A proteomic approach to paclitaxel chemoresistance in ovarian cancer cell lines

A proteomic approach to paclitaxel chemoresistance in ovarian cancer cell lines

Biochimica et Biophysica Acta 1794 (2009) 225–236 Contents lists available at ScienceDirect Biochimica et Biophysica Acta j o u r n a l h o m e p a ...

708KB Sizes 4 Downloads 84 Views

Biochimica et Biophysica Acta 1794 (2009) 225–236

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / b b a p a p

A proteomic approach to paclitaxel chemoresistance in ovarian cancer cell lines Michela Di Michele a,⁎, Anna Della Corte a, Lucia Cicchillitti b, Piero Del Boccio c,d,e, Andrea Urbani c,d,e, Cristiano Ferlini b, Giovanni Scambia b, Maria Benedetta Donati a, Domenico Rotilio a a

Research Laboratories, “John Paul II ” Centre for High Technology Researches and Education in Biomedical Sciences, Catholic University, Campobasso, Italy Department of Oncology, “John Paul II ” Centre for High Technology Researches and Education in Biomedical Sciences, Catholic University, Campobasso, Italy Centro Studi sull'Invecchiamento (Ce.S.I.), Chieti, Italy d Dipartimento di Scienze Biomediche, Università “G. D'Annunzio” di Chieti-Pescara, Italy e IRCCS-Fondazione Santa Lucia Centro di Ricerca sul Cervello, Roma, Italy b c

a r t i c l e

i n f o

Article history: Received 23 June 2008 Received in revised form 29 August 2008 Accepted 18 September 2008 Available online 10 October 2008 Keywords: 2D-DIGE Drug resistance Ovarian cancer Paclitaxel Proteomics

a b s t r a c t Ovarian cancer is the leading cause of gynaecological cancer mortality. Paclitaxel is used in the first line treatment of ovarian cancer, but acquired resistance represents the most important clinical problem and a major obstacle to a successful therapy. Several mechanisms have been implicated in paclitaxel resistance, however this process has not yet been fully explained. To better understand molecular resistance mechanisms, a comparative proteomic approach was undertaken on the human epithelial ovarian cancer cell lines A2780 (paclitaxel sensitive), A2780TC1 and OVCAR3 (acquired and inherently resistant). Proteins associated with chemoresistance process were identified by DIGE coupled with mass spectrometry (MALDITOF and LC-MS/MS). Out of the 172 differentially expressed proteins in pairwise comparisons among the three cell lines, 151 were identified and grouped into ten main functional classes. Most of the proteins were related to the category of stress response (24%), metabolism (22%), protein biosynthesis (15%) and cell cycle and apoptosis (11%), suggesting that alterations of those processes might be involved in paclitaxel resistance mechanisms. This is the first direct proteomic comparison of paclitaxel sensitive and resistant ovarian cancer cells and may be useful for further studies of resistance mechanisms and screening of resistance biomarkers for the development of tailored therapeutic strategies. © 2008 Elsevier B.V. All rights reserved.

1. Introduction Ovarian cancer is characterised by rapid growth and division of one of the three major cell types present in the ovary (germ, stromal and epithelial cells) [1]. The most common form is epithelial ovarian cancer, which represents the leading cause of mortality due to gynaecological malignancies and accounts for 4% of all cancers among women from industrialised countries [2]. It is characterised by a presentation at an advanced stage and poor survival, mostly due to the absence of major specific symptoms in the early stage [3]. At present the most effective strategy for the treatment of advanced ovarian cancer is based on cytoreductive surgery followed by chemotherapy containing taxanes. Among taxanes, paclitaxel (Taxol), is an extremely effective anticancer agent used to treat ovarian cancer and a wide range of tumor types, such as breast, lung, head and neck cancer [4]. Although the majority of advanced ovarian carcinomas

⁎ Corresponding author. Laboratory of Analytical Techniques and Proteomics, “John Paul II” Center for High Technology Researches and Education in Biomedical Sciences, Catholic University, Largo A. Gemelli 1, 86100 Campobasso, Italy. Tel.: +0039 0874 312285; fax: +0039 0874 312710. E-mail address: [email protected] (M. Di Michele). 1570-9639/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.bbapap.2008.09.017

initially respond successfully to taxane-based chemotherapy, longterm applications may lead to the occurrence of drug-resistance [5]. Several mechanisms have been implicated in resistance to paclitaxel, such as the selective expression of beta-tubulin isotypes and the overexpression of P-glycoprotein [6], however the overall molecular mechanisms of paclitaxel resistance remain to be elucidated. Paclitaxel resistance represents a pleiotropic phenomenon characterised by changes in expression and function of several strictly interconnected proteins. A proteomic approach, which enables to overview a huge number of gene expression products simultaneously, would appear as a powerful tool to get insight into the complex mechanisms of intrinsic or acquired resistance to paclitaxel. Proteomics has been extensively adopted to study protein expressions in different types of human cancers and in resistance to several drugs [7–10]. The most widely used technique for large scale protein expression analysis and for quantifying various proteins in different samples is 2DE. DIGE overcomes many of the limitations associated with traditional 2DE, such as reproducibility, limited dynamic range and high time/labor costs, thus allowing for more accurate and quantitative proteomics studies [11]. This innovative technology uses three spectrally-resolvable fluorescent dyes to label two samples and a pool of equal amounts of all the experimental samples to be run

226

M. Di Michele et al. / Biochimica et Biophysica Acta 1794 (2009) 225–236

together on the same gel [12]. The use of a pooled internal standard favours inter-gel matching and minimizes gel-to-gel variability, allowing the detection of small differences in protein expressions with statistical confidence. Moreover a strong correlation has been reported between quantitation by 2D-DIGE and metabolically stable isotope labelling [13]. The aim of the present study was to characterise the molecular basis of resistance to paclitaxel using human epithelial ovarian cancer cells. The proteomes of one sensitive and two paclitaxel resistant (natural and induced) ovarian cancer cell lines were compared by 2DDIGE coupled with MS analysis. Proteins differentially expressed in paclitaxel sensitive and resistant ovarian cancer cells are likely to be involved in pathways that modulate the mechanisms of chemoresistance and are potential biomarkers for treatment response and targets for new therapeutic approaches.

