Upregulation of Myt1 Promotes Acquired Resistance of Cancer Cells to Wee1 Inhibition Free

Adavosertib (also known as AZD1775 or MK1775) is a narrow spectrum inhibitor of the protein kinase Wee1 that has single-agent clinical activity in multiple solid tumors, including sarcoma, glioma, head and neck cancer, and ovarian cancer (1, 2).

Wee1 activity is crucial for maintaining the S- and G₂–M-phase DNA damage checkpoints (3–5) and as such adavosertib sensitizes cancer cells to genotoxic treatments including ionizing radiation, gemcitabine, cisplatin, and camptothecin (6–10). On its own, adavosertib treatment forces S-phase HeLa (cervical cancer cells) and breast cancer cells to directly enter mitosis (10–13). This causes premature condensation of underreplicated chromosomes, leading to double-stranded breaks at the centromeres (centromere fragmentation; refs. 11, 14, 15). Subsequently, these cells arrest and die in prometaphase or following mitotic slippage (11). Nevertheless, some tumors do not respond to adavosertib in the clinic (1, 16); and the mechanisms underpinning clinical resistance are unknown.

In eukaryotes, Wee1 and the related Myt1 kinase (PKMYT1; ref. 17) exhibit functionally redundant roles in the inhibition of the mitosis-promoting complex—Cdk1/cyclin B (18–21). Wee1 phosphorylates Cdk1 on Y15, whereas Myt1 phosphorylates Cdk1 on both T14 and Y15 (17, 21). When cells are ready to enter mitosis, the phosphatase Cdc25C removes these inhibitory phosphates from Cdk1 (22, 23). Cdk1 is rephosphorylated by Wee1 during mitotic exit—again inhibiting its activity (11, 24–26).

Because Wee1 and Myt1 exhibit functional redundancy in Cdk1 inhibition, compensatory Myt1 activation is a candidate mechanism for adavosertib resistance. However, several studies show that knockdown or inhibition of Wee1 alone is sufficient to abrogate the S- and G₂–M DNA damage checkpoints and to cause cells to prematurely enter mitosis (8, 27–29). In contrast, the loss of Myt1 (in the presence of Wee1) neither affects the timing of mitosis nor abrogates DNA damage checkpoints (8, 27–29). These observations led some researchers to conclude that Myt1 is not required for Cdk1 inhibition in cancer cells. However, a more recent study showed that Myt1 is essential for cell survival in a subset of glioblastoma cells that have downregulated Wee1 expression (30). In these glioblastoma cells, loss of Myt1 induced a mitotic arrest followed by cell death (30). In addition, Chow and Poon reported that the combined knockdown of Wee1 and Myt1 causes more HeLa cells to enter mitosis with damaged DNA compared with Wee1 knockdown alone (8). Furthermore, Myt1 knockdown enhances adavosertib-induced cell killing in cell lines derived from brain metastases (31).

Although adavosertib is in clinical development in multiple cancer types, ongoing trials include patients with advanced and metastatic breast cancer. Given the high breast cancer incident and mortality in North America and Europe (32), we studied adavosertib resistance in breast cancer models and report that Myt1 upregulation mediates intrinsic and acquired adavosertib resistance through the inhibition of ectopic Cdk1 activity.

HeLa cells were received directly from the ATCC, whereas breast cancer cells were received from Dr. Roseline Godbout, University of Alberta, Edmonton, Alberta, Canada (also purchased from the ATCC) who amplified, aliquoted, and then froze cells at early passage (P3) in liquid nitrogen. A P3 aliquot was subsequently received and then reamplified, aliquoted, and frozen by our laboratory (P6-9). Cell lines were tested for Mycoplasma contamination by DAPI staining and confocal imaging. HeLa, MDA-MB-468, MDA-MB-231, SK-BR-3, and BT-474 cells were grown as a monolayer in high-glucose DMEM supplemented with 2 mmol/L l-glutamine and 10% (vol/vol) FBS. T-47D and MCF7 cells were grown in high-glucose DMEM supplemented with 2 mmol/L l-glutamine, 10% (vol/vol) FBS, and 0.01 mg/mL insulin. MCF10A and HME-1 cells were grown in Mammary Epithelial Growth Medium supplemented with SingleQuots (Lonza; CC-3150). Cell lines were maintained in culture for a maximum of 2 months (∼20–25 passages). MDA-MB-231 cells expressing mCherry-H2B and EGFP-tubulin were generated as outlined by Moudgil and colleagues (33). MDA-MB-231 cells were transfected with mClover3-10aa-H2B (34). HeLa cells were transfected with mClover3-10aa-H2B (34) and tdTomato-CENPB-N-22 (35). All cell lines were cultured in a humidified incubator at 37°C with 5% CO₂.

Cells were synchronized in G₁–S phase by double thymidine block as outlined in Moudgil and colleagues (33). Cells were treated with 2 mmol/L thymidine for 16 hours with an 8-hour release interval between thymidine treatments. For cell synchronization described in Figs. 4 and 6A–C and Supplementary Fig. S4 (measuring premature entry into mitosis), cells were released from the second thymidine block for 4 hours and then subjected to indicated treatments.

Adavosertib (Chemietek; 955365-80-7), RO-3306 (Sigma; SML0569), and thymidine (Sigma; T1895) were prepared as 10 mmol/L solutions in DMSO and then stored at −20°C.

siRNA for Wee1 5′-CAUCUCGACUUAUUGGAAAtt-3′ (Ambion; siRNA ID: s21), Myt1 5′-GGACAGCAGCGGAUGUGUUtt-3′ (Ambion; siRNA ID: s194985), and 5′-GCGGUAAAGCGUUCCAUGUtt-3′ (Ambion; siRNA ID: s194986) and a scrambled control siRNA 5′-UGGUUUACAUGUCGACUAA-3′ from Thermo Fisher Scientific were used at a concentration of 20 nmol/L (unless otherwise indicated) with 0.2% Lipofectamine RNAiMax (Thermo Fisher Scientific) for 24 hours prior to treatments with adavosertib. In the siWee1 dilution assay, the amount of RNAiMax Lipofectamine used was fixed at 0.2%. Knockdown efficiency was analyzed by Western blot and normalized to tubulin or actin 48 hours after transfection.

