Genome-Wide CRISPR-Cas9 Screen Reveals Selective Vulnerability of ATRX-Mutant Cancers to WEE1 Inhibition

Over 20 years ago, germline variations in ATRX, a member of the SWI/SNF chromatin remodeling superfamily, were found to underlie a rare developmental disorder, X-linked mental retardation with alpha-thalassemia (ATR-X syndrome; ref. 1). More recently, genomic studies (2–5) have revealed that ATRX is frequently mutated in a variety of human cancers, including pancreatic neuroendocrine tumors (PanNET), glioma, liver cancer, and neuroblastoma. Although ATR-X syndrome-related germline variations are mainly missense in the histone methylation recognition domain and the ATPase domain (6), the cancer-associated somatic ATRX mutations are predominantly truncating mutations, resulting in loss of ATRX expression (7), indicating that ATRX functions as a tumor suppressor gene in these cancers.

Although genomic sequencing has enabled the rapid discovery of the mutations driving the development of human cancer, significant challenges remain regarding the effective translation of these data into viable therapies. Most driver mutations are damaging the functions of suppressor genes and cannot serve as a direct therapeutic target. One strategy for developing targeted therapies against such loss-of-function mutations is “synthetic lethality.” Synthetic lethality refers to the relationship between 2 genes for which loss of both genes results in cell death but loss of either one alone does not. For example, poly(ADP)-ribose polymerase 1 (PARP1) and BRCA1/2 are synthetic lethal because loss of both genes abolishes 2 parallel DNA damage repair pathways and causes cell apoptosis (8).

Motivated by the clinical success of PARP inhibitors in treatment of BRCA1/2-deficient tumors, scientists have performed genome-wide screens in human cancer cells to identify synthetic lethal partners relevant to other frequently mutated tumor suppressor genes. In such screens, genes essential for each cell line are first identified by high-throughput perturbation of genes using RNA interference (RNAi) or CRISPR-Cas9 genome editing. Synthetic lethal interactions are then inferred by comparing the gene dependency profiles across a panel of either independently derived cell lines with or without mutations at a specific locus, or paired isogenic lines with engineered mutations. Such screens have been used to identify potential pharmacologic targets that enable selective killing of cancer cells with deficiencies in tumor suppressors TP53 or ARID1A (9, 10), or even with activation of “undruggable” oncogenes including RAS or MYC (11, 12).

Among ATRX-mutant tumor types, hepatocellular carcinoma (HCC) is the 5th most common cancer in the world (13) and glioma is the most common primary malignant brain tumor in adults. There are currently no effective targeted therapies available for HCC or glioma, making them among the most lethal tumors in the world. Here, we performed a genome-wide pooled CRISPR-Cas9–based screen in paired isogenic HCC cells with or without engineered mutations in ATRX. The G₂–M checkpoint control kinase, WEE1, was one of the several candidate genes identified as potentially lethal in the context of ATRX deficiency. The lethality of WEE1 loss was further investigated with small molecule inhibitors of WEE1 through the treatment of engineered ATRX-mutant HCC cell lines and xenografts, as well as primary cell lines from naturally occurring gliomas. The results support inhibition of WEE1 as the basis for the development of new therapeutic strategies to treat multiple tumor types harboring ATRX mutations.

All animal procedures were approved by and performed according to the animal ethical committee at the Cancer Hospital, Chinese Academy of Medical Sciences (Beijing, China).

PLC/PRF/5, HuH-7, and HeLa cell lines were obtained from the National Infrastructure of Cell Line Resource (Beijing, China) and are routinely authenticated via their short tandem repeat profile (latest verification for PLC/PRF/5 and HeLa: April 2019; latest verification for HuH-7: November 2018). PLC/PRF/5, HuH-7, HeLa, and HEK293T cells were cultured in DMEM supplemented with 10% FBS and penicillin (100 U/mL)/streptomycin (0.1 mg/mL; Thermo Fisher Scientific). After thawing, cells were used for up to 15 passages. Glioma cell line U251MG was a kind gift from Alan Meeker (JHMI) and matches the ECACC STR profiling reference. Primary glioma cell lines derived from IMA, 12-0160, 13-0302, and 08-0537, were obtained from the Preston Robert Tisch Brain Tumor Center at Duke University, Durham, NC (latest verification were done in April 2019). The 3 primary glioma cell lines were maintained under adherent conditions on laminin-coated plates in human neural stem cell medium, which consisted of NeuroCult NS-A NSC proliferation medium supplemented with EGF (20 ng/mL), FGF (10 ng/mL), and heparin (0.0002%; Stem Cell Technologies). IMA was maintained in 50% neural stem cell medium and 50% DMEM with 10% FBS and 1% penicillin/streptomycin. U251MG was cultured in RPMI1640 with 10% FBS and 1% penicillin/streptomycin. All cell lines were tested negative for Mycoplasma.

AZD1775 (S1525), roscovitine (S1153), hydroxyurea (S1896), and gemcitabine (S1714) were purchased from Selleckchem.

