Introduction

Gynecological malignancies have a significant impact on the health of women globally. Epithelial ovarian cancer stands out as one of the most lethal gynecologic malignancies, especially in 75% of cases representing the most lethal histotype of high-grade serous ovarian carcinoma (HGSOC)1,2. While most patients with HGSOC initially respond to platinum-based chemotherapy, up to 80% will develop resistance and experience recurrence3,4. As one of the most aggressive histological subtypes of endometrial cancer, uterine serous carcinoma (USC) comprises only 10% of endometrial cancer cases but is responsible for 40% of related deaths5. These cancers are often diagnosed at advanced stages with high recurrence rates and poor overall prognosis. Treatment options are limited for HGSOC and USC patients who are inherently refractory to or have progressed on chemotherapy, and the survival rate has not significantly improved over the past 4 decades6, highlighting the urgent need for novel, well-tolerated therapeutic agents capable of generating durable responses.

Amplification of CCNE1, the gene that encodes Cyclin E1, occurs frequently in many human cancers, including HGSOC and USC, and has been associated with poor patient outcomes and resistance to chemotherapy7,8. As a key cyclin regulating cell cycle progression through the G1 phase to S phase, Cyclin E1 levels peak at the G1/S boundary. It complexes with and activates CDK2 to promote transition from G1 to S phase, initiation of DNA replication, and progression of DNA synthesis. Activation of Cyclin E1/CDK2 complexes accelerates normal cell cycle progression and has been demonstrated to result in premature S phase entry, increased replication stress, and accumulation of DNA double-strand breaks (DSBs)9,10. Cyclin E1 has also been shown to be important for centrosome duplication, and its overexpression can induce mitotic defects11,12. Cyclin E1 expression in cancer cells is subject to multiple levels of regulation through different mechanisms, including amplification of the CCNE1 gene, increased transcription caused by alterations in the RB/E2F pathway13, and reduced protein degradation due to inactivating mutations in FBXW7, an E3 ligase that targets Cyclin E1 for proteolysis14. Although CCNE1 amplification is observed in approximately 20-25% of advanced ovarian cancer patients, Cyclin E1 protein overexpression ranges from 25% up to 90%, which is likely due to the variable methodologies and scoring cutoffs7,15,16,17,18,19,20,21. Similarly, even though 44% of USC patients were reported to have CCNE1 amplification, Cyclin E1 protein overexpression is anticipated to be even higher due to the frequent presence of FBXW7 mutations in this tumor type14,22.

Additional dysregulation of the G1/S checkpoint due to mutations in critical regulators, such as TP53, RB1 and CDKN2A, may increase the dependency on an intact G2/M checkpoint to allow for adequate DNA repair and subsequent survival of cancer cells with Cyclin E1/CDK2 activation8. Importantly, CCNE1 amplification co-occurs with TP53 alterations in many cancers including endometrial, ovarian, pancreatic, and breast cancers7. TP53 mutation is nearly ubiquitous in HGSOC and USC, occurring in >90% of patients, making these cancer types ideal indications for agents targeting key proteins involved in G2/M checkpoint, such as WEE1 inhibitors.

WEE1 is a tyrosine kinase that functions through inhibition of both CDK1 and CDK2 activity by phosphorylation at tyrosine 15 (Y15). It is involved in several stages of cell cycle regulation, including intra-S, G2/M and M23. Additionally, WEE1 regulates the G1/S cell cycle transition and protects stalled replication forks by inhibiting CDK2 activity24,25. It has also been reported that WEE1 can directly regulate Mus81-Eme1 endonuclease activity at stalled replication forks to prevent the formation of DSBs26,27,28. As a crucial component of the G2/M checkpoint, WEE1 prevents cells from entering mitosis in response to cellular DNA damage through inhibitory phosphorylation of CDK127,29. In normal cells, DNA damage activates cell cycle checkpoints that arrest cells at G1 and G2 phase to repair DNA, however G1/S checkpoint defects often exist in cancer cells, preventing adequate repair. Consequently, functional G2/M checkpoints are critical to prevent the accumulation of intolerable levels of DNA damage and circumvent mitotic catastrophe and apoptosis30. Therefore, abrogation of the G2/M checkpoint by selective WEE1 inhibitors is an attractive strategy that may selectively drive cancer cells to enter unscheduled mitosis without adequate DNA repair and trigger cell death, while having limited effects on normal cells that have a functional G1/S checkpoint. Due to the important role of WEE1 in S phase, including the S phase DNA damage checkpoint, WEE1 inhibition may interrupt DNA replication and result in replication stress and DNA damage. As a result, we hypothesized that cancer cells with higher baseline replication stress and a dysregulated G1/S checkpoint caused by Cyclin E1/CDK2 activation will be more sensitive to WEE1 inhibition.

Several WEE1 inhibitors have been developed and are currently being evaluated in clinical trials for patients with solid tumors, including adavosertib, azenosertib, debio0123, and IMP7058, with adavosertib being the most well-studied. Preclinical studies suggest Cyclin E1 overexpression sensitizes triple-negative breast cancer cells to the WEE1 inhibitor adavosertib31. In a Phase 2 trial evaluating adavosertib as a monotherapy, promising clinical activity was seen in patients with refractory epithelial ovarian cancer harboring CCNE1 amplification (CN > 7), with an overall response rate (ORR) of 36%32. In another Phase 2 study in subjects with recurrent, platinum-resistant HGSOC with Cyclin E1 protein overexpression without CCNE1 amplification, adavosertib monotherapy demonstrated an ORR of 43%21. These data provide clinical proof of concept that high Cyclin E1 expression may be a sensitizing biomarker for WEE1 inhibitors. However, clinical development of adavosertib was halted despite its clinical activity in several settings33,34,35,36,37. In the present study, we sought to evaluate whether high Cyclin E1 expression can predict response to azenosertib, a novel, orally bioavailable WEE1 kinase inhibitor with improved selectivity38. Our data demonstrates that high Cyclin E1 levels and corresponding CDK2 activation are associated with sensitivity to azenosertib as a single agent in vitro and in vivo. High Cyclin E1 levels also increased sensitivity to the combination of azenosertib with chemotherapy agents. CDK2 kinase activity was required for Cyclin E1-mediated sensitivity to azenosertib, consistent with the requirement of CDK2 activity for Cyclin E1 function. Additionally, we characterized the underlying mechanism of action and found that azenosertib treatment increased existing replication stress in cancer cells with high Cyclin E1/CDK2 activation, resulted in DNA damage and eventually apoptosis. Notably, we provide preliminary clinical evidence that Cyclin E1/CDK2 activated cancers are sensitive to azenosertib monotherapy. Currently, azenosertib is being evaluated in several clinical trials with the aim of prospectively understanding the relationship between Cyclin E1 expression and azenosertib anti-tumor activity.

Methods

Reagents and cell lines

Azenosertib was synthesized at Zentalis Pharmaceuticals. Azenosertib in powder form was stored at room temperature and protected from light. Azenosertib was formulated in 100% DMSO and aliquoted for long-term storage at –80 °C. RPMI-1640 (#30-2001), EMEM (#30-2003), McCoy’s 5 A (#30-2007) were obtained from ATCC (Virginia, VA), RPMI (#11875-093), DMEM (#12430-054), DMEM/Ham’s F12 (#11320-033), Media 199, Earle’s Salts (#11150-059), sodium pyruvate (#11360-070), sodium bicarbonate 7.5% solution (#25080-094), puromycin (#A11138-03), GlutaMAX™ supplement (#35050-061), MEM Non-Essential Amino Acids solution (#11140-050), penicillin and streptomycin (#15140-122), FBS (#A38400-02) were obtained from Gibco/Thermo Fisher Scientific (Waltham, MA), MCDB 105 (#117-500) was obtained from Cell applications (San Diego, CA), human insulin (#I9278), bovine insulin (#I0516) were obtained from Sigma-Aldrich (St. Louis, MO). OV90 (RRID: CVCL_3768), ES2 (RRID: CVCL_3509), COV362 (RRID: CVCL_2420), TYK-nu (RRID: CVCL_1776), KURAMOCHI (RRID: CVCL_1345), CAOV3 (RRID: CVCL_0201), OAW28 (RRID: CVCL_1614), and OVCAR3 (RRID: CVCL_0465) cell lines were obtained from the ATCC. The OVCAR4 cell line (RRID: CVCL_1627) was obtained from Sigma-Aldrich. Human cancer cell lines were maintained at 37 °C in a humidified incubator at 5% CO2. Cell lines were carried for no more than 15 cell passages.

