Overcoming Platinum Resistance in Preclinical Models of Ovarian Cancer Using the Neddylation Inhibitor MLN4924 Free

Epithelial ovarian cancer is the most lethal female reproductive malignancy, resulting in more than 22,000 new cases and 15,000 deaths in the United States each year (1). Approximately 85% of ovarian cancers have already metastasized beyond the ovary at the time of diagnosis and require adjuvant chemotherapy following surgical excision. Compared with other solid tumors, ovarian cancer is a relatively chemoresponsive disease, and the vast majority of advanced stage epithelial ovarian cancers initially respond to adjuvant platinum-based chemotherapy. However, the inevitable development of chemoresistant disease in the vast majority of patients remains a major obstacle against improving outcomes for patients with ovarian cancer. Thus, improving response and overcoming resistance to chemotherapy are of paramount importance in this disease.

NEDD8 (Neural precursor cell Expressed, Developmentally Downregulated 8) is a small ubiquitin-like molecule, involved in regulatory posttranslational modification of proteins. A new class of antineoplastic agents targeting the NEDD8 conjugation pathway (also referred to as neddylation) has been recently discovered (2). The first in class compound MLN4924 is a small-molecule inhibitor that selectively inactivates NEDD8-activating enzyme (NAE), the E1 involved in this pathway (3). Neddylation is a key regulatory mechanism required for the activation of the cullin RING ligase (CRL) family of E3 ubiquitin ligases. Hence, by inhibiting neddylation, MLN4924 antagonizes the ubiquitylation activity of the downstream CRLs and results in the abnormal accumulation of their substrates. These substrates include CDK inhibitors (p21 and p27), DNA replication licensing factor CDT1 (4), as well as many other proteins (5).

The objective of our investigation was to evaluate the activity of MLN4924 against chemoresistant and chemosensitive ovarian cancer cells. We found that MLN4924 as a single agent exhibits moderate activity against different histologic subtypes of ovarian cancer. However, the combination of MLN4924 with cisplatin or carboplatin resulted in synergistic cytotoxicity. The mechanisms for this synergy do not involve increased DNA re-replication or platinum-DNA adduct formation. Inhibition of CUL3 (but not other cullins) also had a platinum-sensitizing effect s uggesting that this pathway may be implicated in platinum-resistant ovarian cancer.

Following Institutional Review Board approval, deidentified primary human ovarian cancer cell lines of various histologic subtypes were established as described previously (6). SKOV3 and ES2 cell lines were purchased from a vendor that verifies their identity using genomic fingerprinting (American Type Culture Collection) and gown in McCoy's 5A media with 10% FBS. The platinum-resistant sublines, SKOV3-PR and ES2-PR, were created by sequentially treating the respective parental cell lines with escalating dose of cisplatin and allowing recovery between treatments (4-hour treatment followed by 2–3 weeks of recovery to simulated clinical chemotherapy administration). The SKOV3-PR25 and ES2-PR20 lines were subsequently exposed to 25 and 20 μmol/L cisplatin every 2 months, respectively, to ensure that the chemoresistant phenotype is maintained. Cell lines were tested at least once a year and found to be negative for mycoplasma infection.

Short-term viability was assayed using the water soluble tetrazolium assay (WST-1, Roche) according to the manufacturer's recommended protocol. Cells were assayed 24 or 72 hours following start of treatment. Combination index was calculated using CalcuSyn software (Biosoft.com) according to the methods of Chou and Talalay (7). Long-term effects on cell survival were investigated using a focus formation assay. Preliminary experiments were carried out to optimize cell density and drug concentration. Duration of drug exposure was kept constant at 24 hours after which cell were washed with PBS and incubated in complete media. Vehicle-treated cells were plated at lower density to avoid over-confluence. Media were changed every 2 to 3 days until cells were judged to be ready for staining (usually 10–12 days after plating). Foci were fixed and stained using a solution of 20% formaldehyde, 80% methanol, and 0.25% crystal violet, and rinsed with deionized water. Cell staining was captured with a high-resolution CCD camera (FluorChem, Alpha Innotech) and quantified using densitometry software (AlphaEaseFC, Alpha Innotech). An empty well from each plate was used for background subtraction. Comparison of single with combination drug effects was made using the Student t test and two-way ANOVA with P < 0.05 considered significant.

