Inhibition of Aberrantly Overexpressed Polo-like Kinase 4 Is a Potential Effective Treatment for DNA Damage Repair–Deficient Uterine Leiomyosarcoma

Uterine sarcoma is rare, and uterine leiomyosarcoma (LMS) is the most common subtype. Recurrence after surgical treatment is frequent, even in patients with stage I disease, and conventional adjuvant radiotherapy and chemotherapy are generally ineffective (1, 2). Recent studies have shown that uterine LMS frequently exhibits genetic alterations in TP53, RB1, ATRX, and PTEN (3–6). Although knowledge of these may assist pathologic diagnosis, their contribution to therapy is currently limited to clinical trials or preclinical experiments (7). Others have reported that a subset of uterine LMS harbors genetic alterations involving the key homologous recombination (HR) repair pathway genes BRCA2, ATR, ATM, CHEK2, and BRCA1 (3–5, 8–11), and patients with such tumors have a worse clinical outcome. Therefore, the discovery of new treatment strategies would be valuable (12).

Uterine sarcomas, particularly uterine LMS, are genomically unstable tumors. Genomic instability may be measured by a HR deficiency (HRD) score. In a study by The Cancer Genome Atlas on DNA damage repair pathway alterations among different malignancies, it was revealed that sarcomas had relatively high HRD scores (13). Thus, genomically unstable tumors may be more sensitive to DNA damage inducers.

Polo-like kinase 4 (PLK4) regulates centriole duplication and mitotic progression, and its transcription is tightly coupled to the cell cycle (14). Dysregulation of PLK4 function causes genomic instability (10, 15–21). PLK4 also phosphorylates CDC25C (22), CHK1, and CHK2 (23), which are signal transducers of DNA repair pathways. The loss of PLK4 in the fibroblasts of patients with Seckel syndrome causes a reduction in phosphorylated CHK1 and CHK2. Thus, PLK4 is closely associated with DNA damage repair (24).

CFI-400945 is a highly selective oral PLK4 inhibitor developed by two of the researchers (TWM and MRB) and preclinical studies have demonstrated its effectiveness (20, 25, 26). Phase I clinical trials have shown that it was well tolerated, causing only dose-dependent neutropenia (25). The drug has recently entered several phase II clinical trials. At low doses, CFI-400945 inhibits the trans-autophosphorylation of homodimeric PLK4 partners at a conserved phosphor-degron (27). A lack of degron phosphorylation protects PLK4 from proteasomal-mediated degradation, which is a self-regulated negative feedback mechanism that induces centrosome overduplication, chromosome missegregation, and mitotic catastrophe (20, 26). At high doses, CFI-400945 caused cytokinesis failure, production of aneuploid and polyploid cells, and continuous accumulation of DNA damage, leaving affected cells vulnerable to mitotic catastrophe (21, 28, 29). The efficacy of CFI-400945 improves in the presence of DNA damage inducers or other conditions that result in DNA repair deficiency (26, 29, 30).

DNA double-strand breaks cause ataxia telangiectasia mutated (ATM) autophosphorylation at S1981 and activate HR repair (31). Cells defective in HR repair are hypersensitive to ionizing radiation and drugs that induce double-strand breaks because the effects of error-prone nonhomologous end-joining (NHEJ) DNA repair are conducive to this scenario (32). Conversely, tumor cells may also be sensitized to double-strand breaks inducers if an ATM inhibitor, such as AZD0156, is used in combination (33, 34). A phase I clinical study demonstrated that AZD0156 is generally well tolerable with dose-limiting toxicity (35).

We hypothesize that PLK4 is a pharmacologic target in uterine sarcoma, particularly, uterine LMS. CFI-400945 induces chromosome missegregation and increases DNA damage, and tumors with high genome instability and DNA repair defects should respond to this therapeutic approach. This study evaluated PLK4 expression, demonstrated the frequency of genetic DNA repair defects, and determined the HRD scores of our in-house clinical uterine LMS cases. The in vitro and in vivo effects of CFI-400945 were tested on all publicly available myogenic gynecologic sarcoma cell lines, and any additional benefits were measured under different DNA repair defect conditions, including pharmacologic inhibition of ATM and BRCA2 knockout (KO). The underlying DNA repair mechanisms were also analyzed.

Clinical samples were collected from female patients treated at Queen Mary Hospital, The University of Hong Kong. All were uterine LMS diagnosed by World Health Organization Classification criteria (1, 2, 36). Non-gynecologic soft tissue sarcomas were excluded. The use of human tissue for research was approved by the ethics review board of The University of Hong Kong according to the Declaration of Helsinki guidelines (UW19-066), with written informed consent obtained from each patient prior to surgical operations.

Frozen samples were uterine LMS (n = 25) and matched myometrium (n = 14), and uterine leiomyoma (n = 10) and matched myometrium (n = 10). DNA and total RNA were extracted using optimal cutting temperature resins (Sakura Finetek, Torrance, 4583) and an AllPrep DNA/RNA/miRNA Universal Kit (Qiagen, Hilden).