same gel were added to an equal volume of 2× sample buffer (7 M urea, 2 M thiourea, 4% CHAPS, 30 mM Tris pH 8.0, 52 mM DTT, 1% IPG buffer 3–10 and a trace of bromophenol blue) and mixed. The first dimension was carried out on an IPGphor system (Amersham) using pH 3–10 gel strips of 13 cm. The IEF was performed at 20 °C under the following conditions: 12 h at 30 V, 2 h at 300 V, 3 h at 1000 V, 2 h at 8000 V in gradient and 3 h at 8000 V. IPG strips were then incubated for 15 min in equilibration buffer (50 mM Tris–Cl pH 8.8, 6 M urea, 30% glycerol, 2% SDS) with 6.5 mM DTT and then in equilibration buffer with 2% iodoacetamide for an additional 15 min. The second dimensional separations were carried out on 10% SDSpolyacrylamide gels at 20 °C at 2 mA/gel for 1 h and at 10 mA/gel for about 16 h, until the bromophenol blue dye front reached the end of the gels. 2.4. Image analysis

2. Materials and methods 2.1. Chemicals and reagents General chemicals were obtained from Sigma-Aldrich (Poole, UK) while reagents for 2D-DIGE experiments were purchased from Amersham Biosciences (Uppsala, Sweden), unless otherwise indicated. Bradford assay kit was from BioRad (Hercules, CA, USA). The mouse anti human voltage dependent anion-selective channel protein 1 antibody was produced by Calbiochem (La Jolla, CA), the mouse anti human annexin V, clathrin heavy chain 1 and ubiquitin protein ligase E1 antibodies by Abcam (Cambridge, UK) and the pan B tubulin by Covance (California, USA). 2.2. Cell lines and paclitaxel cytotoxicity The human ovarian cancer cell lines A2780 and paclitaxel resistant OVCAR3 were purchased from the European Collection of Cell Cultures (Salisbury, United Kingdom). A2780TC1 is a cell clone derived from A2780 cells chronically exposed to paclitaxel (100 nmol/L) [14]. Growth conditions and paclitaxel toxicity testing were the same as previously described [14]. Briefly, in order to ensure data reproducibility, A2780TC1 cells were cultured in the absence of paclitaxel for two passages before each experiment. Paclitaxel cytotoxicity was assessed in parallel to the preparation of cell lysates for all the cell lines, which were employed at a low number of passages comprised between ten and twelve. Preparation of cell lysates for proteomic analysis was repeated independently four times in identical conditions of cell growth, 48 h after cell plating at a cell density of 20,000 cells/ml in 75 cm2 culture flask in complete media.

Labelled proteins were visualized using the Typhoon Trio imager (Amersham Biosciences). The Cy2, Cy3, and Cy5 components of each gel were individually imaged using excitation/emission wavelengths specific for Cy2 (488/520 nm), Cy3 (532/580 nm) and Cy5 (633/ 670 nm). All gels were scanned directly between the low-fluorescent glass plates with a 100 μm resolution and adjusted photomultiplier tube (PMT) values. Three images per gel were thus obtained. Images were cropped to remove areas extraneous to the gel image using ImageQuant V5.0 (Amersham Biosciences). Gel analysis was performed using DeCyder 2-D Differential Analysis Software v6.5 (Amersham Biosciences). All pooled standard/sample gel image pairs were processed by the DIA (Difference In-gel Analysis) software module to detect and differentially quantify the protein spots in the images. The estimated number of spots for the codetection procedure was set to 3500. The master gel was manually assigned to the gel with the most spots detected. Gel to gel matching of the spot maps from each gel was then performed using the BVA (Biological Variation Analysis) software module, using 20 landmarks. After automatic matching, spots were visually checked for undetected or incorrectly matched spots. For each spot, the software reported the standardized abundance as the ratio of the volume in the Cy3 (or Cy5) sample to the volume of the pooled standard sample labelled with Cy2, where the volumes have been normalized across the gels. Within the BVA module, each comparison was filtered to find the spots having a p-value b 0.05 for the paired T-test. Fold change was calculated as the ratio of the average standardized abundance in pairwise comparisons among the three groups. Finally EDA (Extended Data Analysis) module carried out intra- and inter-gel statistical analyses, performing expression pattern clustering based on PCA and ANOVA.

2.3. 2D-DIGE 2.5. In gel digestion and mass spectrometry Four independent extractions from different cell cultures were carried out for each of the three cell lines. The pellet of 107 cells was solubilized in lysis buffer (7 M urea, 2 M thiourea, 4% CHAPS, 30 mM Tris pH 8.5) containing antiprotease cocktail (Sigma) and incubated 1 h on ice. The lysates were centrifuged at 20 000 ×g for 15 min and the supernatant collected and brought to pH of 8.5 with 0.1 M NaOH to optimize fluorescent tagging. The protein concentration was determined in triplicate by the Bio-Rad Dc protein assay as described by the manufacturers (BioRad). Each sample (50 μg) was labelled with CyDye 3 or 5, while the internal standard, prepared by pooling equal amounts from each of the samples, was labelled with CyDye 2. Cyanine dyes were reconstituted in anhydrous DMF and added to labelling reactions in a ratio of 400pmol CyDye:50 μg protein. Protein labelling was achieved by incubation on ice in the dark for 30 min. The labelling reaction was quenched by the addition of 10 mM lysine (1 μl per 400 pmol dye) and incubation on ice for further 10 min in the dark [15]. Prior to IEF, labelled samples to be separated in the

Spots of interest were manually excised from preparative gels run with 300 μg of proteins and the same procedure as for the DIGE experiments. Proteins were visualized by silver staining [16] to pick spots of interest. The gel pieces were destained by washing twice with 6 mM potassium ferricyanide in 100 g/L sodium thiosulfate and proteins digested overnight with 20 ng/μL trypsin (Sigma) in 40 mM ammonium bicarbonate/10% acetonitrile at 37 °C. The proteolytic peptides were extracted with 50% acetonitrile/0.1% TFA and spotted with α-cyano-4-hydroxycinnamic acid onto a MALDI target. Peptide mass fingerprinting was performed on a Voyager-DE STR MALDI-TOF mass spectrometer (Applied Biosystem, Stafford, USA), operating in positive ion reflectron mode, with an acceleration voltage of 20 kV, a nitrogen laser (337 nm) and a laser repetition rate of 4 Hz. The final mass spectra, measured over a mass range of 800–5000 Da and by averaging 50–200 laser shots, were finally analysed using the DataExplorer software v4.0 (Applied Biosystem).

