OVCA cell lines were obtained from the American Type Culture Collection (Manassas, VA, USA) (CAOV3, OV90, OVCAR3 and SKOV3); from the European Collection of Cell Cultures (Salisbury, England) (A2780CP and A2780S); from Kyoto University (Japan) (CHI, CHIcisR, M41, M41CSR, Tyknu and TyknuCisR); or were kind gifts from Dr Patricia Kruk, Department of Pathology, College of Medicine, University of South Florida, Tampa, FL, and Professor Susan Murphy, Department of OBGYN/Division of Gynecologic Oncology, Duke University, Durham, NC, USA (A2008, C13, CAOV2, HeyA8, IGR-OV1, IMCC3, IMCC5, MCAS, OV2008, OVCA420, OVCA429, OVCA432, OVCA433, FUOV1, PEO1, PEO4, SK-OV-6, T8, TOV-112D, TOV-21-G, Dov13, BG1, Ovary1847, OVCAR10, OVCAR8, OVCAR5, OVCAR4, OVCAR2 and SK-OV-4). Cell lines were maintained in RPMI-1640 (Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (Fisher Scientific, Pittsburg, PA, USA), 1% sodium pyruvate, 1% penicillin/streptomycin (Cellgro, Manassas, VA, USA), and 1% nonessential amino acids (HyClone, Hudson, NH, USA). Mycoplasma testing was performed every 6 months in accordance with the manufacturer's protocol (Lonza, Rockland, ME, USA).

Slides were stained using a Ventana Discovery XT automated system as per the manufacturer's protocol with proprietary reagents (Ventana Medical Systems, Tucson, AZ, USA). Briefly, slides were deparaffinized on the automated system with EZ Prep solution (Ventana). The heat-induced antigen retrieval method was used in Cell Conditioning 1 (Ventana). The mouse monoclonal antibody that reacts to c-Met (#18-7366; Invitrogen) was used at a 1:2,000 concentration in PSS antibody diluent (Ventana) and incubated for 60 min. The Ventana OmniMap anti-mouse secondary antibody was used for 16 min. The detection system used was the Ventana ChromoMap kit, and slides were counterstained with hematoxylin. Slides were then dehydrated and coverslipped as per normal laboratory protocol. The c-Met expression score was defined as the product of intensity and cellularity, where an intensity of 1 was weak; 2, moderate; 3, strong, and a cellularity of 1 ≤33%, 2 =34–65% and 3 ≥66%.

RNA extraction and microarray expression analyses were performed as previously described (13). Briefly, RNA was extracted using the RNeasy kit following the manufacturer's recommendations (Qiagen, Valencia, CA, USA). Quality of the RNA was measured using an Agilent 2100 Bioanalyzer. The targets for Affymetrix DNA microarray analysis were prepared according to the manufacturer's instructions. For OVCA cell lines (n=41), targets were hybridized to customized Human Affymetrix HuRSTA gene chips (HuRSTA-2a520709) [GEO accession number GSE34615 (14)].

Expression data were subjected to background correction and normalization using the Robust Multichip Average algorithm in the Affymetrix Expression Console (http://www.affymetrix.com/products/software/specific/expression_console_software.affx). Pearson's correlation test was performed on individual gene expression and IC₅₀ values. Probe sets with P<0.001 were considered to have significant correlations with IC₅₀ values.

For cells cultured in the presence of cisplatin, resistance was quantified using MTT proliferation assay, in accordance with manufacturer's instructions (Promega, Madison, WI, USA). Cells (5,000/well) were seeded in 96-well plates, allowed to adhere, and incubated with increasing doses of cisplatin for 48 h at 37°C. Following drug incubation, plates were read at 570 nm using a SpectraMax 190 microplate reader (Molecular Devices, USA). For drug combination experiments, CellTiter-Blue cell viability assays were performed using 384-well plates and an automated pipetting station. Briefly, for these assays, 1.2×10³ cells (24-μl volume) were plated in each well and allowed to adhere overnight at 37°C and 5% CO₂. Drug dilutions were then added using the same pipetting station, and the plate was incubated for 72 h. Following drug incubation, fluorescence was measured using a Synergy 4 microplate reader (Bio-Tek Instruments, Inc.). The fluorescence data were transferred to a spreadsheet program to calculate the percent viability relative to untreated cells. Experiments were repeated a minimum of three times.