Cell proliferation assays

Cells were plated in 96-well white-walled, clear-bottom plates (Corning Life Sciences #3903, Corning, NY) for cell viability determination in standard tissue culture conditions. The day after plating, the solution of azenosertib or the indicated drug was dispensed into the 96-well plate using a digital dispenser (Thermo Scientific Multidrop Pico 8 digital dispenser) so that the final concentration of the drug ranged from 1.3 nM to 10 μM. 72 hours later, the CellTiter-Glo (CTG) Luminescent Cell Viability Assay (Promega #G7573, Madison, WI) was used to measure cell viability effects of indicated drugs. Data were analyzed, and IC50 or GR50 values were determined using GraphPad Prism (GraphPad, San Diego, CA). Viable cells were counted at the beginning of the assay using a Countess II FL cell counter (Life Technologies, Carlsbad, CA) using Trypan Blue exclusion so as to plate the same number of viable cells per well at the beginning of the experiment. GR and GRmax were determined by measurement of cell viability before and after 72 hours treatment of azenosertib and the GR calculator (http://www.grcalculator.org). GRmax lies between –1 and 1; negative values correspond to a cytotoxic response (i.e., cell death), a value of 0 corresponds to a fully cytostatic response (no increase or decrease in cell number), and positive values less than one correspond to partial growth inhibition.

Jess simple western analysis

The cells were seeded on a 6 cm plate. The day after seeding, cells were treated with azenosertib or indicated drugs and incubated for indicated times. These cells were collected and lysed on ice with radioimmunoprecipitation (RIPA) lysis buffer (Thermo Fisher Scientific #89901). The cell-driven xenograft (CDX) tumors were homogenized using the BeadBlaster 24 homogenizer (Benchmark Scientific, Sayreville, NJ) in RIPA buffer. The cell and tumor lysates were sonicated, and then centrifuged (15,000 rpm, 15 minutes, 4 °C). Protein concentrations were determined using Pierce BCA Protein Assay Kit (Thermo Fisher Scientific #23227). Western blotting was performed using Jess Simple Western System, an automated capillary-based size-sorting system (ProteinSimple, San Jose, CA). The expression level or phosphorylation signal of target proteins were evaluated according to the manufacturer’s standard method for 12–230 kDa Jess separation module (SM-W004). The protein lysates were mixed with 0.1X Sample buffer and Fluorescent 5X Master mix reagent to achieve a total protein concentration of 2 μg/μL in the final samples. The mixtures were denatured at 95 °C for 5 min. Biotinylated ladder (5 μL) and each protein sample (5 μL) were loaded into individual wells of the sample plate. Primary antibodies were diluted with antibody diluent buffers. The HRP-conjugated anti-rabbit or mouse secondary antibodies were applied according to the instructions of Simple Western kit. All subsequent separation, immunodetection, and analysis steps were performed automatically by the machine. Chemiluminescence reactions with antibodies were measured and their digital blot images were constructed by the Compass software (Version 6.2; ProteinSimple). Quantification by densitometry was performed using the area of targeted proteins and normalized to the total protein. Results are expressed as fold change in the expression of proteins when compared to their expression in control. The antibodies and reagents used in this study are listed in Supplementary Table S1.

Generation of stable cell lines

Isogenic cell lines stably overexpressing CCNE1 were generated using lentiviral plasmid vectors. Lentiviruses targeting CCNE1 and empty vectors were purchased from Cellecta (Mountain View, CA), and used to transduce OV90, COV362, and KURAMOCHI cells. Briefly, 1 × 106 cells were seeded in growth media in T25 flasks, and lentiviruses were added at MOI (multiplicity of infection) = 1. Polybrene (8 μg/mL) was added simultaneously, followed by incubation at 37 °C for 24 hours. The media were removed and replaced with fresh growth media containing puromycin (2 μg/mL). Transduced cells were cultured with puromycin (2 μg/mL) for 2–6 weeks, with media changes every 48-72 hours, before Cyclin E1 overexpression (O/E) was assessed using Jess Simple Western System. All CCNE1 O/E and vector control cells were maintained on puromycin (2 μg/mL) for in vitro assays.

Knockdown using siRNAs

Silencer Select siRNAs for CDK2 (s204 and s205), CCNE1 (s2524 and s2526), and Silencer Select-negative control siRNA No. 2 were purchased from Thermo Fischer Scientific. Transfection was performed using Lipofectamine RNAiMAX (Thermo Fischer Scientific #13778150,) according to the manufacturer’s protocol. The knockdown efficiency of specific siRNAs was confirmed by western blot analysis using Jess Simple Western System for target proteins using total cell lysates harvested on 3 days and 6 days after transfection of siRNAs.

DNA fiber assay

The cells were treated with indicated drugs and performed DNA fiber assay as previously described39. Briefly, the cells were labeled with 20 µM IdU (Sigma‐Aldrich #I7125) for 20 minutes, washed twice with PBS, and labeled with 200 µM CIdU (Sigma‐Aldrich #C6891) for 1 hour at 37 °C. The cells were permeabilized with CSK100 buffer (100 mM NaCl, 10 mM MOPS [pH 7], 3 mM MgCl2, 300 mM sucrose, 0.5% Triton X‐100) for 10 minutes at room temperature. 40 U/mL S1 nuclease (Invitrogen #18001016, Waltham, MA) with S1 buffer (30 mM sodium acetate, 10 mM zinc acetate, 5% glycerol, 50 mM NaCl, [pH 4.6]) was added and incubated for 2 hours at 37 °C. S1 nuclease was removed, and 0.1% BSA/PBS was added. The cells were then collected by scraping. The scraped cells were centrifuged at 7000 rpm for 5 minutes at 4 °C. The supernatant was removed, and the cells were resuspended for 1–2×103 cells/µL in 0.1% BSA/PBS. 2 µL of the cell resuspension was mixed with 8 µL of lysis buffer (200 mM Tris–HCl [pH 7.5], 50 mM EDTA, 0.5% SDS) and waited for 8 minutes on Microslide. The slides were tilted at 15 degrees to allow the fiber to spread along the slide. When the DNA spreads were air dried, the slides were fixed in methanol/acetic acid (3:1) for 5 minutes. The slides were washed with PBS, air dried, and then denatured with 2.5 M HCl for 1 hour. The slides were washed with PBS and incubated in a blocking solution (5% BSA/PBS pre‐warmed at 37 °C) for 45 minutes, and incubated with the primary antibodies: mouse anti‐BrdU (1:20; detects IdU; BD Biosciences #347580; Franklin Lakes, NJ) and rat anti‐BrdU (1:100; detects CldU; Abcam #ab6326, Cambridge, UK) in 1% BSA/PBST (0.05% Tween‐20) overnight at 4 °C. The slides were washed with PBST, and incubated with the secondary antibodies: Alexa Fluor 546-conjugated anti‐mouse (1:100; Invitrogen #A21123) and Alexa Fluor 488-conjugated anti‐rat (1:100; Invitrogen #A21470) and Alexa Fluor 647-conjugated anti-rat (1:50; Invitrogen #A21472) in BSA/PBST for 1 hour at room temperature. After washing with PBST, the slides were mounted in Prolong gold antifade (Invitrogen #P36930). The images were examined and taken by a ZEISS (Jena, Germany) Apotome 3 and analyzed by ImageJ (NIH).

