Targeting Dormant Ovarian Cancer Cells In Vitro and in an In Vivo Mouse Model of Platinum Resistance

Ovarian cancer is the fifth leading cause of cancer-related death among women in the United States with the highest mortality rate of all gynecologic cancers (1). Although 70% of patients with ovarian cancer achieve complete clinical remission following cytoreductive surgery and treatment with a platinum/taxane-based chemotherapeutic regimen, the majority suffer relapse with disease that has acquired or will develop chemoresistance (2–5). New therapies aimed at reducing disease recurrence are urgently needed to improve survival.

Cancer stem-like cells (CSLC) that are able to enter into a state of dormancy and later regenerate tumors may provide a causal explanation for ovarian cancer relapse after initially successful treatments (4, 6). Furthermore, cells with CSLC characteristics are naturally resistant to chemotherapeutic agents, in part due to expression of drug transporters and a slow rate of proliferation (7). In patients with advanced ovarian cancer, ascites fluid often accumulates in the peritoneal cavity and can become enriched with tumor cells that aggregate together to form multicellular spheroids (8). These spheroids contain CSLCs and exhibit resistance to chemotherapy (4, 9). They are also capable of embedding into other tissues and this feature may be responsible for the “grains of sand” texture in the peritoneal cavity in women with advanced stage serous ovarian cancer at the time of initial cytoreductive surgery (10). The embedded cells may enter into a state of dormancy, thus rendering them resistant to commonly used chemotherapeutics that target actively proliferating cells. These dormant cells may later awaken to initiate growth and recurrent disease once conditions are favorable. Thus, targeting these spheroids may offer a plausible approach to achieve long-term remission of women with advanced ovarian cancer. To this end, we grew ovarian cancer cells on nonadherent plates using serum-free spheroid culture conditions, which results in aggregation of cells with spheroid-forming capacity into spheroids (11–13). We then sought to identify novel candidate drugs that are effective at targeting these spheroids.

In the current study, spheroid-forming ovarian cancer cells indeed exhibited characteristics of cellular dormancy, a condition of greatly reduced growth and/or quiescence that is associated with drug resistance and disease recurrence (4, 14). We previously reported that the protein kinase C inhibitor, UCN-01 (7-hydroxystaurosporine; ref. 15), had higher efficacy against slow proliferating and/or quiescent ovarian cancer cells (16). Here, we assessed the efficacy of UCN-01 against ovarian cancer spheroids and compared these results with the efficacy of cisplatin and paclitaxel. In addition, analysis of gene expression profiles in spheroid-forming ovarian cancer cell lines versus cell lines that do not form spheroids suggested that the activity of mitochondrial pathway genes contributes to spheroid-forming capacity. Mitochondria are known to play a central role in regulating cellular metabolism and produce ATP through the citric acid cycle. Oligomycin, an inhibitor of mitochondrial ATP synthase (17), repressed spheroid formation in a dose-dependent manner. In a mouse xenograft model, we found that UCN-01 or oltipraz prolonged the disease-free interval following carboplatin treatment. This markedly contrasted with results from continued carboplatin treatment, in which the mice showed a rapid increase in tumor volume. Collectively, these results show that cells cultured under serum-free spheroid culture conditions are induced into cellular dormancy, and further suggest that genes involved in mitochondrial pathways and function are potential targets for eradication of these dormant ovarian cancer cells.

Ovarian cancer cell lines were from the collection maintained by the Department of Obstetrics and Gynecology at Duke University (Durham, NC) and were genetically authenticated by the DNA Sequencing and Analyses Core at the University of Colorado Denver (Denver, CO) in 2008. Genomic DNA isolated form the individual cell lines was amplified at short tandem repeats using the Applied Biosystems Identifiler Kit and an ABI 3730 capillary sequencer. Freezer stocks were prepared from the authenticated cells and were subsequently used for the experiments described herein. Cells were also Mycoplasma tested using the Lonza MycoAlert PLUS assay by the Duke Cell Culture Facility. Ovarian cancer cell lines were typically used no more than five passages after thawing and were grown in RPMI1640 medium with l-glutamine (Sigma-Aldrich) supplemented with 10% FBS in a humidified incubator with 5% CO₂ at 37°C.

