Amplicon-Dependent CCNE1 Expression Is Critical for Clonogenic Survival after Cisplatin Treatment and Is Correlated with 20q11 Gain in Ovarian Cancer

Genomic amplification of 19q12 occurs in several cancer types including ovarian cancer where it is associated with primary treatment failure. We systematically attenuated expression of genes within the minimally defined 19q12 region in ovarian cell lines using short-interfering RNAs (siRNA) to identify driver oncogene(s) within the amplicon. Knockdown of CCNE1 resulted in G1/S phase arrest, reduced cell viability and apoptosis only in amplification-carrying cells. Although CCNE1 knockdown increased cisplatin resistance in short-term assays, clonogenic survival was inhibited after treatment. Gain of 20q11 was highly correlated with 19q12 amplification and spanned a 2.5 Mb region including TPX2, a centromeric protein required for mitotic spindle function. Expression of TPX2 was highly correlated with gene amplification and with CCNE1 expression in primary tumors. siRNA inhibition of TPX2 reduced cell viability but this effect was not amplicon-dependent. These findings demonstrate that CCNE1 is a key driver in the 19q12 amplicon required for survival and clonogenicity in cells with locus amplification. Co-amplification at 19q12 and 20q11 implies the presence of a cooperative mutational network. These observations have implications for the application of targeted therapies in CCNE1 dependent ovarian cancers.


Introduction
Advanced stage serous tumors account for the majority of invasive ovarian cancers and despite a generally good initial response to cytoreductive surgery and platinum-based chemotherapy, most women face a high risk of recurrence and poor longterm survival [1]. Platinum-based agents, such as cisplatin and carboplatin, are toxic to dividing cells due to the formation of DNA adducts that result in double strand breaks, activating DNA damage-mediated apoptotic signals [2]. Response to chemotherapy is, however, difficult to predict and there are currently no predictive biomarkers for serous ovarian cancers in clinical use. We have previously mapped a region of 19q12 amplification associated with treatment-resistant serous ovarian tumors by performing a genome-wide survey of copy number change [3]. These findings were consistent with previous reports of amplification being associated with poor overall survival [4,5]. Similarly, recurrent amplification of 19q12 has been reported in a variety of cancers including esophageal [6], gastric [7], lung [8] and endometrial tumors [9].
The 19q12 amplification is a high-level focal amplification that targets a cluster of only several genes on chromosome 19. CCNE1 (Cyclin E) has previously been suggested as the target of amplification in ovarian cancer [4, 10,11], however a systematic analysis of known genes within the amplicon has not been performed. Furthermore, whilst CCNE1 amplification likely provides an oncogenic stimulus through activation of the cell cycle, it is not obvious how it may contribute to primary chemotherapy resistance. For example, over-expression of CCNE1 in vitro renders ovarian cancer cells more sensitive to platinum agents, presumably due to increased proliferation [12]. It is possible that the biological consequence of 19q12 amplification is not limited to over-expression of CCNE1, and that other genes in the amplicon contribute to tumor growth or progression. Furthermore, other co-existing mutational events elsewhere in the cancer genome may cooperate or enhance the oncogenic effect of CCNE1 over-expression.
We performed an siRNA knockdown screen of all annotated genes within and immediately flanking the 19q12 amplicon in ovarian cancer cell lines with or without regional amplification. We found CCNE1 to be the only gene target within the amplicon that reduced cell viability in the amplicon-containing OVCAR-3 cell line after siRNA knockdown. CCNE1 knockdown induced cell cycle arrest and apoptosis, while also impairing clonogenic survival after cisplatin treatment, despite increasing in vitro drug resistance in a short-term cytotoxicity assay. In a disease setting, these results suggest that treatment failure in CCNE1 amplified tumors may relate to rapid repopulation of the tumor after chemotherapy and not cellular drug resistance specifically. We also found TPX2 amplification and over-expression to be significantly correlated with CCNE1 copy number status implying the presence of a cooperative mutational network between these genes.