Preliminary experiments were carried out to determine optimal knockdown conditions. Cells were transfected 24 hours after plating with 15 nmol/L oligonucleotides (Dharmacon On Target Plus SmartPool, sequences available upon request) using RNAiMAX reagent (Invitrogen) in serum-free media. On the day after transfection, equal volume of media + 20% FBS was added to each well. Drug treatment was conducted 48 to 56 hours following siRNA. Duration of drug exposure was 24 hours after which the cells were washed with PBS and fed with complete media. Cell survival assays were then conducted as described above.

Cells were trypsinized and fixed with 70% ethanol overnight at 4°C. Following fixation, cells were stained in propidium iodide solution (0.05% NP-40, 50 ng of propidium iodide/mL, and 10 μg of RNase A/mL). The labeled cells were analyzed by a FACScalibur flow cytometer (Becton Dickinson) and FlowJo software (Tree Star). Experiments were carried out in triplicate and repeated at least once. Statistical comparisons were made using the Student t test.

The competitive ELISA was conducted as previously described (8, 9) with some modifications. ES2 and SKOV3 cells were treated with 2 μmol/L of cisplatin with or without 1 μmol/L of MLN4924 for 24 hours, following washing, cells were further incubated for 48 hours to allow DNA damage repair. Cells were washed twice with PBS and genomic DNA (test DNA) was isolated using Qiagen DNeasy Kit. Highly cisplatin-modified DNA (sink DNA) was prepared by incubating salmon sperm DNA with 1 mmol/L of cisplatin in vitro and used for coating plate. The highly cisplatin-modified sink DNA was denatured by boiling for 5 minutes and quick chilling on ice before equal amounts of DNA were coated in the wells of a 96-well ELISA plate in Pierce DNA Coating Solution. ELISA plates were blocked with 2% rabbit serum in PBS for 30 minutes at 37°C. Test (or standard) DNA was boiled and diluted (50 mmol/L NaCl, 50 mmol/L NaP, pH7.4). Sixty microliters of rat anti-cisplatin–modified DNA antibody (CP9/19, Abcam) diluted at 1:400,000 [0.2% bovine serum albumin (BSA), 0.2% Tween-20, 90 mmol/L NaCl in PBS) was mixed with 500 ng of test DNA in 60 μL. After preincubation for 30 minutes at 37°C, 80 μL of the antibody-DNA mixture was transferred to ELISA plates precoated with cisplatin-modified sink DNA and incubated for 1 hour at 37°C. Plates were then washed with PBST, and anti-rat horseradish peroxidase-conjugated antibody (1:2000 in 1% BSA, 0.2% Tween-20 in PBS) was added to measure the amount of anti-cisplatin-DNA antibody that bound to the sink DNA in each well. In this assay, the amount of antibody bound to the well, is inversely proportional to the quantity of cisplatin-adducts present in the test DNA used in the preincubation reaction. Following a 30-minute incubation at 37°C and PBST washes, Ultra TMB (3,3′,5,5′-tetramentylbenzidine, Pierce) was added and absorbance at 450 nm was measured. For quantification, serially diluted cisplatin-modified DNA were used to make a standard curve to calculate the relative amount of cisplatin-adducts in the test DNA. All samples were assayed in triplicate and repeated at least twice.

Protein extraction and immunoblotting were conducted as described previously (6). The following primary antibodies were used: p-CHK1, CUL4A, UBE2M, GAPDH (Cell Signaling), CHK1 (Novus Biologicals), CDT1, FANCD2 (Thermo-Fisher), CUL1 (Santa Cruz), CUL2 (Invitrogen), CUL3, CUL5, and CUL7 (Bethyl), beta actin (Sigma), CUL4B (Epitomics). Where indicated, quantitation of Western blot analyses was conducted using QuantiScan software version 3.0 (Biosoft.com).