For whole-exome sequencing (WES), exome capturing was performed using xGen Exome Research Panel v1.0 based on standard protocols. Briefly, 550 ng genomic DNA was fragmented into 350 to 700 bp insert sizes using the Diagenode Bioruptor Pico system, according to the KAPA Hyper Prep Kit (KR0961-V6.17) protocols for library construction. A volume of 300 ng per library DNA from 12 samples was normalized and combined into a single pool for downstream target enrichment through hybridization with the capture probes of targeted regions. Paired-end sequencing (2 × 151 bp) was performed using Illumina NovaSeq 6000. The mean sequencing coverage for DNA was approximately 157.

For transcriptome sequencing (RNA sequencing), ribosomal RNA depletion was performed by using QIAseq FastSelect RNA Removal Kit. The cDNA libraries were prepared by NEBNext Ultra II Directional RNA Library Prep Kit (Cat# 7760S) using 0.5 μg RNA, according to the manufacturer’s protocol. cDNA libraries of a further five samples were constructed by Roche KAPA mRNA HyperPrep Kit (KR1352-v5.17) by using 1,000 ng RNA and following the manufacturer’s protocol. All adapter-ligated libraries were enriched by 10 cycles of PCRs. Finally, the libraries were sequenced by using an Illumina NovaSeq 6000 (2 × 151 bp).

Sequence reads were aligned to human genome build 37 (hg19) using BWA (version 0.7.10) and Samtools (version 0.1.19, RRID: SCR_002105). Duplicated reads were removed from the Binary Alignment Map (BAM) files using Picard (version 1.73, RRID: SCR_006525). These BAM files were recalibrated using Genome Analysis Toolkits (version 3.8-1, RRID: SCR_001876) with default parameters, and single-nucleotide polymorphisms (SNP) and insertions–deletions (indel) were called using GATK Mutect2 (37, 38). GATK VariantFiltration was used to apply filtering criteria to the remaining variants with a quality of depth of 2 and a Fisher strand score of 60 for single-nucleotide variants and 200 for indels. The generated VCF files were filtered by target bed region of the xGen Exome Research Panel (V1) and annotated by ANNOVAR (version April 2018, RRID: SCR_012821) to predict the functional consequences of the discovered variants (39), as well as to obtain allele frequencies from the dbSNP, dbNSFP, and 1000 Genomes projects. Mutations were filtered by removing variants with high frequency in public germline databases (ExAC EAS and gnomAD exome all cohorts) and an in-house database to derive the list of somatic mutations.

The SNP pileup was focused on the target bed region of the xGen Exome Research Panel (V1), and the copy number was called using PureCN (version 1.6.0) with default parameters (40). Copy-number variation status was defined as follows: homozygous deletion when TCN < 0.5; hemizygous deletion when TCN between 0.5 and 1; amplification when TCN equal to or more than 6. Structural variant was considered present if there were one or more breakpoints in a gene, or if there was fusion transcript detected involving a gene based on the RNA fusion analysis.

The allele-specific copy number profile was produced from paired tumor normal exome sequencing data using Sequenza (version 3.0.0, RRID: SCR_016662; ref. 41). The HRD score was calculated using scarHRD (version 0.1.0; ref. 42) using segments produced by Sequenza as the sum of the telomeric allelic imbalance (43), loss of heterozygosity, and large-scale state transition scores (44, 45). The HRD scores of uterine LMS and LM were compared.

Sequence reads were quantified using Salmon (version 1.6.0, RRID: SCR_017036; https://github.com/COMBINE-lab/salmon/releases). Differential expression analysis of nontumor and tumor tissues was performed using DESeq2 (version 1.39.7, RRID: SCR_015687; https://github.com/thelovelab/DESeq2). Normalized gene counts were plotted using ggplot2 (version 3.3.5, RRID: SCR_014601; https://sourceforge.net/projects/ggplot2.mirror/files/v3.3.5/).

Fusion events were identified using STAR-Fusion (version 1.1.0) with the following default parameters: —FusionInspector inspect —annotate —examine_coding_effect —extract_fusion_reads. CTAT_lib (version July 2017) was used as a reference sequence with GENCODE version 19 transcript annotations.

Total RNA was extracted using TRIzol (Thermo Fisher Scientific, Waltham, MA) and reverse-transcribed into cDNA using ProtoScript II Reverse Transcriptase and oligo-d(T)₂₃-VN (New England Biolabs, Ipswich, MA). cDNA was diluted with water, mixed with GoTaq qPCR Master Mix (2×; Promega, Madison, WI), and quantified using a LightCycler 480 II system (Roche Diagnostics).