Focal amplification of 19q12 is common to various tumor types
We first sought to compare the minimal region of chromosomal gain at 19q12 across multiple tumor types. We obtained data from SNP-based high-resolution copy number studies including 15 tumor types [9,13] for a comparison with our findings [3] ( Figure 1A). Minimal targeted regions of amplification were defined by GISTIC, an analysis tool that assesses the statistical significance of copy number events based on frequency and amplitude [14]. Significant amplification of 19q12 was present in one third of the cancer types analyzed. Of the tumor types with 19q12 amplification, approximately 25% of individual samples showed copy number gain, except for endometrial tumors where a higher frequency was observed (,45%) [9]. Minimal amplicon boundaries were found to target a region less than 2 Mb in size, centered at approximately at 35.0 Mb on chromosome 19. In both ovarian tumor data sets analyzed [3,13] the minimal mapped region of gain incorporated the same five genes (POP4, PLEKHF1, C19orf12, CCNE1 and C19orf2), with similar overlapping regions detected in endometrial and breast tumors. In contrast, the minimal region mapped in non-small cell lung tumors incorporated only CCNE1 while a broad region was mapped in esophageal tumors (,9.5 Mb), spanning 110 annotated genes. Copy number  [8,13,40,41,42]; 2 ovarian [3], 3 [13]; 4 endometrial [9]; 5 breast [13,43,44]; and 6 esophageal tumors [42]. Frequency of occurrence and genes present within peak boundaries indicated. change of the 19q12 locus showed a degree of tumor specificity, in that the amplification was not seen in 10 other tumor types for which substantial data was available, including small cell lung, hepatocellular, colorectal and prostate cancer (data not shown). We also note that amplification of 19q12 has been identified by cDNA array-CGH analysis of gastric tumors [15,16], however this tumor type was not included as our analysis was limited to highresolution SNP copy number data.
To identify cell lines that were representative of primary tumors for functional studies we analyzed high-resolution SNP copy number data for 22 ovarian cancer cell lines at chromosome 19q12 (Sanger Cancer Genome Project Archive) and identified seven cell lines (OVCAR-3, OVCAR-4, Kuramochi, RMG-I, Caov-4, EFO-21, OVCAR-8) that had overlapping amplification at 19q12 ( Figure 1B and Figure S1). Of the seven cell lines, OVCAR-3 contained a focal, high-level amplification that best recapitulated data from primary ovarian tumors ( Figure 1C). Quantitative-PCR (qPCR) demonstrated that the five genes within the region of high-level amplification in OVCAR-3 (POP4, PLEKHF1, C19orf12, CCNE1 and C19orf2), but not flanking genes (UQCRFS1 and ZNF536), were over-expressed relative to the SK-OV-3 control cell line lacking 19q12 amplification ( Figure 1D). Of the five genes, PLEKHF1 and CCNE1 showed the highest expression. The 19q12 amplicon can therefore be mapped to a region spanning 34.4-35.4 of chromosome 19 and involving 5 annotated genes, each of which is over-expressed in OVCAR-3.
Cells lines with amplification at 19q12 are specifically sensitive to CCNE1 knockdown Short interfering RNA (siRNA) were used to knockdown the expression of the seven genes in and adjacent to the high-level 19q12 amplicon in OVCAR-3 and SK-OV-3 plus GAPDH and non-silencing (NS) controls. A schematic of the experimental design used for the combined siRNA knockdown strategy and subsequent drug treatment protocol is shown in Figure 2A. Transcript levels for all targeted genes were efficiently and specifically reduced up to 96 hours after siRNA transfection, with the exception of ZNF536, which flanked the region of amplification ( Figure 2B). By interrogating data obtained from an earlier study [17] we found ZNF536 expression to be low or absent in primary serous ovarian tumors, which may explain an unobservable reduction in gene expression (data not shown). Gene expression was also monitored in the presence of cisplatin, in preparation for functional experiments. PLEKHF1, CCNE1, C19orf2 and ZNF536 were slightly up-regulated by cisplatin treatment in one or both cell lines however siRNA knockdown was still effective ( Figure 2B).
Of the genes tested, only CCNE1 knockdown showed a significant reduction in cell viability in OVCAR-3 (to approximately 60% of NS control cells; p,0.01; Figure 2C). CCNE1 knockdown had no effect on SK-OV-3 cells. Given its role in G1/S transition, we expected that depletion of Cyclin E1 protein would induce G1 arrest (see below) and result in an increase of apoptotic cells. TUNEL staining of untreated cells was comparable (SK-OV-3, 0.1%; OVCAR-3, 0.2%) ( Figure 2D), while CCNE1 knockdown resulted in a significant increase in apoptosis only in OVCAR-3 (p,0.0001). Reduction in protein abundance was also validated in both cell lines after gene knockdown ( Figure 2E).
Having identified specific sensitivity of OVCAR-3 to CCNE1 knockdown, we aimed to validate this finding in independent cell lines and determine whether the observed effect was amplicon dependent. We therefore broadened the analysis to include an additional three cell lines with amplification at 19q12 (OVCAR-4, Kuramochi and OVCAR-8) and a further unamplified line (IGROV-1). A statistically significant correlation between the copy number and gene expression of CCNE1 was found across all lines. However, OVCAR-8 did not show increased CCNE1 expression relative to gene amplification ( Figure S2 A). CCNE1 expression and Cyclin E1 protein levels were efficiently reduced in each line relative to base-line expression by siRNA-mediated knockdown ( Figure S2 B and C). As observed in OVCAR-3, CCNE1 knockdown specifically reduced viability in the additional lines with 19q12 amplification, and only marginally in the unamplified line IGROV-1 ( Figure 2F). Thus ovarian tumor cells with amplification at 19q12 are specifically sensitive to depletion of Cyclin E1, compared with unamplified lines.