Given the paucity of data on the effects of MLN4924 on ovarian cancer, we began by investigating the effect of MLN4924 as a single agent in both established and primary ovarian cancer cell lines. As previously shown, treatment with MLN4924 led to loss of NEDD8-conjugated (neddylated) cullins and inhibition of their ubiquitin ligase activity with resulting accumulation of cullin substrates (Fig. 1A). The SKOV3 and ES2 ovarian cancer cell lines were found to be much less sensitive to MLN4924 than some previously investigated cell lines such as HCT116 colon cancer cells (Fig. 1B). We also tested a panel of nine primary high-grade serous ovarian carcinoma cell lines and found good agreement between dose–response curves of the primary and established ovarian cancer cell lines (Fig. 1B).

MLN4924 has been shown to induce DNA re-replication and DNA damage by inhibiting the CRL4-CDT2 ubiquitin ligase and stabilizing the DNA replication licensing factor, CDT1 (10). Because inhibition of CRL4-CDT2 is toxic to ovarian cancer cells (data not shown), we hypothesized that combining MLN4924 with platinum drugs (most active chemotherapy against ovarian cancer) may result in additive efficacy. We began by testing a panel of ovarian cancer cell lines, including several primary lines established in our lab. This panel included high-grade serous and clear cell carcinoma lines as these represent the two most common histologic subtypes of ovarian cancer and are characterized by distinct molecular and clinical features (11). In contradistinction to these cancer lines, we also investigated two primary serous borderline tumor (also known as low malignant potential or LMP) cell lines. Serous borderline tumors are indolent, slow-growing ovarian neoplasms that are thought to represent the precursors for low-grade serous carcinomas (12). From a molecular standpoint, borderline and low-grade tumors are characterized with rare p53 mutations (which are ubiquitous in high-grade serous carcinomas), and approximately 50% harbor KRAS or BRAF mutations (12). High-grade serous ovarian cancers typically start out chemosensitive, but develop resistance over time. In contrast, low-grade serous cancers are much less chemosensitive (13), and serous borderline tumors which can metastasize and recur more than a decade after diagnosis are not considered to respond to chemotherapy.

Short-term cell viability was evaluated using the water soluble tetrazolium (WST)-1 cell viability assay. Compared with either drug alone, the combination of MLN4924 and cisplatin resulted in significantly lower viability in serous, clear cell, and borderline ovarian cancer cell lines (Fig. 1C and D). Similar results were obtained with the combination of carboplatin and MLN4924 (Fig. 1E). These results led us to conduct drug interaction studies according to the methods of Chou and Talalay (7). In this analysis, a combination index of 1.0 signifies additive effect, whereas values less than and more than 1.0 indicate synergy and antagonism, respectively. MLN4924 and cisplatin showed significant synergy in both SKOV3 cells with combination indices ranging between 0.2 and 0.8 for the majority of drug ratios tested (Fig. 1F). Synergy was also observed in OVT84, a primary clear cell ovarian carcinoma cell lines (Fig. 1F) and ES2 cells (data not shown).

The fact that the synergistic effect of platinum and MLN4924 was observed not only in the high-grade serous and clear cell ovarian cancer lines but also in the slow-growing serous borderline tumors (OVT50 and OVT80 in Fig. 1C and D) was intriguing. This suggested that the synergistic action of MLN4924 and cisplatin was not dependent on growth rate or p53 mutations. As noted above, serous borderline tumors are also inherently much more chemoresistant than high-grade serous carcinomas. Their response led us to hypothesize that the combination of MLN4924 and platinum may also show efficacy against ovarian cancer cell lines with acquired chemoresistance. We tested this hypothesis in 2 in vitro models of chemoresistant ovarian cancer. First, we used primary high-grade serous ovarian cancer cell lines derived from patients who had persistent tumor after 4 cycles of neoadjuvant carboplatin and paclitaxel chemotherapy. Chemotherapy produces logarithmic killing of chemosensitive cancer cells, thus, residual tumor following chemotherapy is composed of a chemoresistant cancer cell population. Figure 2A shows the results of dose–response studies in two chemoresistant primary ovarian cancer cell lines, OVT100 and OVT106. Addition of MLN4924 to cisplatin resulted in the marked sensitization of both cell lines to the cytotoxic effects of cisplatin. Similar results were observed when carboplatin was used instead of cisplatin (data not shown).