Publicly available myogenic gynecologic sarcoma cell lines from ATCC—SK-UT-1 (RRID: CVCL 0533) and SK-LMS-1 (RRID: CVCL_0628)—were maintained in minimum essential medium (Thermo Fisher Scientific, 61,100,061) with 10% (v/v) FBS (Thermo Fisher Scientific, 102,701,06). From the Japanese Collection of Research Bioresources Cell Bank, SKN (RRID:CVCL 3167) was maintained in Ham F-12 Nutrient Mixture (Thermo Fisher Scientific, 21,700,075) with 10% (v/v) FBS. Human primary uterine smooth muscle cells (HUtSMC, PCS-100-460-011) were obtained from ATCC and maintained in vascular cell basal medium supplemented with a vascular smooth muscle cell growth kit (ATCC PCS-100-042 and 030). A short-tandem repeat-profiling service based on ANSI/ATCC ASN-0002-2011 (Genetica, Burlington) was used to authenticate cell lines. Cell lines were used at <20 passages and were regularly screened for Mycoplasma contamination. IHC expression of the myogenic markers smooth muscle actin and desmin was demonstrated on SK-UT-1, SKN, and SK-LMS-1 cell pellets. Other publicly available gynecologic cell lines were deemed unsuitable for our specific research purposes. It is worth noting that SK-UT-1B, although similar to SK-UT-1, was derived from malignant Müllerian mixed tumor and the cells exhibited glandular formation in culture, indicating characteristics of adenocarcinoma. Another cell line, MES-SA, was determined to be undifferentiated uterine sarcoma and has predominantly been used for studying endometrial stromal sarcoma.

CFI-400945 was synthesized at the Campbell Family Institute for Breast Cancer Research of the Princess Margaret Cancer Centre, Toronto (46). For in vitro experiments, 10 mmol/L CFI-400945 was dissolved in dimethyl sulfoxide (DMSO; D2650; Sigma-Aldrich Corp., St. Louis, MO) and diluted in cell culture medium at different working concentrations. AZD0156 (HY-100016), an ATM inhibitor, was purchased from MedChemExpress (Monmouth Junction, NJ). For in vitro experiments, 10 mmol/L ADZ0156 was dissolved in DMSO and diluted in cell culture medium. For the in vivo experiments, CFI-400945 was dissolved in water, while AZD0156 was dissolved in a solution containing 10% (v/v) DMSO and 90% (v/v) of 30% (v/v) Captisol solution (Ligand Pharmaceuticals, Inc., San Diego, CA). In the vehicle group, a mixture of 10% (v/v) DMSO and 90% (v/v) of 30% (v/v) Captisol solution was administered.

Cell lines were cultured at an initial density of 500 cells/well for 6 days in 96-well cell culture plates (3599; Corning) containing 200 µL of CFI-400945 or ADZ0156. Cells were fixed with trichloroacetic acid (sc-203414; Santa Cruz Biotech, Dallas, TX) and stained with sulforhodamine B (SRB; sc-253615; Santa Cruz Biotech). Absorbance was measured at 570 nm using a Tecan Infinite F200 Plate Reader (Tecan Group, Ltd., Männedorf). Any drug synergism between CFI-400945 and ADZ0156 was determined according to the Chou–Talalay method and calculated using CompuSyn (version 1.01, RRID: SCR_022931; https://en.freedownloadmanager.org/Windows-PC/CompuSyn-FREE.html; ref. 47).

Plasmid DNA expressing kinase-dead PLK4 (pEGFP-C3-PLK4 K41M-3xFLAG; RRID: Addgene_69838) were kindly provided by Michel Bornens (Institut Curie, Paris, France). Transfection was performed using Lipofectamine 3000 Reagent (Thermo Fisher Scientific).

Caspase 3/7 activity in the treated cell lines was measured using a caspase-Glo 3/7 Assay System (Promega). Luminescence was measured using a Tecan Infinite F200 Plate Reader (Tecan Group, Männedorf). The luminescence values of the drug-treated samples were normalized to those of the DMSO controls, and the relative fold changes were calculated.

Cell lines were incubated for 3 days in 90-mm cell culture plates (SPL Life Science, Gyeonggi, 20,101) containing 10 mL of CFI-400945. Cells were then trypsinized in 2 mL of TrypLE Expression solution (Thermo Fisher Scientific), washed once with 10 mL of PBS, and stained with Annexin V/PI (BioLegend, San Diego, CA) at room temperature (20°C–25°C) for 15 minutes. Acquisition was performed using a BD LSR Fortessa flow cytometer (BD, East Rutherford, NJ). Data were analyzed using FlowJo version 10.5.3 (FlowJo LLC, Ashland, RRID: SCR_008520).

Cells were seeded on glass coverslips, fixed with 4% (v/v) paraformaldehyde (J19943; Thermo Fisher Scientific) for 10 minutes, and treated with methanol at −20°C for 5 minutes. The cells were then blocked with 5% (v/v) BSA (A-420-250; GoldBio, St. Louis, MO) in PBS and stained with mouse anti–α-tubulin (RRID: AB_477583; Sigma-Aldrich Corp.) and rabbit antipericentrin (RRID: AB_304461; Abcam, Cambridge, UK) and detected with anti–mouse Alexa Fluor 488 (RRID: AB_2633275; Thermo Fisher Scientific) and anti–rabbit Alexa Fluor 555 (RRID: AB_2633281; Thermo Fisher Scientific) secondary antibodies. Samples were mounted on glass slides with ProLong Gold Antifade Mountant (P36983; Thermo Fisher Scientific). Immunofluorescence images were acquired using an LSM800 confocal laser scanning microscope at 40× (CLSM; Carl Zeiss AG, Oberkochen) or a Ti2-E fluorescence microscope (Nikon Corp., Tokyo, Japan).