CCNE1 knockdown reduces acute sensitivity to cisplatin
Given the association of 19q12 amplification with primary treatment failure [3] and poor outcome [4,5] we sought to explore the effect of gene knockdown on drug sensitivity. Although our analysis had identified a specific dependency on CCNE1 in amplified lines, we first re-assessed all seven amplicon-associated genes for impact on chemotherapy response in OVCAR-3 and SK-OV-3 using a 72-hour cytotoxicity assay. Cells were treated at slightly above a pre-determined IC50 (see Methods S1) and viability measured. Knockdown of genes within the amplicon did not significantly impact on cisplatin sensitivity of either cell line. Unexpectedly, cisplatin-induced cytotoxicity in OVCAR-3 treated cells was attenuated by CCNE1 inhibition ( Figure 3A). We performed a dose-response analysis to characterize further the effect of cisplatin treatment after CCNE1 knockdown. A statistically significant shift in the dose-response curve was observed in OVCAR-3 (p,0.05; Figure 3B) but not SK-OV-3 further demonstrating resistance of OVCAR-3 to cisplatin upon CCNE1 inhibition. Consistent with this finding, cisplatin had no differential effect on cell viability in CCNE1 knockdown cells as compared with siRNA controls in two of the three additional 19q12 amplified lines (Kuramochi and OVCAR-4) ( Figure 3C). In contrast, gene knockdown enhanced the effect of cisplatin on cell viability in both cell lines with low baseline CCNE1 expression (OVCAR-8 and the 19q12 unamplified control line IGROV-1).
To investigate the cisplatin resistant phenotype observed in OVCAR-3 cells with CCNE1 knockdown, we used flow cytometry to analyze cell cycle distribution following cisplatin treatment. Cisplatin treatment of OVCAR-3 resulted in a prolonged S-phase ( Figure 4A) whereas SK-OV-3 cells arrested predominately in the G2 phase of the cell cycle ( Figure 4B). Consistent with the requirement for Cyclin E1 in the G1/S transition [18], CCNE1 siRNA knockdown induced G1 arrest in OVCAR-3, most evident in the presence of cisplatin ( Figure 4A). By contrast, the cell cycle distribution of SK-OV-3 was unaltered by inhibition of CCNE1 expression with or without cisplatin ( Figure 4B). These observations were consistent in CCNE1 amplified Kuramochi and OVCAR-4 cells ( Figure S3). Partial G1 arrest was also observed in 19q12 unamplified IGROV-1 cells after CCNE1 knockdown, however only in response to cisplatin treatment. As observed in the viability assay, the CCNE1 amplified but low-expressing OVCAR-8 cells behaved similarly to control lines. We therefore concluded that the dependency of OVCAR-3, Kuramochi, and OVCAR-4 on high CCNE1 expression resulted in a cell cycle arrest after gene knockdown, including in the presence of cisplatin, most likely accounting for the apparent cisplatin resistance observed in shortterm viability assays.
To understand the longer-term impact of CCNE1 depletion on cell survival after cisplatin treatment we assayed the clonogenicity of lines with and without amplification of the 19q12 locus (see schematic Figure 5A). CCNE1 knockdown profoundly reduced the clonogenic capacity of OVCAR-3 but not SK-OV-3 in the absence of drug ( Figure 5B and 5C). In contrast to increased resistance to cisplatin in short-term cytotoxicity assays, CCNE1 attenuation reduced clonogenic survival of OVCAR-3 cells after cisplatin treatment ( Figure 5B). We did not explore the in vivo effects of CCNE1 knockdown in ovarian tumor lines with 19q12 amplification, as we were unable to generate viable lines with stable lentivirus integration of short hairpin RNA (shRNA) directed to CCNE1 (data not shown).
CCNE1 amplification is predictive of poor outcome in primary tumors CCNE1 copy number was measured in 43 primary ovarian tumors from patients with advanced-stage, serous invasive disease. In addition, we included data from 52 tumors from our previous genomic analysis of platinum-resistance in ovarian cancer [3] and obtained matching gene expression data for all samples [17]. All patients underwent primary surgery followed by platinum-based chemotherapy. Clinical information used to correlate CCNE1 Patients were stratified based on CCNE1 copy number status as assessed by qPCR (see Materials and Methods). No difference was noted between clinical characteristics of each group apart from age, with younger patients over-represented in the CCNE1 unamplified group (Table 1 and Figure 6A). CCNE1 gene expression showed a strong correlation with copy number ( Figure 6B) and both were correlated with progression-free survival (PFS) ( Figure 6C & D). CCNE1 copy number, but not gene expression, was also associated with overall-survival (OS) ( Figure 6E & F). The most significant correlation observed was degree of CCNE1 gain and PFS ( Figure 6C), such that all cases with high-level amplification showed progressive disease within approximately twelve months from diagnosis (mean PFS of 10.7 months; Table 1).

