Synthetic lethal interaction between WEE1 and PKMYT1 is a target for multiple low-dose treatment of high-grade serous ovarian carcinoma

Abstract Ovarian cancer is driven by genetic alterations that necessitate protective DNA damage and replication stress responses through cell cycle control and genome maintenance. This creates specific vulnerabilities that may be exploited therapeutically. WEE1 kinase is a key cell cycle control kinase, and it has emerged as a promising cancer therapy target. However, adverse effects have limited its clinical progress, especially when tested in combination with chemotherapies. A strong genetic interaction between WEE1 and PKMYT1 led us to hypothesize that a multiple low-dose approach utilizing joint WEE1 and PKMYT1 inhibition would allow exploitation of the synthetic lethality. We found that the combination of WEE1 and PKMYT1 inhibition exhibited synergistic effects in eradicating ovarian cancer cells and organoid models at a low dose. The WEE1 and PKMYT1 inhibition synergistically promoted CDK activation. Furthermore, the combined treatment exacerbated DNA replication stress and replication catastrophe, leading to increase of the genomic instability and inflammatory STAT1 signalling activation. These findings suggest a new multiple low-dose approach to harness the potency of WEE1 inhibition through the synthetic lethal interaction with PKMYT1 that may contribute to the development of new treatments for ovarian cancer.


INTRODUCTION
High-grade serous carcinoma (HGSC) is both the most prevalent and most fatal subtype of ovarian cancer. Standard therapy for HGSC consists of cytoreducti v e surgery, followed by chemotherapy with DNA-damaging agents, such as platinum drugs, either alone or in combination with taxane drugs ( 1 , 2 , 3 ). While primary tumours usuall y respond favourabl y to the treatment, over 80% of cases relapse and more than half of these acquire resistance to the tr eatment. Mor eover, the r elapsed HGSC is typically fast-growing and invasi v e ( 1 , 2 , 3 ). Platinumr esistant tumours ar e subjected to salvage therapy with DNA-damaging drugs, including do x orubicin, topotecan, etoposide, vinorelbine and gemcitabine. Nonetheless, the response rate at this stage is only ∼10-15% with a median progr ession-fr ee survival of 3-4 months, underscoring the urgent need for better treatments ( 1 , 2 , 3 ). Recently, targeted maintenance treatment with the angiogenesis inhibitor bevacizumab and the poly(ADP-ribose) polymerase (PARP) inhibitors has demonstrated benefits to ovarian cancer patients, e v en w hen accounting for the ad verse ef fects associa ted with trea tment ( 4 ). Howe v er, intrinsic or acquir ed r esistance to PARP inhibitors occurs in most HGSC pa tients, ultima tely limiting the efficacy of this approach ( 5 ).
A proposed therapeutic option for HGSC is the use of WEE1 inhibitors ( 6 ). WEE1 catalyses inhibitory phosphorylation on tyrosine 15 of both CDK1 and CDK2 and thus limits the CDK activity ( 7 , 8 ). WEE1 inhibition deregulates CDK activity and exacerba tes DNA replica tion stress to intolerab le le v els, effecti v ely killing the cells in the process termed replica tion ca tastrophe ( 9 , 10 ). Mechanistically, unrestricted CDK activity leads to excessive replication origin firing, which in turn results in depletion of protecti v e replication protein A (RPA). RPA exhaustion marks a point of no return, when unprotected replicons collapse in lethal genome-wide DNA breakage ( 9 , 11 ). Moreover, WEE1 inhibition also overrides the G2 / M DNA damage checkpoint forcing cells to enter mitosis with unr epair ed DNA, which triggers cell death ( 12 ). Stringent G2 / M checkpoint control is especially vital for cancer cells, as they frequently lose the ability to arrest their cell cycle and repair DNA in the G1 DNA damage checkpoint ( 13 , 14 ).
The most studied inhibitor of WEE1 is adavosertib (AZD1775, MK1775), which has been the focal point of multitude clinical studies ( 15 ) (ClinicalTrials.gov). In particular, positi v e outcomes for HGSC were reported in phase II clinical trials by combining adavosertib and gemcitabine treatment, exploiting high levels of replication stress in HGSC ( 6 ). Howe v er, despite the years of testing, adavosertib has not reached clinical use, which is mainly due to significant adverse effects when used in combination with chemotherapy ( 16 , 17 ). Of note, se v eral other WEE1 inhibitors are currently evaluated in clinical trials, including ZN-c3 (Zentalis), Debio0123 (Debiopharm), IMP7068 (Impact Therapeutics) and SY4835 (Shouyao Holding) (ClinicalTrials.gov).
An attracti v e strategy to limit unacceptable toxicity is to use a lower dose of inhibitors but at same time target multiple proteins of a single signalling pathway to still achie v e a complete pathway inhibition. This multiple low-dose therapy also reduces cancer selecti v e pressure against a single target that may result in treatment resistance ( 18 , 19 , 20 ). A prime candidate for synergistic effect with WEE1 is the kinase PKMYT1. PKMYT1 phosphorylates CDK1 at threonine 14 and thus inhibits CDK activity ( 21 ). WEE1 and PKMYT1 have been originally described to be synthetically lethal in fission yeast ( 22 , 23 ). The synthetic lethal interaction has been further confirmed in CRISPR-Cas9 screens in human cancer cells ( 24 ). Moreover, upregulation of PKMYT1 has been shown to promote resistance of cancer cells to WEE1 inhibition ( 25 ).
PKMYT1 as an anticancer target is much less studied than WEE1; howe v er, Repar e Therapeutics has r ecently identified a first-in-class PKMYT1 inhibitor, RP-6306 ( 26 ). RP-6306, used as single compound, showed promising in vitro results for ovarian cancer cells ( 27 ), and is currently being fast tracked to clinical trials (Clinical-Trials.gov; NCT05147350, NCT05147272, NCT04855656, NCT05605509 and NCT05601440). Notably, PKMYT1 expr ession was r eported to be upr egulated in ovarian cancers and correlated with poor prognosis, making it an appealing therapeutic target ( 28 ). The availability of a PKMYT1 inhibitor prompted us to investigate the synergistic potential of its combined application with WEE1 inhibition.