Immunofluorescence (IF) and immunohistochemistry (IHC) staining

For IF staining of cells, 3.5 × 104 MCF10A cells were plated into an 8-well Culture Slide (Falcon #354118, Corning, NY). With indicated incubation and treatment (figure legends), the cells were fixed in 4% formaldehyde for 15 minutes at room temperature and rinsed with PBS. Before staining with antibodies, nonspecific binding was minimized by blocking with Blocking Buffer (Cell Signaling Technology #12411) for 1 hour. The cells were incubated overnight at 4 °C with the primary antibodies diluted in Antibody Dilution Buffer (Cell Signaling Technology #12378). The cells were then treated with secondary antibodies: Alexa Fluor 555-conjugated anti-mouse or Alexa Fluor 647-conjugated anti-rabbit in antibody dilution buffer for 1–2 hours. After washing with PBS, cells were mounted in Prolong gold antifade (Invitrogen #P36930). For IF staining of formalin-fixed, paraffin-embedded (FFPE) tissue samples of SKOV3 and OVCAR3 xenografts, 5 µm sections were deparaffinized using xylene twice for 10 minutes, 100% ethanol twice for 10 minutes, 95% ethanol for 5 minutes, 70% ethanol for 5 minutes, and 50% ethanol for 5 minutes, and then left in PBS. Tissues were antigen retrieved in a pressure cooker using Citrate (pH 6) and DAKO antigen retrieval (pH 9) buffers. Tissues were washed twice in PBS and then were further permeabilized with 0.4% Triton X-100 in PBS for 1 hour. The excess permeabilization solution was removed by washing three times in PBS, and then a hydrophobic barrier was placed around the mounted tissue. Tissues were blocked with 3% BSA in PBS prior to antibody labeling. The tissues were then incubated with primary antibodies pRPA32 S4/S8 (1:500) and γH2AX S139 (1:250) on a rocker for 2 hours at room temperature in a humidity chamber. After three PBS washes, secondary antibodies were applied: Alexa Fluor 647-conjugated anti‐rabbit and Alexa Fluor 555-conjugated anti‐mouse. After three final PBS washes, all tissue samples were postfixed with 4% PFA for 10 minutes at room temperature. Hoechst staining solution (1:5000) was used and mounted in Prolong gold antifade. Sample images were acquired using a spinning disk confocal microscope with a 63x oil objective. A minimum of 100 tumor cells were used for downstream analysis. Raw image files from Zen (.czi) were imported into custom ImageJ macro software for image segmentation. WEKA machine learning plugin was used for both nuclei and foci segmentation. Downstream analysis of extracted cell and foci ROIs was analyzed using R. pRPA32 S4/S8 positivity was defined as a minimum of 6 foci greater than 200 nm in diameter, and γH2AX S139 positivity was defined as a minimum of 3 foci greater than 200 nm in diameter. For IHC staining of tumors (mice, patients), the samples were collected after each experiment and immediately fixed in 10% neutral-buffered formalin for 24 hours and then stored at 70% ethanol until histology work started. The formalin-fixed samples were paraffin-embedded, and CAP/CLIA-validated Cyclin E1 IHC assay was conducted at Discovery Life Sciences (Newtown, PA). IHC staining of formaldehyde-fixed, paraffin-embedded skin punch biopsy specimens with pCDK1Y15 antibodies (1:100; LifeSpan Biosciences #LS-C178124, 1:50; Cell Signaling Technology #4539) was conducted by QualTek Molecular Laboratories (Discovery Life Science, Goleta, CA). Immunoreactivity of IHC staining was evaluated using histochemical scoring (H-score). The H-score incorporates both the staining intensity and the percentage of stained cells at each intensity level. The intensity levels are classified as follows: 0 (null/negative staining), 1+ (low/weak staining), 2+ (medium/moderate staining), and 3+ (high/strong staining). H-score is calculated by using a formula: H-score = (1 x percent of cells at 1+) + (2 x percent of cells at 2+) + (3 x percent of cells at 3+), resulting in a score range from 0 to 300.

High-content imaging analysis

OVCAR3 and Kuramochi cells were seeded in 96-well plate overnight and treated with azenosertib at 0.1, 0.3, and 1 µM and 0.1% DMSO used as a control. After the treatments, cells were fixed in 4% paraformaldehyde for 20 minutes, washed, permeabilized in 0.5% Triton X-100 in PBS for 20 minutes on ice, blocked with 3% BSA in TBS for 1 hour, incubated with primary antibodies: α-tubulin mouse antibody (1:500; Cell Signaling Technology #3873) or phospho-Histone H3 rabbit antibody (1:500; Cell Signaling Technology #3377) and secondary antibodies: Alexa Fluor 488-conjugated anti-mouse (1:1000; Thermo Fisher) and Alexa Fluor 647-conjugated anti-rabbit (1:1000; Thermo Fisher) in 3% BSA in TBS for 1 hour at room temperature each. After washing with TBST, cell nuclei were stained with DAPI (Invitrogen) in PBS for 20 minutes at room temperature. Images were acquired in confocal mode by Operetta CLS High-Content Imaging System (PerkinElmer Life Sciences, Boston, MA) at 40x water objective. Approximately, 0.7–1×104 cells per condition were analyzed for evaluating mitotic and nuclear phenotypes. Harmony high-content analysis software (PerkinElmer Life Sciences, Boston, MA) was first used for detecting the total number of cells, DNA content on a single-cell basis, micronuclei, and phospho-Histone H3-positive cells. PhenoLOGIC Machine Learning algorithm (PerkinElmer Life Sciences, Boston, MA) was then applied to identify abnormal mitotic cells from total mitotic cells via calculating signal intensity and texture feature of Alexa Fluor 488 for α-tubulin, Alexa Fluor 647 for phospho-Histone H3, and DAPI in the phospho-Histone H3-positive cells. This study was performed by the contract research organization, Bioduro-Sundia (Shanghai, China).

In vivo studies

All animal studies were performed with contractors, drug names were masked. Mice were not anesthetized during the process of inoculation, drug treatment and tumor or body weight measurement. When tumor volume reached >2000 mm3 or showing obvious signs of ulceration or animals showing ≥20% body weight loss or obvious signs of severe distress, animals will be humanely sacrificed by carbon dioxide followed by cervical dislocation to ensure death. Confounders were not controlled. In vivo studies in cancer cell line-derived xenograft (CDX) models were performed at Pharmaron, Inc. and animal use was approved by Pharmaron’s Institutional Animal Care and Use Committee (IACUC) in Pharmaron following the guidance of the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). Six- to 8-week-old female NOD/SCID mice (~18–22 g) obtained from GemPharmatech Biotech Co., Ltd. were used in the efficacy and PK/PD studies. Mice were quarantined and acclimatized to the laboratory environment for 7 days before the study and housed in an environmentally monitored, well-ventilated room maintained at a temperature of (23 ± 3 °C) and relative humidity of 40% - 70%, were exposed to 12 hours of light and dark cycles and were supplied with food and water ad libitum. Mice were injected subcutaneously with tumor cells in 200 μL of media and Matrigel mixture (1:1 ratio) in the right flank with 2 × 107 cells. Mice health was monitored daily, and caliper measurements began when tumors were palpable. The major outcomes includes tumor volume and body weight changes after drug treatment. Tumors were measured twice a week, and tumor volume (TV) measurements were determined utilizing the formula A × B2/2 in which A and B are long and short diameters of a tumor, respectively. The TVs are used for the calculation of the tumor growth inhibition (TGI, an indicator of antitumor effectiveness) value using the formula: (1−(Td–T0)/(Cd–C0)) × 100%. Td and Cd are the mean tumor volumes of the treated and control animals, and T0 and C0 are the mean tumor volumes of the treated and control animals at the start of the experiment. Treatment started when tumors reached an average tumor volume of ~200 mm3 for the efficacy study, and 300 to 400 mm3 for PK/PD studies. Mice with tumor size too big or too small were excluded. Based on the tumor volume and body weight, mice were randomized into treatment groups with n = 10 or 8 mice per group as shown in each Figure legend. Total mice of 40 or 32 were used for the whole study. No a priori calculations were performed, group size was chosen based on historical knowledge of model characteristics. Mice were treated with vehicle (20% Hydroxypropyl-Beta-Cyclodextrin in ddH2O) or drugs orally or intraperitoneally as described in Figure legends. Mean TV change (TV ± SEM) of different treatment groups was compared with the vehicle control group in Fig. 4, as well as between each treatment group in Fig. 5. Body weight changes were monitored twice a week. The experiment was terminated when the mean TV exceeded 2000 mm3 or severe body weight loss. In vivo studies in patient-derived xenograft (PDX) models were conducted at XenoSTART (San Antonio, TX) according to the guidelines approved by the IACUC of XenoSTART. The doses and treatment regimens for each group are described in detail in the respective figures. Tumor fragments (approximately 70 mg, ST633, and ST2086) were implanted subcutaneously into the flank region of 6−12 weeks old female athymic nude mice (Crl:NU(NCr)-Foxn1nu) (Charles River Laboratories). N = 8 mice in each group, TV and animal weight data were collected twice a week electronically using a digital caliper and scale, respectively. Mean TV change of different treatment groups was compared with the vehicle control group in supplementary Fig. 4. Mice were housed under standard conditions. All animal studies were carried out under protocols approved by the XenoSTART IACUC Committee. Once tumors reached a TV of 150–300 mm3, animals were matched by TV and randomized to control (untreated) and treatment groups (n = 5 mice per group). Mice were dosed by oral gavage for 4 weeks. Tumors samples were collected once tumors reached a TV of 1 cm3 for gross and histological biomarker analyses as well as gene expression profiling.