Cytosolic protein was extracted in extraction buffer (10 mmol/L 4-[2-hydroxyethyl]-1-piperazineethanesulfonic acid, pH 7.9, sc-29097; Santa Cruz Biotechnology), 10 mmol/L KCl (P5405; Sigma-Aldrich Corp.), 1.5 mmol/L MgCl₂ (sc-203126; Santa Cruz Biotechnology), 10% (v/v) glycerol (G5516; Sigma-Aldrich Corp.), and 0.2% (v/v) Tween-20 (P9416; Sigma-Aldrich Corp.). The nuclear fraction was extracted using high-salt extraction buffer (50 mmol/L Tris-HCl, pH 8.0; 15,504,020; Thermo Fisher Scientific), 2.5 mol/L NaCl (S3014; Sigma-Aldrich Corp.), and 0.05% (v/v) Tween-20 (P9416; Sigma-Aldrich Corp.) from the insoluble fraction generated by cytosolic protein extraction. Anti-ATM (RRID: AB_725574; Abcam), anti–ATM p-S1981 (RRID: AB_1640207; Abcam), anti-PLK4 (RRID: AB_1129730; Santa Cruz Biotechnology), anti-H2AX (RRID: AB_10971675; Abcam), anti–H2AX-pS139 (RRID: AB_1640564; Abcam), anti-BRCA2 (RRID: AB_2797730; Cell Signaling Technology), and anti-Lamin B (RRID: AB_648156; Santa Cruz Biotechnology) were used for detection.

Cells were treated with 50 nmol/L CFI-400945 plus 400 nmol/L AZD0156 for 3 days, harvested, and embedded in low-melting agarose. The agarose-embedded cells were lysed in comet assay lysis buffer (10 mmol/L Tris-HCl pH 10; 15,504,020; Thermo Fisher Scientific), 100 mmol/L EDTA (15,576,028; Thermo Fisher Scientific), 1% (v/v) Triton X-100 (T8787; Sigma-Aldrich Corp.), and 2.5 mol/L NaCl (S3014; Sigma-Aldrich Corp.) at 4°C for 1 hour, and the DNA was denatured in unwinding solution (300 mmol/L NaOH, Santa Cruz Biotechnology; sc-203387) and 1 mmol/L EDTA (Thermo Fisher Scientific; 15576028) at 4°C for 20 minutes. Equilibration was conducted in Tris/borate/EDTA buffer (15,581,044; Thermo Fisher Scientific), and single-cell electrophoresis was performed at 25 V/3 mA for 5 minutes. The cells were then stained with FxCycle PI/RNase Staining Solution (F10797; Thermo Fisher Scientific), and fluorescence images were captured using a Nikon Ti2-E fluorescence microscope (Nikon Corp.) and analyzed using CometScore (version 2.0; http://rexhoover.com/cometscoredownload.php).

The use of live animals for research was approved by the animal ethics review board of The University of Hong Kong (CULATR 4888-18). Low-passage SK-UT-1 cells (<5) were mixed with Matrigel (356234; Corning) at a 1:1 ratio and subcutaneously injected into the flanks of female Balb/c nude mice aged 6 to 8 weeks. Mice with palpable tumors were evenly distributed into treatment groups according to tumor size. The allocation of these groups to five treatment regimens was randomized, and assignment was blinded. Drugs were administered orally in the following regimens for 3 weeks: vehicle, 5 mg/kg CFI-400945, 7.5 mg/kg CFI-400945, 10 mg/kg AZD0156, and 5 mg/kg CFI-400945 combined with 10 mg/kg AZD0156. Previous studies have shown that CFI-400945 at these dosages was physiologically effective in serum (25, 26). Body weights were determined once every 3 days, and tumor lengths and widths were measured using a digital caliper. Tumor volumes (mm³) were estimated as length × width²/2. At the end of the experiment, tumors and major internal organs were harvested and processed into formalin-fixed, paraffin-embedded (FFPE) sections for microscopy.

IHC staining was performed on FFPE tumor sections with heat‐mediated antigen retrieval. Antibody detection was performed according to the manufacturer’s instructions. Staining for Phosphohistone-H3 (PHH3; anti-PHH3. Ser10; RRID: AB_331535; Cell Signaling Technology, Danvers, MA) was done according to a previously described method (48). Immunostaining of ATM (Clone Y17; RRID: AB_725574; Abcam) and p-ATM (Clone EP1890Y; RRID:AB_1640207; Abcam) was done according to a previously described method, and expression was evaluated by a H-score which was calculated by multiplying the percentage of positive cells (0–100) by the intensity (0–3; ref. 49).