To test the potential platinum-sensitizing effect of MLN4924 in an isogenic model of chemoresistant cells, we generated platinum-resistant sublines of SKOV3 and ES2 cells. These resistant cell lines were designated as SKOV3-PR and ES2-PR, respectively, followed by the highest concentration (μmol/L) of cisplatin, the cells were able to tolerate (e.g., SKOV3-PR25). The cytotoxicity of cisplatin and MLN4924 both alone and in combination was assessed using a focus formation cell survival assays (Fig. 2B). As expected, SKOV3-PR25 and ES2-PR20 cells showed 2- to 3-fold enhanced survival following cisplatin treatment. However, the combination of cisplatin with MLN4924 sensitized these cells, confirming the synergistic effect observed in the parental cell lines and the chemoresistant primary ovarian cancer cells. Thus, in both models of chemoresistant ovarian cancer, the addition of MLN4924 to platinum was able to sensitize the cells and result in significant cytotoxicity. The longer term focus formation assays also revealed that single-agent MLN4924 had a significant cytotoxic effect by itself which was not apparent in the short-term cell viability assay (WST-1) results (Fig. 2B and C vs. Fig. 1C and D), suggesting delayed cytotoxic effects (see the discussion of time course below). Furthermore, the ES2-PR20 cells were more sensitive to the effects of MLN4924 by itself compared with the parental ES2 cells (Fig. 2B), suggesting that cullins or other neddylation-sensitive pathways may be implicated in the development of platinum resistance in this model.

The kinetics of NAE inhibition by MLN4924 are rapid, and accumulation of CRL substrates is readily detectable at 1 to 4 hours posttreatment (3). We, therefore, investigated whether shorter exposure to MLN4924 either alone or prior, during, or after cisplatin treatment may be equally as efficacious as coadministration with cisplatin for 24 hours. As shown in Fig. 2C, short-term (4 hours) treatment with MLN4924 was no different than the vehicle control, when used as a single agent. Furthermore, neither pre- nor posttreatment with 4 hours of MLN4924 appreciably altered the cytotoxicity of cisplatin (Fig. 2C). Thus long-term (24 hours) treatment with MLN4924 is required for both single-agent effect and synergy with cisplatin, at least in this model system.

Published work by our group and other investigators had revealed that in HCT116 colon cancer cells, the toxicity of MLN4924 is, in large part, due to induction of inappropriate DNA replication (DNA re-replication; refs. 3, 10). One possible explanation for the enhanced cytotoxicity of the MLN4924-platinum combination would be if this combination increased the DNA re-replication induced by MLN4924 alone. Consistent with previous reports, MLN4924 induced DNA re-replication (as evidenced by >4N DNA content) by 72 hours posttreatment in SKOV3 cells (Fig. 3A). Cisplatin treatment resulted in a G₁/early S phase block, which was also the dominant effect in the combination group (Fig. 3A). Similar findings were observed in the ES2 cells (data not shown). Thus, the synergistic action of MLN4924 and cisplatin is not due to enhanced DNA re-replication.

The fluorescence-activated cell sorting (FACS) analysis also revealed sub-G₁ peaks in the both the single-agent cisplatin and MLN4924, and the combination-treated cells consistent with apoptosis that was detectable 72 hours after treatment initiation. To better understand the nature and timing of both DNA re-replication and apoptosis induced by MLN4924 and cisplatin, we investigated the protein levels of CDT1 (marker for MLN4924 effect and abnormal replication), γ-H2AX (marker for DNA damage), and cleaved PARP (marker for apoptosis) over a 96-hour period. Treatment with MLN4924, both alone and in combination with cisplatin, resulted in early abnormal CDT1 accumulation during the first 24 hours after treatment initiation (Fig. 3B and Supplementary Fig. S1). This was followed by DNA damage as evidenced by γ-H2AX accumulation, with a notable increase between 24 hours and 48 hours. Apoptosis was most evident 48 to 96 hours posttreatment initiation in both the MLN4924 and cisplatin alone groups. Furthermore, the combination treatment resulted in enhancement of DNA damage and accelerated apoptosis (Fig. 3B and Supplementary Fig. S2).