Amplification of 19q12 is correlated with gain at 20q11
Having identified CCNE1 as a critical driver within the 19q12 amplicon in ovarian cancer, we reasoned that other mutations elsewhere in the genome might interact with Cyclin E1 or be associated with drug resistance. For example, mutations that enhance the effect of or allow tumors to tolerate CCNE1 overexpression may co-occur with 19q12 gain. We obtained both SNPbased copy number and gene expression data on 157 high-grade serous invasive tumors from The Cancer Genome Atlas Project (TCGA) for analysis. Firstly, we examined gene expression of genes whose protein products are required for processing of Cyclin E1 to active low molecular weight forms (ELA2 and CAPN2) [19,20] or its degradation (FBXW7) [18]. However, no statistically significant positive correlation between candidate gene expression and CCNE1 status was observed (data not shown). We then correlated 19q12 gain with all other gains and losses within 0.1 Mb segments of the genome (see Methods S1). The top three correlated regions of copy number change were on 20q11, 1p36 and 6q27 ( Figure 7A). The most significant associated gain was localized to a 2.5Mb region on chromosome 20 ( Figure 7B) and has been validated in a separate unbiased analysis of genome-wide correlations of gain and loss in ovarian tumors [21].
In order to narrow gene candidates within the three regions most likely to interact with CCNE1, we next correlated the expression of genes within each region with CCNE1 expression ( Table 2). Expression of TPX2 was most significantly associated with CCNE1 expression, and was also correlated to its own amplification status ( Figure 7C). The relationship between CCNE1  and TPX2 gene expression was further validated in a second independent data set ( Figure 7D). The strong association between TPX2 and CCNE1 amplification and expression was intriguing given that TPX2 is a centromeric protein required for mitotic spindle function during cell division [22].