One million uterine SK-UT-1 cells were transiently transfected by electroporation with 1 µg each of pCAGGS-I-SceI-Trex2 (RRID: Addgene_44024; courtesy of Jeremy Stark, City of Hope, Department of Cancer Genetics) and pDRGFP (Addgene plasmid #26475; RRID: Addgene_26475; courtesy of Maria Jasin, Sloan Kettering Institute) or pimEJ5GFP (RRID: Addgene_44026; courtesy of Jeremy Stark). The iRFP670-expressing plasmid piRFP670-N1 (RRID: Addgene_45457; courtesy of Vladislav Verkhusha, Albert Einstein College of Medicine) was used as internal transfection efficiency control in all cases. The transfected cells were then treated with CFI-400945, AZD0156, combination of CFI-400945 and AZD0156, or DMSO. Cells were harvested 3 days after treatment. The proportion of GFP+ cells was determined by flow cytometry. The GFP+ rates indicating successful DNA repair by HR (pDRGFP) and NHEJ (pimEJ5GFP) were normalized to the proportions of RFP+ cells.

BRCA2 was knocked out in SK-UT-1 and SKN cells by transecting the cells with one or two sgRNA-Cas9 plasmids using Lipofectamine 3,000 transfection reagent (Invitrogen) according to the manufacturer’s instructions, followed by puromycin (InvivoGen) selection. The Cas9-containing plasmid pSpCas9(BB)-2A-Puro (PX459) V2.0 was a gift from Feng Zhang (Addgene plasmid #62,988; http://n2t.net/addgene:62,988; RRID:Addgene_62988). Two sgRNA sequences targeting BRCA2 (Exon 11: 5'-CAC CGC TGT CTA CCT GAC CAA TCG A-3', Exon 13: 5'-CAC CGT CTT ACC GAA AGG GTA CAC-3') were selected from the Human CRISPR Knockout Pooled Library (GeCKO v2). Successful disruption of the RAD51-interaction domain, which is important for HR repair, of BRCA2 in monoclones was validated by Sanger sequencing (Supplementary Fig. S1A and S1B).

Appropriate methods were used to analyze the data, ensuring transparency and reproducibility. Unless otherwise specified, the Welch t test was performed in R (version 4.1.3); for others, Prism (version 8.2.1, macOS, RRID: SCR_002798) was used. The half maximal inhibitory concentration (IC₅₀) of sarcoma cell lines was expressed as mean ± standard deviation based on three sets of independent experiments. One-way ANOVA and the Tukey honestly significant difference test were used to compare reverse transcription quantitative PCR results between LMS and HUtSMC cell lines and analyze the results of the comet assay in vitro and in vivo studies to compare the mean tumor volume between different groups. Two-way ANOVA and the Dunnett test were used to analyze the results of the caspase 3/7-apoptotic caspase activity assay. Kruskal–Wallis one-way ANOVA and a Dunn test were used to compare the intensity ratios of γH2Ax/DAPI staining in LMS cell lines. A Student t test was performed to compare HRD scores between LMS and LM clinical samples. For in vivo experiments, the number of mice used was determined by power calculations. According to our in vitro SRB data results, eight mice were sufficient to achieve power >90% at α = 0.05 using effect sizes calculated from the mean viability of single and combined treatments with PLK4 and ATM inhibitors. The Student t test was used to compare the tumor volume between the combined treatment and low-dose CFI-400945 in vivo, and abnormal mitoses and PHH3-positive cells in vivo. For all tests, P < 0.05 was considered significant.

Details of small variants, copy number alterations, structural variants, and fusion transcripts of DNA repair genes covered in the current study are available in Supplementary Data S1, and their corresponding BAM files were deposited in the European Nucleotide Archive with an accession number PRJEB75545 (https://www.ebi.ac.uk/ena/browser/view/PRJEB75545). Raw data of the study are available upon reasonable request from the corresponding author.

The ages of patients from whom uterine LMS samples were collected ranged from 42 to 83 years (median, 50). Tumor size ranged from 5 to 25 cm (mean, 14.28). At the last follow-up (median, 22 months), 23 patients (88%) were either dead or alive with disease (Supplementary Table S1). WES and transcriptome analysis of tumors showed loss-of-function alterations in key genes involved in DNA repair (Fig. 1A; Supplementary Data S1). Compared with LM, LMS had a significantly higher mean HRD score (38 vs. 9, P < 0.001; Fig. 1B). Compared with matched myometrium (Fig. 1C a and b), ATM and p-ATM protein expression was significantly reduced or lost in LMS (P < 0.05; Fig. 1C c and d). These results indicate that uterine LMS has defective DNA repair.

PLK4 mRNA was upregulated in clinical uterine LMS samples, a finding consistent with that of a previous report (4). Similarly, the cell lines (SK-UT-1, SKN, and SK-LMS-1) expressed higher levels of PLK4 mRNA than HUtSMCs (P < 0.0001; Fig. 1D and E).

SK-UT-1, SKN, SK-LMS-1, and HUtSMCs were treated with CFI-400945 for 6 days, with DMSO as a control. There was a dose-dependent decrease in cell proliferation (Fig. 2A). Annexin V/propidium iodide (PI) flow cytometry revealed increased cell death and apoptosis (Fig. 2B). Caspase 3/7 activity significantly increased in the three sarcoma cell lines compared with controls (P < 0.05; Fig. 2C).