Organoid drug sensitivity assay
The organoids were dissociated in TrypLE Express solution (Thermo Fisher) as described previously ( 35 ). Dissociated cells were resuspended in 7.5 mg / ml BME-2 matrix gel at 2-5 × 10 4 cells / ml and seeded in 10 l droplets to individual wells of 96-well CellCarrier Ultra plates (PerkinElmer). Once settled, 200 l of the organoid-specific growth medium containing 5 M ROCK inhibitor Y-27632 (HY-10583, MedChemExpress) was added. After 4 days, the growth medium was changed to 200 l of medium containing adavosertib or RP-6306 at the indicated concentrations, or 25 M staurosporine as a positi v e control for cytotoxicity. After 7 days, organoids were stained using Hoechst 33342 (1 g / ml) and CellTox Green (Promega, 1 / 20 000) dyes for 8 h prior to imaging at an Opera Phenix (PerkinElmer) confocal screening microscope. The fraction of dead organoids was discriminated by the CellTox Green signal per organoid from the confocal images analysis using Harmony software (PerkinElmer).

Drug sensitivity assay
Drugs diluted in dimethyl sulfoxide (DMSO) to the desir ed concentrations wer e dispensed at 30 nl volume to 384well black pla tes (Corning, ca t#3864) using an Echo 550 acoustic liquid handler (Labcyte). Cell killing benzethonium chloride (100 M) and compound vehicle (DMSO, 0.1%) were used as positi v e and negati v e controls, respecti v ely. Cells were diluted to medium at the desired number per ml and the suspension was dispensed to the predrugged plates at 30 l. Alternati v ely, drugs were dispensed by manual pipetting into 96-well plates (Greiner-BIO) and cells were dispensed at the desired number at 100 l. After 5 days of incubation at 37 • C, 10 l / 30 l for 384-w ell / 96-w ell plates, respecti v ely, of phospha te-buf fered saline (PBS) containing 4 g / ml Hoechst 33342 and 1 / 10 000 CellTox Green dyes was added for 1 h prior to ima ging. Ima ges were obtained automatically with the ScanR acquisition software controlling a motorized Olympus IX-83 wide-field microscope, equipped with a Lumencor SpectraX light engine and Hamamatsu ORCA-FLASH 4.0, using an Olympus Uni v ersal Plan Super Apo 4 ×/ 0.16 AIR objecti v e.