The lentiCRISPR v2 plasmid (Addgene Plasmid #52961) containing ATRX single-guide RNA (sgRNA; sequence: TGGACAACTCCTTTCGACCA) was transfected into PLC/PRF/5 and HuH-7 cells using Neofect (Neo Biotech). Single-cell colonies were selected, and knockout (KO) status was validated by Western blot analysis and Sanger sequencing. For Sanger sequencing, genomic DNA was extracted from PLC/PRF/5 parental and ATRX KO cells. PCR was performed to amplify the region flanked by the target site, and the 736 bp PCR products were subcloned into pCloneEZ vector (CloneSmarter), transformed into competent cells, and at least 20 clones were picked separately for Sanger sequencing. The sequences were aligned with DNAMAN multiple sequence alignment tools. The primers used for amplifying the flanked region of the target site were as follows:

• ATRX_forward: 5′-CCGTGACTCAGATGGAATGGA-3′

• ATRX_reverse: 5′-GGTTACAGAGCCAGAACAGG-3′

Low passage HEK293T cells (4 × 10⁷) were seeded onto 500 cm² culture plates 1 day before transfection. When cells reached 80% to 90% confluence, a mixture containing the following library and packaging plasmids was transfected into HEK293T cells with Neofect: GeCKO v2 library (63 μg; Addgene plasmid 1000000048), psPAX2 (45 μg; Addgene plasmid #12260), and pMD2.G (18 μg; Addgene plasmid #12259). Culture medium was changed 12 hours after transfection. The supernatant containing lentivirus was collected 60 hours posttransfection, filtered through a PVDF filter membrane (0.45 μm; Millipore SteriCup 250 mL; Millipore) to remove cells, aliquoted, and stored at −80°C.

PLC/PRF/5 parental cells or PLC/PRF/5 ATRX KO cells (200 × 10⁶) were transduced with lentivirus from the human GeCKO v2 library at an MOI of ∼0.3 and an average of 500-fold coverage of the library. As the human GeCKO v2 library contains 123,411 sgRNAs, at least 6 × 10⁷ cells need to be infected to achieve 500-fold coverage. At 48 hours postinfection, cells were selected in puromycin (2 μg/mL; Thermo Fisher Scientific) for 2 days. At this time point, 6 × 10⁷ transduced cells were harvested as reference samples (T0), and the remaining cells were passaged every 3 days. A final pool of cell populations was collected at day 21 (T21), after approximately 14 population doublings. Cell pellets harvested from T0 and T21 were stored at −80°C for isolation of genomic DNA (gDNA), which was performed with the Blood & Cell Culture DNA Maxi Kit (Qiagen).

Two rounds of PCR were performed for identification of sgRNA inserts. In the first round PCR, 200 μg of gDNA was used as template to achieve an average of 250-fold coverage of the sgRNA library assuming 10⁶ cells contain 6.6 μg of gDNA. We performed 67 separate PCR reactions (3 μg gDNA/50 μL reaction volume) for each sample using NEBNext High-Fidelity 2 × PCR Master Mix (New England Biolabs) and combined resulting amplicons.

The second-round PCR was performed to attach Illumina adaptors and index samples. A volume of 2 μL of PCR products from the first round PCR were used as templates for each 50 μL reaction volume. The following primers were used:

• PCR1_forward:

• 5′-AATGGACTATCATATGCTTACCGTAACTTGAAAGTATTTCG-3′;

• PCR1_reverse:

• 5′-GTGACTGGAGTTCAGACGTGTGCTCTTCCGATCTACTGACGGGCACCGGAGCCAATTCC-3′.

• PCR2_forward:

• 5′-AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACGCTCTTCCGATCTTCTTGTGGAAAGGACGAAACACCG-3′;

• PCR2_reverse:

• 5′-CAAGCAGAAGACGGCATACGAGATXXXXXXGTGACTGGAGTTCAGACGTG-3′ (XXXXXX represents a 6-bp index).

Amplifications were carried out for 20 cycles for the first-round PCR and 10 cycles for the second-round PCR. The resulting amplicons were pooled and gel purified using the QIAquick Gel Extraction Kit (Qiagen). Combined elutions were purified on AMPure XP beads (Beckman Coulter), quantified on a Bioanalyzer 2100 (Agilent), and sequenced on a HiSeq X10 platform (Illumina).

At a basic level of interpretation, the abundance of the sgRNAs targeting individual candidate genes will differ significantly between ATRX wild-type (WT) and KO cells because of lethality. Therefore, the fundamental goal is to generate a count of reads from the sequencing data for each sgRNA in the human GeCKO v2 library. To generate these counts, the sgRNA sequence containing reads were first selected from the paired end sequence files using the FASTX-Toolkit (http://hannonlab.cshl.edu/fastx_toolkit/index.html) with the fastx_barcode_splitter.pl command. The sgRNA sequences were clipped out with the fastx_trimmer command, leaving only the sgRNA sequence for further analysis. The sgRNA reads were mapped to the GeCKOv2 sgRNA library with Bowtie (version 1.1.2; ref. 14), which aligns short sequencing reads, with tolerance of a single nucleotide mismatch. For each sgRNA in the library, the count of mapped reads was calculated for each time point, T0 or T21. The raw sgRNA count file was uploaded to the Model-based Analysis of Genome-wide CRISPR/Cas9 Knockout (MAGeCK) algorithm to generate a maximum-likelihood estimation (mle) of gene essentiality score (β score) for each gene using the MAGeCK mle module (15). The ATRX synthetic lethal genes were filtered as those with β score < 0 and a FDR < 0.3 in the ATRX KO group, but with a P value of > 0.05 in the ATRX WT group.