Assays for colony formation in soft agar were performed as described previously (18). Briefly, 2× RPMI media was prepared from powder and supplemented with FBS and antibiotics (Invitrogen). A 1% agarose solution was made with the RPMI media using low melting temperature agarose (Invitrogen; #16520). One mL of 0.5% agarose was placed into each well of 6-well tissue culture dishes and overlaid with 1 mL of 0.33% agarose prepared in 1× RPMI and containing 2 × 10⁴ cells. After 3 weeks' incubation at 37°C in a humidified chamber with 5% CO₂, colonies larger than 100 μm in diameter were counted. The colony number formed for each cell line was determined by averaging the number of colonies >100 μm that were counted in 10 to 20 microscopic fields at 100× magnification. Data were analyzed using GraphPad Prism software with a P value less than 0.05 considered statistically significant.

Single dissociated ovarian cancer cells derived from cell lines were plated at 500 cells/mL in 2 mL serum-free DMEM-F12 (GIBCO; Invitrogen; #11320033) supplemented with 5 μg/mL insulin (SAFC Biosciences; #91077C), 20 ng/mL human recombinant EGF (Invitrogen; #11376454001), 10 ng/mL bFGF (R&D Systems; #223-FB-010), and 0.4% BSA using 6-well ultra-low attachment plates (Corning Life Sciences; #CLS3471–24EA). After 7 to 10 days in culture, spheroids larger than 50 μm in diameter were counted. To assess the ability of spheroids to undergo serial passaging, dissociated ovarian cancer cells from spheroids were plated at 5 × 10³ cells/mL. Cells grown in serum-free spheroid culture were enzymatically dissociated by incubation in a trypsin-EDTA solution (0.025%) for 10 minutes and underwent passaging every 7 to 10 days. Fresh medium (500 μL/well) was added twice a week. Cell viability was measured at each passage using trypan blue staining. The experiments were performed at least in triplicate.

Tumors were obtained from 8 patients diagnosed with advanced ovarian cancer at the time of initial cytoreductive surgery at Duke University Medical Center, after receiving approval from the Duke University Institutional Review Board and following provision of written informed consent. Fresh tumors were minced and suspended in serum-free media with growth factors as described above. To further dissociate the tumors, 300 U/mL of both collagenase (Sigma-Aldrich; #C0130) and hyaluronidase (EMD Biosciences; #389561) was added followed by overnight incubation at 37°C in a humidified chamber with 5% CO₂. The cell suspension was filtered using a 40-μm cell strainer, and red blood cells were removed using Histopaque 1077 (Sigma-Aldrich; #10771). The cells were resuspended in the media described above and plated on ultra-low attachment plates. To determine the ability of the cells within spheroids to undergo self-renewal, spheroids generated by eight ovarian cancer cell lines and seven primary tumors were dissociated following three to five passages. Cells were plated at single-cell dilution into 96-well ultra-low attachment plates under serum-free spheroid culture conditions described above. Microscopic visualization confirmed the presence of one cell per well in the 96-well plates.

HEYA8 cells (3 × 10³ or 6 × 10³) were seeded into individual wells of 96-well tissue culture plates in regular media or into individual wells of 96-well ultra-low attachment plates in serum-free spheroid culture media, respectively. To compare cell proliferation activity between these two conditions, a proliferation rate was calculated using the data obtained from MTT colorimetric assays (CellTiter 96 AQueous One Solution Cell Proliferation Assay; Promega; #G3582). Data from five replicates for at each time point were analyzed using GraphPad Prism software with a P value less than 0.05 considered statistically significant.

For Ki-67 staining, plates were collected after 7 days of culture and fixed with 1:1 acetone:methanol. Spheroid cells were prepared for immunostaining on slides using cytospin centrifugation as described previously (16). The cells were incubated with a monoclonal antibody to human Ki-67 antigen, clone MIB-1 (diluted 1:100; DAKO; #M724029) at 4°C overnight. Immunodetection was carried out using the streptavidin-biotin based Multi-Link Super Sensitive Detection System (4plus Universal HRP Detection System from Biocare Medical; #HP504US).

