parent to better reflect the very different cell cycle profiles shown in Fig. 1C. Treated clones OVCAR8 AZ_300:2 and OVCAR8 AZ_300:3 showed a similar altered pattern of cell cycle control gene expression compared to the treated OVCAR8 parent as seen in the comparison between ES-2 clones and the ES-2 parent (Supplementary Fig. 4). Fig. 4 is adapted from the KEGG cell cycle pathway to show gene expression changes between parent and resistant clones in key parts of the pathway. Changes in gene expression levels for the entire cell cycle control pathway for which expression data were obtained are shown in Supplementary Figs. 5-7. Fig. 4A shows the comparison between AZD1775-treated and untreated ES-2 parent. As would be expected for untreated actively cycling cells, increased expression of cyclins (CCNE and CCNA)  Table 4 for the gene lists. and cyclin dependent kinases (CDK2) and reduced expression of CDK inhibitors (notably CDKN1A and CDKN2B) was observed compared to the treated parent. Reduced expression of the TGFβ pathway in untreated compared to treated parent was indicated by low expression of SMADs2-4, which slow cell cycle progression by inducing expression of CDKN1A and CDKN2B. Expression of SKP2, which encodes an E3 ubiquitin protein ligase mediating degradation of CDKN1A-C, was strongly increased in untreated compared to treated ES-2 parent.
Comparison of treated resistant ES-2 clones with treated ES-2 parent revealed a number of differences (Fig. 4B). There was decreased expression of most cyclin and CDK genes in resistant clones. There was also downregulation of many other genes involved in active cycling, growth and cell division. Unlike the treated parent, the TGFβ pathway was activated in both clones, although the mechanism of activation between clones appeared different. Increased expression of TGFβR1 receptor was found in ES-2 AZ_1000:2_250:3, while ES-2 AZ_1000:2_250:7 has increased TGFβ ligand expression. Increased SMAD3 expression was highest in ES-2 AZ_1000:2_250:3. Strongly increased expression in both resistant clones of CDKN1A and CDKN2B was the other striking change. Expression of both these CDKN genes is efficiently induced by SMAD3/4 [39]. Increased CDKN inhibitor expression was consistent with reduced expression of CDKs and cyclins. There was no evidence for the postulated involvement in resistance of increased expression of WEE1, PKMYT1, or CDC14A in the ES-2 clones. Increased expression of TP53 and reduced expression of CDC25A-C indicated some increased expression of DDR pathway genes in the clones ( Supplementary  Fig. 6). The ability of p53 to activate CDKN1A and GAD45 in the DDR is inoperative as a result of TP53 mutation found in both ES-2 and OVCAR8. However, note that the increased activity through the TGFβ pathway fulfils the same function.
Cell cycle control gene expression comparison between treated resistant OVCAR8 clones and treated OV-CAR8 parent showed less marked changes and more heterogeneity than seen with ES-2 clones (Fig. 4C). Decreased expression of most CDK and cyclin genes was again observed, but there was increased expression of CCNE1-2 and CDK6 in OVCAR8 AZ_300:2 and OVCAR8 AZ_300:3. TGFβ pathway expression was again elevated with increased expression of TGFβR1 and TGFβ2 in OVCAR8 AZ_300:4 and elevated SMAD3/4 in both OVCAR8 AZ_300:2 and OVCAR8 AZ_300:3. The elevated CDKN1A and CDKN2B expression observed in the ES-2 clones was not seen in OVCAR8 clones. Unlike the situation in ES-2, all OV-CAR8 clones had increased WEE1 expression and OV-CAR8 AZ_300:2 and OVCAR8 AZ_300:3 also showed in- (C) Gene expression changes between three treated OVCAR8 clones resistant to 300 nM AZD1775 compared to the treated OVCAR8 parent. For each gene box, expression in OVCAR8 AZ_300:2 is on the left, OVCAR8 AZ_300:3 in the centre and OVCAR8 AZ_300:4 to the right. Blue, reduced expression in the OVCAR8 clone compared to the parent; yellow, increased expression. creased expression of PKMYT1, indicating their possible involvement in resistance. As with ES-2 clones, there was no change in expression of CDC14A. TP53 and GAD45A DDR genes showed increased expression in clones OVCAR8 AZ_300:2 and OVCAR8 AZ_300:3 (Supplementary Fig.  7). DNA replication (ORC) and chromosome segregation (MCM) genes showed increased expression in resistant OVCAR8 clones, whereas these genes were strongly downregulated in ES-2 clones compared to the treated parent. This, together with increased expression in OV-CAR8 clones of genes involved in progression through G1 into S-phase, is hard to reconcile with the longer cell doubling times seen in both ES-2 and OVCAR8 clones (Supplementary Table 3).