Using this approach, we identified 58 potential synthetic lethal genes in the context of ATRX mutation from the 2 replications. We used PANTHER, an online tool to calculate overrepresented gene clusters. A single gene, RGPD6, was not recognized by the PANTHER database. Therefore, we used the remaining genes (n = 57) to calculate the enrichment using Fisher exact test, and the P values were adjusted using the Bonferroni correction.

PLC/PRF/5 parental and PLC/PRF/5 ATRX KO cells were transfected with WEE1 or VCP or PKMYT1 siRNAs (50 nmol/L final concentration; RiboBio) using Lipofectamine RNAiMAX (Thermo Fisher Scientific) according to the manufacturer's instructions. Knockdown efficiency was tested by qRT-PCR 24 hours after transfection. siRNA sequences used for WEE1, VCP, and PKMYT1 were the following:

• siWEE1-1: 5′-UUCUCAUGUAGUUCGAUAUUU-3′;

• siWEE1-2: 5′-UAAUAGAACAUCUCGACUUAU;

• siVCP-1: 5′-CCUGAUGUGAAGUACGGCAAA;

• siVCP-2: 5′-GAUGGAUGAAUUGCAGUUGUU;

• siPKMYT1-1: 5′-GGACAGCAGCGGAUGUGUU-3′;

• siPKMYT1-2: 5′-GGAACCUCCUCAGCCUGUU-3′;

• siNC (negative control): 5′-UUCUCCGAACGUGUCACGU-3′

For cell viability assay, PLC/PRF/5 and HeLa cells were transfected with either negative control siRNA (siNC) or ATRX siRNAs (siATRX-1 and siATRX-2; 50 nmol/L final concentration; RiboBio). Forty-eight hours after transfection, knockdown efficiency was tested by Western blot analysis. siRNA sequences used for ATRX were the following:

• siATRX-1: 5′-GCAGAUUGAUAUGAGAGGAAU-3′;

• siATRX-2: 5′-CGACAGAAACUAACCCUGUAA-3′.

Total RNA was isolated using TRizol reagent (Ambion), and 2 μg of total RNA was used to prepare cDNA using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific) according to the manufacturer's instructions. qRT-PCR was performed in triplicate for each target sequence using LightCycler 480 SYBR Green I Master reaction mix (Roche) on a LightCycler 480 Real-Time system (Roche). The following primers were used.

• WEE1_forward: 5′-CAATTACTGAAAGCAATATGAAG-3′

• WEE1_reverse: 5′-ACATGAGAATGCTGTCCAAG-3′

• VCP_forward: 5′-CTCATCTACATCCCACTTCCT-3′

• VCP_reverse: 5′-CGTTCTCGCCTAATCTCAC-3′

• PKMYT1_ forward: 5′-CTACTTCCGCCACGCAGAA-3′

• PKMYT1_reverse: 5′-CGCACCTTGAAGACCTCTCC-3′

Cells were lysed in RIPA buffer (Beyotime Biotechnology) containing protease inhibitor and PhosStop cocktails (Roche). Cell lysates were separated on an SDS-polyacrylamide gel and transferred to a PVDF membrane (Millipore). The following antibodies were used for immunoblotting: anti-ATRX (HPA001906, Sigma-Aldrich); anti-γH2A.X (05-636; Millipore); anti-phospho-RPA32 (S4/S8; A300-245A-M, Bethyl); anti-GAPDH (TA-100; Zhongshanjinqiao Biotechnology); anti-cleaved caspase 3 (9661); anti-β-actin (4970);anti-phospho-ATM (Ser1981; 5883); anti-ATM (2873); anti-phospho-CHK2 (Thr68; 2197); anti-CHK2 (6334); anti-phoso-CDK1 (Tyr15; 4539); anti-phoso-CDK1 (Thr14; 2543); anti-CDK1 (9116); anti-Wee1 (13084); anti-PKMYT1 (4282; Cell Signaling Technology).

For inhibitor studies, PLC/PRF/5 parental cells (2500/well) and ATRX KO single colonies were seeded in 96-well plates for overnight incubation and treated with different doses of AZD1775 for 3 days. Cell viability was assessed by measuring the absorbance at 450 nm with the Cell Counting Kit-8 (CCK-8; DOJINDO). Glioma cells (2,500 cells/well or 1,000 cells/well for U251) were incubated overnight in 96-well plates and subsequently exposed to increasing doses of AZD1775 in DMSO or DMSO as vehicle control. Viable cells were measured after 5 days using the CellTiter-Glo assay (Promega) according to the manufacturer's instructions. Each AZD1775 dose was tested in triplicate, and the average luminescence is presented as a percent of the DMSO control. The IC₅₀ value was derived from the curve fitting of the dose–response data using GraphPad Prism v6.0. The normalized growth rate inhibition (GR) metric was calculated according to the protocol in Hafner and colleagues (16).

The clonogenic survival assay was performed according to the protocol as described previously (17). PLC/PRF/5 cells (400 cells/well) or HuH-7 cells (5,000 cells/well for the ATRX WT group; 10,000 cells/well for the ATRX KO group for its very low plating efficiency) were plated in triplicate on a 6-well plate, incubated for 24 hours, and treated with AZD1775 at the indicated concentrations for 12 days. Surviving colonies were fixed with formalin and stained with 0.1% crystal violet. Colonies containing more than 50 cells were counted. For the sites with colony overlapping, the number of overlapping colonies was estimated based on the shape and size of the colony cluster. The relative colony-forming efficiency was determined by normalizing the colony numbers to the untreated control.