Centrosomes were labeled with antipericentrin antibody to investigate the effect of CFI-400945 on sarcoma and HUtSMCs cell lines. After 3 days of treatment with low-dose (20–100 nmol/L) CFI-400945, PLK4 was partially inhibited and stabilized, which resulted in supernumerary centrosomes. At higher dosages (500 nmol/L) of CFI-400945, PLK4 was fully inhibited, and centrosomes remained mostly duplicated (20, 26, 50). All CFI-400945 doses caused nuclear enlargement and multilobulation in sarcoma cell lines as well as supernumerary centrosome numbers (Fig. 2D). CFI-400945 induced a dose-dependent increase in DNA content in the sarcoma cell lines (Supplementary Fig. S2A and S2B).

To examine the DNA damage response via HR repair to CFI-400945 treatment in myogenic gynecologic sarcoma cell lines, Western blotting of ATM and p-ATM was performed. p-ATM expression increased dose-dependently in response to CFI-400945, indicating an increase in HR repair response (Fig. 3A). Another Western blot was performed to investigate the effects of CFI-400945 on downstream targets of the HR repair pathway. In response to DNA damage, downstream targets of HR repair pathway were activated. Conversely, when the cell lines were treated with both CFI-400945 and AZD0156, phosphorylation of these downstream targets was decreased, indicating that HR repair was suppressed by pharmacologic ATM inhibition (Fig. 3B). To investigate the role of concurrent ATM inactivation in cell proliferation, myogenic sarcoma cell lines were treated with CFI-400945 and AZD0156 and evaluated by the SRB assay. The synergistic effect of the combination of CFI-400945 (25–200 nmol/L) and AZD0156 (400–800 nmol/L) was proven using the Chou–Talalay method (51), as indicated by a combination index of <1 (Fig. 3C).

H2AX was quantified by immunofluorescent staining with H2AX-S139 of the myogenic gynecologic sarcoma cell lines. In the enlarged and multilobated cells, fluorescence signals were concentrated in the segmented nuclei in which DNA repair occurred (Fig. 4A). Similar to the Western blot results, CFI-400945 triggered the repair response, as confirmed by increased γH2AX levels (P < 0.0001). In turn, CFI-400945–dependent induction of H2AX phosphorylation (S139) was reduced by AZD0156 (P < 0.001; Fig. 4B).

To quantify the effects of the two drugs on DNA damage, comet assays were performed, which revealed a greater amount of tail moment in the CFI-400945–treated SK-UT-1 cells. The addition of AZD0156 further inhibited HR, resulting in further accumulation of DNA damage (P < 0.001; Fig. 4C).

To examine the mechanism of DNA repair under different treatment conditions, a DNA damage repair reporter assay based on I-SceI was performed. A higher normalized GFP+ score in the DR-GFP and EJ5-GFP reporter assays indicated that DNA damage induced by CFI-400945 was repaired primarily through HR (P < 0.05). When ATM was inhibited by the addition of AZD0156, HR repair activity was decreased, and the major DNA repair activity shifted to error-prone NHEJ repair (Fig. 4D).

The in vivo antitumor effect of CFI-400945 was dose-dependent, and tumor size was reduced in both the low-dose (5 mg/kg) and high-dose (7.5 mg/kg) treatment groups compared with the vehicle (P < 0.05; Fig. 5A). The addition of 10 mg/kg AZD0156 to the low-dose CFI-400945 marginally reduced the tumor volume further without affecting body weight (Fig. 5B–D), and no gross or microscopic evidence of toxicity involving major internal organs was found (Supplementary Fig. S3A–S3H; Supplementary Fig. S4A–S4H). Hematoxylin and eosin (H&E) sections of the drug-treated tumors showed cellular enlargement with frequent multilobulated nuclei and atypical mitoses (Fig. 5E a). Increases in aberrant mitoses and reduced IHC PHH3 counts indicated diminished cell proliferation (Fig. 5E b and F).

Annexin V/PI flow cytometry demonstrated that CFI-400945 induced a higher apoptotic cell death rate in BRCA2-KO cells than in wild-type (WT) cells (Fig. 6A). In BRCA2-KO cells, CFI-400945 induced DNA damage, which was not effectively repaired and consequently accumulated, as demonstrated by enhanced immunofluorescent staining of nuclear γH2AX (P < 0.05; Fig. 6B). In the comet assay of SK-UT-1, the olive tail moment of WT versus BRCA2-KO cells was 4.52 versus 18.70 (P < 0.05). For SKN, the olive tail moment of WT versus BRCA2-KO cells was 7.78 versus 18.37 (P < 0.05). In the DNA repair reporter assay, the repair of DNA damage induced by CFI-400945 was reduced in both the HR and NHEJ pathways (Fig. 6C).