A total of 2 × 10⁶ MDA-MB-231-fluc2-tdT cells were mixed with Matrigel and 1X PBS (1:1) and injected using a 1 cc syringe with 26G needle in 50 μL volume orthotopically into inguinal mammary fat pad of 6- to 8-week-old female NSG (NOD/SCID gamma) mice procured from Dr. Lynne Postovit's breeding colony at the University of Alberta. Tumor volume was measured every 4 days with a Vernier caliper, and volume was calculated as [length × (width)²]/2. When tumors reached a volume of about 150 mm³, mice were randomly segregated into two groups (n = 3 per group). Mice received daily treatment with either vehicle (0.5% methylcellulose dissolved in sterile water) or 60 mg/kg adavosertib via oral gavage for 26 days. Tumors were harvested 24 hours after last drug treatment and fixed with 10% formalin for 48 hours prior to embedding. All animal work was approved by the Cross Cancer Institute Animal Care Committee in accordance with the Canadian Council on Animal Care guideline.

Total RNA was isolated from frozen breast tumor biopsies, and gene microarray analysis, data processing, and reverse transcription PCR were processed (as outlined in refs. 36–38). One Wee1 (A_23_P127926) and two Myt1 primers (A_24_P105102 and A_23_P398515) were available for analysis. Myt1 primers were then averaged together after confirming that mRNA detection was similar by comparative analysis. DNA microarray data are deposited in NCBI's Gene Expression Omnibus, accession number GSE22820 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?).

Cells were harvested and processed for Western blot as described previously by Famulski and colleagues (39). Protein extracts were separated on 12% polyacrylamide gels for 7 to 15 minutes at 200 V. PageRuler Plus Prestained protein ladder (Thermo Fisher Scientific; 26619) was used as a molecular weight marker. Proteins were transferred on to nitrocellulose for 7 to 15 minutes at 25 V by Trans-Blot Turbo Transfer System. Membranes were blocked with Odyssey blocking buffer (LI-COR Biosciences). Membranes were probed with the following primary antibodies: anti-Wee1 antibodies (Cell Signaling Technology; 4936; 1:1,000 dilution), anti-Myt1 antibodies (Cell Signaling Technology; 4282; 1:1,000 dilution), anti-Cdc25C antibodies (Cell Signaling Technology; 4688; 1:11,000), anti-Cdk1 antibodies (Santa Cruz Biotechnology; sc-54; 1:500 dilution), anti-phospho-tyrosine 15 Cdk1 (Cell Signaling Technology; 9111; 1:1,000 dilution), anti-phospho-threonine 14 Cdk1 (1:1,000 dilution), anti-tubulin antibodies (Sigma; T5168; 1:4,000 dilution), anti-PARP antibodies (Cell Signaling Technology; 9542; 1:1,000 dilution), anti-pT320-PP1Cα antibody (Abcam; ab62334; 1:30,000), and anti-GST antibody (Rockland; 600-401-200; 1:2,000 dilution). Membranes were then incubated with Alexa Fluor 680–conjugated anti-rabbit (Thermo Fisher Scientific; A21109; 1:1,000 dilution), anti-mouse (Thermo Fisher Scientific; A21057; 1:1,000 dilution), IR800 anti-mouse (LI-COR Biosciences; 827-08364; 1:1,000 dilution), and IR800 anti-rabbit (LI-COR Biosciences; 926-32211; 1:1,000). Membranes were scanned by the Odyssey Fc (LI-COR Biosciences) and then analyzed by Image Studio Lite software.

Cells were processed for immunofluorescence as previously described by Chan and colleagues (39). Cells were seeded on to coverslips at a density of 5 × 10⁴ cells/mL in a 35-mm dish. Following cell synchronization, cells were treated with either DMSO or adavosertib at indicated concentrations. Treatments were maintained for 4 hours, and 0.1% DMSO was used as a control in all experiments. siRNA transfections were performed as outlined in the RNAi section. DNA was stained with 0.1 μg/mL DAPI. Coverslips were stained with the following antibodies: anti-phospho-Ser10 Histone H3 (PH3) antibodies (Abcam; ab5176; 1:1,000 dilution), anti-tubulin antibodies (Sigma; T5168; 1:4,000 dilution), and anti-centromere antibody sera (gift from M. Fritzler, University of Calgary, Calgary, Canada; 1:4,000 dilution). Coverslips were mounted with 1 mg/mL Mowiol 4-88 (EMD Millipore) in phosphate buffer, pH 7.4. Alexa Fluor 488–conjugated anti-mouse and anti-rabbit (1:1,000 dilution; Molecular Probes) and Alexa Fluor 647–conjugated anti-human (1:1,000 dilution; Molecular Probes) secondary antibodies were used to visualize protein localization. Images were captured at ×63 magnification using a Zeiss LSM 710 Meta Confocal Microscope (Carl Zeiss). The pinhole diameter was set at 1 airy unit for all channels, and the exposure gain for each channel was kept constant in between image acquisition of all samples.