To investigate whether quiescent cancer cells in serum-free spheroid culture can resume growth in regular media, cells from cell lines and three primary tumors that underwent serial passage in serum-free spheroid culture were collected and plated on tissue culture–treated plates with regular media. Immunostaining with a pan-cytokeratin cocktail antibody, AE1/AE3 (Dako; #GA05361-2), was used to confirm the epithelial nature of the cancer cells. The Mouse Super Sensitive Negative Control, clone HK119-7M (BioGenex) was used as a negative control.

UCN-01 was kindly provided by the Cancer Therapy Evaluation Program of the National Cancer Institute (Rockville, MD). Cisplatin, paclitaxel, and oligomycin were purchased from Sigma-Aldrich. Cells in regular media were seeded at 3 × 10³ cells per well into tissue culture–treated plates, and cells in serum-free sphere culture were plated at 6 × 10³ cells per well in 96-well ultra-low attachment plates. Drugs were added 24 hours later, and the cells were incubated in the presence of the drugs for an additional 24 hours (oligomycin) or 48 hours (all other drugs). Six wells were used for each drug concentration. The cell survival ratio was determined by quantifying the reduction in growth compared with mock-treated cells at 48 hours using a standard MTT colorimetric assay (Promega AQueous One, Promega; #G3582). The assay was repeated four times.

We analyzed our previously generated Affymetrix HG-U133A DNA microarray data for ovarian cancer cell lines (Gene Expression Omnibus: GSE25429). Data analysis was performed using Partek Genomics Suite software. Affymetrix CEL files were normalized using the Robust Multichip Average method (19). To investigate gene expression profiles that allow cancer cells to survive serum-free spheroid culture conditions, we identified differentially expressed genes between cell lines that have low (n = 4) versus high (n = 5) capacity to survive under these conditions. We used Student t tests with an unadjusted P value cutoff of 0.05, which resulted in the identification of 2,799 probes. Supervised clustering was performed as described previously (20). Bioinformatics analyses of the genes differentially expressed between cell lines with low and high potential to survive serum-free spheroid culture conditions were done using DAVID 6.8 (21, 22).

CAOV2 ovarian cancer cells were transduced with the dual reporter lentivector pGreenFire1 (pGF1-CMV; positive control; System Biosciences; #TR039PA-1) according to the protocol from the manufacturer. This vector encodes the copepod GFP and firefly luciferase (LUC) under a constitutively active CMV promoter and contains the puromycin selectable marker. Expression of GFP enables detection and isolation of cells harboring the construct and expression of LUC allows for monitoring of tumor growth and response using noninvasive live imaging. CAOV2-GFP/LUC cells were sorted for GFP positivity by flow cytometry. The GFP-positive cells were expanded in cell culture prior to injection into nude mice.

The Duke University Institutional Animal Care and Use Committee approved this research. To generate xenograft tumors, 3.5 × 10⁵ CAOV2-GFP/LUC cells were injected intraperitoneally (i.p.) into each of 40 healthy 6- to 7-week-old female athymic Crl:NU(NCr)-Foxn1^nu mice (Charles River Laboratories). Live imaging was performed following administration of isoflurane using an XGI-8 Gas Anesthesia System. Tumor formation and regression with treatment were monitored using the IVIS 100 Optical Imaging System (Xenogen, Caliper Life Sciences) in the Duke Cancer Institute's Optical Molecular Imaging and Analysis Core. Living Image 2.6.1 (Caliper Life Sciences) software was used for quantitative analysis. Tumor flux measurements, defined as photon/sec/cm²/steradian, were the raw output values from the luminescent imaging data.

