Approximately 20% of high-grade serous ovarian cancers (HGSOC) have CCNE1 amplification. CCNE1-amplified tumors are homologous recombination (HR) proficient and resistant to standard therapies. Therapy resistance is associated with increased numbers of polyploid giant cancer cells (PGCC). We sought to identify new therapeutic approaches for patients with CCNE1-amplified tumors. Using TCGA data, we find that the mTOR, HR, and DNA checkpoint pathways are enriched in CCNE1-amplified ovarian cancers. Furthermore, Interactome Mapping Analysis linked the mTOR activity with upregulation of HR and DNA checkpoint pathways. Indeed, we find that mTOR inhibitors (mTORi) downregulate HR/checkpoint genes in CCNE1-amplified tumors. As CCNE1-amplified tumors are dependent on the HR pathway for viability, mTORi proved selectively effective in CCNE1-amplified tumors. Similarly, via downregulation of HR genes, mTORi increased CCNE1-amplifed HGSOC response to PARPi. In contrast, overexpression of HR/checkpoint proteins (RAD51 or ATR), induced resistance to mTORi. In vivo, mTORi alone potently reduced CCNE1-amplified tumor growth and the combination of mTORi and PARPi increased response and tumor eradication. Tumors treated with mTORi demonstrated a significant reduction in ALDH+ PGCCs. Finally, as a proof of principle, we identified three patients with CCNE1 amplified tumors who were treated with an mTORi. All three obtained clinical benefits from the therapy. Our studies and clinical experience indicate mTORi are a potential therapeutic approach for patients with CCNE1-amplified tumors.
Approximately half of high-grade serous ovarian cancers (HGSOC) exhibit homologous recombination (HR) DNA repair deficiency (1). HR-deficient (HRD) tumors often harbor mutations of BRCA1 or BRCA2, which recruit RAD51 to DNA double-strand breaks to initiate HR (2). HRD tumors are typically more responsive to platinum-based chemotherapy. A concerted effort to develop therapies for HRD tumors led to the development of PARP inhibitors (PARPi), which have demonstrated significant benefit for patients with BRCA-mutated tumors (3–5).
In contrast, HR-proficient (HRP) HGSOCs typically respond poorly to platinum chemotherapy and PARPi (6). Cyclin E1 (CCNE1) amplification occurs in a major subset of HRP HGSOC (7–9). CCNE1 amplification is nearly mutually exclusive, with mutations in BRCA1 and BRCA2 genes (7). CCNE1 amplification is commonly associated with primary therapy resistance and poor survival (10, 11).
CCNE1-amplification has been linked with an increase in the number of polyploid/multinucleate giant cancer cells (PGCC; refs. 12, 13). PGCCs are reported to have increased rates of mutagenesis, increasing tumor heterogeneity and accelerating resistance to chemotherapy (14–16) and radiotherapy (14). Polyploidy has been linked with a cancer stem-like state, promoting tumor initiation and growth (17, 18). HR activity is essential for survival and proliferation of PGCCs (19).
CCNE1 binds cyclin-dependent kinase 2 (CDK2) to regulate the cell cycle. Thus, CDK2 inhibitor therapy for CCNE1-amplified tumors has been proposed (12, 20). However, CDK inhibitors have proven toxic, with minimal efficacy, in clinical trials (21). Alternatively, proteasome inhibitors were explored to inactivate the HR pathway. These studies have not yet shown an improvement in the treatment of recurrent ovarian cancer (22), and combined CDK2-proteasome inhibitor treatment did not show any synergistic effect (20).
Here, we evaluate mTOR as a therapeutic target for CCNE1-amplified tumors. The mTOR pathway regulates protein synthesis, and mTOR inhibitors (mTORi) are clinically available (23–25) We show that the mTOR, HR, and DNA damage checkpoint pathways are enriched in CCNE1-amplifed ovarian tumors. We find that inhibition of the mTOR pathway downregulates proteins in the HR and DNA-checkpoint pathways. This is associated with a loss of PGCCs and a decrease in stem-like cancer cells. Consistent with downregulation of HR proteins, mTOR inhibition sensitized CCNE1-amplified cancer cells to PARP inhibition, significantly restricting tumor growth in vivo.