Quantitative image-based cytometry
Cells growing on either 96-well microplates (Greiner-BIO) or 12 mm coverslips were treated with different combinations of drugs for indicated time intervals. After the treatment, the medium was quickly removed and the cells were incubated in pre-extraction buffer (25 mM HEPES, pH 7.5, 50 mM NaCl, 1 mM EDTA, 3 mM MgCl 2 , 300 mM sucrose and 0.5% Triton X-100) on ice for 2 min and immediately fixed in formaldehyde 4% (VWR) for 10 min at the room temperature. For the analysis of the micronuclei number and cGAS, the pre-extraction step was omitted. Primary antibodies ( ␥ H2AX 1:300, Cell Signaling Technology, cat#2577; RPA 1:300, Millipore, cat#MABE285; FOXM1pT600 1:1000, Cell Signaling Technology, cat#14655; cGAS 1:300, Cell Signaling Technology, cat#15102) were diluted in filtered DMEM containing 10% FBS and 5% bovine serum albumin (BSA; Sigma). Incubations with the primary antibodies were performed at room temperature for 1 h. Microplates were washed three times with 0.05% PBS-Tween 20 and incubated in DMEM / FBS / BSA containing secondary fluorescently labelled antibodies (Alexa Fluor dyes 1:1000; Thermo Fisher Scientific) and DAPI (0.5 mg / ml; Sigma-Aldrich) for 1 h at room temperature. Images were obtained automatically with the ScanR acquisition software controlling a motorized Olympus IX-83 wide-field microscope, equipped with a Lumencor SpectraX light engine and Hamamatsu ORCA-FLASH 4.0. Olympus PlanC N 10 ×/ 0.25 AIR objecti v e was used to capture ␥ H2AX, RPA and quantitati v e image-based cytometry (QIBC) data. Micronuclei images were obtained with a 0.75 AIR UP-lanSApo 40 ×/ 0.95 AIR objecti v e. Images were processed and quantified using the ScanR image analysis software for total nuclear pixel intensities for DAPI (arbitrary units: AU) and mean (total pixel intensities divided by nuclear area) nuclear intensities (AU) for ␥ H2AX, chromatin-bound RPA and FOXM1pT600. Micronuclei were segmented based on DAPI channel within a cytoplasmic mask surrounding the nucleus. Similarly, cGAS intensity was determined within the cytoplasmic mask. Further analysis and data visualization was then carried out with Tibco Spotfire software (Tibco, RRID: SCR 008858). Representati v e images were processed using ImageJ / Fiji ( RRID:SCR 002285 , https:// imagej.net/ ).

Western blotting
Cells were lysed in RIPA (Sigma) buffer containing EDTAfree protease inhibitor cocktail (Roche) and phosphatase inhibitors (Roche). Lysates were treated with benzonase nuclease (Sigma-Aldrich) for 30 min on ice. Lysates were centrifuged for 15 min at 20 000 × g at 4 • C. Protein concentration was then measured with the Bradford assay and adjusted accordingly to ensure equal loading. Lysates were mixed with 4 × Laemmli sample buffer (Sigma) and boiled for 10 min at 95 • C. Samples were run on NuPAGE Bis-Tris 4-12% gels according to manufacturer's instructions. Proteins were then transferred to a nitrocellulose membrane and blocked with PBS + 0.1% Tween 20 + 5% milk powder (Sigma) and incubated overnight with primary antibodies a t 4 • C . The membrane was then washed 3 × 5 min in PBS + 0.1% Tween 20 and incubated with secondary HRP conjugated antibodies for 2 h at room temperature. Membranes were again washed 3 × 5 min with PBS + 0.1% Tween 20 and incubated with Classico / Crescendo Western HRP substrate (MilliporeSigma) for 2 min. Chemiluminescence signal was detected using a Bio-Rad ChemiDoc Touch Imaging System. The following primary antibodies were used:

Cell fractionation
Cells were grown in 10 cm dishes, treated as indicated, washed three times with ice-cold PBS and harvested. The soluble fractions were extracted by incubation in ice-cold nuclear buffer (10 mM HEPES, pH 7, 200 mM NaCl, 1 mM EDTA, 0.5% NP-40) supplemented with protease and phosphatase inhibitors (Roche) for 10 min on ice and centrifuged at 2000 × g for 6 min. The remaining pellet was rinsed once with ice-cold washing buffer (10 mM HEPES, pH 7, 50 mM NaCl, 0.3 M sucrose, 0.5% Triton X-100) supplemented with protease and phosphatase inhibitors (Roche), which was removed by centrifugation at 1400 × g for 6 min. Chromatin fractions were extracted by incubation in RIPA buffer (50 mM Tris-HCl, pH 8, 150 mM NaCl, 1% IGEPAL CA-630, 0.1% SDS, 0.1% sodium deoxycholate) supplemented with protease and phosphatase inhibitors (Roche) and benzonase nuclease (Sigma) for 30 min on ice and clarified by centrifuga tion a t maximum speed.

Drug-drug interaction analysis
To assess the outcome of the drug combination treatment, we applied the zero interaction potency (ZIP) synergy model ( 30 ). The ZIP scor e r eflects the additional cell line response induced by the combinatorial treatment compared to the expected response based on the two single compounds. A ZIP score ≥10 is considered synergistic and a score ≤−10 represents antagonism. ZIP scores were calculated for each combination in the dose matrix by SynergyFinder 2.0 ( 31 ) ( https://syner gyfinder.or g/ and https://synergyfinder.fimm.fi) and plotted as synergy landscapes using RStudio ( RRID:SCR 000432 , https://www.rproject.org ) and ggplot2 package ( https://ggplot2.tidyverse. org ) ( 32 ).

Statistical analysis
Statistical analyses were conducted using GraphPad Prism (v.9.5.1). For multiple comparisons, statistical significance (adjusted P values) was calculated using the two-way analysis of variance (ANOVA), Tukey multiple comparison test, Welch ANOVA test with Dunnett's multiple comparison test or unpaired Student's t test. Results are reported as non-significant at P > 0.05, and with increasing degrees of significance symbolized by the number of asterisks: (*) 0.01 < P ≤ 0.05, (**) 0.001 < P ≤ 0.01, (***) 0.0001 < P ≤ 0.001 and (****) P ≤ 0.0001. Statistical details for each experiment can be found in the corresponding legend.

In vivo drug tolerance study
All experiments were carried out under authorization and guidance from the Danish Inspectorate for Animal Experimentation under license number 2021-15-0201-00993, the animal use protocol number P22-560 specifically applicable to the experiments described in the study. Female 7-weekold mice of NGX (NOD-Prkdc scid-IL2rg Tm1 / Rj) strain (Janvier Labs) were randomized in four treatment cohorts of six animals. The mice were housed in individually ventilated cages with a humidity of 55% ± 10%, a temperature of 22 ± 2 • C and a dark / light cycle of 12 h / 12 h with light from 6:00 to 18:00. Adavosertib (Repare Therapeutics) and RP-6306 (Repare Therapeutics) were formulated in 1% DMSO and 0.5% methylcellulose and administered by oral gavage at 15 and 5 mg / kg, respecti v ely, alone or in combination. For intermittent 21-day dosing, the drugs, the combination or the vehicle was given twice daily, with 8 h interval, for 5 days a w eek, follow ed by 2 tr eatment-fr ee days. Animal weight was measured twice weekly, and the overall animal condition was monitored daily. At the end of the treatment, all mice were humanely sacrificed and li v er w eights w ere measured.

