The authors have declared that no competing interests exist.
Conceived and designed the experiments: WC JWH SMM MAM. Performed the experiments: SMM DHM WC MAM JEM. Analyzed the data: WC SMM MAM ASB. Contributed reagents/materials/analysis tools: DHM ASB. Wrote the paper: WC SMM MAM ASB.
The majority of ovarian tumors eventually recur in a drug resistant form. Using cisplatin sensitive and resistant cell lines assembled into 3D spheroids we profiled gene expression and identified candidate mechanisms and biological pathways associated with cisplatin resistance. OVCAR-8 human ovarian carcinoma cells were exposed to sub-lethal concentrations of cisplatin to create a matched cisplatin-resistant cell line, OVCAR-8R. Genome-wide gene expression profiling of sensitive and resistant ovarian cancer spheroids identified 3,331 significantly differentially expressed probesets coding for 3,139 distinct protein-coding genes (Fc >2, FDR < 0.05) (
High Grade Serous Ovarian Cancer (HGSOC) is the most lethal form of ovarian cancer with approximately 16,000 new cases in the United States each year with 5 year survival rates <30% [
Myriad mechanisms of platinum therapy resistance have been identified including changes in cisplatin transport and trafficking, disruption of apoptosis, increased tolerance of cisplatin-DNA adducts, and increased DNA repair in response to cisplatin-DNA interactions [
Experimental models have not recapitulated the many features exhibited in tumors including intercellular communication and the influence of the microenvironment [
Towards these goals, we derived a new OVCAR-8 cisplatin resistant cell line (OVCAR-8R) and used genome wide expression analysis to discover genes differentially expressed in the sensitive and resistant cells as spheroids. Genes differentially expressed between the parental and resistant OVCAR-8 cells are enriched for markers of the mesenchymal state and are associated with survival. Despite significant expression changes of cisplatin transporters, OVCAR-8R spheroids did not show significantly different intracellular platinum concentration or transport properties compared to the parental OVCAR-8 spheroids. We applied multiple methods to evaluate how similar the expression changes adapted by OVCAR-8R may be reflected in HGSOC tumors. A pathway and a direct evaluation of a set of genes both indicated that many features of OVCAR-8R spheroids model HGSOC tumors. These data indicate that the OVCAR-8R spheroid model captured critical aspects of cisplatin resistance relevant to ovarian cancer patients.
Cisplatin (cis-diamminedichloroplatinum(II)) was purchased from Sigma-Aldrich.
The human ovarian adenocarcinoma cancer cell line OVCAR-8 cell line was purchased through the National Cancer Institute Developmental Therapeutics Program’s tumor repository program. OVCAR-8 was made resistant
Cells were plated in 96 well plates and treated 24 h later with the indicated concentrations of cisplatin. Viability was measured after 96 h of treatment using the WST-1 reagent (Roche), according to the manufacturer’s instructions.
Spheroids were generated by seeding OVCAR-8 and OVCAR-8R cells in low attachment agarose gel molds with hemispherical recesses. Gel casts were created by pouring 2% agarose into 3D Petri Dish casting molds (Microtissues, Providence, RI) [
Net cisplatin uptake was determined by measuring platinum content of ovarian cancer cells before and after incubation with cisplatin. Equal numbers of OVCAR-8 and OVCAR-8R cells were plated in T75 flasks. Cells were treated with 5 μM cisplatin for 3 hrs. After cisplatin treatment, cells were washed once with HBSS, and then with PBS lacking calcium and magnesium. Cells were then dissociated in PBS containing no calcium or magnesium and containing 5 mM EDTA, and centrifuged at 229 x g. The pellet was dissolved in 20 mM Tris, pH 7.6 and the cells lysed by multiple freeze-thaw cycles at -80°C.
Platinum was measured by using an Agilent 7500CE ICP-MS (Agilent Technologies, Palo Alto, CA) at the Interdisciplinary Center for Plasma Mass Spectrometry at the University of California at Davis. Cell lysates and platinum standards were introduced using a MicroMist Nebulizer (Glass Expansion, Pocasset, MA) into a temperature controlled spray chamber. Platinum standard solutions, were diluted from standardized platinum stock solutions (SPEX CertiPrep, Metuchen, NJ) to concentrations from 0.01 to 300 ppb in 3% nitric acid (Fisher Scientific, trace metal grade) in deionized water (Millipore). Cisplatin content, calculated from platinum concentrations, was normalized to protein concentration (Bio-rad assay, Bio-rad. Hercules, CA), and background readings, derived as platinum content of cells prior to cisplatin incubation, were subtracted. Net cisplatin uptake (expressed as μg cisplatin/g protein) was first determined as mol platinum per g protein and converted to weight of cisplatin by multiplying by the ratio of molecular weights of cisplatin (300.1 g/mol) to platinum (195.1 g/mol).
