Characterization of Ovarian Cancer Ascites on Cell Invasion, Proliferation, Spheroid Formation, and Gene Expression in an In Vitro Model of Epithelial Ovarian Cancer^1,2

Cell Culture, Clinical Material, and Patients

The OV-90 cell line was maintained in OSE media consisting of 50:50 medium 199:105 (Sigma-Aldrich, St. Louis, MO) supplemented with 10%fetal bovine serum(FBS), 2.5 µg/ml amphotericin B and 50 µg/ml gentamicin [18]. Following appropriate consent, ascites were collected at the time of clinical intervention at the Centre Hospitalier de l'Université de Montréal (Montreal, QC, Canada). Ascites were centrifuged at 2500 rpm for 5 minutes. The acellular fractions were stored at -20°C and tested within 6 months of reception. The protein concentration in ascites fluid was measured by the Bradford assay (Bio-Rad, Hercules, CA). Histopathology, grade, and stage of ovarian tumors were assigned according to the International Federation of Gynecology and Obstetrics criteria. Of the 54 ascites included in our study on invasion, two thirds were from patients diagnosed with papillary serous adenocarcinomas and most presented with stage IIIC and grade 3 disease (Table 1). One third of samples were from patients who had already received chemotherapy prior to surgery. The presence of neoplastic cells in ascites was determined from pathology reports. Table 2 describes the 28 ovarian cancer patients diagnosed with accompanying ascites that were included in the survival analysis.

In Vitro Invasion Assay

Cellular invasion was assayed by determining the ability of cells to invade a synthetic basement membrane (Matrigel; Becton-Dickinson, Bedford, MA). Polycarbonate membranes (8-µm pore size) of the upper compartment of Transwell culture chambers were coated with 0.4 µg/ml Matrigel. The upper compartment was filled with OSE media containing 1% FBS and the lower compartment was filled with OSE media either with no serum, with 5% FBS, or with 5% of the indicated ascites. For inactivation, ascites were heated for 10 minutes at 100°C to denature proteins. Ovarian cancer cells were trypsinized and resuspended in OSE media containing 1% FBS. The cell suspension (20 × 10³ cells/well) was placed in the upper compartment. Then, cells were incubated at 37°C and allowed to invade through the Matrigel barrier for 24 hours. Following incubation, membranes were fixed with methanol and stained (Giemsa; Sigma-Aldrich). Noninvading cells were removed using a cotton swab, whereas invading cells on the underside of the membrane were counted using an inverted microscope. All experiments were performed at least twice.

Cell Proliferation

Two thousand cells were plated either with no serum, with 5% FBS, or with 5% of the indicated ascites in six-well plates and incubated at 37°C. At defined intervals, cells were trypsinized and cell viability was assessed by a Trypan Blue exclusion assay. Cell numbers were evaluated using a hematocytometer. Each experiment was performed in triplicate.

Spheroid Formation

Spheroids were formed using a modification of the hanging droplet method [19]. Briefly, 4 × 10³ cells were resuspended in 15 µl of OSE media supplemented either with 5% FBS, with 5% of the indicated ascites, or without serum, and then placed on the cover of a 150-mm tissue culture plate. The cover was placed over a plate that contained 15 ml of OSE to prevent dehydration of the hanging droplet. Spheroid formation was monitored after 4 days and representative spheroids were photographed.

RNA Extraction

Total RNA was extracted with a reagent (TRIzol; Gibco/BRL, Life Technologies, Inc., Grand Island, NY) as recommended by the manufacturer. RNA was extracted from tumor cells grown to 80% confluence in 100-mm Petri dishes. RNA quality was assessed using a 2100 Bioanalyzer with the RNA 6000 Nano LabChip kit (Agilent Technologies, Mississauga, ON, Canada) according to the manufacturer's protocol.

Microarray Analysis

Hybridization assays and data collection were performed at the McGill University and Genome Quebec Innovation Centre (Montreal, Canada). Briefly, 20 µg of total RNA from each sample was reverse-transcribed using an oligo-dT primer containing a T7 RNA polymerase binding site. In vitro transcription was performed on this cDNA and the resulting cRNA was biotinylated through incorporation of biotinylated dUTP and dCTP. Samples were fragmented in 40 mM Tris-acetate, 100 mM potassium acetate, and 30 mM MgCl₂ (pH 8.1) at 95°C to reduce secondary structure. A total of 15 µg of cRNA was hybridized to an Affymetrix HG-U133A GeneChip array (Santa Clara, CA), washed, stained, and scanned with a Hewlett Packard Gene Array scanner (Palo Alto, CA) and .CEL files were normalized based on a quantile method.