Statistical analysis

Unless stated otherwise, all data were statistically analyzed using the GraphPad Prism software (Version 10). Drug interactions were evaluated using the SynergyFinder Plus computational tool with the Loewe additivity (Loewe) model. Loewe values > 10 indicate synergistic effects of the drugs, Loewe values from −10 to 10 indicate additive effects, and Loewe values <-10 indicate antagonism. Multiple comparisons were assessed by one-way ANOVA and grouped data were analyzed by two-way ANOVA. Data were presented as mean ± SEM or violin plots. All data were analyzed using GraphPad Prism 10.0. Results were considered statistically significant if P ≤ 0.05 (*), P ≤ 0.01 (**), P ≤ 0.001 (***).

Clinical trials

ZN-c3-001 clinical trial (NCT04158336) is being conducted in accordance with U.S. ethical guidelines (i.e., U.S.Common Rule) and the Declaration of Helsinki, Good Clinical Practice, and all federal, state regulatory guidelines and local Ethics Committees (University of Texas MD Anderson Cancer center; University of Chicago Medical Center; The University of Arizona Cancer Center; Karmanos Cancer Institute; Mount Sinai and Next Oncology). This is a Phase 1 open-label, multicenter study of azenosertib (also named ZN-c3) monotherapy in subjects with solid tumors which consists of Dose Escalation, a Food Effect Cohort, and Dose Expansion. Azenosertib was administered continuously or intermittently once daily or twice daily in 21-day cycles to patients enrolled on the protocol. Pre-treatment (archival) tumor tissues collected from patients were used for downstream Cyclin E1 IHC assay at QualTek Molecular Laboratories or next-generation sequencing at Foundation Medicine. Skin punch biopsies were performed as surrogate tissue to study azenosertib target engagement. Paired skin punch biopsies (pre-treatment and on-treatment on Cycle 1 Day 15, 3–4 hours post-treatment) were collected from 22 subjects. pCDK1Y15 IHC was performed at QualTek Molecular Laboratories. All patients included in the clinical trial provided written informed consent prior to study enrollment.

Results

Cyclin E1 expression level predicts sensitivity to azenosertib in ovarian cancer cell lines

In order to determine whether Cyclin E1/CDK2 activation via increased Cyclin E1 protein expression level is associated with sensitivity to WEE1 inhibition by azenosertib, we first characterized Cyclin E1 protein levels in nine representative HGSOC cell lines by western blot. As shown in Fig. 1a, Cyclin E1 protein level was found to be high (Cyclin E1high) in OVCAR3, OVCAR4, OAW28, and CAOV3 cell lines and low (Cyclin E1low) in OV90, ES2, COV362, TYK-nu, and KURAMOCHI cell lines (Fig. 1b). The growth inhibition effect of azenosertib was evaluated in these cell lines using the CellTiter-Glo assay. Azenosertib treatment resulted in significant cytotoxicity in all four Cyclin E1high cell lines with maximum growth rate inhibition (GR) less than -0.5. In contrast, azenosertib showed either inhibitory (partial growth inhibition) or cytostatic effect (growth arrest with no or limited cell killing) in Cyclin E1low cells, except COV362, which demonstrated a cytotoxic effect40 (Fig. 1c). The sensitivity of COV362 might be explained by other mechanisms, such as homologous repair deficiency (HRD) due to BRCA1 mutation as suggested by previous studies41. Importantly, of the 4 Cyclin E1high cell lines, only OVCAR3 has CCNE1 gene amplification (CN = 12) (Fig. 1c), suggesting high Cyclin E1 protein level, irrespective of gene amplification status, is predictive of response to azenosertib. No other important components of the cell cycle machinery such as PKMYT1, Cyclin A2 and Cyclin B1, or WEE1 signaling such as WEE1, phosphor-CDK1 and phosphor-CDK2, demonstrated strong predictive value for azenosertib sensitivity (Fig. S1A–C).

Fig. 1: Cyclin E1high HGSOC cell lines are more sensitive to azenosertib in vitro.
figure 1

a Jess analysis showing cyclin E1 expression level in nine high-grade serous ovarian cancer (HGSOC) cell lines. b Quantification of Cyclin E1 level normalized by Vinculin using densitometry. c CellTiter-Glo (CTG) assay to evaluate the growth rate inhibition (GR) performed on nine HGSOC cell lines grown in 2D culture conditions in a 3-day assay. GR and GRmax were determined by measurement of cell viability before and after 72 hours treatment of azenosertib and the GR calculator (http://www.grcalculator.org). (left) The GR values were generated in at least two independent experiments. The representative GR curve is shown. (right) The GR plot in combination with Cyclin E1 expression level showing the correlation of the level of Cyclin E1 and cytotoxic effect (lowering GRmax). (bottom) Summary of the average GRmax and the copy number of nine HGSOC cell lines used in the study.

Full size image

Azenosertib treatment results in stronger growth inhibition and elevated replication stress in Cyclin E1high ovarian cancer cells compared to Cyclin E1low cells

Due to the critical role of Cyclin E1 in G1/S cell cycle progression and DNA replication initiation, high levels of Cyclin E1 can increase replication stress and promote DNA damage and genomic instability9,10,11. Cyclin E1high cancer cells have an increased reliance on the G2/M checkpoint to resolve DNA damage before entering the next stage of cell cycle. WEE1, as the key regulator of G2/M, may be indispensable for Cyclin E1high cancer cells. To further understand how WEE1 inhibition by azenosertib affects cells with different Cyclin E1 levels, we compared downstream signaling changes in Cyclin E1high (OVCAR3 and OVCAR4) and Cyclin E1low (KURAMOCHI) cells after treatment. As expected, azenosertib treatment resulted in a dose-dependent decrease of pCDK1Y15 in all cell lines, regardless of response to azenosertib, suggesting no difference in target engagement between these cell lines (Fig. 2a). Importantly, azenosertib induced a noticeably higher level of replication stress and DNA damage in Cyclin E1high cells compared to Cyclin E1low cells, as demonstrated by increases in pCHK1S345 and γH2AXS139 (Fig. 2a, Supplementary Fig. S2A). Additionally, increased micronuclei formation was observed following azenosertib treatment in OVCAR3 cells compared to Kuramochi cells. The presence of micronuclei is a known marker of DNA damage42. (Supplementary Fig. S2B). We also observed more dramatic changes in cell cycle-specific markers after azenosertib treatment in Cyclin E1high cells, as indicated by increased pHH3 and Cyclin B1 (M phase), and decreased Cyclin E1 (lowest in G2/M). This suggests azenosertib treatment promotes cells with damaged DNA to enter M phase prematurely. Analysis of cells in M phase (pHH3 positive) showed that more cells displaying abnormal mitosis were detected in OVCAR3 compared to KURAMOCHI cells treated with azenosertib (Supplementary Fig. S2C). In addition, azenosertib treatment led to increased apoptosis in Cyclin E1high but not in Cyclin E1low cells as indicated by cleaved caspase 3/7 (Fig. 2a).