M. Di Michele et al. / Biochimica et Biophysica Acta 1794 (2009) 225–236

All mass spectra were calibrated with a calibration kit (Applied Biosystem) and internally calibrated using trypsin autolysis digestion products. For MS/MS analysis, peptide containing solutions were injected using a CapLC system (Micromass, Waters) coupled on-line with a nano-ESI-Q-TOF instrument (Micromass, Waters). The sample was first concentrated into a Waters Symmetry300 C18 5 μm OPTI-PAK Trap Column, 0.35 × 5 mm by a carrier solvent (water with 0.2% of formic acid) at 20 μl/min for 3 min and after eluted at 0.2 μl/min (using a pre-column splitter) on a C18 PepMap column (LC Packings, Dionex) 5 μm, 75 μm inner diameter × 250 mm, with a water/acetonitrile gradient in the presence of 0.2% of formic acid. The solutions used as mobile phases were Buffer A (2%ACN, 0.1% formic acid) and buffer B (2% water, 0.1% formic acid in ACN); the gradient was 8–50% B in 40 min, 50–80% B in the next 10 min and 8% B in the final 10 min to equilibrate the system. Peptides were analysed with a Q-TOF mass spectrometer (Micromass, Waters) equipped with a nano-Lock-Spray source, using a [Glu1]-fibrinopeptide B 500 nM in H2O/ACN (1:1) with 0.2% of formic acid, as a reference compound for on-line recalibration data using its doubly charged ion at m/z = 785.84. A 2.6 kV tension was applied on the nanoESI capillary. Acquired MS/MS data were converted in a compatible format (Pkl Files) by ProteinLynx 2.0 software (Micromass). Mascot software (Matrix Science, London) was employed for protein database searching. The searches were performed using the NCBI database and the following standard parameters: Homo sapiens; tryptic digest with a maximum of one missed cleavage; carboxyamidomethylation of cysteine, partial methionine oxidation and modification of glutamine to pyroglutamic acid and a mass tolerance of 100 ppm. For MS/MS data, the search was performed with the following additional criteria: maximal tolerance for MS/MS data of 0.3 Da, searching peptide charge of 2+ and 3+. Identifications were accepted based on a tripartite evaluation that takes in account significant MASCOT Mowse scores, spectrum annotation and observed vs. expected migration on 2D gel. A MASCOT score of 64 corresponds to p b 0.05 for mass finger print experiments while a MASCOT score of 37 corresponds to p b 0.05 for MS/MS sequencing; these thresholds were chosen as the cut-off for a significant hit. All identified proteins were manually categorized using the information provided by the Gene Ontology project [17] and literature.

227

Table 2 Experimental design of 2D-DIGE experiment Gel

CyDye3

CyDye5

CyDye2

1 2 3 4 5 6

WT 1 TC 1 OV 1 WT 3 TC 4 OV 4

OV 3 WT 2 TC 2 TC 3 OV 2 WT 4

Pooled standard Pooled standard Pooled standard Pooled standard Pooled standard Pooled standard

Each gel contained the pooled standard and two other samples. The three samples (WT = A2780, TC = A2780TC1, OV = OVCAR-3 epithelial ovarian cancer cell lines) were analysed in quadruplicate by running 6 gels, in order to maximize the likelihood of detecting any sample-to-sample variation. The pooled internal standard comprises equal amounts of proteins from each of the 12 samples.

(1:3000) at room temperature. The bands were detected using the enhanced luminescence kit ECL+ (Amersham Pharmacia Biotech, Arlington Heights, IL). 2.7. Network analysis Regulated proteins identified by 2D-DIGE were analysed further by Ingenuity Pathway Analysis (IPA; Ingenuity Systems, Mountain View, CA; www.ingenuity.com). It builds hypothetical networks from DIGEidentified proteins and other proteins, on the basis of a regularly updated database that consists of millions of individual relationships between proteins, known from literature. The interactions used for the network comprise direct relationships that describe physical (binding) and functional (phosphorylation, proteolysis, …) interaction between

2.6. Western blot analysis Proteins were extracted from A2780, A2780TC1 and OVCAR3 epithelial ovarian cancer cell lines as previously described. Equal amounts of protein extracts (10 μg) were run on 10% SDS-PAGE gels and transferred onto nitrocellulose membranes. The membranes were incubated in 3% BSA in Tris-buffered saline-Tween 20 (TBST) for 1 h at room temperature to block non-specific protein binding sites and then in anti voltage-dependent anion-selective channel protein 1 (1:1000), annexin V (1:1000), clathrin heavy chain 1 (1:500), ubiquitin protein ligase E1 (1:3000) and pan B-tubulin (1:3000) antibodies overnight at 4 °C. Subsequently membranes were washed with TBST and incubated 2 h with goat anti-mouse IgG conjugated with horseradish peroxidase

Table 1 Paclitaxel toxicity profile of A2780, A2780TC1 and OVCAR-3 cell lines

A2780 A2780TC1 OVCAR-3

IC50 (nmol/L)

RI

2.7 ± 1.9 10 027 ± 3195 26.7 ± 2.1

/ 3713 9.9

The growth inhibition effect of paclitaxel after 72 h of culture of the three cell lines is reported as IC50 values (concentration half-maximally reducing the growth of cancer cells ± SD) and Resistance Index (RI), calculated by dividing the IC50 value of the resistant cell lines over that of drug sensitive A2780.

Fig. 1. DIGE master map containing all the spots detected on human epithelial ovarian cancer cell lines A2780 (paclitaxel sensitive) A2780TC1 (paclitaxel acquired resistant) and OVCAR3 (paclitaxel inherently resistant). After labelling, proteins were separated on 13 cm, pH 3–10 linear strips in the first dimension and on a 10% polyacrylamide SDS PAGE in the second dimension. Protein spots marked on the maps were differentially expressed (pb 0.05) in pairwise comparisons between the three cell lines and identified by MS.

228

M. Di Michele et al. / Biochimica et Biophysica Acta 1794 (2009) 225–236

Fig. 2. Representative overview of 2D DIGE maps of A2780, A2780TC1 and OVCAR-3 human epithelial ovarian cancer cell lines. IEF as the first dimension separation was over a pH range 3–10, while the second dimension separation was carried out by SDS-PAGE in a 10% polyacrylamide gel, as described in detail in Materials and methods section. Magnification represents an example of a protein spot (1385) down regulated in A2780TC1 and OVCAR-3 paclitaxel resistant cell lines compared to the A2780 paclitaxel sensitive cell line.

proteins. IPA computes a score for each possible network according to the fit of that network to the inputted proteins. The score is calculated as the negative base-10 logarithm of the p value that indicates the likelihood of the inputted proteins in a given network being found together as a result of random chance. Networks with scores of 10 or higher (negative log of the p value) have a high confidence of not being generated by random chance alone and were the only considered in the present work.

and Resistance Index (RI), calculated by dividing the IC50 value of the resistant cell lines over that of drug sensitive A2780. Both A2780TC1 and OVCAR3 cell lines resulted to be resistant to paclitaxel, being nearly 10-fold and 3700-fold respectively resistant to paclitaxel when compared to A2780.