Materials and Methods
Cell culture, antibodies, and reagents
HGSOC cell lines NIHOVCAR3, CAOV3, and HEY1 were purchased from ATCC. COV318 was purchased from Sigma. OVCAR4 was obtained from the NCI Developmental Therapeutics Program. COV504 cells were a gift from Deborah Marsh (University of Sydney). Cell line authentication was carried out every 4 years and mycoplasma testing were performed by-monthly. Cell line passage 1 to 20 was used. All cells were cultured in RPMI1640 medium, with 10% FBS and 1% penicillin/streptomycin, at 37°C with 5% CO2. All antibodies used are listed in Supplementary Table S4.
IncuCyte live-cell imaging and cell-growth analysis
All HGSOC lines were seeded at 3,000 cells/well except HEY1 (1,000 cells/well) in a 96-well plate overnight. Cells were treated with vehicle or mTORi Torin1(20 nmol/L) or RAD001(20 nmol/L; each from Cayman Chemical) for live-cell imaging and cell proliferation using IncuCyte. For dual therapy with PARPi, olaparib (10 μmol/L; Selleckchem) was used. Four images/well were taken every 4 hours, and cell confluence was recorded. CellTiter-Glo (Promega) was used for viability.
shRNA-mediated knockdown of ATR using lentiviral packaging plasmids Pspax2 and pMD2.G into HEK293T was as described (26). Lentivirus was generated with either scrambled control or ATR shRNA constructs (gift from Malek laboratory, University of Michigan), NIHOVCAR3 cells were transduced with lentivirus and selected with puromycin (1 μg/mL). for siRNA-mediated Gene knockdown, NIHOVCAR3 or COV504 cells were plated at 50,000 cells/well in 6-well plates for 22 to 24 hours. Cells were then transfected with 25 nmol/L RAD51 siRNA or CCNE1 siRNA using FlexiTube siRNA from Qiagen. Target sequences appear in Supplementary Table S3. Cells were collected after 48 hours and counted using a Moxi cell counter (Orflo Technologies). Protein was extracted using RIPA buffer (Thermo Fisher Scientific) with protease inhibitors, followed by Western blotting analysis.
NIHOVCAR3 cells 2.5×105/well were seeded overnight and transfected with pcDNA3-EGFP alone (5 μg) or co-transfection of pcDNA3-EGFP (2.5 μg), with pcDNA3-ATR (2.5 μg) or pCMV-hRAD51 (2.5 μg) Addgene #13031, #31611 and #125570 respectively, using electroporation (200V). Alternatively, FuGENE 6 (Promega) reagent was used to deliver the plasmids. Successful transfection was confirmed using fluorescent microscopy, FACS, and Western blot. Cells were trypsinized after 22–24-hour transfection, counted, and replated 3,000/well in 96-well plates for IncuCyte live-cell imaging or CellTiter-Glo assay.
DNA content and PGCC measurement
1×106 cells were treated with vehicle or mTORi (RAD001 or Torin1, 10 nmol/L each) for 3 days, fixed with ethanol or 4% paraformaldehyde, permeabilized with 0.2% Triton X-100, and stained with Hoechst 33342 (10 μg/mL; Fischer). DNA content was measured by FACS following exclusion of cell aggregates by FSC-A versus FSC-W gating. Cells with DNA content >4n were counted as PGCCs.
DR-GFP reporter assay
U2OS cells stably expressing the DR-GFP reporter (282C DRGFP U2OS, a gift from Jeremy Stark) were seeded at 24,000 cells per well in 6-well plates overnight. The cells were transfected with 1.9 μg pCMV-I-SceI in Lipofectamine 2000 (Invitrogen; ref. 27). After 3 hours, the transfection media was replaced with fresh DMEM with/without mTORi (10 nmol/L Torin1). Following 24 hours of treatment, mTORi-containing DMEM was replaced with fresh DMEM; 48 hours post-Torin treatment, the cells were collected and fixed for flow cytometry. GFP+ cells were quantified using a Beckman Coulter CytoFlex flow cytometer, with 10,000 events acquired for each condition. Experiments were performed in triplicate.