Attempting to overcome resistance
On the basis of our RNA-Seq results, where TGFβ signalling was upregulated in resistant ES-2 and OVCAR8 clones, we postulated that a targeted inhibitor blockade of this pathway could overcome AZD1775 resistance. Rep-Sox, a TGFβ receptor1 inhibitor with good specificity and potency [40], was chosen to test our hypothesis. Growth assays were performed on resistant ES-2 and OVCAR8 clones and on parental lines. Resistant OVCAR8 clones (OVCAR8 AZ_300:3 and OVCAR8 AZ_300:4) showed no increased sensitivity to the inhibitor (data not shown). Fig. 5 shows growth assays where ES-2 was compared with resistant clones ES-2 AZ_1000:2_250:3 and ES-2 AZ_1000:2_250:7 in the presence of the inhibitor. IC50 determinations for ES-2 AZ_1000:2_250:3 showed 20-fold increased sensitivity to RepSox compared to both ES-2 parent and ES-2 AZ_1000:2_250:7 and so identified a potential route to overcome AZD1775 resistance in this clone.

Discussion
PARP inhibitors have provided a major therapeutic advance for women with HRR-deficient HGSOC [5], although the appearance of resistance can continue to limit survival [6]. The widespread occurrence of TP53 mutation, leading to reliance on the G2 checkpoint, makes AZD1775 inhibition of WEE1 kinase together with DNA damaging agents a particularly attractive therapeutic option for all molecular HGSOC subtypes. While WEE1 kinase inhibition in combination with PARP inhibitors has the additional advantage of inducing both DNA replication stress and nucleotide resource depletion [41]. Our aim was to identify resistance mechanisms to AZD1775 in both HRR-proficient (ES-2) and -deficient (OVCAR8) HGSOC cell lines and then find strategies to overcome resistance that could be deployed in the clinic. The maximum AZD1775 plasma concentration reported in patients with solid refractory tumours is 1650 nM [42]. Thus, the 5-fold increase in IC50 values for AZD1775 that we observed from~200 nM in ES-2 up to >1000 nM in resistant clones is in the physiologically relevant range to make the difference between cancers respond- ing or being refractory to AZD1775 therapy. We postulated that resistance could arise through changes in the DDR and cell cycle control pathways (see Fig. 6). During our study a different resistance mechanism was described in SCLC, involving increased expression of AXL and MET receptors signalling down to increase CHK1 activation and restore G2 arrest [21]. We showed clearly that this mechanism was not operating in HGSOC.
WEE1 gene mutation, preventing inhibitor binding, or increased WEE1 gene expression could overcome AZD1775 inhibition. Sequencing the WEE1 gene from resistant clones excluded involvement of coding region mutation in resistance. Western blotting and RNA-Seq for ES-2 clones showed no consistent increase in WEE1 kinase expression, but an increase at both protein and RNA levels was seen in OVCAR8 clones. Although this could contribute to the development of resistance in OVCAR8, the demonstration of continuing target engagement of WEE1 kinase by AZD1775 in ES-2 and OVCAR8 clones shows that other changes must also be involved. Continuing target engagement is also consistent with failure to observe increased MDR1 expression in resistant clones. Resistance to some chemotherapeutics and PARP inhibitors commonly results from overexpression of the MDR1 drug transporter [30], but we found no evidence that this was responsible for resistance to AZD1775 in our cells.