IHC was performed on formalin-fixed, paraffin-embedded tissue samples using standard procedures as previously described (40). Briefly, 4 μm slices were sectioned on precleaned Colorfrost Plus microscope slides (Thermo Fisher Scientific) using a microtome (Leica). Tissue samples were baked at 60°C for 2 hours and deparaffinized 3 times in xylene for 10 minutes each and subsequently rehydrated in a gradient of ethanol washes (100%, 80%, and 50%). Tissue sections were subjected to antigen retrieval in a pressure cooker using 0.05% citraconic anhydride antigen retrieval buffer (pH 7.4). Tissue samples were blocked with 4% BSA for 30 minutes and incubated with primary antibody against Myt1 (1:50; Cell Signaling Technology; 4282) overnight at 4°C. Next day, endogenous peroxidase activity was blocked for 30 minutes using 3% H₂O₂, followed by incubation with anti-rabbit horseradish peroxidase–labeled secondary antibody (Dako EnVision+ System; K4007) for 1 hour at room temperature in the dark. Samples were incubated with DAB + substrate chromogen (Dako) for brown color development, counterstained with hematoxylin, and mounted with DPX mounting medium (Sigma). Images were captured using the Zeiss Axioskop2 plus upright microscope (Carl Zeiss) equipped with AxioCam color camera. Images were then analyzed using IHC profiler plugin (40) for ImageJ as described previously (41). Briefly, images were color deconvoluted to unmix pure DAB (3,3′-diaminobenzidine) and hematoxylin-stained areas using the nuclear-stained image option in the IHC profiler plugin. DAB-stained (brown) nuclei were marked using the threshold feature of ImageJ and assigned an automated score using the IHC profiler macro. The automated score is calculated based on the following formula:

Wherein the score of the zone is assigned as 4 for the high-positive (+3) zone, 3 for the positive (+2) zone, 2 for the low-positive (+1) zone, and 1 for the negative (+0) zone (40).

Cells were seeded in a 35-mm glass bottom dishes (MatTek Corporation). Glass bottom plates were placed on a motor-controlled stage within an incubator chamber maintained at 37°C and 5% CO₂. Live cell imaging was carried out with the Ultraview ERS Rapid confocal imager (PerkinElmer) on an Axiovert 200M Inverted Microscope (Carl Zeiss) and a CMOS camera (ORCA-Flash4.0, Hamamatsu) using the 40× objective lens. Images were captured at 5-minute interval for 24 hours using the Volocity software. The 488 nm and 561 nm lasers were set at 20% power and 200 ms exposure time. Movie files were exported as OME-TIFF files and further processed in Imaris 9.0.1 (Bitplane) for background subtraction and noise reduction.

Images were taken with a high-content–automated microscopy imaging system (MetaXpress Micro XLS, software version 6, Molecular Devices, as outlined in Lewis and colleagues; ref. 11). Briefly, MDA-MB-231 and HeLa cells were seeded onto a 96-well plate at a density of 4,000 cells per well. Single images were captured in each well with a 20× (NA 0.75) objective with the equipped siCMOS camera using bandpass filters of 536/40 nm and 624/40 nm. On average 200 cells per well were imaged. The images were then manually analyzed with the MetaXpress software using the mCherry-H2B to mark changes in DNA organization. Mitotic timing was calculated by the interval between nuclear envelope breakdown (indicated by the first evidence of chromosome condensation) to the onset of anaphase (or chromosome decondensation in the case of mitotic slippage). Only cells that entered mitosis were analyzed for mitotic timing experiments, and the fates of the mitotic cells (and resulting daughter cells) were tracked for the duration of the experiment (48 hours). Cell death was determined by the formation of apoptotic bodies, loss of cell attachment, and/or loss of membrane integrity.

The phospho-specific antibody against pT14-Cdk1 was generated by immunizing rabbits with a synthetic peptide phosphorylated at the T14 residue (conjugated to KLH). Sera were first depleted of antibodies against the unphosphorylated epitopes with a nonphosphorylated peptide column. Cdk1-pT14 antibodies were then affinity-purified with a pT14 peptide column. Specificity of the antibodies was demonstrated by no signal in Western blot of mutant myt1 fly extracts.

Cells were seeded into 96-well plates and transfected with siRNAs against Wee1, Myt1, and scrambled control for 24 hours. Cells were then treated with increasing concentrations of adavosertib (16–4,000 nmol/L, 1:2 serial dilution). After 96-hour treatment, media were aspirated and then cells were stained with 0.5% crystal violet for 20 minutes as outlined by Bukhari and colleagues (13). Crystal violet was then removed, and plates were rinsed 3 times with water and left to air-dry for 24 hours. Crystal violet within stained cells was resuspended in 100% methanol. Absorbance at 570 nm was measured using FLUOstar OPTIMA microplate reader (BMG Labtech). Percent cell survival was calculated by subtracting blank wells and then normalizing DMSO controls to 100%. The first point on each curve represents 0 nmol/L adavosertib. Graphs were plotted using GraphPad Prism V7.

Cdk1 kinase assays were completed as outlined in Lewis and colleagues (42, 43). Briefly, 20 ng of glutathione-S-transferase (GST) fused with a 9 amino acid PP1Cα peptide (GRPITPPRN) was combined with 2,000 cells in 2x Cdk1 phospho-buffer (100 mmol/L β-glycerophosphate, 20 mmol/L MgCl₂, 20 mmol/L NaF, and 2 mmol/L DTT) and 400 μmol/L ATP and then incubated at 37°C for 15 minutes. Reactions were then terminated with Laemmli sample buffer (Bio-Rad; 161-0747) and then analyzed by Western blot. The ratio of pT320-PP1Cα to GST, minus “no lysate” control, was calculated for each sample. DMSO was set to one for adavosertib serial dilution and HeLa for baseline Cdk1 experiment.