Total RNA was isolated from spheroids using the Trizol reagent (Invitrogen, Carlsbad, CA). Microarrays were processed at the Yale Center for Genome Analysis facility. RNA was fragmented and labeled with the Affymetrix GeneChip Whole Transcript Target Labeling Assay and hybridized to the Affymetrix Human Gene 1.0 ST Arrays according to recommended Affymetrix protocols (Affymetrix, Santa Clara, CA). Signals were calculated by Robust Multichip Analysis (RMA) using Expression Console software (Affymetrix, version 1.1). Genes with low signals, defined as the lowest quartile in both cisplatin sensitive and resistant cells, were excluded from further analysis. Raw and processed microarray data were deposited into the NCBI Gene Expression Omnibus database (GSE45553).
Total RNA was isolated as described above and reverse-transcribed to cDNA as described previously [
The Significance Analysis of Prognostic Signatures (SAPs) code and ovarian tumor datasets were downloaded from dryad [
Hierarchical clustering was performed in Gene-E using Pearson correlation to calculate distances [
OVCAR-8 cells were chosen for this study because they readily form spheroids [
Serial exposure of OVCAR-8 cells to sub-lethal concentrations of cisplatin resulted in significant, lasting changes in cisplatin resistance (
Cell viability of cisplatin resistant and sensitive cells grown as monolayers (A) or spheroids (B), expressed as a percentage of untreated viability, following exposure to varying concentrations of cisplatin for 96 h. Statistical comparisons were performed at each dose using a two-tailed Student’s t-test. Asterisks (**, p<0.01; ***, p<0.001) indicate significant difference in viability between cisplatin-sensitive (dashed line) and resistant cells (solid line) at the same cisplatin concentration (mean ± SE, n = 9). (C+D). Photomicrographs of cisplatin resistant (C) and sensitive (D) cells under phase contrast illumination. Resistant cells display stronger adhesion to the substrate while sensitive cells do not. Photomicrographs of spheroids from resistant (E) and sensitive (F) lines.
Eight genes were selected with a wide range of expression differences between the resistant and sensitive cells to test in an orthogonal assay (
Genes in the microarray dataset with differences in expression between sensitive and resistant ovarian cancer spheroids. (A) qPCR and microarray expression levels correlate. qPCR relative expression was calculated relative to GAPDH. Microarray scores are RMA. Trend line was determined by linear regression (r2 = 0.71, p = 0.009). (B) The fold changes for the selected transcripts.
Drug transporters that mediate intracellular drug concentration have been investigated as drivers of chemoresistance [
The complicated mixture of differential expression of known cisplatin transporters did not immediately suggest that cisplatin transport was responsible for the observed resistance. To test whether drug transport and subsequent changes to intracellular concentration were driving resistance in OVCAR-8R spheroids, we measured intracellular platinum concentration in the spheroids using a cisplatin uptake assay by mass spectrometry. No difference in net uptake of cisplatin in cisplatin-sensitive (25.2 ± 8.2 μg cisplatin/g protein) spheroids, compared to cisplatin-resistant spheroids (26.5 ± 7.7 μg cisplatin/g protein) was observed. These observations indicate that even though expression of many drug transporters was down-regulated in resistant cells, the intracellular platinum concentration remained unaffected.
As OVCAR-8R spheroids did not demonstrate reduced intracellular platinum concentrations, other potential resistant pathways were evaluated. To handle intracellular platinum concentrations, sequestration can be mediated by platinum binding proteins including Glutathione-S Transferases and metallothioneins [
To identify pathways that may be mediating resistance, gene set enrichment analysis, (GSEA), was performed [
Reactome nucleotide metabolism gene set (NES = 1.4, FDR = 0.18) including NT5E, recently linked to cisplatin resistance [
In order to test the hypothesis that the OVCAR-8R cells are more mesenchymal compared to OVCAR-8 cells, we evaluated whether the global gene expression program was indicative of a more mesenchymal phenotype. Mesenchymal cancer cells can be identified by examination of expression signatures indicative of mesenchymal states [
(A) GSEA enrichment shows that the EMT hallmark gene set is significantly biased towards OVCAR-8R spheroids compared to OVCAR-8. (B) Key mesenchymal markers are significantly up-regulated in OVCAR-8R compared to OVCAR-8 spheroids. (C) The Taube EMT signature separates the OVCAR-8R and OVCAR-8 spheroids by hierarchical clustering using Gene-E with Pearson correlation calculated distances. Each row represents a gene. R is the OVCAR-8R and S is the parental OVCAR-8 spheroids. Red indicates higher risk, blue indicates lower risk.
Alternatively, EMT could be driven by YAP1, another transcriptional driver that increases cancer stemness and EMT in multiple systems [
An increased mesenchymal state has been associated with shorter survival in ovarian cancer [
(A) SAPS analysis across 1735 ovarian tumor datasets suggests that both the Fc3 and Fc4 resistance transcript lists are significantly associated with shorter survival. (B) The Fc3 and Fc4 resistance gene sets were strongly enriched for shorter survival in the top 5% of all 5357 gene sets tested. Statistics for the resistance and representative strongly enriched gene sets are shown. The HOX13_01 gene set is shown as a positive control to compare the statistics.