The principal mechanism of platinum-induced cytotoxicity is the formation of DNA-platinum adducts resulting in intra- and interstrand cross-links. The cell's proficiency at repairing these cross-links determines whether cell cycle is restored or apoptosis is induced (14). Thus, increased apoptosis in the MLN4924-platinum combination group may result from increased DNA-platinum adduct formation or interference with DNA adduct repair processes. To test this possibility, we measured DNA-platinum adduct levels using a competitive ELISA in SKOV3 and ES2 cells treated with cisplatin and MLN4924 alone or in combination (Fig. 3C). As expected, platinum-DNA adduct levels in MLN4924 only treated cells were at background levels. Cisplatin treatment for 24 hours resulted in the robust induction of adducts in both cell lines. Forty-eight hours after cisplatin had been removed (72 hours after start of treatment), adduct levels decreased by approximately 50% indicating DNA damage repair. Addition of MLN4924 had little effect on the maximal quantity or the decrease in platinum-DNA adducts (Fig. 3C). These results suggest that MLN4924's effect on cisplatin toxicity is not mediated through increased cisplatin-DNA adduct formation or impaired DNA adduct repair.

Given the lack of detectable differences in DNA adduct formation between cisplatin alone and cisplatin plus MLN4924-treated cell, we investigated possible differences in downstream checkpoint activation. In addition, the Fanconi anemia pathway is instrumental to DNA cross-link repair (15), and MLN4924 was recently reported to suppress FANCD2 monoubiquitination induced by cisplatin (16). We, therefore, also investigated FANCD2 monoubiquitination and CHK1 activation at 4- and 24-hour time points in our system. In both SKOV3 and ES2 cells, FANCD2 monoubiquitination increased between 4-hour and 24-hour time points in all treatment groups (Fig. 3D and E). Checkpoint activation, as detected by phosphorylated CHK1 (Ser345), was detectable following both MLN4924 and cisplatin treatment in ES2 (Fig. 3D and E) and SKOV3 cells (data not shown). Thus, MLN4924 did not appreciably interfere with FANCD2 monoubiquitination or checkpoint activation under treatment conditions where it significantly augmented cisplatin-induced DNA damage and apoptosis.

Cullin neddylation is a key activating modification of the CRL class of E3 ubiquitin ligases. As such, CRLs are the best known downstream effectors of MLN4924, and many of the effects of MLN4924 have been shown to be mediated by the inhibition of specific CRLs. We hypothesized that the synergy between MLN4924 and platinum may be mediated, at least in part, through one or more of the CRLs. The cullin family consists of 7 major members: CUL1, CUL2, CUL3, CUL4A, CUL4B, CUL5, and CUL7 (17). All have been shown to undergo neddylation except for CUL7 (18). Using an siRNA screening approach, we investigated whether the inhibition of cullins 1–7 affected cisplatin-induced cytotoxicity in SKOV3 and ES2 cells. Cell viability following treatment with cisplatin and cullin-targeting siRNA was tested using focus formation assays and normalized to the effect of each cullin's siRNA alone (Fig. 4A and B). Knockdown of CUL3 produced the greatest enhancement of cisplatin toxicity, reducing cell survival by approximately 35% to 45%. The knockdown of CUL5 was associated with a more modest 25% decrease in SKOV3 cells but not in ES2 cells. In contrast, depletion of CUL1 and several other cullins showed an antagonistic effect on cisplatin-induced cell death (Fig. 4B). Of note, knockdown of CUL7, which is not neddylated and therefore not subject to MLN4924 inhibition, did not appreciably alter cisplatin-induced toxicity in SKOV3 cells and antagonized cisplatin effect in ES2 cells (Fig. 4B). Finally, to investigate the possibility that MLN4924's enhancement of cisplatin cytotoxicity reflects the cumulative inhibition of other CRLs in addition to CRL3 and CRL5, we also conducted siRNA against UBE2M (aka UBC12). UBE2M is the E2 enzyme in the NEDD8 conjugation pathway and transfers the activated NEDD8 to cullins 1–4 (19, 20). As expected, knockdown of UBE2M phenocopied MLN4924, resulting in approximately 45% reduction in cell viability following cisplatin treatment (Fig. 4B). Notably this enhancement of cisplatin toxicity is in the same range as that observed with siCUL3 (35%–40%). Together these results support a model in which CRL3 inhibition may represent a major contributing mechanism underlying the synergy between MLN4924 and cisplatin.