3. Results

To study the mechanisms of resistance to paclitaxel in ovarian cancer the proteome of the paclitaxel sensitive (A2780) and resistant (A2780TC1 and OVCAR3) human epithelial ovarian cancer cell lines were analysed. In particular, 2D-DIGE was used to identify differentially expressed proteins among the three cell lines. Extracts from A2780, A2780TC1 and OVCAR3 ovarian cancer cells were labelled with Cy3 or Cy5 dyes and subjected to 2-DE as detailed in Table 2. Experimental design has been set up taking in account major criteria indicated by Karp et al. [20]. The quantitative comparison for a protein of the two samples coresolved on the same gel was between the Cy3 or Cy5 signals and the Cy2 internal standard signal for this protein, not between the Cy3 and Cy5 signals directly. The intra-gel ratios for each resolved protein (Cy3:Cy2 and Cy5:Cy2) were then normalized to the cognate ratios from the other gels. To avoid any possible bias derived from the labelling efficiency, half of the samples of each group were labelled with Cy3 dye and the other half with Cy5 dye. However, in our

3.1. Paclitaxel toxicity profile of A2780, A2780TC1 and OVCAR3 cell lines To study the mechanisms of resistance to paclitaxel in ovarian cancer, three epithelial ovarian cancer cell lines were used in this work. A2780 was chosen as a paclitaxel sensitive human epithelial ovarian cancer cell line. It was compared by a proteomic approach to two paclitaxel resistant human epithelial ovarian cancer cell lines: A2780TC1, the paclitaxel resistant counterpart of A2780, used as a model of acquired resistance, and OVCAR3, another ovarian cancer cell line, naturally resistant to paclitaxel and thus analysed as a model of inherent resistance, as demonstrated in our laboratory and by others [14,18,19]. The growth inhibition effect of paclitaxel in these two resistant cell lines is reported in Table 1 as IC50 values (concentration half-maximally reducing the growth of cancer cells)

3.2. Proteomic pattern of paclitaxel sensitive and resistant human epithelial ovarian cancer cells

Fig. 3. Multivariate analysis of DIGE/MS results. Principle Component Analysis (PCA) reduces the dimensionality of a multidimensional analysis, and displays the two principle components that can distinguish between the two largest sources of variation within the dataset. (A) Hierarchical clustering of the 12 independent samples based on the global expression patterns of modulated proteins detailed in Table of Supplementary material. Hierarchical clustering of individual samples is shown on top and clustering of individual proteins shown on the left, with relative expression values displayed as an expression matrix (heat map) using a relative scale ranging from −0.5 (green) to +0.5 (red). (B) Scatter plot of A2780 (paclitaxel sensitive), A2780TC1 and OVCAR3 (paclitaxel acquired and inherent resistant) cell lines. Each data point in the PCA scatter plots describes the global expression values for the subset of proteins listed in Table of Supplementary material. PCA discretely clustered the 12 individual Cy3- and Cy5-labelled DIGE expression maps into the three ovarian cancer cell lines.

M. Di Michele et al. / Biochimica et Biophysica Acta 1794 (2009) 225–236

229

230

M. Di Michele et al. / Biochimica et Biophysica Acta 1794 (2009) 225–236

Fig. 4. Distribution of the spots differentially expressed (p b 0.05) between pairwise comparisons of A2780 paclitaxel sensitive (WT), A2780TC1 paclitaxel acquired resistant (TC) and OVCAR3 (OV) inherently resistant cancer cell lines, as detected by DIGE.

study no appreciable differences between spot patterns and volumes of Cy3 and Cy5 labelled proteins were found (data not shown). A representative DIGE gel is reported in Fig. 1, showing the master gel containing all spots detected on the three cell populations, while Fig. 2 shows representative imaged 2D patterns of proteins extracted from each cell line. About 2600 protein spots were resolved and the protein patterns were visually similar in the three cell lines. The four replicates performed for each ovarian cancer cell line showed wellresolved spots, with high reproducibility and without streaking and high background. Reproducibility was confirmed when each sample was run once more on 2D preparative gels to collect protein spots for mass spectrometry analysis. Most of the ovarian cancer cell proteins showed molecular masses between 30 and 100 kDa and a pI ranging from 4 to 8 (Fig. 1). 3.3. Identification of proteins differentially expressed in paclitaxel sensitive and resistant human epithelial ovarian cancer cell lines After DIGE, the Cy2, Cy3, and Cy5 channels of each gel were individually imaged and the comparisons of protein expression in 2D images were carried out using Decyder-DIA software. To reduce the intrinsic variability associated to the different samples, more stringent criteria were chosen: only proteins present at least in three of the quadruplicate gel images were considered for statistical analysis; moreover spots of interest identified were manually verified to have the three-dimensional profile characteristics of a protein spot and spots with volume measurements close to the background and with no defined shape as precipitated dye, dust particles and bubbles were eliminated. It is worth mentioning that the statistics were calculated on all spots, and manual verification was only used for determining spots for identification. Running a filter on the original spots to exclude artefacts before performing the statistical analyses would have reduced the number of differentially expressed spots. ANOVA was performed on the normalized abundances of matched spots to compare similarity among the three cell lines according to expression patterns, by applying the false discovery rate (FDR) correction method [21]. Expression and identification information for all individual proteins in the hierarchical clustering analysis (grouped into horizontal bars in the expression matrix and related via a similar dendrogram on the left) are reported in Fig. 3A. The resulting heat map clearly classified the 12 samples in three major clusters, each corresponding to one ovarian cancer cell line. In particular, the dendrogram separated into 2 main branches, paclitaxel sensitive (A2780) and resistant ovarian cancer cells, and the branch of resistant cells into natural (OVCAR3) and acquired resistant (A2780TC1) cell lines sub branches. These grouping assignments were reiterated in an unsupervised Principle Component Analysis (PCA) of the protein expression patterns within each sample (Fig. 3B). In particular the first principal component, which distinguished 82% of the variance, clearly separates the proteome data of the paclitaxel sensitive from the paclitaxel resistant cell lines, and the second component, with additional 5% of the variance, distinguished between the two resistant