The current study evaluated the effects of PLK4 inhibition in uterine sarcomas. Because PLK4 mRNA was overexpressed in our clinical uterine LMS samples, which was consistent with another dataset (4), we concluded that PLK4 inhibition was a potential pharmacologic target. To determine the sensitivity of uterine LMS to DNA-damaging inducers, such as the oral PLK4 inhibitor CFI-400945, we studied the DNA repair status of uterine LMS in our clinical samples. Similar to other reports, we demonstrated that a subset of uterine LMS was defective in DNA repair (3–5, 8–11, 52). In our cohort, the most common genetic alterations involved in the HR pathway were BRCA2 (48%), RAD51B (48%), ATM (28%), and CHEK2 (24%). At the protein level, we observed a significant reduction or complete loss of ATM and p-ATM protein expression in 94% (65/69) and 60.9% (42/69) of uterine LMS, respectively (49). The HRD score of uterine LMS was also higher than that of LM, which indicates that uterine LMS is a genomically unstable tumor (13). In addition, this score predicts sensitivity to DNA-damaging drugs (53).

Similar to clinical tumor samples, PLK4 mRNA expression in the three gynecologic sarcoma cell lines was higher than in the smooth muscle cell line. The effects of CFI-400945 on these sarcoma cell lines were similar to those observed previously, which included chromosome missegregation, intensified DNA damage, and apoptosis (17, 20, 26, 29). Our in vivo experiments confirmed the efficacy of a low-dose regimen of CFI-400945 as a single agent. Prior clinical trials have shown dose-dependent neutropenia with high doses of the drug (25); however, in our experimental model, no microscopic evidence of toxicity was observed in the major organs of the mice treated with a high-dose regimen (Supplementary Fig. S3A–S3H).

Because CFI-400945 treatment induced excessive DNA damage in malignant cells, factors that further increased DNA error accumulation or impaired DNA repair would likely increase drug sensitivity (29). The treatments administered on myogenic gynecologic sarcoma cell lines demonstrated that CFI-400945 induced HR repair response by promoting ATM phosphorylation dose-dependently and triggered the phosphorylation of downstream CHK2 and H2AX. Conversely, PLK4 interacts directly with other targets in cell signaling pathways (22, 23) although the precise mechanisms in relation to DNA repair require further investigation. Additional studies could improve the selection of drug combinations for tumors with specific genetic backgrounds and in cases of drug resistance.

ATM inhibitors can enhance the sensitivity of tumors to DNA-damaging agents (29, 54, 55). In our study, we observed increased DNA damage when CFI-400945 was coadministered with AZD0156. By suppressing HR repair and reducing the phosphorylation of ATM, CHK2, and H2AX, the localization of phosphorylated H2AX to DNA was decreased, thereby disrupting the double-strand break repair response (56). To further investigate the underlying mechanisms of this synergistic effect, we transiently transfected DNA repair reporter plasmids to quantify HR and NHEJ repair activities in CFI-400945–treated cells (Fig. 4D). ATM normally promotes HR repair by inhibiting NHEJ repair through the release of the Ku and DNA-PKcs components of the NHEJ mechanisms (32, 57, 58). When ATM was inhibited in the combined treatment of AZD0156 and CFI-400945, NHEJ repair was upregulated. Overall, the observed synergism in the combined treatment can be explained by a shift in primary repair activity from HR to NHEJ, which is a more error-prone mechanism.

In our in vivo study of the SK-UT-1 cell xenograft model established in Balb/c nude mice, oral administration of CFI-400945 significantly reduced tumor size at both low (5 mg/kg) and high doses (7.5 mg/kg; Fig. 5A). The efficacy of low-dose CFI-400945 treatment was further enhanced with the coadministration of AZD0156, which induced no undue toxicity (Fig. 5B–D). Reduced PHH3-positive cell counts in drug-treated tumors also confirmed reduction in proliferation (Fig. 5E–F). These results demonstrated that CFI-400945 is effective for treating myogenic gynecologic sarcoma, with a synergistic effect when combined with an ATM inhibitor, even at a low dose.

As BRCA2 is one of the most frequently altered genes among key DNA damage repair genes in our in-house cohort of uterine LMS as well as in other datasets (3, 12), we generated BRCA2-KO myogenic gynecologic sarcoma cell lines to mimic the effects of BRCA2 deficiency and tested the efficacy of CFI-400945 under these conditions. The BRCA2-KO cell lines underwent a higher cell death rate than the WT cell lines (Fig. 6A). Due to defective HR repair, a greater amount of accumulated DNA damage was observed in the KO cell line (Fig. 6B). Intriguingly, both HR and NHEJ repair activities were diminished in BRCA2-KO cells (Fig. 6C). Although some studies have reported inconsistent findings on the effect of BRCA2 deficiency on HR and NHEJ functions (59, 60), these differences may be attributed to variations in the types of cancers studied, their diverse genomic profiles, and the choice of cell lines used. Overall, our results for the BRCA2-KO cell lines demonstrated the potential of CFI-400945 as a therapeutic agent for BRCA2-mutated uterine LMS. Further in vivo studies on CFI-400945 and BRCA2 mutations in uterine LMS could provide additional support for our findings.