To test acquired resistance to the Wee1 inhibitor adavosertib, we derived adavosertib-resistant cell lines from HeLa (cervical cancer) and MDA-MB-231 (breast cancer) cell lines. We and others have previously shown that HeLa and MDA-MB-231 are highly sensitive to adavosertib as single-agent treatment (10, 11, 13). We selected for resistant cells by culturing these cell lines in medium containing 500 nmol/L adavosertib for approximately 2 months (Fig. 1A). The resulting cell populations were tested for adavosertib sensitivity by crystal violet assays (13). In the case of both MDA-MB-231 and HeLa, adavosertib-selected cells (denoted as “R500”) showed much higher resistance to the Wee1 inhibitor compared with mock-treated (Parental, “P”) cells (Fig. 1B and C). Selection increased the IC₅₀ from 305 to 1,090 nmol/L in HeLa and from 349 to 1,217 nmol/L in MDA-MB-231. Because the activities of two related kinases Myt1 and Wee1 were found to be redundant in Cdk1 regulation in various organisms/tissues (19–21, 30), we hypothesized that upregulation of Myt1 could underline Wee1 inhibitor resistance. We therefore compared Myt1 protein levels in the derived adavosertib-resistant cell lines with the parental cell lines. Indeed, resistant cell populations had increased Myt1 levels: 2.0-fold in HeLa and 3.1-fold in MDA-MB-231 relative to parental cell populations. To test whether decreasing Myt1 levels could resensitize resistant cells to adavosertib, we used two different siRNAs to knockdown Myt1 (siMyt1 #1 and #2) and compared adavosertib sensitivity with siRNA scrambled control–treated cells (siSc). Myt1 knockdown in R500 HeLa cells decreased the IC₅₀ from 1,261 nmol/L in siSc-treated cells to 283 and 373 nmol/L in siMyt1 (#1 and #2)-transfected cells, respectively (Fig. 1D). In R500 MDA-MB-231 cells, Myt1 knockdown decreased the IC₅₀ from 1,447 nmol/L (siSc) to 163 and 371 nmol/L for siMyt1 #1 and #2, respectively (Fig. 1E). These data suggest that Myt1 upregulation is a driver of resistance in the R500 HeLa and MDA-MB-231 cell lines.

We previously tested the efficacy of adavosertib treatment in an orthotopic breast cancer xenograft model (13). In that study using a luciferase-labeled MDA-MB-231 cell line, mice were treated with 60 mg/kg of adavosertib for 26 days. Although adavosertib treatment caused significant tumor growth delay, no tumor shrinkage was observed (13). This indicates that the tumors had at least partially acquired adavosertib resistance. To investigate whether increased Myt1 expression contributed to tumor resistance, we harvested MDA-MB-231–derived tumors from NSG (NOD/SCID gamma) mice immediately after a 26-day treatment with 60 mg/kg of adavosertib. Tumor slices were analyzed for Myt1 expression and Wee1 inhibition by IHC (Fig. 1F; n = 3 per group). Tumors treated with adavosertib showed increased Myt1 expression (+2; positive) compared with vehicle control–treated mice (+1; low positive). These results strongly support Myt1 upregulation as a mechanism for acquired adavosertib resistance in vivo.

To confirm that upregulation of Myt1 can confer adavosertib resistance, we transiently overexpressed Myt1 tagged with GFP (GFP-Myt1) or GFP alone in HeLa and MDA-MB-231 cells and then measured cell sensitivity to adavosertib by crystal violet assays (Fig. 2A). GFP-Myt1 overexpression resulted in a modest but significant increase in the adavosertib IC₅₀s for both cell lines relative to GFP-transfected controls 2.2-fold and 1.7-fold increase in HeLa (P < 0.001; two-way ANOVA) and MDA-MB-231 (P = 0.0079; two-way ANOVA), respectively. Although HeLa exhibited a greater increase in the adavosertib IC₅₀ compared with MDA-MB-231, HeLa also exhibited higher GFP-Myt1 expression levels compared with MDA-MB-231.

Because increased Myt1 expression induced adavosertib resistance in HeLa and MDA-MB-231, we wondered whether endogenous Myt1 expression in other cell lines could be used as a biomarker to predict adavosertib sensitivity. Adavosertib sensitivity was screened in a panel of breast cell lines (five cancer and two nontumorigenic; Supplementary Table S1). HeLa and MDA-MB-231 cells were included as positive controls for adavosertib sensitivity. Cell lines were transfected with siMyt1 or siSc and then treated with adavosertib for 96 hours. Adavosertib treatment reduced cell number in a dose-dependent manner in all nine cell lines tested, and as expected Myt1 knockdown (confirmed by immunoblot) further reduced cell number (Fig. 2B–E; Supplementary Fig. S1A–S1E). We calculated IC₅₀ values for each cell line (Supplementary Table S2). HeLa, MDA-MB-231, and HME-1 cell lines had IC₅₀ values in the 300 nmol/L range, but the remaining 6 cell lines (MCF10A, MDA-MB-468, SK-BR-3, MCF7, T-47D, and BT-474) had IC₅₀ values that were approximately 2 to 4 times higher. To investigate how the high IC₅₀ values correlated with Myt1 levels, Myt1 protein levels (normalized to total protein content) were quantified in each cell line (Fig. 3A; top) and then plotted against the corresponding IC₅₀ values. Linear analysis revealed a strong correlation between calculated IC₅₀ values and Myt1 protein levels (Fig. 3B; top plot; R² = 0.6903) in agreement with our prediction. We also tested if the levels of Wee1 or Cdc25C (the phosphatase responsible for reversing Cdk1 phosphorylation, by Wee1 and Myt1) correlated with cell sensitivity to adavosertib, but no significant correlations were observed (Fig. 3A and B). These data strongly support that Myt1 levels can be used as predictive biomarkers for cell sensitivity to adavosertib.

In addition to Wee1, adavosertib also exhibits activity against Polo-like kinse-1 (Plk1) in some cell types including small-cell lung carcinoma cells (44, 45). However, adavosertib treatment induces premature mitosis in cells consistent with the inhibition of Wee1, whereas inhibition of Plk1 induces a G₂ arrest (46). To confirm that the reduced cell survival observed following adavosertib/siMyt1 treatment was dependent on loss of Wee1 activity and not an off-target effect of adavosertib, we substituted the treatment with adavosertib for a validated siRNA-targeting Wee1 (siWee1; ref. 11). Select cell lines (HeLa, MDA-MB-231, MDA-MB-468, and T-47D) were transfected with increasing amounts of siWee1 in the background of either siSc or siMyt1 (Supplementary Fig. S2A–S2D). siWee1 transfection decreased cell survival in a dose-dependent manner. However, cells transfected with siMyt1 + siWee1 had fewer surviving attached cells compared with siSc + siWee1 controls in all four cell lines (P < 0.0001; two-way ANOVA). These data corroborate that the observed cell death with adavosertib and siMyt1 is due to loss of Wee1 and Myt1 activity.