Gene expression profiles were analyzed using R (www.r-project.org), a statistical programming language, and Bioconductor [20], an open source software library for the analyses of genomic data based on R. Background subtraction, normalization (quantile normalization), and expression value calculations were performed using the justGCrma function available as part of Bioconductor's gcrma package. Bioconductor's genefilter package was used to filter out genes with insufficient variation in expression across all samples tested. Expression values retained after this filtering process presented intensities greater than 100 U in at least two samples and a log base 2 scale of at least 0.2 for the interquartile range across all tested samples. Differentially expressed genes were identified using the limma package, which estimates the fold-change between predefined groups by fitting a linear model and using an empirical Bayes method to moderate standard errors of the estimated log-fold changes for expression values from each probe set.

Quantitative PCR

cDNA synthesis was prepared using the SuperScript First-Strand Synthesis System for reverse transcription-polymerase chain reaction (Invitrogen Life Technologies, Carlsbad, CA) according to the manufacturer's instruction. Reverse transcription-polymerase chain reaction was performed on 2 µg of total RNA using 2.5 µl of the random hexamer solution. Samples were diluted 1:10 in water prior to Q-PCR. Positive and negative controls were included in all experiments.

Q-PCR was performed using the Rotor-gene 3000 (Corbett Research, Montreal Biotech Inc., Montreal, QC, Canada). Quantitect SYBR Green PCR (QIAGEN Inc., Mississauga, ON, Canada) was used for labeling in a final volume of 25 µl containing 5 µl of sample cDNA and 10 pg of the different primers and reactions performed as described by the manufacturer. Experiments were repeated at least twice. Serial dilutions (1:5) were performed to generate a standard curve for each gene tested to define the efficiency of the Q-PCR reaction and a melt curve was done to confirm the specificity of the reaction. We used the Pfaffl analysis method to measure the relative quantity of gene expression [21]. The algorithm is defined by R = (E_target)^{ΔCp target(control - sample)}/(E_ref)^{ΔCp ref(control - sample)}, where R is the relative expression ratio, E is the efficiency of the PCR reaction, and ΔCp is the difference of the C_t (crossing point of the sample at a given threshold). The reference gene, ActinB, was selected based on its stable expression in all samples by microarray analysis. Moreover, Q-PCR confirmed its appropriateness because no significant statistical differences were noted among the samples. The first sample (with 5% FBS) served as the reference sample in each experiment. The mean value of the C_t from replicates was used to calculate R. Marker expression was evaluated in the OV-90 cell line either with no serum, with 5% FBS, or with 5% of the indicated ascites, under the same conditions used to evaluate the invasion potential of the OV-90 cell line. For each marker, a Pearson correlation was calculated between the scored invasion result (1 < 100% and 2 ≥ 100% of invasion) and the scored genes expression (1 < median and 2 ≥ median).

Statistical Analysis

Univariate Cox proportional hazard regression, Kaplan-Meier survival plots, and log-rank tests were performed to determine the significance of markers' ability to predict the survival of EOC patients (Table 2). The expression threshold used in the log-rank test that gave the best sensitivity-specificity values was established based on the receiver operating characteristics (ROC) curves. We used the ROC and SURVIVAL packages from R version 2.4.0 (Vienna, Austria).

The Spearman correlation coefficient test (two-tailed) was used to estimate the correlation between the invasion rates and clinical data and gene expression. Statistical analyses were performed with SPSS software 11.0 (SPSS Inc., Chicago, IL).