Drug combination experiments were analyzed for synergistic, additive, or antagonistic effects using the combination index method developed by Chou and Talalay. For the application of this method, the drug concentration dilutions of the drugs were used at fixed dose ratios based on the IC₅₀ values of each drug obtained from preliminary experiments (e.g., 50:1, 2:5, 1:250). Briefly, the dose-effect curve for each drug alone is determined based on experimental observations using the median-effect principle and is compared to the effect achieved with a combination of the two drugs to derive a combination index value. This method involves plotting dose-effect curves, for each agent and their combination, using the following median-effect equation: fa/fu = (D/Dm)m, where D is the dose of the drug, Dm is the dose required for a 50% effect (equivalent to IC₅₀), fa and fu are the affected and unaffected fractions, respectively (fa=1-fu), and m is the exponent signifying the sigmoidicity of the dose-effect curve. The computer software program XLfit was used to calculate the values of Dm and m. The combination index values used for the analysis of the drug combinations were determined by the isobologram equation for mutually nonexclusive drugs that have different modes of action: combination index = (D)1/(Dx)1 + (D)2/(Dx)2 + (D)1(D)2/(Dx)1(Dx)2, where (Dx)1 and (Dx)2 in the denominators are the doses (or concentrations) for D1 (drug 1) and D2 (drug 2) alone that gives x% inhibition, whereas (D)1 and (D)2 in the numerators are the doses of drugs 1 and 2 in combination that also inhibited x% (i.e., isoeffective). Combination indices <1, equal to 1, and >1 indicate synergism, additive effects and antagonism, respectively.

The PCA methodology was used to derive a gene expression signature with a corresponding score. First, the data were reduced into a small set of uncorrelated principal components. This set of principal components was generated based on its ability to account for variation. The first principal component (1st PCA) is used to represent the overall variability in the data. The gene expression score is equal to ∑w_ix_i, a weighted average expression among the differentially expressed genes, where x_i represents gene i expression level, w_i is the corresponding weight (loading coefficient) with ∑w²_i = 1, and the w_i values maximize the variance of ∑w_ix_i. Directional signs of PCA scores are recognized to be arbitrary and can vary between software and the algorithm used to calculate the PCA model (15); however, this does not affect the interpretation of the PCA model and can be easily solved by multiplying both scores and loadings by −1, a 180° rotation. Details of this methodology have been reported by our group previously (13).

PCA models were then explored for associations with overall survival from OVCA (using median PC1 score as the threshold to define high vs. low pathway score) using a dataset for which both gene expression and overall survival data were available, the publicly available Australian (Aus) Dataset (Affymetrix U133Plus GeneChips, n=218).

Since overexpression of HGF has been associated with poor clinical outcome in OVCA (5–7), we evaluated the expression of c-Met, the only known HGF receptor, in OVCA cells and primary OVCA tumor samples. Pathological scoring of c-Met immunostains (cellularity × intensity) in 79 primary OVCA samples and 41 OVCA cell lines indicated the presence of c-Met expression in 92% (73/79) (Table I) and 83% (34/41) (Table II) of samples, respectively. c-Met expression in primary OVCA did not appear to be associated with stage, response to primary platinum therapy, or CA125 levels (Table I).

To determine whether HGF signaling influenced OVCA cell line chemosensitivity, we evaluated the activity of MK8033, an inhibitor of HGF/c-Met signaling. MK8033 exhibited strong anti-proliferative effects against a panel of OVCA cell lines (Fig. 1). Furthermore, when combined in constant molar ratio of 1 with carboplatin plus paclitaxel (molar ratio of 20,000:1), MK8033 significantly lowered the carboplatin-paclitaxel IC₅₀ in the majority of cell lines tested (Fig. 1). Combination index values calculated using the Chou-Talalay isobologram equation indicated synergistic activity between MK8033 and carboplatin-paclitaxel for almost all OVCA cell lines tested, as indicated by combination index values of <0.7 (Fig. 1).

Pearson's correlation test of IC₅₀ values and Affymetrix HuRSTA genechip expression data identified 47 genes to be correlated with MK8033 plus carboplatin-paclitaxel sensitivity [false discovery rate (FDR) of <0.2] (Table III). PCA modeling of these genes indicated that the expression signature was associated with overall survival from OVCA (AUS dataset, n=218, P=0.013) (Fig. 2).