Pharmacokinetics
Whole blood samples were collected 30 min and 8 h after the first drug trea tment. Immedia tely after collection, 20 l of blood was mixed with 60 l of 0.1 M citrate buffer (0.1 M trisodium citrate, pH 7.4) and stored at -80 • C before the analysis. All samples were quantified using a re v ersed-phase liquid chromato gra phy gradient coupled to electrospray mass spectrometry operated in positi v e mode. Pharmacokinetic parameters were calculated using non-compartmental analysis.

Combined inhibition of WEE1 and PKMYT1 synergizes in killing of cancer cells
WEE1 inhibition has emerged as a strategy to eliminate cancer cells; howe v er, adv erse effect concerns hav e warranted further preclinical investigations. We noted that combined WEE1 and PKMYT1 genetic ablation was lethal in a glioma setting ( 24 ). To evaluate the potential synergistic effect of co-targeting WEE1 and PKMYT1, we conducted a dose response matrix for cell viability with the WEE1 inhibitor adavosertib in combination with the PKMYT1 inhibitor RP-6306 ( Figure 1 A and B). The U2OS cell line was selected as a model since it has been well characterized for WEE1 and CDK functions, and notab ly, inv estigated in detail for the effects of adavosertib treatment ( 33 ). At 100 nM concentra tion, the combina torial trea tment led to ef ficient killing of most U2OS cells (Figure 1 B). In stark contrast, treatment with 100 nM of single compound treatment had negligible effect on cell viability (Figure 1 B). A synergy analysis re v ealed a strong synergistic interaction between the inhibitors (Figure 1 C). Furthermore, these findings were reproduced using the combination of RP-6306 with another clinically relevant WEE1 inhibitor, ZN-c3 (Supplementary Figure S1A and B). The efficacy of used inhibitors was validated by western blots, assaying the substrates of WEE1 and PKMYT1 --CDK1pY15 and CDK1pT14, respecti v ely (Supplementary Figure S1C and D).
In contrast to normal cells, cancer cells are characterized by multiple genetic alterations in dri v er genes that promote high rate of proliferation and impose replication stress ( 34 , 35 ). The high le v el of r eplication str ess r enders cancer cells particularly dependent on safeguarding mechanisms and these can be exploited by cancer treatments. To assess w hether the m ultiple low-dose a pplication of WEE1i and PKMYT1i pr efer entially eliminated high r eplication str ess cells and less so their normal counterparts, we conducted dose response matrix in normal BJ fibroblast cells with doxy cy cline-inducib le oncogene HRAS(G12V) (   6 NAR Cancer, 2023, Vol. 5, No. 3 first confirm the impact of HRAS(G12V) induction, we monitored prolifer ation r ates that indeed were elevated by activating HRAS signalling (Supplementary Figure S1E). Moreover, normal non-induced BJ fibroblasts were indeed mor e r esistant to the tr ea tment and exhibited cell dea th only in higher doses compared to U2OS cells (Figure 1 D). This was particularly marked for the response to the PKMYT1 inhibitor. Accordingly, the synergy score was comparably lower in the non-induced setting, and the inhibitors synergized only in a very limited concentration window (Figure 1 E). Importantly, HRAS(G12V) induction sensitized the BJ cells to the tr eatment (Figur e 1 F and G). To further investigate the impact of combinatorial dosing on normal cells, we administered adavosertib and RP-6306 by oral gavage to mice. The achie v ed plasma le v els (when corrected to plasma protein binding) of both compounds were in a concentr ation r ange where we observed synergistic effect in killing cancer cells in vitro (Supplementary Figure  S2A-C). Notably, compounds had little impact on mouse bodyweight and li v er size when administered individually or jointly (Supplementary Figure S2D and E). This suggests that when administered at low doses, the combinatorial dosing is tolerated in mouse models.