Fig. 2: Cyclin E1 overexpression sensitizes HGSOC cell lines to azenosertib, resulting in greater growth inhibition and replication stress.
figure 2

a Jess analysis of Cyclin E1high OVCAR3, OVCAR4, and Cyclin E1low KURAMOCHI cell lines treated with different doses of azenosertib for 16 hours. Biomarkers of target engagement (p-CDK1Y15), cell cycle (Cyclin E1, A2, B1, pHH3S10), DNA damage response (pCHK1S345, γH2AXS139), and apoptosis (Cleaved caspase-3 and 7) were analyzed. b The plot showing the decrease of IC50 and GR50 of OV90 stably expressing Cyclin E1 relative to that of vector control cells determined by measurement of cell viability. c Jess analysis of isogenic pair cell lines of OV90 stably expressing Cyclin E1 treated with different doses of azenosertib for 16 hours. Biomarkers of target engagement, DNA damage response, and mitosis were analyzed. d DNA fiber length analysis of isogenic pair of OV90 control cells (OV90/Vector) and overexpression cells (OV90/Cyclin E1) treated with 200 nM of azenosertib and 0.5 mM of hydroxyurea (HU) for 24 hours. Statistical significance was calculated using two-way ANOVA. * P < 0.05, ****P < 0.0001.

Full size image

We next evaluated the effect of Cyclin E1 overexpression on azenosertib sensitivity in isogenic cell lines. Cyclin E1 overexpression via lentiviral transduction was achieved in three Cyclin E1low cell lines (OV90Cyclin E1, KUROMOCHICyclin E1 and COV362 Cyclin E1). All three Cyclin E1 overexpressed cell lines demonstrated increased sensitivity to azenosertib compared to their respective isogenic control cell line (Fig. 2b and Supplementary Fig. S2D, E). Notably, baseline pCDK1Y15 was considerably higher in OV90Cyclin E1 cells than control cells, suggesting Cyclin E1 overexpression rendered the cells more reliant on the G2/M checkpoint (Fig. 2c). Consequently, inhibition of WEE1 by azenosertib resulted in higher levels of replication stress and DNA damage in OV90Cyclin E1 cells than empty-vector control cells, as demonstrated by increased pCHK1 and γH2AX (Fig. 2c). We also assessed replicating DNA fiber length using a DNA fiber assay, a method to monitor replication fork dynamics and measure replication stress39. Hydroxyurea was used as positive control as it is known to be a replication stress inducer43,44. Compared with control cells, DNA fibers in OV90Cyclin E1 cells were shorter, indicating Cyclin E1 overexpression alone increased baseline replication stress. Azenosertib treatment reduced the length of DNA fibers in both cell lines (Fig. 2d; Supplementary Fig. S2F), supporting previous reports that WEE1 inhibition resulted in replication stress23. The significantly shorter length of DNA fibers in OV90Cyclin E1 cells compared to control cells suggests a model wherein azenosertib treatment exacerbates existing replication stress beyond threshold levels, ultimately leading to cell death.

Sensitivity of Cyclin E1high HGSOC cells to azenosertib is dependent on CDK2

To further evaluate the role of Cyclin E1 in sensitizing ovarian cancer cells to WEE1 inhibition, we examined the activity of azenosertib after CCNE1 knockdown by siRNA in two Cyclin E1high cell lines, OVCAR3 and OVCAR4. Knockdown of CCNE1 desensitized both cell lines to azenosertib (Fig. 3a, b), confirming the importance of Cyclin E1 in response to WEE1 inhibition. Since Cyclin E1 needs to bind and activate CDK2 to enable cell cycle progression, we next investigated whether CDK2 is required for Cyclin E1-mediated azenosertib sensitivity in ovarian cancer cell lines. Knockdown of CDK2 by two independent siRNAs also desensitized OVCAR4 cells to azenosertib, with IC50 increasing 6-14-fold (Fig. 3c, d). Similar results were also observed in OVCAR3 cells after CDK2 knockdown (Supplementary Fig. S3A, B). Downstream signaling analysis demonstrated CDK2 knockdown antagonized the azneosertib-induced changes in cell cycle markers Cyclin E1 and p-HH3 without interrupting azenosertib target engagement (pCDK1Y15) in OVCAR4 cells (Fig. 3e). Moreover, the simultaneous addition of a selective CDK2 kinase inhibitor, tagtociclib, antagonized the growth inhibition by azenosertib in OVCAR3 and OVCAR4 cells, indicating kinase activity of CDK2 is critical for azenosertib sensitivity (Supplementary Fig. S3C). Consistent with the effect in cell proliferation, the knockdown of CDK2 significantly decreased the level of DNA damage and apoptosis induced by azenosertib as indicated by decreased γH2AX and cleaved caspase 3 (Fig. 3e). Collectively, these results show that CDK2 is essential for Cyclin E1-mediated sensitization to azenosertib in ovarian cancer cells.

Fig. 3: Sensitivity of cyclin E1high HGSOC cells to azenosertib is dependent on CDK2/Cyclin E1.
figure 3

a Jess analysis showing Cyclin E1 expression in OVCAR3 and OVCAR4 cell lines transfected with two different CCNE1 siRNAs. b Dose response of OVCAR3 (top) and OVCAR4 (bottom) cell lines transfected with CCNE1 siRNAs to azenosertib for 72 hours. Statistical significance was calculated using two-way ANOVA. ****P < 0.0001. c Jess analysis showing CDK1 and CDK2 expression in OVCAR4 cell line transfected with two different CDK2 siRNAs. d Dose response of OVCAR4 cell line transfected with CDK2 siRNAs to azenosertib for 72 hours. Statistical significance was calculated using two-way ANOVA. ****P < 0.0001. e Jess analysis of indicated antibodies with OVCAR4 cell lines transfected with two different CDK2 siRNAs and treated with different doses of azenosertib for 16 hours.