Tissue microarray (TMA), FACS, IHC, immunofluorescence, and immunoblotting and qRT-PCR: ovarian ALDH FACS analysis and TMA preparation were as described (26, 28). TMA slides were deparaffinized, antigen retrieved in citrate buffer (10 mmol/L sodium citrate, pH 6.0) treated with 0.6% H2O2, then incubated with anti-CCNE1 antibody (Sigma, 1:500) overnight, followed by biotinylated anti-rabbit antibody and DAB staining (Vector ABC IHC Kits; Supplementary Table S4).
The number of PGCCs for tumors with high CCNE1 expression (n = 10, defined as >90% of nuclei/tumor islet CCNE1+) was compared with low/normal CCNE1 expression (n = 10, <50% nuclei/tumor islet expressing CCNE1), and correlated with the presence of polypoid giant cells. When tumor islet could not be clearly distinguished, three random high-power fields per sample were analyzed. Percentage of PGCCs were quantified in tumor islets as number of PGCCs/total tumor cells/high-powered field. At least three high-powered fields per sample were analyzed, and results compared via unpaired t test.
For immunofluorescence assay, fluorophore-conjugated secondary antibodies were used following overnight primary antibody incubation, with DAPI for nuclear staining. Images were acquired via Leica DM4B. ImageJ and GraphPad Prism (8.0.2) were used to analyze number and area of cells with positive staining. Ten high-power fields were used per sample group. Antibodies are described in Supplementary Table S4. ImageJ was used for densitometry analysis, cell number quantification, and cell-size measurement.
For RAD51 and γH2AX staining of cultured cells, NIHOVCAR3 cells were plated on coverglass in 12-well plates overnight. Cells were treated with mTORi Torin1 or RAD001 (10 nmol/L each) or with cisplatin (300 ng/mL) for 3 days, fixed in 4% paraformaldehyde, permeabilized with 0.2% Triton X-100, followed by immunofluorescent analysis with RAD51 and γH2AX antibodies.
For qRT-PCR and Western blotting analysis, HGSOC lines were seeded at 2.5 × 105 cells/plate except HEY1, 1.25 × 105 cells/plate, in 60 mm plates and treated with RAD001 (20 nmol/L), olaparib (10 μmol/L), or in combination for 2 days. Cells were treated with cisplatin (1 μg/mL) for 2 days. Cells were lysed in 300 μL NP40 buffer (Invitrogen, FNN0021) with 0.1% SDS for Western blotting. RNA was isolated using RNeasy Mini-Kit (Qiagen) followed by RT-PCR using SuperScript III First-Strand Synthesis System (Invitrogen). Antibodies, primer sequences for qRT-PCR are listed in Supplementary Tables S3 and S4.
In vivo tumor growth and treatment
All studies were performed with the approval of the University of Pittsburgh IACUC. NIHOVCAR3 or OVCAR4 cells (1 × 106) were injected subcutaneously. When tumors were detectable (∼2 to 20 mm3), mice were randomized into four groups (n = 10 tumors/group), then treated with DMSO control (5 days per week), RAD001 (mTORi, 10 mg/kg, 2 days/week, with DMSO on the other 3 days), PARPi (50 mg/kg, 5 days/week), and combined (mTORi+PARPi). For treatment of established tumors, the PARPi-treated group, when tumors reached the termination point (tumor size ∼1,000–2,000 mm3/mouse), mice were treated with mTORi alone or combined with PARPi at the doses indicated above.
Ovarian TCGA data and statistical analyses
We used cBioPortal (https://www.cbioportal.org) for ovarian patient survival analysis (29), comparing CCNE1-altered (CCNE1:AMP, EXP>2 EXP Z-score >2) and the remaining (unaltered) cases using RNA-seq root mean square error. Gene correlation analyses were performed by comparison between CCNE1-amplified and/or overexpressed (CCNE1:AMP EXP Z-score >2) and the rest of the sample, using both RNA-seq data and reverse-phase protein expression array. Genes significantly corelated with CCNE1 amplification and overexpression were extracted and further analyzed for pathway enrichment analysis, using Reactome (https://reactome.org). Statistical analysis was conducted using GraphPad Prism (8.0.2) and Excel. t test was used for two-sample comparisons; ANOVA was used for group comparisons. The Gehan–Breslow–Wilcoxon test was used for survival analysis. P < 0.05 was considered statistically significant.