WEE1i and PKMYT1i co-inhibition exacerbates replication stress and triggers replication catastrophe
Next, we aimed to characterize in detail the mechanism of action of the WEE1i and PKMYT1i combination. WEE1 and PKMYT1 suppress replication stress, and they also guard against pr ematur e mitotic entry e v en in cells e xperiencing genotoxic challenges through replication stress ( 12 , 27 ). Thus, we reasoned that combined WEE1i and PKMYT1i could elevate replication stress to intolerable le v el. To assess replication stress le v els, we employed QIBC to measure accumulation of single-stranded DNA by quantifying the le v els of chromatin-bound RPA and phosphorylation of histone H2AX on serine 139 ( ␥ H2AX) as a marker of DNA damage ( 9 ). The dose response matrix of acute treatment (4 h) displayed a synergistic effect of adavosertib and RP-6306 in inducing the replication stress, which in higher inhibitor concentra tion propaga ted into replication catastrophe ( Figure 2 A and B, and Supplementary Figure  S3A). We observed the same impact for the combination of ZN-c3 and RP-6306 (Supplementary Figure S3B-D).
Gi v en the role of WEE1 and PKMYT1 as master regulators of CDK activity, we reasoned that induction of replication stress and catastrophe corresponded to increased le v els of CDK activity in S phase. Indeed, we observed that CDK activity, measured as phosphorylation of the CDK substrate FOXM1 at threonine 600 ( 38 ), increased ra pidl y upon the combined treatment and correlated with induction of replication stress as measured by QIBC (Figure 2 C-E). Moreover, w e w er e able to r e v erse the replica tion ca tastrophe phenotype by inhibition of CDK activity with CDK1i RO-3306 (Figure 2 C-E) ( 39 ). Complementary to QIBC, western blot analysis of cellular fractionates showed increased CDK activity, as measured by pan-CDK substrate, and increased chromatin loading of RPA, which indicates replication stress. Of note, CDK1i RO-3306 moderately inhibits also CDK2 ( 39 ). Gi v en this limitation and the overlapping substrates and roles of both CDK1 and CDK2, we refrain to pinpoint the observed effect exclusively to either CDK1 or CDK2, albeit it is clearly dri v en by CDK acti vity. We also observed increased phosphorylation and activation of markers of the DNA damage response such as CHK1 phosphoryla ted a t serine 345, CHK2 phosphoryla ted a t threonine 68 and ␥ H2AX (Figure 2 F). Collecti v ely, the data indicated a marked replication stress and replication catastrophe response to combined WEE1 and PKMYT1 inhibition that likely explains the major treatment lethality in cancer cells. Moreover, cells that do not die in imminent replication catastrophe will be forced by high CDK activity into premature mitotic entry with high le v els of r eplication str essgenerated DNA damage resulting in further loss of cell fitness.

Combined WEE1 and PKMYT1 inhibition increases genomic instability and activates a cGAS-STING response
Increased le v els of replication stress have been associated with exacerbation of genome instability and formation of micronuclei after mitotic progression with DNA damage ( 40 , 41 ). To test micronuclei formation in our system, we treated the U2OS cells with combinatorial low-dose WEE1i and PKMYT1i for 3 days and assessed the percentage of cells with micronuclei. We observed a significant increase in their formation with combinatorial WEE1i and PKMYT1i treatment in doses as low as 33 nM ( Figure  3 A and B). The incr eased pr esence of micronuclei is linked to activation of innate immunity and the clearance of tumours in vivo ( 40 ). Mechanistically, this is mediated by the cGAS-STING pathway and subsequent activation of STAT signalling response. Accordingly, we observed increased accumulation of cGAS ( Figure 3 A and C), activation of its downstream effector TANK-binding kinase 1 (TBK1pS172) and ele vated mar ker of STAT1 activation (STAT1pY701) (Figure 3 D). Taken together, these data demonstra te tha t the WEE1i and PKMYT1i combination activates cGAS-STING response.