For cell-cycle analysis, cells were harvested by trypsinization, washed with PBS, fixed in cold 70% ethanol, incubated with 50 μg/mL RNaseA at 37°C for 30 minutes, and then stained with 50 μg/mL propidium iodide. For HuH-7 cells, DAPI (1 μg/mL) was used to stain the DNA without RNase A incubation. Cells were analyzed with a BD Flow Cytometer. ModFit software was used for processing the data. Apoptotic assays were performed using the FITC Annexin V Apoptosis Detection Kit (556547; BD Biosciences) according to the manufacturer's instructions. The apoptotic cells were analyzed with a BD Flow Cytometer. Data were analyzed with the FlowJo software.

The neutral comet assay was performed using the CometAssay HT Kit (4252-040-K; Trevigen) according to the manufacturer's instructions. Briefly, cells (1 × 10⁵/mL) were mixed with molten LMAgarose (at 37°C) at a ratio of 1:10 (v/v), and a volume of 30 μL of this mixture was immediately transferred onto the sample area of a CometSlide. After the agarose/cells was evenly dispersed, slides were placed flat at 4°C in the dark for 30 minutes in a high humidity environments. The cells were then lysed overnight by immersing slides into lysis buffer. After lysis, the slides were rinsed in distilled water and immersed in neutral electrophoresis buffer for 30 minutes before application of an electric field. An electric field (typically 1 V/cm) were applied to the cells for 45 minutes at 4°C, and cells were stained with SYBR gold (S11494; Thermo Fisher Scientific) for 30 minutes in the dark and photographed using a ZEISS microscope with an attached camera. The comets were analyzed using CASP software (18).

Cells (1 × 10⁵) were seeded onto Fisherbrand microscope cover glass (Thermo Fisher Scientific), which were placed in 12-well cell culture plates, incubated overnight, and subsequently exposed to 200 nmol/L AZD1775 or DMSO as vehicle control. After 24 hours, cells were fixed with 4% paraformaldehyde, permeabilized with 0.5% Triton X-100, blocked with normal goat serum, and incubated with Alexa Fluor 488 conjugated anti-phospho Histone H2A.X (Ser139) antibody (05-636-AF488; Millipore) for 1 hour at room temperature. Nuclei were counterstained with DAPI (1 μg/mL; Sigma-Aldrich). Images were obtained using laser confocal microscopy (Olympus FV1000).

Cells were incubated in medium with 10 μmol/L BrdUrd for 30 minutes at 37°C, harvested by trypsinization, and fixed in ice-cold 70% ethanol overnight. To enable access of the antibody to incorporated BrdUrd, cells were treated with 2 mol/L HCl containing 0.1 mg/mL pepsin for 20 minutes at RT. Cells were rinsed once with PBS, incubated with APC-labeled anti-BrdUrd antibody (BD Biosciences) at room temperature for 30 minutes in the dark, rinsed again with PBS, incubated with 50 μg/mL RNase A at 37°C for 30 minutes, and then stained with 50 μg/mL propidium iodide. For HuH-7 cells, DAPI (1 μg/mL) was used to stain the DNA without RNase A incubation. Samples were analyzed using a FACS Scan (Becton Dickinson). The results were analyzed with NovoExpress software from ACEA Biosciences.

PLC/PRF/5 parental cells or PLC/PRF/5 ATRX CRISPR KO cells (2 × 10⁶ cells per mouse) were subcutaneously inoculated into female BALB/C nude mice ages 6 to 8 weeks (Huafukang). Tumor volumes were calculated with caliper measurements using the following formula: (length × width²)/2. When the mean tumor volumes reached 150 to 200 mm³, mice bearing tumors derived from PLC/PRF/5 ATRX WT cells or ATRX KO cells were divided into 2 subgroups and orally administered with either AZD1775 (50 mg/kg body weight) or vehicle control once daily for 15 days. The in vivo doses used in our study are following those used in previous studies (19–21). Using the body surface area (BSA) scaling method (22) recommended by the U.S. FDA, we translate the animal dosage to the human clinical trials. This calculation results in a human equivalent dose of ∼4 mg/kg, which translates into a 240 mg dose of AZD1775 for a person weighing 60 kg. This dose is within a clinically achievable range as this drug has been used at a dose of 225 mg twice per day over 2.5 days per week for 2 weeks per 21-day cycle in phase I clinical trials (23, 24). We also evaluated the effect of AZD1775 on the body weight of mice administered orally with a dose of 50 mg/kg once daily. The body weight of mice treated with AZD1775 was reduced less than 10% compared with the vehicle control group. Nude mice were housed in a pathogen-free environment and were given food and water ad libitum for the duration of experiments.

Data analyses were performed using GraphPad Prism v6.0 and Microsoft Excel. Statistical significance was determined by the 2-tailed unpaired Student t test. P values are reported in the graphs. *, P < 0.05; **, P < 0.01; and ***, P < 0.001. n.s. denotes not significant.