Full size image

Increased basal replication stress in a Cyclin E1high xenograft model is enhanced by azenosertib and associated with greater in vivo anti-tumor activity

To assess the antitumor activity of azenosertib in vivo, immunocompromised mice bearing OVCAR3 (Cyclin E1high) and SKOV3 (Cyclin E1low) subcutaneous xenografts were treated with azenosertib monotherapy orally once daily (QD) for 28 days at different doses. Cyclin E1 expression levels in tumor samples were examined by immunohistochemistry (IHC) (Supplementary Fig. S4A). We observed significant tumor growth inhibition by azenosertib in a dose-dependent manner in the OVCAR3 model, with a dose of 80 mg/kg resulting in 88% TGI. However, azenosertib did not significantly inhibit tumor growth in SKOV3 model at any doses tested, though there was a trend towards inhibition at the highest dose (Fig. 4a). All treatments were well tolerated in the mice as no significant body weight loss was observed (Supplementary Fig. S4B). Additionally, azenosertib anti-tumor activity was evaluated in two patient-derived xenograft USC models. Consistent with our observations in HGSOC models, dose-dependent inhibition of tumor growth was observed in the Cyclin E1high model (ST633) but not in the Cyclin E1low model (ST2086) (Supplementary Fig. S4C–E). Target engagement of azenosertib was measured by pCDK1Y15 reduction in tumor samples with or without azenosertib treatment. Baseline pCDK1Y15 levels were found to be higher in the Cyclin E1high OVCAR3 model than the Cyclin E1low SKOV3 model, suggesting a dependency of OVCAR3 tumors on the G2/M checkpoint (Fig. 4b). Although target engagement was confirmed in both models, azenosertib treatment resulted in higher levels of DNA damage in the OVCAR3 model as demonstrated by stronger γH2AX level by IHC (Fig. 4c). Furthermore, immunofluorescence staining of markers of single strand breaks and replication stress (pRPAS4/S8 foci), and DNA damage (γH2AX foci) revealed that SKOV3 tumors had lower baseline γH2AX foci than OVCAR3 tumors (Fig. 4d, e). γH2AX and pRPA/γH2AX dual-positive foci were significantly induced by azenosertib treatment in OVCAR3 tumors compared to SKOV3 tumors (Fig. 4d, e). Together, these data suggest WEE1 inhibition via azenosertib exacerbates the already higher levels of DNA damage and replication stress observed in Cyclin E1high tumors, resulting in more pronounced anti-tumor effects compared to Cyclin E1low tumors.

Fig. 4: Greater anti-tumor effects of azenosertib in a Cyclin E1high tumor model are associated with increased replication stress.
figure 4

a Azenosertib was administered via daily oral gavage at the doses indicated to NOD/SCID mice bearing established cyclin E1low SKOV3 (left) and cyclin E1high OVCAR3 (right) xenografts (n = 10/group). No animals were excluded from the analysis. Dosing was initiated when tumors were approximately 200 mm3. Azenosertib was administered to mice daily until day 28. Data are shown as mean tumor volume ± SEM. Statistical significance was calculated using two-way ANOVA followed by Bonferroni post-test. *P < 0.05, **P < 0.01. b, c Immunohistochemistry (IHC) analysis of pCDK1Y15 (b) and γH2AXS139 (c) from the SKOV3 and OVCAR3 xenograft tumors treated with azenosertib for 5 days and collected at 12 hours after the last dose. The Y axis presents the average H-score of 2-3 tumors. d Representative images of immunofluorescence (IF) staining of pRPA32S4/S8 and γH2AXS139 for MCF10A cell line samples treated with gemcitabine for 24 hours and xenograft tumor samples of SKOV3 and OVCAR3 treated with azenosertib for 5 days and collected at 4, 12, and 24 hours after the last dose. MCF10A cells treated with gemcitabine were used as a positive control as gemcitabine is known to induce replication stress. e Quantification of γH2AX foci-positive cells in (d) was shown in top plot. Ratio of dual positivity of pRPA32S4/S8 and γH2AXS139 was shown in bottom plot.

Full size image

Azenosertib and chemotherapy show synergistic effects in Cyclin E1high ovarian cancer models in vitro and in vivo

Chemotherapy is widely used and remains the standard of care for most cancer types, including ovarian cancer and USC. Despite its initial effectiveness, relapse is common and represents a significant unmet medical need. Most chemotherapy agents work by inducing DNA damage, replication defects, or microtubule impairment, providing a strong mechanistic rationale for combining these agents with azenosertib. Therefore, we evaluated the combination of azenosertib with the chemotherapy agents gemcitabine, oxaliplatin, and paclitaxel and contrasted the effects in Cyclin E1high and Cyclin E1low cell lines. Cell viability was measured after monotherapy and combination treatment in a matrixed format, and the synergy score was calculated with the SynergyFinder tool (https://synergyfinder.org). All three combination treatments resulted in greater synergistic effects in Cyclin E1high OVCAR3 cells compared to Cyclin E1low OV90 and TYK-nu cells (Fig. 5a). To determine the mechanistic basis for synergy, downstream cell cycle and DNA damage signaling changes after treatment were analyzed. Paclitaxel is known to arrest the cell cycle at mitosis by stabilizing microtubules, inhibiting depolymerization, and preventing separation of chromosomes45. Indeed, single-agent paclitaxel treatment significantly increased pHH3 and decreased Cyclin E1, indicating M phase arrest (Fig. 5b). The treatment also induced γH2AX and cleaved caspase-3 levels, which was further increased by co-treatment with azenosertib. Interestingly, combination treatment resulted in more apoptosis (cleaved caspase 3) but not much difference in γH2AX level in Cyclin E1high cells compared with Cyclin E1low cells (Fig. 5b; Supplementary Fig. S5A), possibly suggesting an alternative mechanism beyond DNA replication stress or DNA damage. It has been reported that carboplatin treatment can induce Cyclin E1 expression in triple negative breast cancer cell lines31. We observed that carboplatin treatment increased Cyclin E1 expression in a dose- and time-dependent manner in OVCAR3, OV90, and KURAMOCHI ovarian cancer cells (Supplementary Fig. S5B, C). Furthermore, carboplatin treatment increased pCDK1Y15 levels in all cell lines, suggesting that cancer cells showed increased WEE1 activity in response to carboplatin exposure. Combination treatment with azenosertib further increased DNA damage and apoptosis, as demonstrated by increases in γH2AX and cleaved caspase 3, respectively. These effects were more pronounced in Cyclin E1high or Cyclin E1-overexpressing cells (Fig. 5c; Supplementary Fig. S5D). Similarly, gemcitabine, which depletes nucleotide pools in S phase, also increased Cyclin E1 expression and pCDK1Y15 levels, along with increased DNA damage markers. Combination treatment of azenosertib and gemcitabine resulted in significantly more apoptosis in OVCAR3 cells than OV90 cells as demonstrated by increases in cleaved caspase 3 (Fig. 5d).

Fig. 5: Increased synergy between azenosertib and chemotherapy was observed in Cyclin E1high HGSOC cells.
figure 5

a The drug synergistic effect based on Loewe model in Cyclin E1high OVCAR3 (left), Cyclin E1low OV90 (center), and Cyclin E1low TYK-nu (right) cell lines of the combination of azenosertib and oxaliplatin (top), paclitaxel (middle), and gemcitabine (bottom). Loewe values > 10 indicate synergistic effects of the drugs, Loewe values from −10 to 10 indicate additive effects, and Loewe values <-10 indicate antagonism. b Jess analysis of OVCAR3 and OV90 cell lines treated with azenosertib in combination with different doses of paclitaxel for 48 hours with the indicated antibodies. c Jess analysis of OVCAR3 and OV90 cell lines treated with azenosertib in combination with different doses of carboplatin (Carbo) for 48 hours with the indicated antibodies. d Jess analysis of OVCAR3 and OV90 cell lines treated with azenosertib in combination with gemcitabine (Gem) for 48 hours with the indicated antibodies. 123 or 370 nM of azenosertib and 22 or 5 nM of gemcitabine were treated in OVCAR3 and OV90 cell lines, respectively. e The indicated doses of azenosertib, paclitaxel, or a combination were administered to mice bearing the OVCAR3 cell line xenografts (n = 8/group). Statistical significance was calculated on day 20 using Mann–Whitney U test. **P < 0.01, ***P < 0.001 compared to Vehicle treatment, +++P < 0.001 compared to azenosertib or paclitaxel single treatment. f The indicated doses of azenosertib, carboplatin, or a combination were administered to mice bearing the OVCAR3 cell line xenografts (n = 8/group). Statistical significance was calculated on day 28 using two-way ANOVA followed by Bonferroni post-test. *P < 0.05 **P < 0.01 compared to Vehicle treatment, ++P < 0.01 compared to azenosertib or carboplatin single treatment. g The indicated doses of azenosertib, paclitaxel, or a combination were administered to mice bearing the A2780 cell line xenografts (n = 8/group). Statistical significance was calculated on day 10 using Dunnett T3 test. +P < 0.05 compared to azenosertib single treatment. h The indicated doses of azenosertib, carboplatin, or a combination were administered to mice bearing the SKOV3 cell line xenografts (n = 8/group). Tumor growth curves were shown in e–h. TGI, Tumor Growth Inhibition. No animals were excluded from the analysis.