The study was conducted in accordance with ethical guidelines and regulations of reginal, United States, and international regulatory requirements. Written informed consents were obtained from all eligible patients prior to the initiation of treatment. Patients were enrolled in a clinical trial approved by the University of Pittsburgh IRB (NCT01031381). Institutional review board approval was granted prior to the start of the trial. All patients discussed here had recurrent, advanced, HGSOC. CCNE1-amplification analysis was based on commercially performed genomic tumor testing ordered by the patient provider.
The data generated in this study are available upon request from the corresponding author.
CCNE1 overexpression/amplification is associated with PGCCs and induction of HR and DNA checkpoint pathway gene expression
CCNE1 amplification has been linked with the presence of PGCCs (12). To confirm this in CCNE1-amplified HGSOC, we evaluated CCNE1 protein expression in a human ovarian cancer tissue microarray and in the human protein atlas (26, 30). CCNE1 IHC identified a subset of “CCNE-high” tumors with >90% of tumor cell nuclei staining intensely CCNE1-positive. The percentage of PGCCs in CCNE1-high and CCNE1-normal/basal tumors was significantly higher than in tumors having basal CCNE1 expression (Fig.1A and B). This was confirmed using a public TMA dataset (https://www.proteinatlas.org/ENSG00000105173-CCNE1/pathology/ovarian+cancer#; Supplementary Fig. S1).
We next analyzed RNA-seq data, DNA copy number, and mutation profiles of 308 HGSOC ovarian cancers in the TCGA dataset (29). Comparing survival outcomes of patients with CCNE1-amplified and/or overexpressing (EXP Z score >2, “altered”) tumors to the remaining cohort (unaltered) showed that CCNE1 amplification/overexpression is associated with worse overall and disease-free survival (Supplementary Fig. S2A). Comparing gene expression in the CCNE1 mRNA-amplified/overexpressing group we observed that numerous genes were upregulated in the HR (e.g., RAD51/RAD51C/RAD54L, BRCA1/BRCA2), ATR signaling, and cell-cycle checkpoint pathways (Fig.1C; Supplementary Table S1).
To corroborate, we evaluated expression in HGSOC cell lines with known CCNE1 status: NIHOVCAR3, OVCAR4, and COV318 (CCNE1-amplified), HEY1 (CCNE1-WT), and COV504 (CCNE1-heterozygous deletion [Hetdel]). CCNE1 mRNA expression in these lines correlated with CCNE1 copy number (Fig. 1D; Supplementary Figs. S2B and S2C). Using these lines, we compared the expression of RAD51, ATR, and CHK2, three key genes which correlated with CCNE1 expression in the TCGA dataset. All three genes were significantly correlated with CCNE1 expression (Fig. 1E–G; Supplementary Fig. S2B). We also examined expression of ATM and checkpoint kinase 1 (CHK1), proteins which play important roles in DNA damage checkpoint response (31, 32). For both, significantly higher mRNA expression was noted in CCNE1-amplified cells compared with Hetdel, but no significant difference compared to WT (Supplementary Figs. S2B and S2D–S2E).
mTOR inhibition downregulates HR and DNA-damage checkpoint pathways in CCNE1-amplified HGSOC
We next confirmed upregulation of CCNE1-correlated genes at the protein level in cell lines in the presence/absence of cisplatin. NIHOVCAR3 (CCNE1-amplified) demonstrated consistently high expression of RAD51, ATR, CHK2, and pATR and pCHK2 (activated form) compared with HEY1, CAOV3 (CCNE1-WT), and COV504 (CCNE1-Hetdel; Fig. 2A). Little change was observed in KU70 and KU80, two proteins of the NHEJ pathway (Fig. 2A; Supplementary Fig S3). The addition of cisplatin generally increased expression of these proteins across cell lines (with the exception of RAD51 in HEY1; Fig. 2A; Supplementary Fig. S3). Total CHK2 and its phosphorylated active form (pCHK2, pT68) are also significantly co-expressed with CCNE1 at protein levels in the TCGA data (Supplementary Table S2).