To screen for synthetic lethal partner genes of ATRX, we first generated isogenic ATRX KO cell models using PLC/PRF/5 and HuH-7, 2 HCC cell lines, which are WT for the gene (25). Both cell lines have TP53 mutations that are known to inactivate the TP53 (Supplementary Fig. S1A and S1B), making them good models for studying ATRX function as ATRX and TP53 mutations often co-occur in tumors (7). We used CRISPR technology (26) to specifically target the ATRX gene locus, introducing inactivating mutations and enabling the generation of isogenic clones with mutated ATRX (ATRX KO). We isolated 2 independent cell clones for PLC/PRF/5 and 1 clone for HuH-7, with complete absence of ATRX protein expression because of frameshift mutations introduced into the coding region of the gene (Supplementary Fig. S1C–S1F). The parental PLC/PRF/5 line and clone sc26, referred to as ATRX WT and ATRX KO, respectively, were used for the screen. Before performing the screen, the sc26 clone was passed 15 passages and no ALT phenotype was observed.

We performed a genome-wide CRISPR-Cas9 screen (27) for essential genes in our ATRX null model system, using the GeCKOv2 sgRNA library (28), which contains 123,411 sgRNAs targeting 19,050 genes and 1,864 miRNAs. We infected ATRX WT and ATRX KO cells with the library at a multiplicity of infection (MOI) of ∼0.3 and selected for sgRNA and Cas9-expressing cells in the presence of puromycin for 2 days (Fig. 1A). The cells were collected at 2 time points, including immediately after puromycin selection (T0) and after approximately 14 population doublings (T21; Fig. 1A).

We used next-generation sequencing to measure the abundance of all sgRNAs in these 2 cell populations at both time points and compared them to calculate the essentiality score for each gene using the algorithm MAGeCK (Supplementary Fig. S2A and S2B; ref. 15). We then searched for genes targeted by the sgRNAs that were represented at nearly the same level at T0 as at T21 in ATRX WT cells, but depleted in ATRX KO cells at T21 compared with T0. Under these stringent criteria, we identified 58 genes, which were essential for viability in ATRX KO cells but nonessential for ATRX WT cells, as potential synthetic lethal partners for ATRX (Fig. 1B; Supplementary Table S1).

From the top 15 candidates that having the most deleterious effects to ATRX KO cells, we chose 3 (WEE1, VCP, or PKMYT1) to validate the screen results. We used RNAi to knockdown the expression of the 3 genes in PLC/PRF/5 ATRX KO and WT cells. Cell growth was evaluated in colony formation assays. siRNAs for VCP, WEE1, and PKMYT1 efficiently reduced corresponding mRNAs to <50% of controls in both cell backgrounds (Supplementary Fig. S3A–S3C). Cell growth was inhibited more significantly by these siRNAs in ATRX KO cells than in the parental cell lines (Supplementary Fig. S3D–S3I). Thus, suppression of these genes selectively inhibited colony formation of ATRX KO cells, indicating that the hits in our screen were true synthetic lethal partners of ATRX.

Among the 58 candidates identified in the screen, several groups of genes encode proteins that are physically interacting in a complex. For example, both proteins of the SMC5/6 complex were identified in the screen (Fig. 1C). The SMC5/6 complex functions in homologous recombination during replication and has been implicated in the maintenance of heterochromatin (29, 30). The synthetic lethality between ATRX and the SMC5/6 complex is consistent with ATRX's known function in heterochromatin formation and gene silencing (31). Similarly, 2 components of the HIRA: ASF1A: UBN1: CABIN1 complex were identified with similar lethality scores in the screen (Fig. 1D). The HIRA complex is a histone 3.3 chaperone required for heterochromatinization of senescent human cells (29). ATRX is involved in deposition of histone H3.3 to telomeres in a HIRA-independent way (32). The synthetic lethal relationship between ATRX and HIRA complex members raised questions on the compensation of the 2 H3.3 deposition pathways in certain contexts.

Because our screens have successfully identified genes associated with well-known ATRX functions, we annotated the functions of all the hits for new mechanistic insights into the activity of ATRX. We found that these 58 genes were enriched in pathways involved in a few cellular processes, including mRNA processing, mitosis, chromatin organization, and cell-cycle regulation (Fig. 1E). Strikingly, 19 of the 58 hits are involved in cell cycle and checkpoint pathways (Fig. 1F). These genes include those recruited to the replication fork complex at intra-S phase checkpoint (TIPIN, TOPBP1, MCM4, SMC5, SMC6, and RRM2; ref. 33), redundant G₂–M checkpoint regulators WEE1 and PKMYT1 (34), proteins critical for centromere and spindle assembly in M phase (SMC1A, CENPN, CDCA8, SKA3, TUBGCP6, TUBB, TUBA1B, and DYNLRB1; refs. 35, 36), and proteins involved in heterochromatin formation (ASF1A, CABIN1, HAT1; refs. 29, 37, 38). These results suggest a central role for ATRX in cell-cycle progression and the possibility that different regulatory points in the cell cycle can be targeted to treat ATRX mutant cancers.