Effect of Ascites on OV-90 Invasive Potential

To address the interactions among elements found within the peritoneal tumor environment, we characterized the effect of a large panel of ascites (Table 1) on the ability of an aggressive EOC cell line (OV-90) to stimulate invasion in an in vitro assay. Media containing 5% of ascites (acellular fraction) from patients with ovarian cancer was added to the lower chamber of Transwell plates containing micropore filters precoated with Matrigel and OV-90 cells were added to the upper chamber. The potential for ascites to affect OV-90 invasion was scored in comparison to media supplemented with 5% FBS (Figure 1A). OV-90 is capable of invasion in Matrigel assay in presence of 5% FBS (Figure 1). A large number of ascites led to an inhibition of OV-90 cell invasion compared to cells in the presence of FBS. A lower, but still important number of ascites was more stimulatory for invasion than the FBS control. We assessed the correlation between the invasion rates of ascites and clinical parameters available for patients: age, grade, stage, presence of neoplastic cells, and chemotherapy, but no significant correlation was observed (Table 1).

The acellular fraction of 10 independent EOC ascites (A1317, A1318, A1322, A1322(2), A1337, A1592, A1835, A1946, A2085, and A2090; Table 1) was selected for further characterization based on their ability to invade. The invasive potential was again compared to a 5% FBS control. In OV-90 cells, it should be noted that there is no difference in the invasion rate with or without 5% FBS and thus the classification of ascites as being either inhibitory or stimulatory is independent of the effect of serum. Inhibitory ascites reduce the invasion rate compared to either no serum or 5% FBS, whereas stimulatory ascites results from a higher number of cells able to cross a Matrigel barrier in comparison to controls. Ascites A1592, A1946, A2085, and A2090 induced an invasion rate greater to the one observed in the presence or the absence of FBS. A1317, A1318, A1322, A1322(2), A1337, and A1835 diminished OV-90 cell invasion potential (Figure 1B). These effects were not due to differences in protein concentration of each ascites because correcting for the amount of protein did not affect the overall patterns observed (data not shown).

Effect of Ascites on OV-90 Growth

To determine the effect of ascites on proliferation rates, 10 ascites were selected and tested for their ability to alter the growth potential of OV-90 cells compared to a 5% FBS control (Figure 2). OV-90 cells were incubated for 2 days in media supplemented either with no serum, with 5% FBS, or with 5% of the indicated ascites and cell growth was evaluated by a Trypan Blue exclusion assay. The highest growth rates were observed with exposure to FBS, as well as A1835, A1946 or A2085, whereas the remaining ascites samples conferred variable but lower growth rates. In particular, OV-90 exhibited lower growth rate in the presence of the A1337 and A1322(2) ascites compared to those observed with no serum. We noted that after 24 hours of incubation, the time used to monitor invasion, no statistically significant differences in the growth rate were observed (with one exception). This suggests that the monitored invasive effects are not a simple reflection of cell growth.

Effect of Ascites on OV-90 Spheroid Formation

We have previously demonstrated that the OV-90 cells are able to grow as large compact spheroids [19]. Because the relation between these three-dimensional structures and invasive potential remains poorly defined, we determined the effect of ascites on the formation of in vitro spheroids. For this purpose, the OV-90 cell line was incubated either in the absence of serum, in 5% FBS, or in 5% of ascites. The formation of spheroids was monitored after 4 days. As shown in Figure 3, the OV-90 cell line formed multiple very small and nonreproducible spheroids from drop to drop in the absence of FBS, which is consistent with previous findings [19]. In the presence of FBS or ascites A1317 and A1592, however, OV-90 formed large (approximately 500 µm in diameter) and compact spheroids. This was reproducible from drop to drop with one unique spheroid of similar size formed in each drop. Although cell scattering around spheroids was observed, it was not significantly different between spheroids generated in the presence of different ascites. The ability for different ascites to induce large and compact spheroid formation was also confirmed with ascites A1318, A1322, A1322(2), A1337, A1835, A1946, A2085, and A2090, which gave similar results (Figure W1). Therefore, the ability of ascites to stimulate spheroid formation did not correlate with the invasion potential of the tested ascites.

Ascites Invasive Phenotype Characterization

Ascites that stimulated or inhibited OV-90 cell invasion were combined to determine which effect is the most predominant (Figure 4A). We selected one stimulatory (A2090) and two inhibitory (A1369 and A1526) ascites. When stimulatory and inhibitory ascites were added together (5% A2090 + 5% A1369 or 5% A2090 + 5% A1526), we observed that the invasion rate was not significantly different compared to the presence of inhibitory ascites alone. This result suggests that the inhibitory ascites have a predominant impact on invasion of OV-90 cells compared to a stimulatory ascites. This effect was not due to an increased concentration of ascites in the assay as adjusting individually tested ascites to 10% had no effect on the invasive phenotype (data not shown).