20q11 amplification renders cells resistant to TPX2 knockdown
To test whether there was a functional dependence on TPX2 in cell lines with amplification of the 20q11 locus and whether it interacts with CCNE1, we assessed TPX2 status in lines used for our knockdown experiments. TPX2 was amplified ( Figure 8A) and over-expressed ( Figure 8B) in all three CCNE1 amplified and overexpressing cell lines (OVCAR-3, OVCAR-4 and Kuramochi) compared to the control line, SK-OV-3. In addition, we further identified OAW-28 as having high-level 20q11 amplification ( Figure 8A, data from Sanger Cancer Genome Project Archive) and the highest level of gene expression across the tested cell lines ( Figure 8B).
In contrast to the close relationship between CCNE1 amplification and cellular sensitivity to gene knockdown, an inverse relationship was observed between TPX2 copy number and the effects of TPX2 siRNA ( Figure 8C). For example, SK-OV-3 cells with low gene expression ( Figure 8B) and minimal detectable TPX2 protein ( Figure 8D), were highly sensitive to gene knockdown, whereas OAW-28 cells were essentially resistant. RT-PCR ( Figure S4) and western blot ( Figure 8D) analyses demonstrated efficient knockdown of TPX2, however protein was still detectable in some cell lines that initially had high levels of TPX2 protein. We also observed that knockdown of CCNE1 resulted in diminished TPX2 gene expression in 19q12 amplified lines ( Figure S4 A) suggestive that TPX2 expression is affected downstream of Cyclin E1.
Given the correlation between CCNE1 and TPX2 amplification, we also examined whether concurrent knockdown of both genes would further diminish viability in cell lines containing both amplifications ( Figure 8C). After we allowed for a competitive effect of combining siRNAs (see Methods) no obvious interaction was observed with simultaneous knockdown of TPX2 and CCNE1. Knockdown of CCNE1 had minimal effect on OAW-28 cell viability despite efficient protein reduction in double knockdown experiments ( Figure 8D). This finding is consistent with our initial results, as OAW-28 cells do not have 19q12 amplification (see Figure S1 for OAW-28 copy number at this locus).