Using the gene list from CCNE1-amplifed and non-amplified tumors, we performed Reactome mapping. We found CCNE1 expression correlated with PI3K/AKT, DNA damage, and checkpoint pathway proteins, and mTOR signaling significantly co-expressed with CCNE1 in ovarian cancer patient specimens (Fig. 2B). The mTOR pathway regulates protein synthesis (25, 33) and is thus a candidate to regulate expression of upregulated HR/checkpoint proteins. We therefore analyzed the impact of mTOR inhibition on RAD51, ATR, and CHEK2 protein expression and phosphorylation. In addition, as downregulation of HR proteins could sensitize otherwise HRP cell lines to PARP inhibition, we tested the impact of dual mTORi and PARPi therapy.
HGSOC lines with differential CCNE1 copy number status were treated with mTORi, PARPi, or dual mTORi/PARPi. mTORi resulted in downregulation of total and phosphorylated (active form) ATR and CHK2, all of which were upregulated in the CCNE1-amplified cell lines (Fig. 2C). In most cell lines, PARPi either had no impact or increased expression of total and phosphorylated ATR and CHK2. Addition of mTORi to PARPi countered PARPi-mediated induction of pATR and pCHK2. The impact of mTOR inhibition on expression of the HR protein RAD51 followed the same trend (Fig. 2C; Supplementary Fig. S4). KU70 and KU80 levels were unaffected by mTORi or PARPi (Fig. 2C; Supplementary Fig. S4). Phosphorylated p70 S6 kinase and phosphorylated-4EBP1, markers of activated mTOR signaling, are higher in both CCNE1-amplified cells and CCNE1 WT HEY1 cells compared with Hetdel COV504. Both are reduced by mTOR inhibition (Supplementary Fig. S4). Suggesting CCNE1 amplification could be increasing mTOR signaling, CCNE1 knockdown in NIHOVCAR3 with two independent siRNAs showed significant decreases of p4EBP1 compared with control siRNA (Supplementary Fig. S4C).
HR repair defect in CCNE1-amplified ovarian cancer cells upon mTORi treatment
RAD51 foci formation is a biomarker for HR repair. Given Western blot-indicated RAD51 expression was high in CCNE1-amplified tumor cells and downregulated with mTORi treatment (Fig. 2A and C), we examined RAD51 foci formation in CCNE1-amplified NIHOVCAR3 cells, alone or treated with mTORi (Torin1 or RAD001). Cisplatin was used as positive control of HR response. Control cells demonstrated modest RAD51 expression and numerous RAD51 foci which colocalized with γH2AX (Fig. 2D). Compared to control, cisplatin treatment significantly induced RAD51 expression and RAD51/γH2AX foci (Fig. 2Di; Supplementary Fig. S5A), whereas mTORi treatment was associated with a broad loss of RAD51 expression and RAD51 foci. RAD51 expression was detected in 37.50% and 58.85% or cells treated with RAD001 and Torin1, respectively, compared with 82% of control cells and 96% of cisplatin-treated cells (Fig. 2Dii; Supplementary Fig. S5A).
To confirm that mTORi results in HR repair defect in CCNE1-amplified cells, we performed a DR-GFP reporter assay in the presence/absence of mTORi, using CCNE1-amplified U2OS cells stably transfected with SceGFP cassette with a single I-SceI site (Fig. 2Ei; ref. 27). Cells were transfected with pCMV-I-SceI to induce double-strand break. Successful HR repair was measured by the production of GFP+ cells. mTOR inhibition significantly reduced HR repair efficiency, with 2.9-fold reduction in cells treated with mTORi compared with control (Fig. 2Eii). Consistent with this, growth of CCNE1-amplified OVCAR3 cells, compared with Hetdel COV504, was more sensitive to RAD51 siRNA knockdown and mTOR inhibition (Fig. 2F; Supplementary Fig. S5B).