Among the 19 cell-cycle regulators identified in the screen, WEE1 is the only one with small molecule inhibitors designed and tested in clinical trials. One of the inhibitors, AZD1775 (MK1775), has passed regulations for patient safety and is currently in several phase II clinical trials in combination with chemotherapies for advanced solid tumors (http://www.clinicaltrials.gov). We therefore investigated the effects of WEE1 small molecule inhibition in ATRX mutated cells for possible translation into the clinic. We treated ATRX WT and ATRX KO clonal cells with different doses of the WEE1 inhibitor AZD1775 for 3 days, and determined IC₅₀ values from cell viability curves (Fig. 2A). Both ATRX KO cell clones exhibited an increased sensitivity to AZD1775 based on the IC₅₀ value of 0.38 and 0.31 μmol/L, which were about 4-fold lower than the IC₅₀ value of 1.4 μmol/L for ATRX WT cells. Because ATRX KO cells had a slightly slower growth rate than ATRX WT cells (Supplementary Fig. S4A), we used a normalized growth rate inhibition (GR) metrics to correct for the confounding effects of growth rate on IC₅₀ (16). The ATRX KO cells had a lower GR₅₀ compared with ATRX WT cells (1.79 μmol/L vs. 0.47 μmol/L and 0.59 μmol/L, ATRX WT vs. ATRX KO sc26 and sc5; Supplementary Fig. S4B), suggesting that they are more sensitive to AZD1775 than the WT cells in 2D cell culture. Similarly, enhanced sensitivity to AZD1775 was observed in ATRX KO cells compared with ATRX WT HuH-7 cells (Fig. 2B), indicating that this effect of ATRX KO was reproducible in a different cell line. Consistently, ATRX depletion by RNAi also increased sensitivity to AZD1775 in PLC/PRF/5 and HeLa cells (Fig. 2C and D; Supplementary Fig. S5A and S5B).

AZD1775 treatment also resulted in significantly reduced clonogenic survival of ATRX KO relative to ATRX WT cells (Fig. 2E and F). After treatment with 50 to 200 nmol/L AZD1775, the number of colonies generated by ATRX KO cells was significantly reduced relative to ATRX WT cells (Fig. 2E and F). Similar results were observed in HuH-7 cells; HuH-7 isogenic ATRX KO clones were also more sensitive to AZD1775 treatment relative to WT cells (Fig. 2G and H). Taken together, this observed increase in sensitivity of ATRX KO cells to AZD1775 was consistent with the synthetic lethal interaction between ATRX and WEE1 identified in our genetic screen.

We next investigated the mechanism by which loss of ATRX sensitized cells to WEE1 inhibition. WEE1 regulates cell-cycle progression by inactivating CDK1 through phosphorylation at tyrosine 15. To confirm that AZD1775 inhibits the kinase activity of WEE1 in PLC/PRF/5, we examined the protein levels and phosphorylation levels of CDK1 by Western blot at different time points. As expected, the Y15 phosphorylation significantly decreased after 1-hour treatment with AZD1775 in both ATRX WT and ATRX KO cells, whereas the CDK1 protein levels did not change (Supplementary Fig. S6A–S6C). As a control, phosphorylation of CDK1 threonine 14, which is mainly controlled by another kinase PKMYT1, did not change after AZD1775 treatment (Supplementary Fig. S6D). Meanwhile, WEE1 or PKMYT1 protein levels did not significantly change during 24-hour treatment of AZD1775 in either ATRX WT or KO cells (Supplementary Fig. S6E and S6F). These results suggested that AZD1775 inhibit WEE1 activity on cell-cycle regulation in both ATRX WT and KO cells. Some ATRX-deficient cells have been shown to exhibit defects in cell cycle with an increase in DNA replication stress (39, 40). Therefore, we investigated whether the enhanced sensitivity of ATRX KO cells to WEE1 inhibition was a result of combined deficiencies in cell-cycle regulation.

To test this hypothesis, we first examined the cell-cycle distribution of ATRX WT and KO cells under AZD1775 treatment as a measure of DNA content with propidium iodide using FACS (Fig. 3A). Consistent with previous reports (39, 40), the ATRX KO cells had a slightly increased proportion of S-phase cells (Fig. 3A). After AZD1775 treatment, a robust increase of cells in S-phase occurred in ATRX KO cells (70.34 ± 1.90% vs. 31.34 ± 2.25%, ATRX KO vs. WT; n = 4; Fig. 3A) was observed. Labeling of cells with BrdUrd supported this finding, as the percentage of nonreplicating cells in S-phase (2N < PI < 4N/BrdUrd^¯) was increased in ATRX KO cells under AZD1775 treatment (2.47 ± 0.42% vs. 11.35 ± 1.46%, ATRX WT vs. ATRX KO; n = 4; Fig. 3B and C). Similarly, proportions of S-phase cells and nonreplicating cells in S-phase were significantly increased in HuH-7 ATRX KO cells treated with AZD1775 (Supplementary Fig. S7A–S7C). The above results indicated that WEE1 inhibition caused S-phase arrest in ATRX KO cells.