To assess if a defined gene expression signature can be derived from the list of differentially expressed transcripts that is associated with patient survival, we identified a 17 gene expression signature that separated OVCAR-8 and OVCAR-8R cells (
Genes up-regulated in OVCAR-8R spheroids are expressed higher in patients with shorter survival and vice-versa. (A) Hierarchical clustering with Pearson correlation distances separates high and low expression in OVCAR-8R and OVCAR-8 spheroids. Hierarchical clustering was performed using Gene-E software [
Gene Symbol | Gene Name | Expression Levels in High Risk Patients |
---|---|---|
APOE | apolipoprotein E | High |
BTG2 | BTG family, member 2 | Low |
CTSA | cathepsin A | High |
EIF1AX | eukaryotic translation initiation factor 1A, X-linked | Low |
FLNC | filamin C, gamma (actin binding protein 280) | High |
GNA12 | guanine nucleotide binding protein (G protein) alpha 12 | High |
IGFBP4 | insulin-like growth factor binding protein 4 | High |
LGI2 | leucine-rich repeat LGI family, member 2 | High |
MEIS1 | Meis homeobox 1 | Low |
NLGN1 | neuroligin 1 | Low |
OLFML3 | olfactomedin-like 3 | High |
PCOLCE2 | procollagen C-endopeptidase enhancer 2 | High |
PLA2G4A | phospholipase A2, group IVA (cytosolic, calcium-dependent) | Low |
PTH2R | parathyroid hormone 2 receptor | High |
RPRM | reprimo, TP53 dependent G2 arrest mediator candidate | Low |
SERPINE1 | serpin peptidase inhibitor, clade E (nexin, plasminogen activator inhibitor type 1), member 1 | High |
SPOCK1 | sparc/osteonectin, cwcv and kazal-like domains proteoglycan (testican) 1 | High |
We developed a new spheroid model of drug resistance and provided evidence supporting its utility as a model for HGSOC tumors. We applied genome-wide expression profiling to gain insight into potential resistance mechanisms in the spheroids and examined which pathways may be relevant to patient tumors. In this model, resistance was not due to changes in drug transport or DNA repair, but rather to sequestration in combination with increased expression of anti-apoptosis pathways, cytokines, and an increased mesenchymal expression profile. Importantly, the changes adapted by the resistant cells in the expression profile identified patients with shorter survival and higher likelihood of relapse. We conclude that multiple mechanisms contribute to the cisplatin resistance of OVCAR-8R spheroids that are relevant to patients.
Previous
We specifically tested if resistance is associated with decreased platinum concentrations in the resistant cells and did not observe a significant difference in intracellular cisplatin concentrations. The gene expression data and platinum uptake assay were consistent with cisplatin resistance being caused by increased sequestration of platinum, through up-regulation of metallothionein and other sulfur rich proteins. Consistent with sequestration, metallothionein I and II isoforms were up-regulated between 3-fold to 7-fold in OVCAR-8R cells, and metallothionein-1E was among the top genes up-regulated (47-fold) compared to parental OVCAR-8 cells. Higher expression of metallothioneins is a known mechanism of cisplatin resistance and their increased expression in OVCAR-8R cells is likely contributing to the observed resistance [
The mesenchymal nature of ovarian cancer cells is most often associated with more drug resistant tumors and shorter survival [
Our observations indicate that the OVCAR-8 spheroids represent a good model to examine cisplatin resistance
A limitation of our study is that we only analyzed one cell line. Despite this limitation, this spheroid model is relevant to ovarian tumors as indicated by the common gene expression changes observed in the model and in ovarian tumors. The observation of the increased expression of many mesenchymal markers, a global gene expression profile associated with survival using a global analysis, SAPS, as well as the derivation of a specific 17 gene expression signature, all support the utility of this spheroid model to investigate mechanisms relevant to patients.
In summary, these observations indicate that the mechanisms of resistance in the OVCAR-8R cell line model are relevant to ovarian cancer patients, and support further investigation into the role of these genes in the development of resistance in ovarian cancer. This study of a spheroid model of ovarian cancer provides a foundation to gain new insights into cisplatin resistance in an
(DOC)
(XLSX)
Four worksheets are provided listing the gene sets enriched in the resistant or parental cells.
(XLSX)
We thank Joel Commisso at the Interdisciplinary Center for Plasma Mass Spectrometry at the University of California, Davis, for his technical expertise. We thank Milton C. Gum for his advice regarding dataset analyses. This work was funded by the NIH NCRR supplement grant P41 RR001395-27S1 (J.W.H.), NSF DBI-1005378 “REU Site: Biological Discovery in Woods Hole”, faculty startup funds from the Office of Research at Oklahoma State University (W.C.), and the Mary Kay Foundation (A.S.B.).