Full size image

To evaluate the anti-tumor activity of azenosertib in combination with chemotherapy in vivo, OVCAR3 (Cyclin E1high, Fig. S4A), SKOV3 (Cyclin E1low, Fig. S4A) or A2780 (previously reported as Cyclin E1low 46) models were utilized. The combination of azenosertib with either carboplatin or paclitaxel significantly improved anti-tumor activity compared with each individual agent in the Cyclin E1high OVCAR3 model (Fig. 5e, f), resulting in substantial tumor regression (p < 0.05). In contrast, our data demonstrate that Cyclin E1low A2780 and SKOV3 models are more resistant than the Cyclin E1high OVCAR3 to all treatments, either alone or in combination. In the Cyclin E1low A2780 model compared to either compound alone, azenosertib combination with paclitaxel induced moderate tumor growth (Fig. 5g). Furthermore, combination of azenosertib with carboplatin demonstrated limited additional anti-tumor activity in the Cyclin E1low SKOV3 model (Fig. 5h). These treatments were within the tolerable range as shown by acceptable body weight change (≤10% reduction in BW) (Supplementary Fig. S5E, F). Collectively, these data demonstrate that azenosertib in combination with chemotherapy may benefit both Cyclin E1high and Cyclin E1low tumor models, however a substantial benefit resulting in regression is only achieved in Cyclin E1high models both in vitro and in vivo.

Azenosertib monotherapy demonstrated anti-tumor activity in patients with Cyclin E1/CDK2-activated cancers

To determine if preclinical results supporting Cyclin E1 as a sensitizing biomarker for azenosertib are translatable into a clinical setting, we analyzed the data from an early cohort of patients enrolled in phase I clinical trial of azenosertib monotherapy in subjects with solid tumors (NCT04158336). Tumor tissue collection was optional in this study and was only available for a subset of patients, so we are highlighting patients with RECIST confirmed partial responses with available archival tissue for assessment of Cyclin E1 protein status and CCNE1 gene copy number. In this study, patients received doses of azenosertib ranging from 350–500 mg QD on an intermittent schedule consisting of 5 days on, 2 days off on a 21-day cycle. Target engagement of azenosertib was evaluated based on changes in pCDK1Y15 from baseline in surrogate skin (hair follicle) tissue from 22 subjects. The results showed a strong correlation between pCDK1Y15 decrease and azenosertib plasma exposure. 100% target engagement (as determined by ≥50% decrease of pCDK1Y15) was achieved at exposures of ~6000 h*ng/mL and above (Fig. 6a). As expected, 4 out of 4 (100%) subjects with a partial response to azenosertib showed evidence of target engagement. We next assessed Cyclin E1 expression in patients that responded to azenosertib treatment. Cyclin E1 protein levels and CCNE1 amplification status were determined by IHC and targeted genomic profiling, respectively, using retrospectively collected archival tissue from each patient. Three patients with both Cyclin E1 protein levels (Fig. 6b) and CCNE1 amplification status available and with objectively confirmed partial responses (PRs) (per RECIST v1.1)47 to azenosertib (Fig. 6c) were selected as case studies to demonstrate the activity of azenosertib in patients with Cyclin E1 alterations46. Additional details are described below and shown in Table 1.

Fig. 6: Azenosertib monotherapy in patients with Cyclin E1high cancers.
figure 6

a Each symbol represents an individual subject from azenosertib phase I clinical trial and shows the percentage of pCDK1Y15 change in surrogate skin tissue (% change from baseline) as quantified by H-score. Green-filled circles indicated subjects that showed partial response to azenosertib. The dotted horizontal line represents the threshold used to define target engagement, defined as ≥50% of pCDK1Y15 on-treatment as compared to baseline. b IHC analysis of Cyclin E1 and H&E staining of archival tumor samples from 3 representative cases. c Azenosertib monotherapy efficacy in patients with Cyclin E1high cancers. Patient 0173-008, HGSOC, week 12 scan indicated a 57% reduction in target lesions. Patient 0171-010, HGSOC, week 12 scan indicated a 42% reduction in target lesion. Patient 0171-019, Uterine Serous Carcinoma, week 12 scan indicated 30% reduction in target lesion.

Full size image
Table 1 Summary information for three responders
Full size table

Patient 0173-008: A 64-year-old female with platinum-resistant HGSOC who received 6 prior lines of treatments, including: 1. Bevacizumab, carboplatin, paclitaxel, and niraparib; 2. Niraparib; 3. Carboplatin and doxorubicin; 4. Topotecan; 5. Batiraxcept; and 6. Senaparib. Archival tissue demonstrated CCNE1 gene amplification (copy number 16), MYC and MDM2 amplification, and TP53 L130V and RAD50 R1238* mutations and confirmed high Cyclin E1 protein level (H-score = 255). Tumor assessment showed a PR with 57% reduction in target lesions at week 12. The patient remained on study treatment for 6 months.

Patient 0171-010: A 67-year-old female diagnosed with platinum-resistant HGSOC who was treated with 5 prior lines of therapies, including: 1. Carboplatin plus paclitaxel as neo-adjuvant/adjuvant therapy followed by niraparib maintenance; 2. Bevacizumab and pegylated-liposomal doxorubicin; 3. Paclitaxel and afuresertib; 4. Atezolizumab and uliledlimab; and 5. Gemcitabine. Archival tissue showed CCNE1 gene amplification (copy number 17), TP53 R175H mutation, KRAS amplification, and high levels of Cyclin E1 protein (H-score = 205). Tumor assessment showed a PR with 42% reduction in target lesions at week 12. The patient had been in the study for 9 months and was still receiving treatment at the time of data cutoff.

Patient 0171-019: A 69-year-old female diagnosed with USC who received adjuvant carboplatin + paclitaxel, followed by lenvatinib + pembrolizumab at the time of recurrence (with unknown response). CCNE1 was not amplified in archival tissue, but a high level of Cyclin E1 protein expression was observed (H-score = 195). TP53 R282W/V272M, PPP2R1A P179R, and NF1 R440* mutations were also shown. Tumor assessment showed a PR with a 30% reduction in target lesions at week 12. The patient remained on study treatment for 3 months.

Discussion

In this study, we evaluated the predictive value of Cyclin E1/CDK2 activation in determining sensitivity to the WEE1 inhibitor azenosertib in models of ovarian cancer and USC. We found high levels of Cyclin E1 protein in cancer cells were strongly associated with sensitivity to azenosertib both in vitro and in vivo. We observed consistent results whether looking at endogenous baseline levels of Cyclin E1 protein expression or in the case of inducible expression of Cyclin E1. Downstream signaling analysis demonstrated Cyclin E1/CDK2 activation increased baseline replication stress as indicated by high levels of γH2AX as well as pronounced effects in a DNA fiber length assay, generally considered the gold standard assay for assessing diverse forms of replication stress. Azenosertib treatment further exacerbated this stress, resulting in DNA DSBs, abnormal mitosis, and ultimately apoptosis. These findings were well translated into an in vivo setting, where we observed significantly higher anti-tumor activity of azenosertib in Cyclin E1high ovarian cancer models and a patient-derived xenograft USC model compared with those Cyclin E1low models. Pharmacodynamic analysis of tumor tissues from these models revealed that azenosertib treatment resulted in significantly higher levels of DNA damage, providing a rationale for the observed higher tumor growth inhibition.