Due to the role of Myt1 in cell-cycle regulation, we suspected that Myt1 promotes adavosertib resistance by maintaining Cdk1 inhibition. Structure–function studies have reported that adavosertib does not strongly interact with Myt1 and is 100 times more selective toward Wee1 (9, 47). To confirm this, we treated MDA-MB-231 and MDA-MB-468 cells with adavosertib and then assayed Myt1 and Wee1 activity by examining Cdk1 phosphorylation. In the presence of adavosertib, the cellular levels of pT14-Cdk1 (a surrogate biomarker for Myt1 activity) remained stable, whereas the levels of pY15-Cdk1 (a surrogate biomarker of Wee1 activity) declined (Supplementary Fig. S3A and S3B). These data show that even 1,000 nmol/L adavosertib does not inhibit Myt1.

To confirm that high Myt1 levels inhibit Cdk1 activity even in the presence of adavosertib, we assayed in vitro Cdk1 activity (43). First, we examined whether transient overexpression of Myt1 could suppress Cdk1 activity in HeLa and MDA-MB-231 cells. Cells were transfected with GFP or GFP-Myt1 and then treated with adavosertib for 4 hours following G₁–S release. Lysates were then prepared and incubated with a recombinant Cdk1 substrate, GST-PP1Cα. Total levels of pT320 GST-PP1Cα (an indicator of Cdk1 activity) were then quantified by immunoblot (Fig. 4A; ref. 43). Adavosertib treatment increased in vitro Cdk1 activity in a dose-dependent manner; however, GFP-Myt1 overexpression (confirm by immunoblot) significantly reduced absolute in vitro Cdk1 activity in both HeLa and MDA-MB-231 cells relative to cells expressing only GFP (Fig. 4B and C).

Next, we tested whether adavosertib-resistant cell lines (SK-BR-3 and BT-474) exhibited less in vitro Cdk1 activity compared with HeLa or MDA-MB-231. We first established the baseline Cdk1 activity in cell lines in the absence of adavosertib. Cell lysates were prepared from cells 4 hours after release from G₁–S arrest and normalized by total protein (Fig. 4A; no transfection). SK-BR-3 and BT-474 had 50% to 60% less in vitro Cdk1 activity compared with the more adavosertib-sensitive HeLa and MDA-MB-231 (Fig. 4D). Next, we tested the effects on Cdk1 activity of combining Myt1 knockdown with adavosertib. HeLa, MDA-MB-231, and two adavosertib-resistant cell lines (SK-BR-3 and BT-474) were transfected with siSc or siMyt1 and then treated with adavosertib (Fig. 4E). Low in vitro Cdk1 activity was observed in siSc- and siMyt1-transfected cells in the absence of adavosertib. Adavosertib treatment in siSc-transfected cells increased Cdk1 activity 2- to 4-fold compared with DMSO. However, the highest Cdk1 activity was observed in cells treated with adavosertib and transfected with siMyt1. Together, these data show that high Myt1 levels suppress Cdk1 activity in the presence of adavosertib.

Subsequently, we treated each cell line with increasing concentrations of adavosertib (Fig. 4F). To compare the absolute change in Cdk1 activity between cell lines, the baseline Cdk1 activity (normalized to total protein; Fig. 4D) was multiplied by the change in in vitro Cdk1 activity within each cell line. We found that resistant cell lines (SK-BR-3 and BT-474) had less absolute Cdk1 activity in the presence of adavosertib compared with that of HeLa and MDA-MB-231 (Fig. 4G). To investigate whether the observed Cdk1 activity correlated with entry into mitosis, adavosertib-treated cells were stained for PH3 (a mitosis biomarker; ref. 48) and analyzed by immunofluorescence microscopy. Consistent with in vitro Cdk1 kinase assay, even at 2,000 nmol/L adavosertib, <10% of SK-BR-3 and BT-474 cells stained positive for PH3 compared with 40% to 50% in HeLa and MDA-MB-231 (Supplementary Fig. S4A and S4B). Together, these data strongly support that high Myt1 expression drives adavosertib resistance by suppressing Cdk1 activity.

Wee1 is required for normal mitotic exit, and the loss of Wee1 causes cells to arrest in mitosis (24–26). Similarly, Myt1 knockout in glioblastoma cells has been previously shown to prolong the duration of mitosis, leading to cell death (30). We therefore wondered whether Myt1 levels may affect how long cancer cells arrest in mitosis following adavosertib treatment. We transfected MDA-MB-231 cells expressing mCherry-H2B and GFP-tubulin with siMyt1 or siSc and then treated cells with adavosertib or DMSO. Next, we measured the duration of mitosis by timelapse microscopy (Fig. 5A and B). No significant differences were observed between siMyt1- or siSc-transfected MDA-MB-231 cells in the absence of adavosertib (Fig. 5B; 0 nmol/L); cells exhibited normal chromosome alignment and mitotic timing. In contrast, adavosertib (in the background of siSc) increased the total time in mitosis compared with DMSO controls. Normal chromosome alignment and segregation were observed in most cells at low concentrations of adavosertib (≤250 nmol/L), but adavosertib concentrations ≥ 500 nmol/L induced abnormal chromosome condensation, which frequently was followed by in mitotic slippage (Supplementary Fig. S5A and S5B; two phenotypes are shown). Combined knockdown of Myt1 with adavosertib (125–250 nmol/L) prolonged mitosis more than adavosertib alone (Fig. 5A; P < 0.0001). siMyt1 combined with adavosertib also increased the percentage of cells that exhibited abnormal chromosome condensation and/or mitotic slippage. No increases in the duration of mitosis were observed after siMyt1 transfection and adavosertib concentrations ≥ 500 nmol/L. To compare the effects on mitosis in MDA-MB-231 with another cell line, we repeated the experiment in HeLa cells expressing GFP-H2B (Fig. 5C). Like MDA-MD-231, adavosertib/siSc increased the duration of mitosis in HeLa compared with DMSO/siSc. Unlike in MDA-MB-231, siMyt1 significantly enhanced the duration of mitosis at all concentrations of adavosertib tested.