Following establishment of a measurable tumor, treatment with carboplatin was performed every 4 days for 3 weeks at 80 mg/kg of mouse body weight. Tumor signal was monitored by IVIS imaging. Twenty-six mice were responsive to the carboplatin treatment and were randomly grouped for secondary treatment (Table 2). Arm 1 mice (n = 5, carboplatin/carboplatin) received continued carboplatin treatment at 80 mg/kg i.p. every 4 days for 2 weeks. Arm 2 mice (n = 5, carboplatin/UCN-01) received UCN-01 treatment at 10 mg/kg i.p. for five days, no treatment for two days, then another cycle of UCN-01 every day for another five days, totaling 10 doses. Arm 3 mice (n = 6, carboplatin/oltipraz) received oltipraz at 250 mg/kg via oral gavage every day for six days, no treatment for one day, then another cycle of oltipraz every day for six more days, totaling 12 doses. Arm 4 mice (n = 4, carboplatin/none) received no further treatment. Arm 5 mice (n = 6, none/none) received no treatment from the beginning to the end of the study. Mice were imaged weekly during the secondary treatment period.

Twenty-two cancer cell lines were used to perform anchorage-independent growth assays. Average numbers of colonies formed per microscopic field that were >100 μm in diameter for these cell lines ranged from 0 to 7.15 (Supplementary Table S1). The cell lines CAOV3, DOV13, and OVCA429 did not form any colonies. OVCA420 only formed colonies <100 μm in diameter and TYK-nu, SKOV8, 41M, OVCA432, OVCAR3, and FUOV1 formed a small number of colonies (>0 but ≤0.1 colonies/microscopic field). Nine cell lines with ≥2.9 colonies/microscopic field were categorized as having high tumorigenic potential, including PEO1, OVCAR8, OVARY1847, OV90, HEYA8, OVCAR2, CAOV2, HEY, and HEYC2.

Nonadherent spheroid culture conditions have been used to enrich for CSLCs from tumor tissue of a variety of malignancies (11–13). Therefore, we tested the ability of the nine cell lines with increased tumorigenic potential to form spheroids using a nonadherent spheroid assay. The numbers of spheroids formed by each of the cell lines and the ability of the spheroids to undergo serial passaging are shown in Supplementary Table S2. Five cell lines were able to undergo passage more than five times and were determined to have increased capacity to form spheroids under serum-free spheroid culture conditions. On the other hand, OVARY1847, OV90, PEO1, and OVCAR8 formed either no or few spheroids, and most were not able to survive beyond two passages. Representative monolayer cells and spheroids of HEYA8 are presented in Fig. 1A. Primary ovarian cancers were also used for nonadherent spheroid assays. A number of spheroids were obtained initially from seven of eight tumors, but only few spheroids, if any, were formed following five passages. Representative spheroids for primary ovarian cancers are also shown in Fig. 1A.

Sustainable spheroid formation beyond 10 passages was not achieved for the ovarian cancer cell lines tested nor for spheroids derived from primary ovarian cancers. However, a small number of single cells remained viable in serum-free spheroid culture even 3 months after initiation of the culture. Surviving primary ovarian cancer cells in serum-free spheroid culture conditions for 83, 90, and 98 days were tentatively transferred into tissue culture–treated dishes. They subsequently adhered to the dishes and resumed growth. Representative staining with a pan-cytokeratin marker is shown in Fig. 1B. To investigate the ability of the cells within the spheroids to self-renew, a characteristic of stem cells, single-cell cloning was performed by serial dilution of spheroid-forming cells under serum-free spheroid culture conditions. No secondary spheroids were formed within the 3-month duration of the experiment.

Serial cell proliferation assays demonstrated that HEYA8 ovarian cancer monolayer cells exhibit rapid growth and became confluent three days after plating, while the growth of HEYA8 cells in serum-free spheroid culture was, as expected, strongly repressed (unpaired t test, P < 0.0001 on days 2 and 3; Fig. 1C). The proportion of proliferation marker Ki-67–positive cells was decreased in the spheroid-forming cells of HEYA8 (cytospin slides prepared on the 7th day in serum-free spheroid culture), as compared with HEYA8 grown as monolayer cells (8% vs. 87%; Fig. 1D). Spheroids from a primary tumor (28th day; just before 4th passage) also exhibited a relatively small population of Ki-67–positive cells (Fig. 1D).