cell lines. PCA and hierarchical clustering results based on the protein expression of the paclitaxel sensitive and resistant ovarian cancer cell lines were consistent between them and indicated a complete and clear separation of the three cell lines, demonstrating high reproducibility between the replicate DIGE samples. Comparisons between pairwise A2780, A2780TC1 and OVCAR3 gels revealed that 172 protein spots varied between cancer cell lines in a statistically significant way (p b 0.05), all further verified to be valid protein spots following manual inspection. In particular 43 of these spots varied among all the three cell lines and 101 protein spots varied between A2780 and A2780TC1 gels, 112 between A2780 and OVCAR3 gels and 86 between A2780TC1 and OVCAR3 samples in a statistically significant way (p b 0.05) (Fig. 4). As expected, the number of modulated spots was higher between paclitaxel sensitive (A2780) and resistant cancer cell lines (A2780TC1 and OVCAR3) and lower between the two paclitaxel resistant cell lines. Among differentially expressed spots, 34% had a b1.5 fold, 41% had a 1.5 b fold b 2, 17% had a 2 b fold b 3 and 7% had more than a 3 fold increase or decrease in protein abundance. The 172 gel spots corresponding to proteins which showed significantly different expression levels (p b 0.05) between the three cell lines were excised and in-gel digested with trypsin. The digests obtained were analysed by MALDI TOF MS and LC MS/MS for protein identification. In particular, most of the proteins were identified by PMF using MALDI TOF MS and only those identifications with MASCOT score N64 were accepted. In some cases, as when MALDI identification was unsuccessful, the MASCOT score was significant but low (comprised between 64 and 70) or proteins were not previously known to be present in ovarian cancer cells (denoted by footnote d in Table of Supplementary material), the protein identification was performed by LC MS/MS. The LC MS/MS identifications coincided with the results obtained by MALDI TOF MS, adding confidence to the identification results. The results of protein identifications are summarized in Table of Supplementary material, in which the spot numbers correlate with those reported in Fig. 1. The molecular weights and isoelectric points for these proteins mostly matched those derived from the spot position on the gel, providing additional validation of the identifications. In some instances, more than one protein spot was

Fig. 5. Functional classification of the identified proteins. Proteins were classified using the information provided by the Gene Ontology project and the scientific literature. Several proteins were associated with more than one function and in such cases one category was chosen arbitrarily.

M. Di Michele et al. / Biochimica et Biophysica Acta 1794 (2009) 225–236

found to correspond to the same protein, consistently with the presence of different post-translationally modified forms of the same protein. Thus, the final number of unique protein identified was 108. About 12% of spots could not be identified probably because they were too weak and did not deliver a sufficient amount of peptides or because the MS data were insufficient for protein identification or because they were new proteins, not previously identified and therefore absent in databases. A biological process clustering of the identified proteins based on the scientific literature was created in order to perceive the distribution of the differentially expressed proteins within important cellular regulatory functions. Proteins were clustered in 10 classes based on biological functions (Fig. 5). When proteins were associated with more than one function, one category was chosen arbitrarily. The most represented classes were those of stress response and chaperones (24%), metabolism (22%), protein biosynthesis and folding (15%), cell cycle and apoptosis (11%) and cytoskeleton and cell structure (7%) (Fig. 5). The “unknown” protein class included one protein that was successfully matched but did not have a known function yet (spot 2533). 3.4. Protein validation by western blot To confirm some of the changes in protein expression in the three epithelial human ovarian cancer cell lines revealed by DIGE/MS analysis, an aliquot of the protein lysates was subjected to onedimensional SDS PAGE and blotted with the antibodies against four differentially expressed proteins (voltage-dependent anion-selective channel protein 1, annexin V, clathrin heavy chain 1 and ubiquitin protein ligase E1). As literature reports a marked alterations of βtubulin isotypes in drug resistance in ovarian cancer [5], the protein βtubulin was used as the internal loading control of the protein concentration in the extracts to detect all tubulin isoforms, being the entity of changes of the specific isoforms negligible respect to the total

Fig. 6. Validation of the 2-D DIGE results by western blot. Conventional 1D SDS PAGE gels were run separately, with protein extracts from paclitaxel sensitive (A2780) and resistant (A2780TC1 and OVCAR3) cell lines. Proteins were transferred onto nitrocellulose membranes and incubated with antibodies against some of the proteins identified in DIGE analysis: clathrin heavy chain 1 (CLTC, 180 kDa), ubiquitin protein ligase E1 (UBE1, 117 kDa), annexin V (ANXV, 38 kDa) and voltage-dependent anion-selective channel protein 1 (VDAC1, 31 kDa). B-Tubulin (TUBB, 55 kDa) was used as the internal control of the protein concentration in the extracts. The degree of differential expression in the three cell lines is shown in histograms to the side of the figure (mean of three replicates ± SE). Intensity of the proteins immunoblotted and patterns of expression deducted from the DIGE (Table of Supplementary material) were highly consistent.

231

amounts of B-tubulins. As expected, no significant change was observed for tubulin among the three cancer cell lines (Fig. 6). The densitometric quantitation of protein patterns in A2780, A2780TC1 and OVCAR3 cell lines showed similar trends with DIGE data for all proteins assayed (Fig 6, Table of Supplementary material). 3.5. Pathway analysis of proteins modulated in paclitaxel sensitive and resistant ovarian cancer cell lines To provide insight into the paclitaxel resistance mechanisms in ovarian cancer cells, pathway analysis was carried out on the datasets of identified proteins using Ingenuity Pathway Analysis software. The proteins undergoing change were analysed in a systematic way using known protein–protein interactions published in the literature to determine the most significant interaction networks, global pathways and functions, ranked by score. From our dataset generated from the proteins listed in Table of Supplementary material, five high ranking networks were identified. Fig. 7 shows the highest scored cluster of “cell death and protein folding and post-translational modification”. The overall score for the depicted network was 58, indicating that the probability of matching the indicated proteins by a purely random event was 10− 58. This cluster consists of a network of 33 proteins, of which 27 were identified by DIGE-based proteomics and 6 were recognized to be related because of their reported interactions with the proteins identified by DIGE. 4. Discussion In the present work, a comparative proteome study using DIGE and MS based analysis was carried out and led to the identification of proteins associated with paclitaxel resistance in ovarian cancer cells, allowing to get deeper insights on the complex resistance mechanism. This may contribute to the development of new protein markers for preventing poor prognosis due to resistance and establishing successful therapeutic strategies. DIGE allowed detection of a wide number of proteins differentially expressed in pairwise comparisons between A2780, A2780TC1 and OVCAR3 cells, which were analysed by ANOVA. Results from hierarchical clustering and PCA analysis were overlapping (Fig. 3), demonstrating a good discrimination of the three cell lines. In particular, the first component clearly separated the proteome data of the paclitaxel sensitive and resistant cell lines, while the second component distinguished between the two resistant cell lines. In this way, we may state with stronger confidence that the differential expression in A2780TC1 and OVCAR-3, compared with A2780 cell line, is due to resistance. Therefore, DIGE based proteomic approach used in our study revealed to be a powerful tool to distinguish among paclitaxel sensitive and resistant protein patterns. Moreover, the number of differentially expressed proteins between pairwise comparisons among the three cell lines indicates that the two paclitaxel resistant cell lines are closer between them, as differing by a smaller number of proteins. Almost the totality of the proteins with significant differential expression ratios were successfully identified by mass spectrometry and grouped on the basis of their biological functions (Table of Supplementary material). Most of the identified proteins have previously been reported to be modulated in chemoresistant ovarian and/or other cancers using other cell lines or clinical tissues, supporting the findings of our study. As the variations in protein expressions observed could be indirect consequences of other changes affecting less expressed and non detectable proteins, a pathway analysis was performed to include these proteins. Ingenuity Pathways Analysis showed that the most significant pathway associated with the highest number of changing proteins was related to cell death, protein folding and post-translational modification (Fig. 7). We focused on proteins which represent