Discussion
We performed the first systematic siRNA knockdown of all genes within the minimally defined 19q12 amplicon in ovarian cancer showing that CCNE1 is the key oncogenic target. Given known roles of Cyclin E1 in cancer, including de-regulation of the cell cycle and promoting genomic instability, it was the likely driver of the 19q12 locus, however other genes had not been excluded. For example, C19orf2, which is immediately adjacent to CCNE1, has recently been annotated to encode URI, an unconventional prefoldin protein. Studies of the C. elegans URI homologue suggest an involvement in chromatin remodeling, preventing and/or repairing endogenous genotoxic DNA damage and maintenance of genome integrity [23]. More recently, URI has been identified as a key inhibitor of protein phosphatase 1c (PP1c) and is involved in regulation of the mTOR/S6K1 survival pathway based on nutrient and growth factor availability [24]. Despite these intriguing biological associations, we found no evidence of URI as a driver of the 19q12 locus. Reduction in cell viability after CCNE1 knockdown was specific to cell lines with 19q12 amplification, with limited or no effect in non-amplified lines, indicating an 'addiction' [25] to CCNE1 deregulation. Our findings validate a recent report showing amplification-specific sensitivity to CCNE1 attenuation in OV-CAR-3 and IOSE-29 cells [26]. Unamplified lines appeared to bypass siRNA mediated G1/S checkpoint arrest and apoptosis, possibly reflecting CCNE1 independent mechanisms of cell cycle de-regulation and distinct oncogenic processes. Interestingly, increased expression of CCNE1 has recently been shown in serous tubal intraepithelial carcinoma (STIC), a proposed precursor for high-grade serous carcinoma [27]. This finding reinforces the significance of CCNE1 de-regulation in ovarian cancer and suggests it is an early requirement in tumor evolution.
19q12 amplification is strongly associated with primary treatment failure in ovarian tumors [3] and is therefore both a potential prognostic marker and therapeutic target. Having identified CCNE1 as the driver of 19q12 amplification we sought to understand how it contributes to primary treatment failure. In short-term cytotoxicity assays, over-expression of Cyclin E1 increases sensitivity to cisplatin [12]. Consistent with this data, we found that siRNA-mediated knockdown of CCNE1 induced G1 arrest in 19q12-amplified cell lines, possibly protecting cells from cisplatin damage occurring through cell cycle progression. These findings, together with the lack of correlation of amplification status with in vitro cisplatin sensitivity of cell lines (Table S1) are at odds with the clinical behavior of primary tumors. Significantly, we found that knockdown of CCNE1 profoundly inhibited clonogenic cell survival and this effect was augmented by cisplatin treatment (Figure 5B & C). Collectively, our findings suggest that 19q12 amplification does not increase resistance to chemotherapy per se but rather may confer a survival advantage post-treatment. In addition to increased cellular division, high CCNE1 expression may assist re-entry into cell cycle from quiescence in surviving cells after chemotherapy [28]. Uniquely amongst cell cycle proteins in Drosophila, Cyclin E1 overexpression has been shown to promote self-renewal of neuroblasts [29] and may translate to an increased clonogenic capacity in tumor cells. Although we were unable to generate stable OVCAR-3 with integrated shRNA directed to CCNE1 to further validate these findings in vivo, these observations suggest that CCNE1 amplification may enhance the ability of tumor cells to repopulate the tumor after the cessation of chemotherapy.
We also sought to identify other mutational events that may enhance the effect of CCNE1 deregulation by seeking regions of copy number change that correlate with 19q12 gain. Previous investigations of correlated gains include a recent study in breast cancer [30] where co-amplification of 8p12 and 11q13 was identified and thought to cooperate functionally through activation of independent oncogenic pathways. Cooperative networks in glioblastoma that are associated with outcome have also been identified through analysis of co-occurring copy number changes [31]. We identified copy number change of three loci at 20q11, 1p36 and 6q27 to be significantly associated with CCNE1 amplification in ovarian tumors.
The 20q11 region contains twelve candidate genes, including TPX2, ASXL1 and BCL2L1 with expression correlated to CCNE1 (Table 2). TPX2 is a microtubule-associated protein downstream of Ran-GTP that triggers microtubule nucleation. It both activates and is a substrate for Aurora-A kinase and is important in mitotic spindle formation and chromosome segregation during cell division [reviewed in 22]. Low copy number gain and protein over-expression has been observed in other tumor types including pancreatic tumors where siRNA knockdown reduced cell growth in vitro, induced apoptosis and sensitized pancreatic cell lines to paclitaxel treatment [32].
We chose to further investigate the role of TPX2 in ovarian cancer given its expression was most significantly correlated with CCNE1. Furthermore, TPX2 has a plausible biological association  with Cyclin E1, both having known cell cycle functions. Unlike CCNE1 however, we did not find evidence of oncogene 'addiction' to TPX2 in cells with amplification. In addition, cells sensitive to CCNE1 knockdown showed no further reduction in viability when simultaneously treated with siRNA against TPX2. Notably, both OVCAR-3 and OVCAR-4 cells showed the greatest reduction in viability after single TPX2 knockdown. Gene expression analysis in 19q12 amplified cells show that TPX2 gene expression is reduced after CCNE1 knockdown ( Figure S4) implying TPX2 acts downstream of CCNE1, and that simultaneous TPX2 knockdown has only a minimal additive effect. The relationship between CCNE1 and TPX2 expression in vitro is consistent with the association identified in primary tumors and although our results suggest that TPX2 acts down stream of CCNE1, it may not be a key driver of the 20q11 amplicon. A systematic analysis of other genes within the amplicon is required, which is often broad and contains a number of potentially important gene targets. For example, the apoptotic regulator BCL2L1, adjacent to TPX2, has been previously suggested as a 20q11 amplification target in cancer [13]. Although not correlated with CCNE1 expression in our analysis, co-amplification of ID1 with CCNE1 may further contribute to cell cycle de-regulation in ovarian cancer. ID1 is involved in proliferation and differentiation, and functions by inhibiting binding and activity of other helix-loop-helix transcription factors. In breast cancer cells, gene knockdown has been shown to decrease CCNE1 expression and Cyclin E1/CDK2 activity [33]. Interestingly, we also note the presence of a noncoding RNA (NCRNA00028) and two micro-RNAs (hsa-mir-1825 and hsa-mir-3193) within the defined region of chromosome 20q11 gain. The potential effect of amplification on their function or subsequent de-regulation of downstream targets was not investigated in our current analysis. However, we are not aware of any evidence that would link these molecules to the regulation of CCNE1.
In primary tumors, CCNE1 copy number was significantly associated with shorter progression-free and overall survival, with tumors bearing high-level amplification showing the worst outcome ( Figure 6). Copy number status was more informative than gene expression; a significant correlation with high expression and PFS was observed, but not for OS. Similar to ERBB2 amplification in breast cancer, high-level CCNE1 amplification may therefore have predominant clinical utility in identifying patients most likely to have a poor response to standard treatment. Interestingly, younger patients were less likely to have CCNE1 amplified tumors (Table 1 and Figure 6A) suggesting a difference between the etiology of these tumors.
Our observations have potential implications for future therapeutic strategies based on targeting of cell cycle deregulation such as via cyclin-dependent kinase (CDK) inhibition. To date, the clinical success of single-agent CDK inhibitors has been disappointing [34]. Limited success may in part relate to off-target drug effects or redundancy between CDKs. It is anticipated that chemotherapeutic agents may sensitize cells to CDK inhibition [35]. However as shown here, interruption of G1/S transition may increase resistance to G1/S-phase targeted treatment and therefore CCNE1/CDK2 targeted therapies may need to appropriately phased with conventional therapy. Additionally, the surprising degree of resistance of cells without 19q12 amplification to CCNE1 knockdown suggests that CDK2 inhibitor clinical trials should incorporate careful pre-selection of ovarian cancer patients and other tumor subtypes associated with Cyclin E1 over-expression, such as triple-negative and basal-like breast tumors [36].