We measured the percentage of cell death in MDA-MB-231 and HeLa cells by timelapse microscopy for up to 48 hours (Fig. 5D and E). Few cell deaths were observed in siSc/DMSO controls for either cell line; however, 18% of MDA-MD-231 and 25% of HeLa cells died during mitosis or after slippage in siMyt1/DMSO-treated cells. Consistent with cell survival assay data (Fig. 2B and C), we observed that adavosertib alone induced cell death in a dose-dependent manner in both MDA-MB-231 and HeLa, which was further enhanced by siMyt1. Mitotic cell death commonly occurred at high concentrations of adavosertib (≥500 nmol/L) in siSc-transfected cells and at lower doses of adavosertib (125–250 nmol/L) in siMyt1-transfected cells. These data suggest that loss of both Wee1 and Myt1 prolongs mitosis and induces mitotic cell death compared with loss of either kinase alone.

Centromere fragmentation occurs when cells enter mitosis without completing DNA synthesis (11, 14). We previously reported that 20% of HeLa cells treated with 1,000 nmol/L adavosertib underwent mitosis accompanied with centromere fragmentation (11); however, at concentrations close to the IC₅₀ of HeLa (250 nmol/L), centromere fragmentation was rarely observed. Here, we show that after Myt1 knockdown, abnormal chromosome condensation and alignment (characteristic signs of centromere fragmentation; refs. 11, 14) are observed in both MDA-MB-231 and HeLa cells treated with only 250 nmol/L adavosertib. HeLa and MDA-MB-231 cells were transfected with siSc or siMyt1, synchronized in G₁–S, and then released into media containing adavosertib or DMSO for 4 hours (11). Cells were then fixed and stained for centromeres, microtubules, and DNA and analyzed by immunofluorescence microscopy (Fig. 6A). In the absence of adavosertib, <2% of HeLa and MDA-MB-231 cells transfected with siSc or siMyt1 entered mitosis; observed mitotic cells exhibited normal centromere localization, chromosome morphology, and mitotic spindles. Treatment with 250 nmol/L adavosertib in the presence of siSc increased the percentage of mitotic cells to 12% and 17% in HeLa and MDA-MB-231, respectively (Fig. 6A and B). Of these mitotic cells observed in the presence of adavosertib alone, most exhibited abnormal chromosome condensation and had centromeres and microtubules that clustered away from the bulk of the chromosomes, consistent with centromere fragmentation (Fig. 6A and B). However, 250 nmol/L Adavosertib in the presence of siMyt1 increased centromere fragmentation 11-fold in HeLa and 4-fold in MDA-MD-231 (Fig. 6B).

To acquire higher-resolution images of the chromosomes from mitotic cells, HeLa cells were prepared for karyotype analysis (Fig. 6C). The chromosomes in DMSO-treated cells that were transfected with siSc or siMyt1 showed no signs of abnormalities. Likewise, most adavosertib/siSc-treated cells had intact chromosomes with only a small number exhibiting signs of chromosome shattering. In contrast, treatment with adavosertib resulted in chromosome shattering in nearly all mitotic spreads from Myt1 knockdown cells.

Finally, to further investigate the phenomenon of centromere fragmentation, a HeLa cell line expressing GFP-H2B and tdTomato-CENP-B was established to monitor chromosomes and centromeres, respectively. Asynchronous cells were transfected with siMyt1 or siSc and then treated with either DMSO or 250 nmol/L adavosertib for 24 hours. No centromere fragmentation was observed in cells treated with siSc/DMSO, siMyt1/DMSO, or siSc/250 nmol/L adavosertib (Fig. 6D and E; Supplementary Table S3). However, combining 250 nmol/L adavosertib with siMyt1 resulted in centromeres clustering away from the bulk of the chromosomes in 26.7% of cells (Fig. 6D; Supplementary Table S3). Cell death was observed in 100% of cells exhibiting centromere fragmentation (Fig. 6E). To confirm that centromere fragmentation was dependent on upregulated Cdk1 activity following loss of Wee1 and Myt1 activity, we treated cells with the small-molecule Cdk1 inhibitor RO3306. We found that RO3306 completely suppressed centromere fragmentation in siMyt1/adavosertib-treated cells (Supplementary Table S3). Collectively, these findings may indicate that the ability of Myt1 to inhibit Cdk1 during DNA replication protects cells from undergoing centromere fragmentation when Wee1 is inhibited by adavosertib.

Because most of our data were derived from breast cancer cell lines, we wanted to know if Myt1 was overexpressed in tumors from patients with breast cancer. We compared Myt1 mRNA levels in breast cancer tissues (176 samples; refs. 36–38) against normal breast tissue (10 samples) by cDNA microarray (36–38) and found that median mRNA levels of Myt1 were approximately 14-fold higher in cancer tissue compared with normal tissue (P = 0.0004; Student t test; Fig. 7A; Supplementary Fig. S6A and S6B). Nevertheless, we noted that Myt1 mRNA levels varied greatly among samples. We then correlated Myt1 expression with overall disease grade, mitotic grade, and hormone receptor status (basal like vs. ER positive; Fig. 7B–D; Supplementary Fig. S6C–S6E). We found that higher Myt1 expression was associated with a higher overall disease grade (P < 0.0001; Student t test), a higher mitotic grade (P < 0.0001; Student t test), and basal-like (triple negative) hormone receptor status (P = 0.009; Student t test). We next evaluated whether Myt1 expression was associated with either disease recurrence (disease-free survival) or overall patient survival (Fig. 7E and F). We therefore dichotomized the samples into high and low Myt1 expression groups with the low group representing the bottom quarter percentile of the samples, which is comparable with Myt1 expression in normal tissue. We found that higher Myt1 expression was associated with both a worse disease-free survival (P = 0.0118; Mantel–Cox test) and overall survival (P = 0.0121; Mantel–Cox test). Because our sample size (n = 176) was relatively small, we accessed the cBioPortal database to compare high and low Myt1-expressing breast cancers to confirm our findings. An analysis of 1,423 additional samples confirmed high Myt1 levels were strongly associated with a lower overall survival (Supplementary Fig. S6F; P < 0.0001; Mantel–Cox test; refs. 49, 50). Previous studies have reported that Wee1 is overexpressed in various cancers including breast cancer (3, 4, 51–54). In this study, no differences in Wee1 expression were observed between breast cancer and normal tissue samples (Supplementary Fig. S6G). Breast cancer cells with Wee1 levels above the median expression were associated with a higher overall and mitotic grade (Supplementary Fig. S6H and S6I), but there were no differences in hormone status, overall survival, and disease-free survival (Supplementary Fig. S6J–S6L). Although it is worth pointing out that we examined only Myt1 mRNA levels, not Myt1 protein expression, these data suggest that Myt1 overexpression may be an important mechanism promoting cancer development.