CCNE1-amplified cells are preferentially responsive to mTORi in an HR/checkpoint protein-dependent manner
It is generally felt that CCNE1-amplified cells are dependent on an intact HR pathway for survival (34, 35). We next determined whether mTOR inhibition could serve as a therapeutic for CCNE1-amplified HGSOC cells. We treated NIHOVCAR3, HEY1, and COV504 cells with mTORi (RAD001 or Torin1) and treatment response correlated with CCNE1 gene dosage, with COV504 (CCNE1-Hetdel) cells showing no response to mTORi, HEY1 (CCNE1-WT) cells showing a modest response, and NIHOVCAR3 (CCNE1-amplified) cell growth most impacted by mTORi (Fig. 3A; Supplementary Fig. S6). The opposite effect was observed when cells were treated with PARPi, with COV504 response > Hey1 > NIHOVCAR3 and OVCAR4 nonresponsive to PARP1 (Supplementary Fig. S6B).
“Activation of ATR” was the pathway most strongly correlated with CCNE1 amplification (Fig. 1C). To determine if ATR downregulation was contributing to response to mTORi, we downregulated ATR expression in NIHOVCAR3, using two ATR shRNA constructs (Fig. 3B). Downregulation of ATR reduced NIHOVCAR3 viability and sensitized cells to mTORi treatment (Fig. 3C). Consistent with impact on cell-cycle checkpoints, mTORi-treated NIHOVCAR3 cells demonstrated cell-cycle arrest (Supplementary Fig. S7).
We next tested the impact of overexpressing either ATR or RAD51 in NIHOVCAR3 cells. Only modest expression of ATR was obtainable; however, RAD51 was strongly expressed (Fig. 3D; Supplementary Figs. S8A and S8B). RAD51 expression was also associated with an increase in ATR expression (Fig. 3D). Induction of ATR expression was associated with an ∼40% increase in resistance to mTORi, whereas RAD51 expression led to complete resistance (Fig. 3E). Similarly, live-cell imaging showed ATR expression led to mild mTORi resistance, whereas RAD51 expression demonstrated profound mTORi resistance (Fig. 3F; Supplementary Fig. S8C).
mTOR inhibition impedes tumor growth of CCNE1-amplified ovarian tumors and instills PARPi response
Next, we assessed the impact of RAD001 (Everolimus) on CCNE1-amplified NIHOVCAR3 tumors (n = 10/group). Given downregulation of HR proteins, we tested the impact of mTORi, alone and in combination with the PARPi olaparib. mTORi demonstrated significant delays in tumor growth as a single agent (Fig. 4Ai). Although PARPi showed no activity as a single agent, mTORi and PARPi demonstrated increased therapeutic activity over mTORi alone (Fig. 4Ai). At the time of euthanasia for the control group, 100% of control and PARPi-treated animals had established tumors (∼600 mm3), whereas 80% of the RAD001-treated group and only 50% of the dual-therapy group had detectable tumors (Fig. 4Aii), suggesting the potential eradication of disease.
We repeated this study using the CCNE1-amplified OVCAR4 model (36). Once again, we observed that (i) PARPi showed no impact on tumor growth, (ii) mTORi demonstrated significant antitumor activity, and (iii) there was increased activity in the dual-therapy group (Fig. 4Bi). At the time the control animals were euthanized (56 days after initiation of therapy), 100% of control and PARPi-treated mice had tumors, 90% of mice in the mTORi group had tumors, and only 20% of animals in the dual-therapy group had detectable tumors (Fig. 4Bii). To further monitor tumor initiation, mTORi- or dual therapy-treated animals were monitored, off therapy, an additional 5 months before being euthanized. At this time, small tumors were detectable in 9/10 mTORi-treated animals (average weight 75 mg) and 6/10 dual therapy–treated animals (average weight 33 mg; Fig. 4Ci–ii).
We also tested the impact of mTORi on established NIHOVCAR3 tumors (∼600 mm3) that were progressing on PARPi therapy. Tumors treated with either mTORi alone or in combination with PARPi demonstrated significant tumor regression (Fig. 4D).