232

M. Di Michele et al. / Biochimica et Biophysica Acta 1794 (2009) 225–236

Fig. 7. Biological pathways associated with modulated proteins identified in paclitaxel sensitive and resistant epithelial human ovarian cancer cells by DIGE/MS analysis. Ingenuity Pathway Analysis software (www.ingenuity.com) was used to map identified proteins onto existing mammalian pathways and networks that associate proteins based on known protein–protein interactions, mRNA expression studies and other biochemical interactions established in the literature. The table reports the significant biological pathway associated with identified proteins. Shown is a graphical representation of the highest scoring of five networks obtained from all identified proteins containing the largest number of focus genes (27). Nodes represent proteins: shaded features depict proteins identified in the present study, whereas un-shaded features depict additional members of these networks and pathways that were not detected by DIGE/MS. Node shapes indicate function: enzymes (diamond), transcription regulators (oval), nuclear receptors (rectangle), cytokines (square), transporter (trapezoid), and “other” (circles). Protein–protein associations are indicated by edges containing single lines, whereas proteins that act upon another protein (controlling their expression) are indicated by arrows. Continuous or dotted line indicates respectively direct or indirect protein interactions. Arrows indicates proteins differentially expressed between A2780 and A2780TC1 or OVCAR3 but not between A2780TC1 and OVCAR3 ovarian cancer cell lines (Table 3). Abbreviations used are: 14-3-3, CBP; ACTB, beta-actin; ALDH7A1, antiquitin; ANXA5, annexin V; Caspase, interleukin 1 converting enzyme; CCT3, T-complex protein 1 gamma subunit; CCT4, T-complex protein 1 delta subunit; CCT6A, Tcomplex protein 1 zeta subunit; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; HYOU1, 150 kDa Oxygen-regulated protein; HSPA8, heat shock 70 kDa protein 8; Hsp90, Hsp84; HSP90AA1, heat shock protein 90 kDa; Hsp70, Hsp70 gene; HSP, heat shock protein; HSPA5, heat shock 70 kDa protein 5; HSPA9, heat shock 70 kDa protein 9B variant; HSPB1, heat shock protein 27 kDa; HSPD1, heat shock 60 kDa protein 1; LMNA, lamin A/C; Mek 1/2, Mkk 1/2; NFkB, NF-KAPPA B; NME1, Nucleoside-diphosphate kinase; PARK7, DJ-1 protein; PRDX3, peroxiredoxin 3; PRDX6, peroxiredoxin 6; PRDX2, peroxiredoxin 2; PPIA, cyclophilin A; Ras, Ras protein; RPSA, 40S ribosomal protein SA; SERPINH1, 47 kDa heat shock protein; SOD1, superoxide dismutase 1; TCP1, T-complex protein 1, alpha subunit; VCP, Transitional endoplasmic reticulum ATPase; VIM, Vimentin.

M. Di Michele et al. / Biochimica et Biophysica Acta 1794 (2009) 225–236

the main nodes of this pathway and on those proteins known to be involved in drug resistance, some of which were further validated by immunoblotting (Fig. 6), and their role in the process of chemoresistance will be discussed. The protein spot number (Fig. 1, Table of Supplementary material) and the abbreviations used in the pathway analysis when the protein is a node of the pathway analysed (Fig. 7) are reported in parenthesis for each protein spot described. Most of the proteins of the highest score network (Fig. 7) belong to the functional class of stress response and chaperones, which resulted to be the most numerous class for proteins identified in our study (Fig. 5). In particular, different members of the heat shock family were found to be modulated, as HSP47(1282), HSP60 (1002; HSPD1), HSP90 (669; HSP90AA1), HSP150 (527), DJ1 (1862; PARK7) and different isoforms of HSP27 (1702, 1782; HSPB1) and HSP70 (701, 789, 806, 869, 873, 906, 911; HSPA5, HSPA8, HSP9). Our findings agree with several works which report that paclitaxel resistance is associated to changes in the expression of heat shock proteins (HSP) in ovarian and other cancer cells [22–25]. As an example, the expression of three isoforms of GRP78 (789, 806, 869; HSPA5) was found to be decreased in both paclitaxel resistant cell lines. This endoplasmic reticulum resident protein has an important protective role in protein folding in stressful conditions, such as glucose starvation and hypoxia in cancer. Thus, GRP78 implication in cancer progression and drug resistance may be related to its protective antiapoptotic activity [26,27]. The class of stress response and chaperones also includes several proteins of the network here analysed associated with the redox level of the cell, such as superoxide dismutase (2412; SOD1), 150 kDa oxygen regulated protein (314, 328; HYOU1) and peroxiredoxin-2, -3 and -6 (2065, 1844, 1883, 1705; PRDX2, PRDX3, PRDX6). Cancer hypoxia conditions may explain the changed expression of these proteins involved in the detoxification of oxygen reactive species and promotion of cell survival [28,29], supporting a close relation of these proteins to apoptosis and the complex phenomenon of drug resistance. Also different subunits of T-complex protein 1 (941, 990, 1028, 1084, 1093, 1097, 1110, 1117, 1120, 1146; CCT3, CCT4, CCT6A, TCP1), included in the same class of stress response and chaperones, were already found to be involved in the process of chemoresistance by Castagna et al. [30]. Several differentially expressed protein spots were identified as proteins and enzymes involved in different metabolic pathways. Among proteins included in the network analysed, three isoforms (1000, 1006, 1147; ALDH7A1) of aldehyde dehydrogenase showed a decreased expression in paclitaxel resistant cell lines. This important enzyme, overexpressed in several tumor tissues, had a determinant role in resistance to multiple chemotherapeutic agents, played through the inactivation of drugs into non cytotoxic metabolites [31,32]. Glyceraldehyde-3-phosphate dehydrogenase (906, 1523, 1626; GAPDH) and alpha enolase (1065, 1198, 1200, 1203, 1225), which showed an altered expression in paclitaxel sensitive vs. resistant ovarian cancer cells, seems to be relevant in the acquisition of drug resistance in cancer cell lines, due to their transcriptional regulation activity more than their function in glycolysis [33]. In the present study the resistance to paclitaxel is associated with the modulation of several proteins of transport functional class. As an example, a decrease in expression of two isoforms of transitional endoplasmic reticulum ATPase (601, 604; VCP) was detected. This protein, which forms a ternary complex with other endoplasmic reticulum associated degradation proteins, is necessary for the export of unfolded proteins from the ER to the cytoplasm, where they are degraded by the proteasome, thus preventing apoptosis [34,35]. Cyclophilin A, of which three isoforms (2510, 2512, 2513; PPIA) were found downregulated in paclitaxel resistant ovarian cancer cell lines compared to the sensitive cell line, belongs to the class of protein biosynthesis and folding. Overexpression of Cyclophilin A protects cancer cells against cellular stresses, including hypoxia and drug treatment, through its antioxidant activity, thus preventing apoptosis [36]. Two isoforms of 40S ribosomal protein (1255, 1398; RPSA),