Our investigation highlights a synergic cytotoxic effect between cisplatin and MLN4924 in both platinum-sensitive and platinum-resistant ovarian cancer cells. These findings are consistent with the recent reports of MLN4924 sensitizing cells to interstrand cross-linking agents (16) and ionizing radiation (21, 22). In contrast with a previous report (16), however, we did not observe any significant inhibition of FANCD2 monoubiquitination or abrogation of CHK1 Ser345 phosphorylation under experimental conditions that resulted in significant DNA damage and cooperative cell death. This discrepancy may be related to differences in cell models, time course, or doses of the drugs used.

Investigations of the mechanisms of antineoplastic action of MLN4924 have implicated a number of CRL substrates in mediating the cytotoxic effects of this drug (23). However, given that inhibition of CRL pathways by MLN4924 can result in the alteration of as many as 20% of all proteasome-regulated proteins (3), we focused our mechanistic studies more upstream on the CRLs themselves. This approach was instructive, in that we observed that inhibition of CRL3 enhanced the cytotoxic effects of cisplatin in both SKOV3 and ES2 cells. Interestingly, we also uncovered that knockdown of CUL1 and other cullins can result in resistance to platinum-induced cell death (in a cell-specific manner), which to our knowledge has not been previously reported. This highlights the fact that while the net effect of MLN4924 in our preclinical investigation is the enhancement of cisplatin toxicity, the biologic contribution of the different CRL pathways and the role of their respective substrates can be variable and at times oppositional.

The implication of CRL3, in particular, is interesting because of the links to several pathways that could influence cell survival. CUL3 interacts with Kelch-like erythroid cell-associated protein-1 (KEAP1) and the RING domain protein RBX1 to form the CRL3-KEAP1 ubiquitin ligase. Substrates for CRL3-KEAP1 include NRF2, a key transactivator of the antioxidant response; IKKβ, a positive regulator of NF-κB; and the BCL2 and BCL-XL antiapoptotic proteins (24). Although it is tempting to speculate how alterations in these or other substrates may regulate sensitivity to cisplatin, as noted for the inhibition of individual cullins, substrate stabilization alone may be correlative rather than causative, and different substrates may have opposing effects. It is also important to note that identification of CRL substrates is an active area of research, and the above list of potential substrates and corresponding pathways is far from complete.

Finally, we showed a synergistic effect between MLN4924 and cisplatin in several in vitro models of platinum/chemoresistant ovarian cancer. Naturally, this leads us to hypothesize that the addition of MLN4924 to platinum-based chemotherapy may increase its efficacy and/or mitigate the invariable development of platinum resistant in patients with ovarian cancer. However, this hypothesis requires rigorous testing in additional preclinical studies and clinical trials.

No potential conflicts of interest were disclosed.

Conception and design: A.A Jazaeri, P.G. Smith, A. Dutta

Development of methodology: A.A Jazaeri

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.A Jazaeri, J. Park, J.L. Bryant

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.A Jazaeri, M.R. Conaway, S.C. Modesitt, M.A. Milhollen, A. Dutta

Writing, review, and/or revision of the manuscript: A.A Jazaeri, S.C. Modesitt, A.J. Berger, A. Dutta

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.A Jazaeri, E. Shibata, J.L. Bryant, A.J. Berger

Study supervision: A.A Jazaeri, P.G. Smith, A. Dutta

The authors thank the members of the Division of Gynecologic Oncology and the Tissue Procurement Facility at the University of Virginia.

This work was financially supported by the American Cancer Society Mentored Research Scholar Grant, MRSG-09-035-01-CCE, University of Virginia's Cancer Center (Women's Oncology Program), and Department of Obstetrics & Gynecology (to A.A. Jazaeri), and a grant from the NIH, R01-CA60499 (to A. Dutta).

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.