Cyclin E1/CDK2 activation frequently co-occurs with TP53 mutations in many cancers, including ovarian and USC22,48. It is well established that TP53 mutation results in G1/S cell cycle checkpoint defects, thereby increasing dependency on G2/M checkpoint. Concurrent Cyclin E1/CDK2 activation further intensifies this dependency by increasing replication stress, presenting an opportunity for G2/M checkpoint targeting agents, such as WEE1 inhibitors. Our data agrees with this hypothesis, showing that ovarian cancer cells with high Cyclin E1 expression were sensitive to the WEE1 inhibitor azenosertib. Conversely, downregulation of either Cyclin E1 or CDK2 significantly reduced sensitivity to azenosertib, highlighting the critical role of Cyclin E1/CDK2 complex activation in mediating the response to this inhibitor. Our preclinical study also provides evidence that WEE1 inhibition by azenosertib increased replication stress, DNA damage, and mitotic defects. Therefore, genetic alterations that disrupt the cell cycle, induce replication stress, or impair homologous recombination repair may predict the response to WEE1 inhibitors, especially when they coexist with TP53 mutation. Other co-occurring genetic events, such as mutations in FBXW7, PPP2R1A, PIK3CA, BRAF, and KRAS, are also under investigation and will be analyzed in both preclinical and clinical studies.

Activation of the Cyclin E1/CDK2 complex due to CCNE1 gene amplification or protein overexpression has been shown to be an oncogenic driver in many types of cancers, particularly gynecological malignancies7. Cyclin E1 protein upregulation can occur through transcriptional and posttranslational regulation in addition to CCNE1 amplification10. Therefore, Cyclin E1/CDK2 activation detected via Cyclin E1 expression defines a broader patient population than gene amplification, and an IHC assay for protein expression can capture the diverse mechanisms of Cyclin E1 deregulation. Alterations in CCNE1 gene or Cyclin E1 protein have emerged as an important biomarker in many cancer types and settings, defining a patient population that responds poorly to chemotherapy and PARP inhibitors49,50. In addition, elevated Cyclin E1 levels have been linked to acquired resistance of cancer cells to a variety of therapeutic agents, such as CDK4/6 inhibitors and trastuzumab51,52,53. The lack of effective therapeutic options for tumors with Cyclin E1 alterations underscores the urgent need for novel treatment strategies.

Several agents targeting important cell cycle regulators including WEE1, CDK2 and PKMYT1, have been developed with the aim of targeting tumors with CCNE1 amplification or Cyclin E1 protein overexpression. Selective CDK2 inhibitors under clinical investigation include PF-07104091, INCB123667, and BLU222. The limitations of CDK2 inhibitors lie in the fact that Cyclin E1 can promote proliferation through CDK2-independent mechanisms54. Additionally, CDK2 activity can be compensated by CDK1, and developing truly selective CDK2 inhibitors has been challenging54,55,56. Therefore, CDK2 inhibitors may only yield limited cytotoxic effects when used as single agents. PKMYT1, another member of the WEE1 kinase family, plays a role in G2/M checkpoint regulation through inhibitory phosphorylation of CDK1. The PKMYT1 inhibitor lunresertib has been shown to be synthetically lethal with CCNE1 amplification in preclinical models46. However, it only demonstrates moderate anti-tumor activity in preclinical models and limited clinical activity as a single agent46,57. In comparison, WEE1 regulates several stages of the cell cycle by inhibiting both CDK1 and CDK2, which can have significant effects on cancer cells. WEE1 inhibitors might be superior to CDK2 inhibitors and PKMYT1 inhibitors in treating tumors with CCNE1 amplification or Cyclin E1 overexpression based on a broader role in cell cycle regulation. Indeed, as shown here, azenosertib demonstrated robust activity in preclinical studies as well as preliminary anti-tumor activity in ongoing clinical studies of ovarian cancers and USCs. Whether Cyclin E1/CDK2 activation has the same predictive value for azenosertib response in other cancers or in the palbociclib-resistant setting warrants further evaluation.

While Cyclin E1 overexpression has been historically associated with resistance to chemotherapy, we found tumor cells with high level of Cyclin E1 protein were more sensitive to both azenosertib monotherapy and the combination with chemotherapy, suggesting that azenosertib may reverse Cyclin E1-mediated chemotherapy resistance. The DNA damage response and the cell cycle are closely linked through cell cycle checkpoints, with the G2/M checkpoint playing a crucial role in preventing cells with damaged or under-replicated DNA to enter mitosis58. Our findings are consistent with a general model wherein WEE1 kinase inhibition exacerbates the effects of DNA-damaging agents by removing the G2/M checkpoint and triggering mitotic catastrophe. Platinum-based chemotherapies directly interact with DNA, forming intra- and interstrand crosslinks, inducing DNA damage and eventually cell death59. Gemcitabine directly inhibits synthesis of the deoxyribonucleotides required for DNA synthesis, thus interrupting DNA replication and DNA repair and inducing replication stress. In our study, both carboplatin and gemcitabine treatment elevated Cyclin E1 levels, resulting in increase of markers of replication stress and rendering cancer cells more dependent on the G2/M checkpoint as indicated by the accompanying increase of pCDK1Y15. This provides a strong rationale for the combination of azenosertib and these chemotherapeutic agents, and potentially other DNA-damaging agents, for the treatment of patients with Cyclin E1/CDK2 activated tumors. The mechanism of synergy for combination of azenosertib and paclitaxel is different from that of carboplatin and gemcitabine, at least in the models tested. The combination of azenosertib and paclitaxel induced more apoptosis in Cyclin E1high cells without introducing more DNA damage, suggesting a mechanism such as enhanced mitotic defects might be involved. Importantly, in an ongoing clinical trial (NCT04516447), the combination of azenosertib with chemotherapy has exhibited promising clinical anti-tumor activity in patients with platinum-resistant ovarian cancers, with the best outcomes observed in the combination of azenosertib and paclitaxel. Further retrospective analysis demonstrated that patients with high Cyclin E1-expressing tumors demonstrated significant improvements in ORR and progression free survival (PFS) compared against patients with low Cyclin E1-expressing tumors60.

Early signs of azenosertib anti-tumor activity were seen in patients with tumors exhibiting CCNE1 amplification and/or high Cyclin E1 protein expression detected by IHC in the ongoing clinical trial (NCT04158336). We also observed robust pharmacodynamic inhibition of pCDK1 in surrogate tissue obtained by skin punch biopsy (Fig. 6a), suggesting effective target engagement by azenosertib at clinically achievable doses. These results provide important proof of concept that patients with Cyclin E1/CDK2 activation may define a relevant population to enrich for clinical benefit from azenosertib. Further comprehensive biomarker analyses of the broader patient population are underway and are expected to shed light on the sensitivity and specificity of Cyclin E1 protein expression and CCNE1 amplification status as predictive biomarkers of azenosertib response. CCNE1 amplification is included on most next-generation sequencing platforms, and IHC for Cyclin E1 offers a facile way to test for Cyclin E1/CDK2 activation in the community setting. Pending prospective validation in future studies, both approaches may provide an avenue for companion diagnostic testing to identify patients who may benefit from azenosertib treatment.

Collectively, our findings suggest a model wherein tumors with activation of Cyclin E1/CDK2 undergo rapid cell cycling and concomitant increases in DNA damage that leads to a replication stress phenotype, particularly in the setting of loss of p53 and compromise of the G1/S checkpoint. Azenosertib treatment removes the normal inhibitory control that WEE1 exerts on CDK1 and CDK2 and further exacerbates the level of cellular replication stress, leading to intolerable levels of DNA damage and ultimately mitotic catastrophe and cell death. Overall, the data presented here highlight Cyclin E1/CDK2 activation as a sensitizing biomarker for azenosertib, supporting the implementation of biomarker-driven clinical trials with azenosertib. Azenosertib may represent a novel treatment option for patients with Cyclin E1 activated tumors with or without CCNE1 gene amplification. Moreover, azenosertib may represent a novel monotherapy treatment option for patients with Cyclin E1/CDK2 activated tumors with or without CCNE1 gene amplification as well as enhance chemotherapy-induced DNA damage and anti-tumor activity across multiple tumor types.