233

another protein belonging to the same class as cyclophylin A and included in the pathway in analysis, were found to be modulated in paclitaxel resistant cancer cell lines. Our result is consistent with literature, which reports the association of 40S ribosomal protein with the multidrug resistant phenotype of cancer cells [37]. Another largely represented functional class is the one of cell death and apoptosis. Among proteins belonging to this class and to the network here analysed, lamins (838, 942, 953, 959; LMNA) are nuclear proteins that play an important role in the complete condensation of chromatin and nuclear shrinkage [38], thus highlighting their importance in the progression of apoptosis. In our study stathmin (2447) resulted to be downregulated in paclitaxel resistant ovarian cancer cell lines. Altered levels of stathmin expression were already reported for paclitaxel resistant ovarian cancer cells and other chemoresistance processes in different cancers [39–41]. The voltage dependent anion-selective channel, VDAC1 (1825), which was downregulated in resistant cells in our study, is an essential player in mammalian apoptosis by regulating mitochondrial membrane permeability and directly or indirectly interacting with various proteins. Through such interactions, VDAC1 probably acts as a convergence point for a variety of cellular life-or-death signals [42,43] and plays a potential role in the chemoresistance. A direct interaction of paclitaxel with VDAC1 has been previously demonstrated [14]. We also identified annexin V (1696; ANXA5), a member of annexin family, Ca2+ dependent phospholipid binding proteins. Annexins are often overexpressed in different cancers and associated with paclitaxel and other drug resistance processes [44]. Among proteins identified in our analysis, ten are proteins implicated in cytoskeleton and cell structure. For example, actin (1337; ACTB) was downregulated in ovarian cancer cells resistant to paclitaxel. It constitutes the framework of the cytoskeletal machinery, playing an important role in apoptosis and being involved in resistance to paclitaxel and other antimicrotubule agents in cancer [45]. Moreover, considering the important effects of drebrin (370) and cofilin 1 (2434) on regulating activity of actin, we speculate they may exert a resistant role through modulating the actin cytoskeleton and further inhibiting apoptotic cell death in response to chemotherapeutic agents [46]. Vimentin (1306; VIM), another key node of the network, plays an important role in the dynamic remodelling of the cell during development of neoplastic phenotype and execution of apoptosis, being one of the substrates cleaved by caspases [47]. Changes of vimentin expression may play an important role in the onset of resistance, possibly representing a molecular mechanism to replace the caspase-cleaved vimentin to counteract programmed cell death. Several proteins involved in nucleic acid synthesis were modulated in paclitaxel resistant cells, probably allowing faster DNA repair. Among these, nucleoside diphosphate kinase A (231; NME1), whose increased level of expression were associated with resistance to initial chemotherapy and reduced overall survival in cancer [48,49]. Subsequently we focused on some proteins whose expression varied between the paclitaxel sensitive and the two resistant ovarian cancer cell lines, but not between the two resistant cell lines (Table 3). In particular, we focused on five of these proteins which showed the highest change in expression between A2780 and A2780TC1 or OVCAR3: galectin-3 (1255), tubulin beta 4 (1185), clathrin heavy chain 1 (196), ubiquitin carboxyl terminal hydrolase isozyme L1 (1866) and mitochondrial isocitrate dehydrogenase (1363). Indeed we assumed that proteins which were more markedly modulated in resistant versus sensitive cell line may represent candidate biomarkers of paclitaxel resistance in ovarian cancer cells with stronger confidence. Galectin-3 (1255) is a multifunctional oncogenic protein which regulates cell growth, adhesion and proliferation, angiogenesis and apoptosis. Its overexpression in response to antitumor agents demonstrates anti-apoptotic effects, contributing to cell survival in

234

M. Di Michele et al. / Biochimica et Biophysica Acta 1794 (2009) 225–236

Table 3 Proteins differentially expressed between paclitaxel sensitive (A2780) and resistant (A2780TC1 inherent and OVCAR3 acquired) human epithelial ovarian cancer cell lines, but not between the two resistant cell lines, obtained after DIGE coupled with MALDI TOF and LC MS/MS mass spectrometry analysis Spot N°a

Protein

Functional class

Accesion Identification MASCOT Seq Observed Theoretical A2780TC1/A2780 code methodb score cov MW MW (kDa)/pI Average p value (%) (kDa)/pI ratioc