As the CDK2 inhibitor dinaciclib (CDK2i) has been reported to have therapeutic efficacy for CCNE1-amplified tumors (20), we compared the impact of RAD001 to dinaciclib in CCNE1-amplified NIHOVCAR3 tumors. Although Dinaciclib showed activity, RAD001 was more effective (Supplementary Fig. S9). Significant toxicity was observed with CDK2i treatment, whereas RAD001 was well tolerated. No synergy was observed with combined treatment (Supplementary Fig. S9).
mTORi reduce pCHK2 and PGCCs in vivo
We next evaluated DNA damage repair/checkpoint proteins in the mTORi-treated tumors. γH2AX foci, indicators of DNA damage (37), were decreased in mTORi-, PARPi-, and dual therapy-treated tumors (Fig. 4E). pCHK2 was diffusely present in control tumor cells and significantly reduced in mTORi-treated and mTORi/PARPi-treated tumors, with a statistically significant reduction in the number of pCHK2+ nuclei per high-power field scored (Fig. 4E and F). Furthermore, we noted that pCHK2 was commonly associated with multinucleate giant cells/PGCCs. Scoring the nuclear size of cells in control and treatment groups confirmed that mTORi reduced both the number of PGCCs and the presence of pCHK2 in these cells (Fig. 4G and H).
Polyploidy can lead to aberrant spindle formation, mitotic catastrophe, and death (38). Cells surviving this mitotic catastrophe result in therapy-resistant cancer stem-like PGCCs (15, 17, 18). DNA damage repair pathways may be essential for PGCCs to survive mitotic catastrophe (38). To determine if mTORi could be impacting mitotic spindles, we stained tumor cells with α-tubulin to evaluate mitotic spindle organization. In general, mitotic cells in CCNE1-amplified tumors had mono-, di-, and multipolar spindle organization (Fig. 5A). In contrast, mTORi-treated tumors demonstrated highly aberrant mitotic spindles (Fig. 5A). Control and mTORi-treated tumors demonstrated a 2.2-fold increase in deranged mitotic spindles (Fig. 5B).
To confirm that the formation of PGCCs is prevalent in CCNE1-amplified tumors, we labeled CCNE1-amplified HGSOC lines NIHOVCAR3 and COV318 with Hoechst staining and used FACS to identify cells with >4n DNA content after excluding cell aggregates. CCNE1-amplified HGSOCs had a significantly higher percentage of PGCCs than CCNE1-WT and CCNE1-Hetdel (Fig. 5C; Supplementary Fig. S10A). Treatment with mTORi resulted in significant polyploidy reduction (Fig. 5C; Supplementary Fig. S10B).
Recent work suggests PGCCs can function as cancer stem-like cells (CSCs; refs. 17, 18). Our group and others have shown that aldehyde dehydrogenase (ALDH) expression is enriched in CSCs (39, 40). We therefore evaluated PGCCs for ALDH1A1 expression. In concordance with a potential stem-like role, we observed that PGCCs from CCNE1-amplified tumors were predominantly ALDH1A1-positive (Fig. 6A, control and PARPi panels, left). Consistent with mTORi reduction in PGCCs, mTORi treatment was associated with a significant reduction in ALDH+ cells and ALDH+ PGCCs (Fig. 6A and B). Flow cytometric analysis of live single-cell suspensions from mTORi-treated tumors with the Alde fluor assay confirmed a reduction in ALDH activity in mTORi and dual mTORi/PARPi-treated tumor cells (Fig. 6C; Supplementary Fig. S11A). Reduction of ALDH activity by mTORi was further confirmed in vitro using two different mTORi inhibitors (Supplementary Figs. S11B–S11D).
mTORi treatment of patients with CCNE1-amplified tumors
Finally, we sought to evaluate the potential for benefits in patients. We recently evaluated the impact of RAD001 and bevacizumab in patients with recurrent HGSOC (41). We identified 4 patients with CCNE1 amplification. One withdrew from the trial and was not evaluable. The others demonstrated clinical benefit with RAD001/bevacizumab therapy; 1 patient had a partial response and two demonstrating stable disease for several months (Fig. 6D and E).
CCNE1 amplification/overexpression is common in many solid tumors, including ovarian cancer (7, 9, 42). CCNE1 amplification in HGSOC is linked to therapy resistance (11). This study builds on prior work indicating that CCNE1 amplification/upregulation is associated with the activation of HR DNA damage repair proteins and DNA damage checkpoint response genes (43). Our findings suggest induction of these pathways in CCNE1-amplified tumors is at least partly dependent on mTOR signaling. As hyperactivation of HR may account for chemotherapy and PARPi resistance in CCNE1-amplified and other HRP tumors (44–47), our results implicate mTOR as a therapeutic target to overcome primary resistance and potentially increase response to PARPi therapy.