WEE1i and PKMYT1i multiple low-dose treatment is efficient against a variety of HGSC cell lines r egar dless of driver oncogene
As described above, combined WEE1i and PKMYT1i application might be suited for aggressi v e har d-to-treat cancers with high prolifer ation r ates, such as HGSC. In addition to high proliferation rate, ovarian cancers almost ubiquitously ov ere xpress PKMYT1, suggesting their dependence on PKMYT1 activity (Supplementary Figure  S4A) ( 25 ). Also, despite the partiall y overla pping roles of WEE1 and PKMYT1, we did not observe that their expression would be upregulated in m utuall y e xclusi v e fashion in tumours (Supplementary Figure S4B (Supplementary Figure S4C) ( 42 ). Regardless of the expression of specific dri v er oncogenes, all ovarian cancer cell lines were eradicated by WEE1i and PKMYT1i following multiple low-dose exposure (Figure 4 A-J). Upon tailored inspection and like U2OS cells, we observed replication stress and induction of replication catastrophe in OVCAR3 (high cyclin E; Supplementary Figure S5A and B), KURAMOCHI (high KRAS; Supplementary Figure S5C and D) and OV-CAR8 (moderately high cyclin E and MYC; Supplementary Figure S5E and F) after the WEE1 and PKMYT1 inhibition in multiple low-dose treatment.
We also noted that all tested ovarian cancer cell lines responded in a similar dose range to U2OS, with COV318 displaying slightly different pattern. COV318 represents a highly heterogeneous cell line and the majority of COV318 population responded with e v en higher sensiti vity than the other HGSC cell lines, though ∼10% of the cells survi v ed the treatment. In addition, we recapitulated the findings of increased micronuclei formation and activation of cGAS-STING response from U2OS in the OVCAR8 cell model (Supplementary Figure S6A-D).

WEE1i and PKMYT1i multiple low-dose approach eradicates patient-derived o v arian cancer organoids
To further assess the potential for translation of the WEE1i and PKMYT1i multiple low-dose approach, we evaluated its efficacy in a set of clinically relevant HGSC patient-deri v ed organoid cultures that retained genetic makeup and heterogeneity of the tumour of origin (summarized in Figure 5 A) ( 29 ). Imaging-based toxicity assay re v ealed dose-dependent synergistic efficacy of the combination in all tested HGSC organoid cultures ( Figure  5 B −F), independent of CCNE1 , MYC or KRAS amplification, the site of origin of the tumour cells or previous exposure to the replication stress-inducing carboplatinbased neoadjuvant chemotherapy (NACT). Importantly, organoid cultures EOC989 and EOC884, which were deri v ed from residual tumour cells from patients treated with chemothera py (N ACT), sho wed prominent syner gistic response to the combined WEE1 and PKMYT1 inhibition. Both WEE1 and PKMYT1 inhibitors are tested in clinical trials in combination with gemcitabine (ClinicalTrials.gov, NCT02101775 and NCT05147272). Ther efor e, we also addressed potential impact of combining gemcitabine with our multiple low-dose approach. We pr e-tr eated select ovarian cancer cells (OVCAR3 and KURAMOCHI) and organoids (EOC884 and EOC989) with gemcitabine for 18 h followed by combined WEE1 and PKMYT1 inhibition.
We observed a considerable impact as cells wer e mor e r eadily eradicated following gemcitabine pr e-tr eatment (Supplementary Figure S7A −D). This suggests potential in combining standard chemotherapy with the WEE1i and PKMYT1i multiple low-dose approach.