CDK1 can also be phosphorylated by PKMYT1 on Y15 as well as T14 [20]. Thus, PKMYT1 could potentially substitute for AZD1775-inhibited WEE1 kinase to maintain where we hypothesised that changes could lead to resistance to WEE1 kinase inhibition. Mechanisms 1 to 7 were postulated prior to commencement of the study, mechanisms 8 and 9 were revealed by the study. The figure summarises the results obtained for the different possible mechanisms. Ticks indicate a change that was observed. Crosses denote a possible change that was not seen. Pink ticks and crosses, resistant ES-2 clones; blue ticks and crosses, OVCAR8 clones. Superimposition of a tick and cross denotes a change only found in some resistant clones from that cell line. The crosses underneath postulated mechanism 6, increased expression (activation) of CHK1, indicate that no evidence was found for increased activation of CHK1 as a result of increased signalling from AXL and MET receptors as seen in SCLC. the ability for G2 arrest. Expression of PKMYT1 was very low in both the ES-2 parent and resistant clones. Western blotting and RNA-Seq showed much higher expression in OVCAR8 with further elevation in most resistant clones, thus supporting the notion that elevated PKMYT1 levels in OVCAR8 clones are involved in resistance.
Major reductions in total CDK1 protein levels occurred in most ES-2 resistant clones. In the absence of WEE1 kinase inhibition of CDK1, following inhibition of WEE1 itself by AZD1775, reducing total CDK1 levels could result in lower levels of active CDK1 being available to drive cells with damaged DNA into premature mitosis. Our conclusion that reduced total CDK1 levels have an important role in AZD1775 resistance in ES-2 clones is sup-ported by a report that inhibition of CDK1 increased viability of leukaemic cells treated with AZD1775 and cytarabine [43].
Cell cycle progression is also dependent on CDC25 phosphatase for removal of inhibitory phosphorylations on CDK1 and CDK2 [44]. The CDK1/cyclin B complex itself mediates activating phosphorylation of CDC25C in a positive feedback loop to reinforce the drive into mitosis [45]. Decreased expression of CDC25 phosphatase could keep the CDK1-cyclinB complex inactive and so preserve G2 arrest. CDC25B and C operate in G2/M, while CDC25A functions in S-phase. Our RNA-Seq data showed decreased expression of all three CDC25 isoforms in ES-2 clones and reduced levels of CDC25B/C in OVCAR8 AZ_300:2 and OVCAR8 AZ_300:3. These decreases could also contribute to resistance by reducing the drive for cell cycle progression.
CDC14A removes the CDK1-mediated phosphorylation of WEE1 kinase that marks it for proteasomal degradation [46] and it also inhibits CDC25 [47,48]. Although CDC14A phosphatase expression could potentially help maintain G2 arrest in the presence of AZD1775, we found no evidence for its involvement in resistance.
While this manuscript was in preparation, increased PKMYT1 expression was shown to be responsible for AZD1775 (adavosertib) resistance in breast cancer cells [49]. Based on this result and a previous finding that PKMYT1 knockdown enhanced AZD1775 sensitivity [50], it was suggested that PKMYT1 levels could be a useful biomarker of response to AZD1775. We would caution that we have seen equivalent levels of AZD1775 resistance arising in two HGSOC cell lines with very different levels of PKMYT1 expression and that other changes, such as reducing the level of total CDK1, seem equally effective in generating AZD1775 resistance. Another recent study found that leukemia cell lines were dependent on increased HDAC and MYC expression for AZD1775-resistance [51]. While it would be valuable, given the molecular heterogeneity we observed between independent resistant ES-2 clones, to investigate the resistance mechanism in additional HGSOC cell lines, this was beyond the scope of the current study.