7-hydroxystaurosporine (UCN-01; 25, 50, and 100 nmol/L) exhibited an enhanced antiproliferative effect against HEYA8 cells in serum-free spheroid culture compared with HEYA8 monolayer cells at each tested dose (unpaired t test, P = 0.0016, P = 0.016, and P = 0.009, respectively; Fig. 2A). The response to cisplatin (25, 50, and 100 μmol/L) did not quite reach statistical significance when comparing the spheroids versus monolayer cells but showed a less antiproliferative effect against spheroids (unpaired t test, P = 0.47, P = 0.08, and P = 0.15, respectively; Fig. 2A). Paclitaxel treatment (10, 50, and 100 nmol/L) did not inhibit, but rather increased cell survival in spheroids (unpaired t test, P = 0.003, P = 0.0002, and P = 0.001, respectively; Fig. 2A). Similar results were obtained for the six additional cell lines that also formed spheroids in serum-free culture conditions (Supplementary Fig. S1).

Gene expression profiling was performed to identify genes and/or pathways that distinguish spheroid-forming cells (n = 5) from the cell lines that do not have spheroid-forming capacity (n = 4). Of 22,277 probes on the Affymetrix U133 HTA gene expression microarray platform, 1,358 probes (1,066 named genes) were significantly increased in cell lines that have high spheroid-forming capacity under serum-free spheroid culture conditions (Supplementary Table S3), while 1,441 probes, representing 1,114 named genes, were significantly increased in cell lines with low potential to form spheroids (Supplementary Table S4). A heatmap of all 2,799 probes with significant differences in expression between the spheroid-forming and non–spheroid-forming groups is shown in Fig. 2B. Functional annotation clustering analysis of the 1,066 genes whose expression is increased in spheroid-forming cells showed the most significant enrichment for genes involved in mitochondrial function (137 genes, 11.6-fold enrichment; FDR = 2.3e⁻¹⁹) and DNA damage/repair (46 genes, 5.9-fold enrichment; FDR = 1.2e⁻⁵; Table 1; full list in Supplementary Table S5). For the 1,114 genes that are more highly expressed in the cells with low spheroid-forming capacity, metal binding (5.6-fold enrichment; 268 genes, P = 1.8e⁻¹⁰, FDR = 2.6e⁻⁷) and cell junctions (4.2-fold enrichment; 67 genes, P = 4.3e⁻⁷, FDR = 6.1e⁻⁴) were the most significant annotations (Table 1; full list in Supplementary Table S6).

Because mitochondrial pathway genes were identified as possible vulnerable targets for eradicating spheroids (Table 1), we tested the impact of oligomycin on spheroids. Oligomycin inhibits mitochondrial function and electron transport by blocking the proton channel (F₀ subunit) for ATP-synthase in complex V. Oligomycin treatment only modestly inhibited growth of both HEYA8 cells grown as monolayers and spheroids at 5 μg/mL but reduced cell survival 81.4% and 70.4%, respectively, at 25 μg/mL (P < 0.0001 for both; Fig. 3A), Cell–cell contact within spheroids was diminished at 5 μg/mL and spheroids were nearly completely dissociated at 25 μg/mL oligomycin (Fig. 3B).

Patients with platinum-resistant/refractory ovarian cancer have poor clinical prognosis. Platinum-resistant ovarian cancer tends to relapse within the first 6 months after the end of first-line chemotherapy (carboplatin and/or paclitaxel; ref. 23). We have shown that UCN-01 and oltipraz function preferentially on slowly proliferating cells in vitro. Previous studies have shown that slow-cycling subpopulations of quiescent cancer cells include cancer stem-like cells and contribute to chemoresistance and recurrence (24). To test our hypothesis that UCN-01 and oltipraz would be effective in treating platinum-resistant/refractory ovarian cancer, we employed a novel mouse model of carboplatin resistance.