Ethics Statement
This study was approved by the Human Research Ethics Committees at the Peter MacCallum Cancer Centre, Queensland Institute of Medical Research, University of Melbourne and all participating hospitals. Written informed consent was obtained from all participants in this study.

Copy Number Data
Microarray data was obtained from Tumorscape (www.broadinstitute.org/tumorscape), the Cancer Genome Atlas Project (TCGA) (cancergenome.nih.gov) for primary tumors and the Sanger Cancer Genome Project Archive (http://www.sanger.ac. uk/genetics/CGP/Archive/) for ovarian cell lines. Data was visualized using Partek Genomics Suite 6.4 (Partek Inc., St Louis, MO) and Tree View [37]. Further copy number analysis is described in Methods S1.

Molecular Methods
Cell line DNA was extracted using a DNeasy Kit (Qiagen, Valencia, CA) and quantitative-PCR (qPCR) analysis of CCNE1 DNA copy number status was performed as described previously [3]. Total RNA was extracted from cell pellets using an RNeasy mini Kit (Qiagen) and reverse transcribed using M-MLV prior to SYBR green qPCR as detailed in Methods S1. Primer sequences to measure gene expression were obtained from qPrimerDepot [38]. Two primer sets were used to assess TPX2 copy number and were designed using Primer3 [39] or obtained elsewhere [32]. All primers are listed in Table S2.

siRNA Transfection
Cells were seeded at a density of 5-10610 3 cells per well in 96 well plates (for viability assays), Lab-Tek II Chamber Slides (Sigma Aldrich, St Louis, MO) (for TUNEL staining) or 3-6610 4 cells in 24 well plates (for all other assays) approximately 20 hours prior to transfection. Selected seeding densities resulted in cell growth as a sub-confluent monolayer at transfection. ON-Target plus siRNA pools (listed in Table S3) and transfection reagents were obtained from Dharmacon (ThermoFisher Scientific, Lafayette, CO). Optimal transfection conditions were determined based on maximal gene knockdown and minimal cytotoxicity as assessed by RT-PCR and MTS cell viability assay (see Methods S1). Nuclear localization of siGLO Green RNA duplex was used to monitor transfection efficiency by fluorescence microscopy 24-48 hours after transfection (data not shown). Transfection reagent and siRNA pools were pre-incubated at room temperature for 15 minutes in serum-free media (20% of final transfection volume) to facilitate formation of lipid-siRNA complexes. Cells were then incubated in final transfection mix containing 0.4% (v/v) DharmaFECT 2 and 50 nM siRNA in antibiotic-free media with serum. In double knockdown experiments, total siRNA amount was normalized between treatment groups by addition of a nonsilencing siRNA up to 100 nM. After 24 hours, media was replaced and following a further 72 hours, cells assayed for viability, cisplatin sensitivity (at a pre-determined IC50 dose, Table S4), cell cycle distribution, clonogenic survival or apoptosis as described below and in Methods S1.