Our study is limited by the small number of publicly available myogenic gynecologic sarcoma cell lines with diverse genetic backgrounds (61). The cell lines used in this study each had variable genomic alterations in genes involved in DNA damage repair (Supplementary Fig. S5; ref. 62), and it was not possible to demonstrate the effects of CFI-400945 on a much desired “HR-proficient model” of myogenic gynecologic sarcoma. The strength of our study was the generation of a BRCA2-KO model because this was clinically relevant in view of the high frequency of BRAC2 genetic alterations in uterine LMS (Supplementary Fig. S6; ref. 4). Future studies should consider using patient-derived xenografts or tumor organoids as representative models. However, obtaining such samples is challenging because of the rarity of uterine LMS and the lack of expertise in generating sarcoma organoids.

In conclusion, our study demonstrated that CFI-400945 effectively inhibits myogenic gynecologic sarcoma by causing double-strand DNA breaks that are mainly repaired through the HR pathway. Coadministration of an ATM inhibitor induced an appreciable drug synergism. BRCA2-KO cell lines were defective in HR and NHEJ repair and were more sensitive to CFI-400945 than BRCA2 WT cell lines. Overall, our findings demonstrated the therapeutic potential of CFI-400945 in myogenic gynecologic sarcoma, including uterine LMS. Genomic profiling of advanced-stage or recurrent uterine LMS may reveal HR defects and guide CFI-400945 therapy.

H.H.Y. Lee reports grants from Hong Kong SAR Food and Health Bureau’s Health and Medical Research Fund and Hong Kong Jockey Club Charities Trust during the conduct of the study. K.L. Chow reports grants from Hong Kong SAR Food and Health Bureau’s Health and Medical Research Fund and Hong Kong Jockey Club Charities Trust during the conduct of the study. H.S. Wong reports grants from Hong Kong SAR Food and Health Bureau’s Health and Medical Research Fund and Hong Kong Jockey Club Charities Trust during the conduct of the study. T.Y. Chong reports grants from Hong Kong SAR Food and Health Bureau’s Health and Medical Research Fund and Hong Kong Jockey Club Charities Trust during the conduct of the study. A.S.T. Wong reports grants from HMRF 08192396 during the conduct of the study. G.H.W. Cheng reports grants from Hong Kong SAR Food and Health Bureau’s Health and Medical Research Fund and Hong Kong Jockey Club Charities Trust during the conduct of the study. J.M.K. Ko reports grants from Hong Kong SAR Food and Health Bureau’s Health and Medical Research Fund and Hong Kong Jockey Club Charities Trust during the conduct of the study. H.C. Siu reports grants from Hong Kong SAR Food and Health Bureau’s Health and Medical Research Fund and Hong Kong Jockey Club Charities Trust during the conduct of the study. K.Y. Tse reports personal fees from Eisai, GlaxoSmithKline Limited, and Zai Lab and grants and other support from Pfizer, AstraZeneca, Zai Lab, and Gilead outside the submitted work. M.R. Bray reports other support from Treadwell Therapeutics outside the submitted work. T.W. Mak reports personal fees from AstraZeneca and other support from Agios Pharmaceuticals and Treadwell Therapeutics outside the submitted work. S.Y. Leung reports grants from Hong Kong SAR Food and Health Bureau’s Health and Medical Research Fund and Hong Kong Jockey Club Charities Trust during the conduct of the study, as well as grants from Pfizer, Merck, and Servier outside the submitted work. P.P.C. Ip reports grants from Hong Kong SAR Food and Health Bureau’s Health and Medical Research Fund and Hong Kong Jockey Club Charities Trust during the conduct of the study. No disclosures were reported by the other authors.

H.H.Y. Lee: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. K.L. Chow: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. H.S. Wong: Formal analysis, validation, investigation, visualization, methodology, writing–review and editing. T.Y. Chong: Formal analysis, validation, investigation, visualization, methodology, writing–review and editing. A.S.T. Wong: Formal analysis, investigation, methodology, writing–review and editing. G.H.W. Cheng: Data curation, software, formal analysis, validation, visualization, methodology, writing–review and editing. J.M.K. Ko: Formal analysis, investigation, methodology. H.C. Siu: Data curation, software, formal analysis, visualization, methodology. M.C.F. Yeung: Software, formal analysis, validation, methodology. M.S.Y. Huen: Methodology, writing–review and editing. K.Y. Tse: Investigation, writing–review and editing. M.R. Bray: Methodology, writing–review and editing. T.W. Mak: Conceptualization, methodology, writing–review and editing. S.Y. Leung: Resources, software, formal analysis, funding acquisition, validation, writing–review and editing. P.P.C. Ip: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

This study was supported in part by a grant from the Hong Kong SAR Food and Health Bureau’s Health and Medical Research Fund (No. 08192396, P.P.C. Ip) and the Hong Kong Jockey Club Charities Trust (S.Y. Leung). The authors would like to express their gratitude to Dr. Jacqueline Mason for her assistance with the acquisition of CFI-400945, as well as to Dr. Tsun Yee Tsang and Annie Chan for technical support.