We used in vivo live imaging to visualize tumor growth and regression in our mouse model. We first generated stable cell lines that express LUC and GFP in CAOV2 ovarian cancer cells. The LUC/GFP–positive cells (CAOV2-GFP/LUC) were enriched by flow cytometry for GFP-positive signals. Forty mice were injected intraperitoneally with 3.5 × 10⁵ CAOV2-GFP/LUC cells. After 7 days, 26 mice had visible tumors by in vivo live imaging with a mean of 3.2 × 10⁷ total photon flux. Six of these mice received no treatment (Arm 5). The remaining 20 of these tumor-bearing mice were treated once every four days with carboplatin at 80 mg/kg for a total of three rounds of treatment, until tumor signal was reduced to an average of 50% total photon flux (total photon flux from 3.2 × 10⁷ to 1.5 × 10⁷). The mice were then assigned to arms for the secondary treatment as indicated in Fig. 4A. The secondary treatment regimen is described in detail in Materials and Methods. The dosing regimen for each mouse is shown in Table 2 and the average total photon flux from each experimental arm is summarized in Supplementary Table S7. The mice receiving carboplatin followed by additional carboplatin (Arm 1) or no additional treatment (Arm 4) showed decreased tumor size from the primary carboplatin treatment, but tumor growth resumed after treatment was terminated (Fig. 4A; Supplementary Fig. S2). We found that mice experienced a reduction in tumor burden after three rounds of treatment with carboplatin, but this improvement was reversed with further carboplatin treatment due to apparent acquired resistance. Mice in the experimental arms that received carboplatin primary treatment followed by UCN-01 (Arm 2) or oltipraz (Arm 3) showed little to no evidence of tumor regrowth (Fig. 4). As expected, mice that received no treatment showed progressive tumor growth throughout the study and reached humane endpoints sooner than all other mice. Continued carboplatin treatment in Arm 1 showed some benefit relative to no treatment or no continued treatment (Supplementary Table S7, Arm 4). At necropsy, the majority of mice receiving secondary UCN-01 or oltipraz treatment were found to have no evidence of remaining disease, which confirmed our observations with live imaging. Similar to tumor distribution in ovarian cancer patients, most tumors were located on the peritoneum or omentum in the mice with detectable disease at necropsy, with fewer tumors attached to the fallopian tubes, ovaries, or intestines. Several mice in Arm 5 (no treatment) and Arm 4 (carboplatin followed by no treatment) developed malignant ascites. Overall, these data indicate that prolonged use of carboplatin in vivo can result in the development of chemoresistance, whereas drugs that target slow proliferating cancer cells (like UCN-01 and oltipraz) may delay disease recurrence.

Nearly inevitable relapse is a major problem in treating patients with advanced ovarian cancer (5). The concept of cancer cell dormancy has been put forward in hematologic malignancies, breast and prostate cancer as an explanation for disease recurrence (14, 25, 26). According to this postulate, dormancy provides a source of latent cancer cells that resist conventional chemotherapy, because such drugs are primarily effective against rapidly proliferating cancer cells (27). Patients with ovarian cancer do not typically receive treatment between the completion of their surgery and primary chemotherapy and the point at which a recurrence is detected. This leaves a time frame during which residual cancer cells could modify their genetic and/or epigenetic profiles to regrow when systemic and/or local microenvironment conditions become favorable (28). Several in vitro models of dormancy have been reported (29, 30), but the molecular features accompanying tumor cell dormancy are poorly understood (14, 25). We showed that ovarian cancer cells that grow in an anchorage-independent manner in serum-free media are able to enter into a dormant state and can exit this state as many as three months later following transfer into regular culture conditions. The cellular morphology of dormant cells was very different from the same cells grown in regular culture, and this change in phenotype may involve gene mutations and/or epigenetic modifications. The peritoneal cavity is a common site of disease recurrence in ovarian cancer (31). Our findings support the notion that a specialized subpopulation of ovarian cancer cells remain viable as dormant cells in the peritoneal cavity, possibly for years, and ultimately cause recurrent disease. The serum-free spheroid culture model can be employed to clarify the signaling pathways required for the survival or reactivation of dormant cancer cells, although a major limitation is that this model may not accurately reflect the state of these cells in vivo.