To examine the biological conditions triggering S-phase arrest in ATRX KO cells under WEE1 inhibition, we first investigated the levels of DNA damage using phosphorylated H2A.X as a marker of stalled replication forks and double-strand breaks (DSB). The percentage of γH2A.X-positive ATRX KO cells increased immensely upon treatment with AZD1775 compared with ATRX WT cells, as measured by FACS analysis in both PLC/PRF/5 and HuH-7 cells (Fig. 3D and E; Supplementary Fig. S7D and S7E). Indeed, these γH2A.X-positive cells were mainly distributed in S-phase (Fig. 3D; Supplementary Fig. S7D), indicating that the S-phase arrest in ATRX KO cells was partly because of the accumulated DNA damage. Immunofluorescence staining was consistent with these results, showing a sharp increase in the percentage of γH2A.X-positive cells compared with controls (16.98 ± 0.57% vs. 55.42 ± 1.51%, n = 4, AZD1775 treated WT vs. KO cells, P < 0.0001; Fig. 3F and G; Supplementary Fig. S8A and S8B). Addition of the CDK1/2 inhibitor roscovitine almost completely repressed the increase of γH2A.X levels in both ATRX WT and KO cells after WEE1 inhibition (Fig. 3H and I; Supplementary Fig. S9A and S9B), suggesting that the increased DNA damage was because of the elevated CDK1/2 activity induced by WEE1 inhibitors.

To determine whether the increase in γH2A.X after WEE1 inhibition was because of an increase in DNA double-strand breaks, we used the neutral comet assay, which is a sensitive method for monitoring DSBs. We observed that WEE1 inhibition caused a significant increase in comet tail moment in ATRX KO cells (Fig. 4A), confirming the increased DSB formation and replication fork collapse in these cells. Consistent with the presence of γH2A.X in ATRX KO cells, the activation of ATM-dependent DSB signaling events were detected at ATRX KO cells without drug treatment, as revealed by the increased levels of CHK2 phosphorylation (Fig. 4B). After AZD1775 treatment, the DNA-PK–dependent phosphorylation of RPA32 on positions 4 and 8 increased more significantly in ATRX KO cells than in ATRX WT cells (Fig. 4C and D), indicating that the ATRX loss augmented accumulation of DSBs.

Because long-term AZD1775 treatment resulted in significantly reduced clonogenic survival of ATRX KO relative to ATRX WT cells (Fig. 2E and F), we examined the possible mechanisms mediating cell lethality. Using FACS, we found that WEE1 inhibition caused a significant increase in the portion of cells undergoing apoptosis (Annexin V staining: 4.06% vs. 15.05% and 19.12%, ATRX WT vs. ATRX KO sc26 and sc5; Fig. 4E and F). Similar results were also observed in HuH-7 cells (Supplementary Fig. S10A and S10B). To determine whether caspase-3 was activated in the process of apoptosis, ATRX KO and WT cells were treated with 200 nmol/L AZD1775 over a 72-hour time course. Western blot analysis revealed a greater increase in the level of activated (cleaved) caspase-3 in ATRX KO cells than in ATRX WT cells, suggesting that cells were committed to apoptosis in a caspase-3–dependent manner (Fig. 4G). Similar changes in the level of γH2A.X were also observed, indicating that the induction of γH2A.X upon AZD1775 preceded apoptosis (Fig. 4G). Taken together, these data demonstrate that DNA damage induced by replication stress contributed to the increased sensitivity of ATRX deficient cells to WEE1 inhibition, which ultimately committed ATRX KO cells to apoptosis (Fig. 5A).

Based on the model, we proposed that other drugs that cause increased replicative stress could also exacerbate ATRX loss. We tested several such drugs and found that ATRX deficiency cells exhibited an enhanced sensitivity to hydroxyurea (IC₅₀: 181.60 ± 3.49 μmol/L vs. 81.08 ± 10.79 μmol/L, ATRX WT vs. ATRX KO, n = 3, P = 0.0009) and gemcitabine (IC₅₀: 0.0418 ± 0.0029 μmol/L vs. 0.0258 ± 0.0015 μmol/L, ATRX WT vs. ATRX KO, n = 3, P = 0.0081) than the WT cells (Fig. 5B and C). These results opened the possibilities for exploiting other replicative stress–inducing chemicals for future drug development for ATRX mutant cancer.

To test the potential therapeutic efficacy of WEE1 inhibition in vivo, we generated a xenograft model in nude mice using PLC/PRF/5 ATRX WT and KO cell lines. Animals were inoculated subcutaneously with cells, and AZD1775 or vehicle control was administered orally for 15 days once tumors reached a volume of nearly 150 to 200 mm³. PLC/PRF/5 ATRX WT xenografts continued to grow, even under AZD1775 treatment, reaching a several-fold increase in volume by the end of the experiment (Fig. 6A). In contrast, tumor growth was significantly inhibited in PLC/PRF/5 ATRX KO xenografts in response to AZD1775 compared with the vehicle control treated group (∼230 mm³ vs. 800 mm³, AD1775 vs. vehicle control; Fig. 6B). Furthermore, by day 5, volumes of ATRX KO xenografts had reached a plateau under treatment (∼230 mm³; Fig. 6B), whereas ATRX WT and vehicle control ATRX KO xenografts continued to increase in both size (Fig. 6A–C) and weight (Fig. 6D).

Growth inhibition in vivo as in vitro was also possibly mediated by increases in DNA damage. Sections from ATRX KO compared with ATRX WT xenografts exhibited increased levels of γH2A.X staining, which were further enhanced under treatment with AZD1775 (15.66% vs. 2.67%, AD1775 treated ATRX KO vs. ATRX WT, P < 0.01; Fig. 6E and F).