Western Blot
Whole cell protein lysates were boiled, resolved by SDS-PAGE using 12.5% (w/v) acrylamide gels and then transferred to PVDF membranes. Blots were blocked in 5% (w/v) non-fat milk powder in PBS-T (0.1% Tween 20 in PBS) and probed overnight at 4uC in 1:500 primary antibody against human Cyclin E1 (clone HE12) (Santa Cruz Biotechnology, Santa Cruz, CA) or TPX2 (clone 18D5) (Biolegend, San Diego, CA). Membranes were washed in PBS-T and incubated with 1:7000 dilution of peroxidaseconjugate secondary antibody for 1 hr at room temperature, washed and developed by chemoluminescence before being exposed to radiographic film. Blots were re-probed with an antibody against a-tubulin to assess protein loading.

Flow Cytometry
Control, transfected and cisplatin treated cells were rinsed in PBS, trypsinized to form a single-cell suspension and fixed in 70% ice-cold ethanol. Cell were pelleted and resuspended in a solution containing 50 mg.mL 21 propidium iodide and 100 U.mL 21 RNAseA (Qiagen) for 30 minutes at room temperature. Up to 10,000 cells were then counted by FACS. Viable cell cycle profiles and percentage of cells in each cell cycle phase was determined using Modfit LT (Beckman Coulter, Brea, CA).

TUNEL Staining
Apoptotic cells were identified using the ApopTagH In Situ Apoptosis Detection Kit (Intergen, Purchase, NY). A minimum of eight, uniformly spaced images were taken per well and positively and negatively stained cells counted by a researcher blinded to the experimental setup.

Clonogenic Survival Assay
Transfected cells were treated with cisplatin for 1 hour then PBS washed, trypsinized to form a single-cell suspension, counted (Coulter Counter) and cell number equalized for each experimental condition. Cells were then seeded at low density in wells of a 6well plate in triplicate and left to form colonies for up to ten days. The number of plated cells differed depending on cell line plating efficiency; 500 and 15,000 cells were used for SK-OV-3 and OVCAR-3 respectively. Cell colonies were then fixed and stained with 20% (v/v) methanol and 0.1% (w/v) crystal violet. Cells were rinsed in water, air-dried and discrete colonies counted using MetaMorph (Molecular Devices, Sunnyvale, CA).

CCNE1 Copy Number and Gene Expression in Primary Tumors
Tumor samples and clinical data were obtained from women with advanced stage, serous invasive disease enrolled through the Australian Ovarian Cancer Study (www.aocstudy.org). This project had institutional ethics review board approval at all participating centers.
Samples were segregated based on CCNE1 copy number level as assessed by qPCR (above) using a log 2 copy number ratio cut-off of $0.5 (,3 copies) for gain and $2 (,8 copies) for amplification (Table 1). Matched expression data from Affymetrix U133 plus 2.0 microarrays was obtained from a previous study [17]. The CCNE1 probe showing the highest signal level (213523_at) was selected for our analysis and showed a significant correlation with gene copy number ( Figure 6B). Samples were segregated into low, medium and high expression of CCNE1, where tumors with high expression were defined as those above the median signal value + [0.5 x median absolute deviation (MAD)] and low expressing tumors where those below the median signal value -[0.5 *MAD]. Statistical analysis was performed in GraphPad Prism (GraphPad Software, San Diego, CA). PFS and OS was calculated from the date of diagnosis (surgery).