Whether or not dormant cancer cells possess CSLC features is currently a topic of debate (14, 32, 33). In breast cancer, a majority of disseminated tumor cells persist in a dormant state in the bone marrow and are distinguished by having a CD44⁺/CD24^−/low cell surface marker profile (32). We sought to enrich for ovarian CSLCs using the serum-free culture model reported previously (13), but failed to reproduce sustainable spheroids with self-renewal capability. This shortcoming may be because the cancer cells investigated did not contain CSLCs as discussed previously (11). CD44 is present on CSLCs of a number of cancer types, including breast, prostate, colon, and ovarian cancer (13, 34–37) as well as in cell lines (11, 38). In our study, CD44 expression was strongly and positively correlated with the capacity for anchorage-independent colony formation and the ability of spheroids to undergo passaging, although CD44 expression was not correlated with spheroid numbers. CD24, a cell surface marker that is absent on breast CSLCs, here discriminated the capacity for spheroid formation among the cancer cell lines. These data also suggest a relationship between CSCL marker expression and the ability to enter into a dormant state. It is well known that spheroid formation is determined by both intrinsic properties of cells and by the cellular microenvironment (39). Accordingly, we found that OVARY1847, OV90, PEO1, and OVCAR8 spheroid cells did not survive over a prolonged period in serum-free spheroid culture, while they could be expanded in regular media (Supplementary Table S2). These findings suggest that the microenvironment indeed plays an important role in the selection of a specific population of cancer cells that enter into cellular dormancy.

Drug resistance remains a major problem in the treatment of recurrent ovarian cancer. Spheroid models have been used primarily as an in vitro model for three-dimensional solid tumors to investigate mechanisms of drug resistance. Spheroid cells are known to become refractory to conventional chemotherapeutic agents, especially to paclitaxel when compared with monolayer cells (40). Therefore, eradication of spheroids within ascites may be quite clinically relevant to decrease the high rate of ovarian cancer recurrence. We previously reported that UCN-01 had higher efficacy against slow-proliferating and/or quiescent cancer cells (16), and that UCN-01 exhibits higher efficacy against ovarian cancer cells grown as spheroids. This finding is in stark contrast to the increased resistance exhibited by the spheroids to carboplatin and paclitaxel.

UCN-01 has been evaluated in clinical trials for various malignancies, including ovarian cancer (41–44). In the ovarian cancer trial, UCN-01 was delivered with topotecan in the recurrent setting, when tumor growth was already evident. To delay or prevent recurrent disease, it is likely important to deliver drugs like UCN-01 or oltipraz when the cells are in a slow-proliferating or dormant state, before the cells begin recurrent growth. This approach has not yet been formally tested in clinical trials with an aim of targeting the residual dormant cancer cell population. Targeting this residual cell population may be a relevant strategy for improving survival of women with high-grade invasive ovarian cancer.

Bioinformatic analyses of the genes showing differential expression between the spheroid-forming and non–spheroid-forming ovarian cancer cell lines showed that genes involved in mitochondria function were elevated and important to the survival of the spheroids in serum-free spheroid culture conditions. UCN-01 is an inhibitor of protein kinase C and cyclin-dependent kinases and is also known to induce mitochondrial damage and apoptosis (45, 46). In support of the microarray data analysis, another mitochondrial inhibitor, oligomycin, reduced cell survival in both monolayer cells and spheroids and also inhibited the ability of the cells in serum-free spheroid culture conditions to maintain their cell–cell contacts and spheroid phenotype. Although one limitation of our study is the potential for off-target effects of the inhibitors used, our results are compatible with another report demonstrating that metabolism-related genes are upregulated in compact spheroids rather than loose cellular aggregates (47). Such structural change induced by mitochondrial inhibitors may improve efficacy of conventional chemotherapeutic drugs by allowing their penetration and enhancing their distribution within spheroids. Because compounds toxic to mitochondria have been developed through clinical trials (48), it may be pertinent to investigate the role of the mitochondrial pathway in dormant cancer cells to determine targetability.