As ATRX mutations are frequently found in human glioma, we investigated whether WEE1 dependency also exists in this tumor type in the context of naturally occurring mutations. Patient-derived primary glioma cell lines with WT or mutant ATRX were treated with increasing doses of AZD1775, and IC₅₀ values were determined from the cell viability curves (Fig. 7). Two independent cell lines, 08-0537 and IMA, harboring ATRX nonsense mutations, exhibited ∼2- to 15-fold increased sensitivity to AZD1775 relative to ATRX WT cell lines (Fig. 7). The differences in sensitivity to AZD1775 still held after adjusting for the differences in growth rate (Supplementary Fig. S11A and S11B). These results indicate that the synthetic lethality between ATRX and WEE1 may be present in different cancer types harboring ATRX mutations. Because AZD1775 exhibits good penetration in brain tumors (41, 42), it could be investigated as a potential therapeutic agents in the treatment of ATRX-deficient malignant glioma.

In this study, we used an unbiased genome-wide CRISPR-Cas9 KO screen to discover synthetic lethal interactions specific to ATRX deficiency, which affects a variety of tumor types (2–5). We identified several candidate synthetic lethal partners specific to ATRX deficiency with this genetic screen. The most promising candidate was WEE1, a kinase involved in cell-cycle regulation and DNA damage response. Furthermore, we demonstrated the efficacy of targeting WEE1 with the small molecule inhibitor AZD1775 in in vitro and in vivo models deficient in ATRX.

ATRX is mutated in the majority of progressive gliomas and secondary GBMs (7). In a recent study, AZD1775 exhibited good penetration across the human blood–brain barrier (BBB) and extensively accumulated in human GBM tumors in an acidic tumor microenvironment to reach potentially therapeutic concentrations (41, 42). This requirement of an acidic basolateral pH is based on the inactivation of ABCB1/ABCG2-mediated efflux clearance and activation of the uptake transporter OATP1A2 (41) of AZD1775. Interestingly, almost all ATRX-mutant gliomas/GBMs harbored the IDH1/2 hotspot mutations (7), which are associated with the production of D-2-hydroxyglutarate (D2HG), the altered catalytic flux and the promotion of the acidic tumor microenvironment (43). In this case, AZD1775 has the potential to be an effective therapeutic choice for a significant number of patients with gliomas/GBM. These results have implications for the treatment of diverse tumor types with ATRX mutations and for the feasibility of functional screens for the identification of new pharmacologic targets in cancer.

Based on the results of our study and others, ATRX loss confers sensitivity to WEE1 inhibition by further destabilizing the genome, which ultimately leads to apoptosis. ATRX is required for the restart of stalled replication forks and recovery from replicative stress in S-phase (40). Our data and previous studies have shown that loss of ATRX alone induces replication stress and delayed S-phase progression (39, 40, 44). WEE1 inhibition also has a direct effect on DNA replication in S-phase through activation of CDK1/2 (45). Previous work has shown that depletion of WEE1 activity results in unscheduled DNA replication, nucleotide shortage, stalling of replication forks and accumulation of DNA damage (46–49). Together our observations supported the model that the synergistic effects of ATRX loss and WEE1 inhibition on DNA replication leads to accumulation of DNA damage and S-phase arrest of cells, and eventually apoptosis (Fig. 5A).

In summary, CRISPR technology coupled with next-generation sequencing made it possible to identify WEE1 and other candidate genes as critical to the survival of ATRX deficient HCC and glioma cells. This experimental strategy that combines genomic with functional data directly addresses the challenges in designing therapeutic agents with high specificity but low toxicity. Although our study was limited to 2 tumor types, our findings are of clinical interest to other cancers with frequent ATRX mutation. Finally, other synthetic lethal candidates identified in our screen may provide additional effective targets or critical insight into the tumor suppressive roles of ATRX in cancer.

No potential conflicts of interest were disclosed.

Conception and design: J. Liang, H. Zhao, S. Liu, H. Yan, Y.-X. Zeng, X. Wang, Y. Jiao

Development of methodology: J. Liang, X. Wang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Liang, H. Zhao, B.H. Diplas, J. Liu, D. Wang, Y. Lu, Q. Zhu, W. Wang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Zhao, S. Liu, X. Wang

Writing, review, and/or revision of the manuscript: J. Liang, H. Zhao, B.H. Diplas, S. Liu, X. Wang, Y. Jiao

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Liang, H. Zhao

Study supervision: X. Wang, Y. Jiao

Other (perform the experiments): J. Wu

We thank Drs. B. Vogelstein, K.W. Kinzler, N. Papadopoulos, and S. Zhou for helpful discussions. We thank Ms. Guangyu Li for data processing of the GR values for cell viability assays. We appreciate Ms. Ran Gao's help on all the clonogenic survival assays and flow cytometry experiments. This work was supported by the National Natural Science Foundation Fund (81472559 to Y. Jiao and 81502420 to J. Liang), the National Key Basic Research Program of China (973 program no. 2015CB553902 to Y. Jiao), the CAMS Initiative for Innovative Medicine (2017-I2M-4-003 to X. Wang; 2017-I2M-4-002 to H. Zhao), the State Key Project on Infection Diseases of China (2017ZX10201021-007-003 to H. Zhao), and the National Laboratory Special Fund (2060204).

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.