DISCUSSION
Her e, we r eport the combinatorial drugging of PKMYT1 and WEE1 kinases, w hich synergisticall y eradicates cancer cells already at low drug dose. Our data highlight the potential for multiple low-dose treatment and support the notion that combining full-dose treatments may not be the only approach when administrating a set of targeted drugs. Synthetically lethal interactions are attracti v e in cancer treatment as they may allow reduced adverse effects while drugging cancer-specific vulnerabilities in a similar manner as the targeting of BRCA deficiency with PARP inhibition ( 43 ). Howe v er, fe w drug candidates and treatments are currently based on synthetic lethality and the multiple low-dose approach is still under de v elopment. The multiple low-dose tr eatment may r epr esent a desir able str ategy also for targeting other checkpoint kinases as Golder et al. demonstrated that multiple low-dose combinatorial drugging of ATR and CHK1 inhibitors proved effective in killing HGSC cells ( 44 ). Our data show that cancer cells or oncogene exposed cells are more sensiti v e than normal cells to the WEE1i and PKMYT1i combina torial trea tment, which suggests the evolution of sensitivity during cancer development. It was recently demonstra ted tha t PKMYT1 inhibition with RP-6306 displays synthetic lethality with CCNE1 amplification based on the marked replication stress induced by CCNE1 ov ere xpression ( 27 ). Indeed, cancer cells generally display elevated replication stress due to activated dri v ers such as CCNE1 , KRAS and MYC ( 34 , 35 ). We observed that oncogenic HRAS(G12V) expression also sensitized cells to PKMYT1 inhibition alone and to combined WEE1i and PKMYT1i treatment. In addition, combined WEE1i and PKMYT1i treatment was effecti v e in killing a di v erse panel of HGSC cell lines and organoids, regardless of their driver oncogene. This suggests that the treatment efficacy is not limited to a particular oncogene, but rather to a more general feature of oncogene-induced replication stress. In support of this notion, we have observed an exacerbation of r eplication str ess and induction of r eplica tion ca tastrophe, w hich likel y mechanisticall y underlay a major part of combina torial trea tment lethality in cancer cells. Moreover, cell cycle control in cancer cells is perturbed by the frequent hits in the p53 and retinoblastoma protein (RB) pathways ( 45 ). This creates a dependence on the remaining cell cycle control mechanisms and causes cancer cell vulnerabilities to targeted trea tments tha t interfere with these mechanisms. Thus, incr eased r eplication str ess and limited cell cycle control mechanisms make cancer cells reliant on G2 phase control to limit detrimental pr ematur e mitotic entry. Deregulated passage through G2-M transition promotes mitotic cell death and it also triggers micronuclei formation, which in turn leads to innate immunity activation such as cGAS-STING-mediated interferon responses ( 40 , 41 , 46 ). In agreement with this finding, we observed induced cGAS-STING signalling at low doses of combinatorial treatment. We also observed that the low-dose combination appeared well tolerated in mouse models, although comprehensi v e additional studies are needed to evaluate the in vivo efficacy of the low-dose combination, as well as tolerability in higher species.
The ubiquitous TP53 mutations in HGSC ( 47 ) as well as alterations in G1 / S transition master regulators [RB1 deficiency ( 48 ) and CCNE1 amplifications ( 49 )] present the highly relevant molecular landscape for exploration of WEE1 and PKMYT1 co-inhibition as a treatment option. It is of interest to assess the potential of the   0  33  100  11  300  0  33  100  11  300  0  33  100  11  300  0  33  100  11   combina tion trea tment to eradica te the residual tumour cells, which lose the sensitivity to the standard chemotherapeutics ( 50 ) and fr equently ar e selected to r estor e the homologous r ecombination r epair pathway ( 51 ), excluding the further possibility to use PARP inhibitors in the treatment. We have employed ovarian cancer organoids to r epr esent three-dimensional cell culture models closely reflecting the primary tissue's biology and pathology. We observed a pronounced synergistic response to the combination in HGSC organoids established from residual tumour samples collected after a few cycles of chemotherapy (EOC884 and EOC989; Figure 5 , and Supplementary Figure S4C and D).
As the preclinical drug testing in organoids helps accurately predict the clinical treatment outcome ( 29 , 52 , 53 , 54 ), the observed efficacy of the multiple low-dose treatment with WEE1 and PKMYT inhibitors suggests that the respecti v e tumours may have responded to the therapeutic combination. Hence, the combination may offer a treatment strategy to a ttenua te the relapses due to chemoresistant disease, which affects up to 70% of HGSC cases ( 1 , 3 ). Even though we focused mostly on ovarian cancer model systems, we have also shown that the multiple low-dose strategy can potentially be effecti v e in other tumour types as indicated by our results in the U2OS cells of osteosarcoma origin. Collecti v ely, our findings argue for a reduced focus on maximal tolerable dose of single targeted cancer drugs and suggest the use of drug combina tions a t low and non-toxic concentrations.

DA T A A V AILABILITY
All data generated during this study are included in this article and its supplementary files. Raw microscopical data are available upon request.