Our results from the xenograft mouse studies suggest that UCN-01 and oltipraz, and similar drugs that target slower proliferating cells, may have efficacy against residual ovarian cancer cells when delivered after completion of the primary chemotherapeutic regimen and prior to emergence of recurrent disease. Continued carboplatin treatment resulted in tumor regrowth, contrasting with the results from treating with UCN-01 and oltipraz, in which tumor cells remaining following completion of the carboplatin regimen were apparently not able to initiate growth in the mice, at least during the duration of the experiment. It is likely important that treatment to target the residual cancer cells commenced soon after the completion of the primary chemotherapy. This may be a critically important but underappreciated point because the numbers of residual cancer cells are at a nadir, and presumably more easily targeted. The prior phase I and II clinical trials in which UCN-01 was tested for efficacy against recurrent ovarian cancer initiated treatment after the recurrent disease was evident (49, 50). The conclusion was that although well tolerated, UCN-01 (and topotecan, given together in those studies) did not show antitumor activity. However, after the disease recurs, the cancer cells are again rapidly dividing, and according to our results here and those we have previously published (16), these actively proliferating cells would have shown resistance to UCN-01. This may explain the failure to achieve a response in those initial trials.

A limitation of our study is the use of select ovarian cancer cell lines that were chosen based on chemoresponsiveness and the ability to form compact spheroids. Verification of findings in additional in vitro and in vivo ovarian cancer model systems should be undertaken to confirm efficacy and generalizability.

In conclusion, our results support that serum-free spheroid culture can be used as a model to examine the mechanisms of cancer cell dormancy. Ovarian cancer spheroid formation is associated with expression of the cancer stem-like cell marker, CD44. The capacity for cells to form spheroids and enter into a dormant state is associated with mitochondrial gene functions, and UCN-01 and oligomycin, two drugs that are able to interfere with mitochondrial function, caused spheroid dissociation while spheroids were highly refractory to cisplatin and paclitaxel. Unlike continued tumor growth observed with prolonged treatment with carboplatin in our xenograft mouse model, UCN-01 and oltipraz were inhibitory to tumor regrowth subsequent to carboplatin treatment. Dormant cancer cells that survive initial chemotherapy are likely to be the source of recurrent disease, and we contend that these cells need to be targeted prior to the onset of recurrent disease. Our findings suggest that targeting aspects of mitochondrial function may be a reasonable strategy for decreasing the high recurrence rate of advanced stage ovarian cancers.

N. Matsumura reports grants from AstraZeneca and personal fees from Takara Bio (outside director) outside the submitted work. S.K. Murphy reports grants from the Department of Defense, Ovarian Cancer Research Program during the conduct of the study. T. Baba reports a grant from the Ovarian Cancer Research Fund. No disclosures were reported by the other authors.

Z. Huang: Data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. E. Kondoh: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing-original draft, writing-review and editing. Z.R. Visco: Data curation, formal analysis, methodology, writing-review and editing, data curation, formal analysis, methodology, writing-review and editing. T. Baba: Data curation, formal analysis, investigation, methodology, writing-review and editing. N. Matsumura: Formal analysis, validation, investigation, writing-review and editing. E. Dolan: Investigation, writing-review and editing. R.S. Whitaker: Resources, data curation, investigation, methodology. I. Konishi: Resources, writing-review and editing. S. Fujii: Resources, writing-review and editing. A. Berchuck: Conceptualization, resources, supervision, writing-review and editing. S.K. Murphy: Conceptualization, resources, formal analysis, supervision, funding acquisition, visualization, methodology, writing-original draft, project administration, writing-review and editing.

We gratefully acknowledge Dr. Seiichi Mori for providing antibodies for flow cytometry and Dr. Jeffrey R. Marks for assistance with micrography. This research was supported by the Department of Defense award W81XWH-11-1-0469 (to S.K. Murphy, Z. Huang, R.S. Whitaker, and A. Berchuck), the Gail Parkins Ovarian Cancer Awareness Fund (to A. Berchuck and S.K. Murphy), and the Ovarian Cancer Research Fund (to T. Baba).

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.