Clinical trials show that adavosertib treatment responses are variable, and some cancers do not respond to adavosertib (1, 16); however, the mechanisms of drug resistance are unknown. We find that cancer cells can acquire resistance to adavosertib through the upregulation of Myt1. HeLa and MDA-MB-231 cell lines, which initially exhibited high sensitivity to Wee1 inhibition, after selection acquired resistance to adavosertib that was marked by a 2- to 3-fold increase in Myt1 expression (Fig. 1B and C). In addition, in vivo experiments using an orthotopic breast cancer xenograft model demonstrated that tumors following 26 days of adavosertib treatment had increased Myt1 expression compared with control-treated tumor tissue (Fig. 1F). Together, these data strongly suggest that Myt1 upregulation is a mechanism for tumor cells to acquire resistance to Wee1 inhibition. However, our data do not exclude the possibility that other proteins or pathways may also promote resistance to adavosertib. Recently, proteomics study showed that primary resistance to adavosertib is also associated with the upregulation of the mTOR pathway in small-cell lung carcinoma cells (55).

To validate that Myt1 upregulation had a direct role in adavosertib resistance, we tested the effects of Myt1 knockdown and overexpression. Myt1 knockdown resensitized resistant HeLa and MDA-MB-231 cell lines to adavosertib (Fig. 1D and E), which suggested that Myt1 upregulation was required to sustain resistance in these cells. Furthermore, Myt1 knockdown also enhanced adavosertib sensitivity in breast cancer cell lines initially exhibiting intrinsic resistance to the Wee1 inhibitor (Fig. 2B–E; Supplementary Fig. S1). Moreover, we were able to induce adavosertib resistance directly in parental HeLa and MDA-MB-231 cells by transiently overexpressing GFP-Myt1 (Fig. 2A). Collectively, these data argue that Myt1 upregulation directly drives adavosertib resistance.

Our results suggest that Myt1 expression level is a candidate predictive biomarker for tumor sensitivity to adavosertib. We compared adavosertib sensitivity in cancer cell lines and identified a strong correlation (R² = 0.69) between adavosertib resistance and Myt1 protein expression (Fig. 3; Supplementary Table S2), in agreement with a study suggesting a negative correlation of Myt1 mRNA levels with Wee1 inhibitor sensitivity (31). Although Wee1 and Cdc25C are important regulators of Cdk1 activity (22, 23, 56), we did not observe any significant correlation between the expression of these proteins and adavosertib sensitivity.

Developmental studies in model organisms have shown that Wee1 and Myt1 are at least partially redundant in the regulation of Cdk1 (18–20). This redundancy is important because ectopic Cdk1 activity is lethal due to the promotion of replication stress, premature entry into mitosis, and the dysregulation of mitotic processes (10, 12, 21). We confirmed that Myt1 retained its activity even in the presence of adavosertib, indicating that Myt1 could compensate in Cdk1 regulation when Wee1 is inhibited (Supplementary Fig. S3; refs. 9, 47). However, we found that in cells with low levels of Myt1, such as HeLa and MDA-MB-231, Myt1 activity is insufficient to suppress ectopic Cdk1 activity when Wee1 is inhibited (Fig. 4E–G). Transient overexpression of GFP-Myt1 in HeLa and MDA-MB-231 reduced in vitro Cdk1 activity relative to GFP controls (Fig. 4B and C), which argues that higher Myt1 levels can induce resistance in these cells by inhibiting Cdk1. Likewise, SK-BR-3 and BT-474 have high Myt1 baseline levels and exhibit low Cdk1 activity in the presence of adavosertib (Fig. 4D–G). Myt1 knockdown enhanced in vitro Cdk1 activity induced by adavosertib in all cell lines tested (Fig. 4E), which further argues that Myt1 at least partially inhibits Cdk1 activity if Wee1 is inhibited. Together, these data corroborate that Myt1 promotes resistance to adavosertib through the inhibition of Cdk1.

Our data suggest Myt1 inhibition of ectopic Cdk1 activity protects cells from mitotic catastrophe, the mode of cell death induced by Wee1 inhibition (10, 11, 13). Mitotic catastrophe is not defined by a molecular pathway, but it can be identified by mitotic abnormalities including premature mitosis, centromere fragmentation (14, 15), and mitotic exit delays (57–59). We observed that the type and incidence of these mitotic abnormalities prior to cell death varied depending on the concentration of adavosertib used and cellular Myt1 levels. Despite HeLa and MDA-MB-231 cells having similar Myt1 protein levels and adavosertib IC₅₀s, we did note that siMyt1 transfection in the presence of adavosertib had a more significant effect on the duration of mitosis in HeLa compared with MDA-MB-231 (Fig. 5). This difference may be explained by other biological differences that may exist between the two cell lines. Apart from Myt1, differences in the expression (or activity) of other cell-cycle regulators may cause longer mitotic delays following adavosertib treatment (Fig. 7G). The cellular levels of mitotic cyclins and their degradation rate in anaphase are known to affect the duration of mitosis in cells treated with other antimitotic drugs (58, 60). The upregulation of mitotic cyclins (cyclin A/B) and the downregulation of Cdk-interacting protein p21 have been suggested to correlate with increased cancer cell sensitivity adavosertib (10). Alternatively, differences in the activity of the upstream kinase/phosphatase signaling pathway that regulates Wee1, Myt1, or Cdc25C could also affect the duration of mitosis in these cells. It is also possible that the loss of Myt1 and Wee1 activity may induce mitotic catastrophe by an unknown mechanism that is independent of Cdk1 activity.