AZD1775 inhibition of WEE1 kinase also inactivates the S-phase checkpoint resulting in faster passage through S-phase and increased replication-associated DNA damage [32][33][34]. Highly dysregulated cell cycle profiles with abnormal S-phase and continued replication without cell division were observed for AZD1775-treated parental ES-2 cells. While, in marked contrast, resistant ES-2 clones showed normal profiles with reduced numbers of S-phase cells. Increased doubling times for resistant clones support the notion that slowing the cell cycle permits them to overcome the effects of WEE1 kinase inhibition. Further support comes from the result with the mitotic marker phosphorylated Histone H3. Highly elevated levels of this marker were present in both parental lines treated with AZD1775, reflecting its role in blocking G2 arrest leading to premature mitosis. Levels of phosphorylated Histone H3 were much reduced in treated ES-2 and OVCAR8 clones. According to our model resistant clones should have lower levels of AZD1775-induced replication-associated DNA damage than parental cells. All resistant ES-2 clones had a much-reduced level of γH2AX, a marker for doublestrand DNA breaks, and of activated DNA-PK, needed for repair of ds DNA breaks. Levels of activated RPA, marker for ss DNA regions [36], were also much lower in all ES-2 treated clones and lower in most OVCAR8 treated clones than the treated parents. Further support comes from a report that p21, encoded by CDKN2A, limits AZD1775induced S-phase DNA damage [52]. CDKN2A is one of the genes induced by increased activity of the TGFβ pathway in our resistant ES-2 clones. Additional support comes from the finding that DNA damage resulting from silencing of WEE1 was reduced by treatment of cells with the CDK inhibitor roscovotine [33].
To attempt to overcome resistance we focussed on inhibiting a pathway found to be upregulated in resistant clones on the basis of our RNA-Seq data: TGFβ signalling into the cell cycle control pathway to cause cell cycle arrest. A TGFβR1 inhibitor proved effective at controlling the growth of one of the resistant clones. Although components of the TGFβ pathway were upregulated in three of the four clones tested, TGFβR1 receptor expression itself was highest in the responsive clone. Additional studies would be needed to demonstrate the value of the TGFβR1 inhibitor in overcoming AZD1775 resistance. For instance, investigating whether TGFβR1 overexpression confers AZD1775 resistance to parental ES-2 cells and whether TGFβR1 knockdown resensitizes the resistant ES-2 clone with the highest level of TGFβR1 overexpression to AZD1775.

Conclusions
We have shown that the mechanism of resistance to WEE1 inhibitor AZD1775 initially reported in small cell lung cancer cell lines [21] does not operate in either of the two main molecular subtypes of HGSOC cell lines. Although overexpression of the membrane glycoprotein MDR1 is a common mechanism of drug resistance in cancer cells, we found no evidence that it was responsible for AZD1775 resistance in our HGSOC cells. Instead AZD1775 resistance in HGSOC cells resulted from changes in the cell cycle control pathway that combine to slow AZD1775-induced cell cycle progression and so reduce accumulation of replication-associated DNA damage. In HRR-proficient ES-2 cells the major change was reduced levels of CDK1, the target for WEE1 inhibition. In HRRdeficient OVCAR8 cells increased levels of another kinase, PKMYT1, that can also inhibit CDK1 appear to have a more important role. It has previously been suggested that PKMYT1 levels could be a useful biomarker of response to AZD1775 [49]. However, we have seen equivalent levels of resistance arising in two HGSOC cell lines with very different levels of PKMYT1 expression and would caution that other changes, such as reducing the level of total CDK1, appear equally effective in generating resistance as high levels of PKMYT1. There was considerable clonal heterogeneity in the cell cycle control gene expression changes seen in resistant cells. We interpret this to reflect the complexity of cell cycle control and the different opportunities that this presents for resistance to arise (see Fig. 6). This clonal heterogeneity could also explain our limited ability to overcome resistance to AZD1775 with an inhibitor targeted against the upregulated TGFβ signalling pathway in resistant clones.

Author contributions
DWM and CG designed the research study. MKG and ATW performed the research. MC provided help and advice on laboratory protocols. MJO'C provided help and advice on AZD1775. MKG, JPT and GRG analyzed the data. MKG and DWM wrote the manuscript. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript.

Ethics approval and consent to participate
Not applicable.