To determine if this predominant effect is protein-dependent, inhibitory ascites were boiled at 100°C for 10 minutes before being tested on OV-90 cells (Figure 4B). The results showed that protein inactivation abolished the inhibitory effect of the two selected ascites (A1369 and A1526). These combined results suggest that inhibition of OV-90 cell invasion is protein-based and that the inhibitory effect is stronger than the stimulatory effect.

Modification of Gene Expression Induced By Ascites in OV-90 Cells

To identify potential molecular players in invasion regulated by ascites, gene profiling was performed using Affymetrix HG-U133A GeneChip arrays with RNA extracted from OV-90 cells after 24 hours of exposure to either no serum, to 5% FBS, or to 5% of 1 of 10 different ascites (A1317, A1318, A1322, A1322(2), A1337, A1592, A1835, A1946, A2085, and A2090).

For supervised analysis purposes, two groups were created for comparison. The group that stimulated invasion (referred here as GSTIMUL group) included samples with no FBS, with 5% FBS, or with 5% of the four ascites that stimulated OV-90 cell invasion (A1592, A1946, A2085, and A2090). An inhibitory group (referred here as GINHIB group) containing the six ascites that inhibited cell invasion (A1317, A1318, A1322, A1322(2), A1337, and A1835) was also defined. Expression analysis identified 243 probe sets to be differentially expressed (P ≤ .1) between the GSTIMUL and GINHIB groups (Table W1).

Differential Expression Validation of Selected Candidates By Q-PCR

As differences in expression are subtle but tended to be statistically significant, it was important to validate the value of the cutoff selected. For this purpose, we selected genes with different P values (Table 3) to test the robustness of their association with invasion. Table 3 describes the seven candidate genes upregulated in the GSTIMUL group [dickkopf homolog 1 (Xenopus) (DKK1), regulator of G-protein signaling 4 (RGS4), interferon-stimulated transcription factor 3, gamma 48 kDa (ISGF3G), tribbles homolog 1 (Drosophila) (TRIB1), MAP kinase phosphatase 1 (MKP1), cyclooxygenase 2 (COX2), and motile sperm domain containing 1 (MOSPD1)] and two candidate genes upregulated in the GINHIB group [plectin 1, intermediate filament binding protein 500 kDa (PLEC1) and myristoylated alanine-rich protein kinase C substrate (MARCKS)] that were selected for further validation. Q-PCR was used to validate the differential expression of selected candidates in RNA derived from OV-90 cells exposed individually to the entire panel of 54 ascites (Table 1). The relative expression ratio (R) of each gene, based on the Pfaffl method (see Materials and Methods section for details), was quantified and, for each experiment, the median ratio was calculated and the results scored. Pearson correlations were then calculated for each candidate correlating ascites invasion effects (stimulatory or inhibitory) and scored gene expression, as shown in Table 4. The scored gene expression of RGS4, ISGF3G, TRIB1, MKP1, MOSPD1, and PLEC1 correlated significantly with the ascites' invasion effects, but not the expression of DKK1, COX2, and MARCKS. These results also suggest that an extensive validation of candidates identified in the supervised analysis by Q-PCR is warranted to uncover the richness of genes implicated in the invasive process.

Survival

To evaluate the prognosis potential of the selected gene candidates RGS4, ISGF3G, TRIB1, MKP1, MOSPD1, and PLEC1, we sought to access the expression of these genes in samples derived from patients with ovarian ascites. Therefore, we extracted expression values from a microarray analysis of RNA extracted from 28 primary cultures derived from the cellular fraction of ascites from ovarian cancer patients (Table 2). Univariate Cox regression analysis showed a strong association only between TRIB1 gene expression and overall survival (P = .0007). Using a threshold determined by ROC analysis, a Kaplan-Meier curve coupled to a log-rank test identified the presence of two patient groups and high TRIB1 expression was associated with a poorer survival rate (log-rank P = .005) (Figure 5).