In the case of SK-BR-3 and BT-474, the amount of Myt1 expressed was enough to prevent premature mitosis in most cells treated with adavosertib. At the highest concentration tested (2,000 nmol/L), less than 10% of SK-BR-3 or BT-474 prematurely entered mitosis. In contrast, 10% to 20% of HeLa and MDA-MB-231 underwent premature mitosis at 250 nmol/L, and 40% to 50% of cells entered mitosis at 2,000 nmol/L adavosertib. Although Myt1 did not protect most HeLa and MDA-MB-231 cells from mitotic catastrophe at high concentrations of adavosertib (500–2,000 nmol/L), Myt1 was essential for cell survival at lower, more clinically relevant concentrations. Cells treated with 125 to 250 nmol/L adavosertib progressed through mitosis significantly slower than those treated with DMSO, but cell death was relatively low (10%–20% cell death at 250 nmol/L; Fig. 5). Myt1 knockdown in combination with 125 to 250 nmol/L adavosertib caused cells to arrest in mitosis 2 to 3 times longer than in the case of adavosertib alone and caused cell death to increase to 75%. Similarly, 10% of HeLa and 17% of MDA-MB-231 cells underwent premature mitosis associated with centromere fragmentation when treated with 250 nmol/L adavosertib (Fig. 6). However, Myt1 knockdown in combination with adavosertib induced mitosis associated with centromere fragmentation in 75% of HeLa and 50% of MDA-MB-231 cells. Together, these data show that Myt1 and Wee1 cooperatively suppress cell death by mitotic catastrophe.

Our findings establish Myt1 level as a candidate predictive biomarker for tumor response after adavosertib treatment. This could have wide-ranging clinical implications as adavosertib enters the clinic. Currently, adavosertib is undergoing phase I/II clinical trials against different cancer types alone and in combination with different anticancer agents. Although tumor response is observed in some patients, cancer progression continues in many other patients treated with adavosertib (1, 61). To our knowledge, clinical studies on adavosertib do not take Myt1 expression levels into account prior to treating patients with this inhibitor. However, Myt1 expression should be assessed in pre- and posttreatment tissue samples from participants in these studies to evaluate if there is a relationship between Myt1 expression and the clinical response to adavosertib. If a correlation between high Myt1 levels and adavosertib resistance is identified, Myt1 expression levels could be used to stratify patients with cancer in a future clinical trial on adavosertib. Our data show that Myt1 is overexpressed in tumor tissue (relative to normal tissue; Fig. 7A). Likewise, Myt1 is also reported to be upregulated in other cancer types such as colorectal cancer and glioblastomas (30, 62). If high Myt1 levels are indeed a predictive biomarker for tumor response to adavosertib, it is unlikely that adavosertib on its own will be effective in targeting these tumor types. Furthermore, the finding that Myt1 overexpression is associated with poor breast cancer prognosis suggests that those patients with breast cancer most in need of new therapies are least likely to benefit from adavosertib.

Combining adavosertib with a small-molecule Myt1 inhibitor will likely prove beneficial in overcoming resistance. However, there are additional reasons why a Myt1 inhibitor may be beneficial in the clinic. Myt1 level is a candidate prognostic biomarker, because high level in breast cancer is associated with a higher tumor grade, higher mitotic grade, triple-negative status, reduced overall survival, and increased disease relapse (Fig. 7B–F). In addition, Myt1 upregulation is associated with cancer cell metastasis and lower overall survival in colorectal cancers (62). Myt1 is also a crucial survival factor in a subset of glioblastomas (30). Together, these data suggest that Myt1 may be a driver of tumor aggressiveness, which further provides a rationale for developing Myt1 inhibitors.

No potential conflicts of interest were disclosed.

Conception and design: C.W. Lewis, W.-S. Choi, J.D. Smith, A.M. Gamper, G.K. Chan

Development of methodology: C.W. Lewis, E. Homola

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.W. Lewis, A.B. Bukhari, E.J. Xiao, W.-S. Choi, J.D. Smith, A.M. Gamper, G.K. Chan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.W. Lewis, A.B. Bukhari, E.J. Xiao, W.-S. Choi, J.D. Smith, A.M. Gamper, G.K. Chan

Writing, review, and/or revision of the manuscript: C.W. Lewis, A.B. Bukhari, J.D. Smith, J.R. Mackey, S.D. Campbell, A.M. Gamper, G.K. Chan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G.K. Chan

Study supervision: A.M. Gamper, G.K. Chan

Other (antibodies for detecting Cdk1 phosphorylation on T14): S.D. Campbell

We thank Michael Weinfeld (and laboratory members), Michael Hendzel, and Chan laboratory members for their helpful discussion. We also thank Xuejun Sun and Geraldine Barron for assistance in the cell imaging facility. C.W. Lewis is supported by the NSERC Alexander Graham Bell Canada Graduate Scholarship and Izaak Walton Killam Memorial Scholarship. A.B. Bukhari is supported by Alberta Cancer Foundation's Dr. Cyril M. Kay Graduate Scholarship. J.D. Smith was supported by the Queen Elizabeth II Graduate Scholarship. E.J. Xiao was supported by a NSERC USRA award. The Chan laboratory was funded by NSERC (RGPIN-2016-06466), CIHR (PJT-159585), and the Cancer Research Society (19060). The Gamper laboratory was funded by a startup grant from the University of Alberta and a CIHR grant (PJT-159585).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.