We found that CCNE1-amplified HGSOC lines and tumors contain a high percentage of PGCCs. This is in-line with findings that overexpression of CCNE1 in mice results in multipolar spindle formation and chromosomal mis-segregation, with increased aneuploidy and polyploidy (48). PGCCs have been linked with increased cancer “stemness,” increased therapy resistance and tumor initiation capacity (18). We found that PGCCs express high levels of the CSC marker ALDH. Consistent with a role for PGCCs in cancer stemness, dual therapy with mTORi and PARPi significantly reduced the number of PGCCs; this was associated with decreased tumor initiation in vivo. Consistent with a role for mTORi in regulating DNA repair/replication and PGCCs having altered DNA repair/replication, mTORi reduced PGCCs in cell lines, regardless of CCNE1 status.
PGCCs in CCNE1-amplified tumors highly express CHK2, which co-stained with DNA damage marker γH2AX, suggesting a role for these proteins in maintaining the viability of PGCCs; treatment with mTORi, which downregulates these proteins, resulted in a significant increase in cells with aberrant mitotic spindles, loss of PGCCs, and a significant delay in tumor growth. We also showed that ovarian PGCCs highly express ALDH. Suggesting a role for ALDH in maintaining DNA fidelity, we found that ALDH inhibitors induce DNA double-strand breaks and synergize with ATR inhibitors (49). The exact role of ALDH and DNA damage response remains to be determined.
The mTOR pathway controls numerous aspects of protein expression. We find that expression of mTOR pathway correlate with CCNE1 expression and is critical for expression of RAD51, ATR, and CHK2. As expected, mTORi-mediated downregulation of these proteins in these otherwise HR-proficient/PARPi-resistant tumors was associated with increased PARPi response in vivo. These data aligns with findings that, like mTORi, ATR inhibitors can significantly impact the growth of CCNE1-amplified tumors and synergize with PARPi (50). Similarly, bromodomain and extra-terminal protein inhibitors, which also downregulate HR and DNA checkpoint proteins, have shown activity in CCNE1-amplified tumors (34, 35).
Finally, mTORi treatment, in a limited number of patients with CCNE1-amplified tumors, was associated with clinical benefit; 1 patient had partial response and two had stable disease (41). It is important not to overinterpret results given the limited number of patients; however, given the generally poor treatment response of these patients, results are encouraging. Although duration of response was limited (3–6 months), it is possible that use of mTORi earlier or combination with PARPi could improve these results.
In conclusion, we find that CCNE1-amplified HGSOC express high levels of HR and DNA damage response proteins, and mTORi treatment results in downregulation of these proteins. Importantly, mTORi treatment of CCNE1-amplified tumors in vivo and in patients is associated with tumor response. Furthermore, mTORi, via downregulation of the HR pathway, allows response to PARPi in these otherwise therapy-resistant tumors. These studies support clinical trials of mTORi in CCNE1-amplified tumors.
R.J. Buckanovich reports being cofounder of Tradewinds Bioscience. None of the work in this manuscript relates to Tradewinds. No disclosures were reported by the other authors.
S. Bai: Data curation, writing–original draft, writing–review and editing. S.E. Taylor: Writing–original draft, writing–review and editing. M.Z. Jamalruddin: Data curation, methodology. S. McGonigal: Data curation. E. Grimley: Data curation. D. Yang: Data curation. K.A. Bernstein: Data curation, writing–original draft, writing–review and editing. R.J. Buckanovich: Data curation, writing–original draft, writing–review and editing.
We would like to thank S. Malik for providing the ATR shRNA constructs and C. Bakkenist for providing the ATR overexpression construct. This work was supported by NIH grants R01CA218026 (to R.J. Buckanovich and S. Bai) and R01ES031796 (to K.A. Bernstein). The work was supported by core supports under the UPMC Hillman Cancer Center core support grant 5P30CA047904 and the Hillman Fellows for Innovative Cancer Research Program.
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.
Note: Supplementary data for this article are available at Molecular Cancer Therapeutics Online (http://mct.aacrjournals.org/).