1255 Laminin binding protein (Galectin-3) Signal transduction P17931

1185

Tubulin beta, 4 variant

Cytoskeleton and cell structure

Q53G92

800

Glycyl tRNA ligase

P41250

1434

Fructose-bisphosphate aldolase A

Protein biosynthesis and folding Metabolism

2065 Peroxiredoxin-2

Stress response and chaperones

P32119

1006 Aldehyde dehydrogenase family 7 member A1

Metabolism

P49419

1385

Cell cycle and apoptosis Metabolism

P62195

Stress response and chaperones Cytoskeleton and cell structure

P30041

1005 Nuclear matrix protein NMP 238

Transcription

Q9Y265

1203 Alpha enolase

Metabolism

P06733

2434 Cofilin (non muscle isoform)

Cytoskeleton and cell structure Metabolism

Q9Y281

Metabolism

P06733

Cell cycle and apoptosis Metabolism

Q5TCJ3

Metabolism

P14618

Metabolism

P29401

Mitochondrial inner membrane protein 2533 Unnamed protein product

Cell cycle and apoptosis Unknown

Q16891

1062 H+-transporting two sector ATPase 196 Clathrin heavy chain 1

Metabolism Transport

Q5QNZ2 Q00610

1866

Ubiquitin carboxyl-terminal hydrolase isozyme L1

P09936

1363

Isocitrate dehydrogenase [NAD] subunit alpha, mitochondrial

Protein biosynthesis andfolding Metabolism

26S protease regulatory subunit 8

1000 Aldehyde dehydrogenase family 7 member A1 1705 Peroxiredoxin-6 1179

1199

Tubulin beta chain

1198

Polypeptide Nacetylgalactosaminyltransferase 13 Alpha enolase

953

Lamin A/C isoform CRA_b

2070 Butyrophilin subfamily 2 member A1 precursor 1026 Pyruvate kinase isozymes M1/M2 1017

Transketolase

656

P04075

P49419

Q9H4B7

Q8IUC8

Q7KYR7

/

P50213

MALDI TOF MS LC MS/MS MALDI TOF MS LC MS/MS MALDI TOF MS LC MS/MS MALDI TOF MS MALDI TOF MS LC MS/MS MALDI TOF MS LC MS/MS MALDI TOF MS MALDI TOF MS MALDI TOF MS MALDI TOF MS LC MS/MS MALDI TOF MS MALDI TOF MS MALDI TOF MS MALDI TOF MS MALDI TOF MS MALDI TOF MS LC MS/MS MALDI TOF MS MALDI TOF MS MALDI TOF MS MALDI TOF MS MALDI TOF MS MALDI TOF MS LC MS/MS MALDI TOF MS MALDI TOF MS MALDI TOF MS

OVCAR3/A2780 Average ratioc

p value

236

16

46/4.5

32/4.8

− 3.0

0.000033 −4.3

0.000005

230

18

51/4.8

51/4.8

− 2.4

0.000033 −2.6

0.000005

160

6

78/6.2

78/5.9

− 2.3

0.000173

−3.3

0.000024

123

48

42/8.4

40/8.4

− 2.0

0.000033 −1.6

0.000094

156

19

19/5

22/6.8

− 1.8

0.000105

−1.6

0.000071

230

13

61/7

56/6.2

− 1.8

0.000170

−2.1

0.000094

85

41

46/8.2

46/8.2

− 1.6

0.000131

−1.5

0.000210

88

35

62/7.3

56/6.2

− 1.6

0.000037 −1.9

0.000172

114

52

23/6.4

25/6

− 1.5

0.000197

−1.4

0.000198

201

9

50/4.8

50/4.8

− 1.5

0.000054 −1.3

0.000188

167

50

62/6.8

51/6

− 1.5

0.000179

−1.7

0.000065

115

39

56/6.9

47/7

− 1.3

0.000131

−1.9

0.000099

86

47

17/8.7

19/8.3

− 1.3

0.000257

−1.5

0.000041

65

12

58/7.2

60/6.1

− 1.3

0.000277

−1.7

0.000142

183

59

53/7.1

47/7

− 1.2

0.000203 −1.8

0.000099

315

19

75/7

65/6.4

1.2

0.000126

1.5

0.000012

66

24

23/7.3

34/7

1.2

0.000327

2.1

0.000013

98

36

58/8.6

58/8

1.4

0.000045

1.3

0.000160

78

15

68/7.9

68/7.9

1.5

0.000183

1.5

0.000112

109

36

93/6

80/5.7

1.5

0.000123

1.3

0.000095

87

68

15/6.7

13/9.5

1.5

0.000173

2.1

0.000014

280 95

14 19

60/9.2 185/5.8

60/9.2 190/5.5

1.6 1.6

0.000173 0.000033

1.4 1.3

0.000133 0.000027

68

38

25/5.6

25/5.3

1.8

0.000252

1.8

0.000099

64

25

44/6.2

40/6.5

2.4

0.000179

1.6

0.000250

a

Spot numbers correspond to those included in the 2D image (Fig. 1). The protein identification was performed by MALDI TOF MS and/or LC–MS/MS. In cases when both were performed, MASCOT score and sequence coverage refers to the highest score identification. c Average ratio between human epithelial ovarian cancer cell lines (A2780, A2780TC1, OVCAR3) calculated considering 4 replica gels. Only statistically significant ratios (p b 0.05) are reported; n.c. = not changed in a statistically significant way (p N 0.05). b

several types of cancer cells [50]. This is in contrast with our results which report a strong down-regulation of galectin-3 (1255) in paclitaxel resistant cell lines. However, dual activities dependent on cellular location have recently been demonstrated: cytoplasmic expression of galectin-3 prevents the apoptotic death of cancer cells

induced by a drug, while the cells expressing nuclear galectin-3 show enhanced drug-induced apoptosis [51], thus supporting our finding. One isoform of β-tubulin (1185), a cytoskeleton protein, showed a decreased expression in paclitaxel resistant cell lines. This is not surprising if considering that one of the main mechanisms of action of

M. Di Michele et al. / Biochimica et Biophysica Acta 1794 (2009) 225–236

paclitaxel in ovarian cancer is linked to the binding of specific βtubulin subunits, thus blocking cell cycle and inducing cell death. It is well known that changes in isotypes of β-tubulin class are involved in paclitaxel resistance in epithelial ovarian tumors [5,52]. Clathrin heavy chain, belonging to the functional class of transport, significantly increased in paclitaxel resistant ovarian cancer cells. Its involvement in chemoresistance process was already been found by Schmidt et al. [53], who also found this protein overexpressed in a paclitaxel-resistant cancer cell lines. Moreover, in our study, clathrin was indicated as a potential biomarker of paclitaxel resistance by the statistical analysis performed through Decyder software, indicating a putative role for this protein in the prediction of drug resistance in ovarian cancer. Ubiquitin protein ligase 1 (1866), upregulated in paclitaxel resistant cell lines, is a member of a large family of proteins engaged in the regulation of many target proteins, catalyzing the ubiquitination of protein substrates for targeted degradation through the 26S proteasome (1243, 1385), as well as for non proteolytic regulation of their functions or subcellular localizations [54]. Ubiquitin–proteasome pathway is the predominant mean of proteolysis in eukaryotic cells and its functionality is critical in the degradation of abnormal proteins that result from oxidative damage. This complex can therefore contribute to the pathological state of cancer, in which some regulatory proteins are either stabilized due to decreased degradation or lost due to accelerated degradation [54,55]. Isocitrate dehydrogenase was markedly upregulated in paclitaxel resistant compared to sensitive cell lines. Its role in chemoresistance still remains to be elucidated, however it was already found to be involved in drug resistance in ovarian [56] and other types of cancer [57]. To our knowledge, this is the first direct proteomic comparison of paclitaxel sensitive and resistant ovarian cancer cell lines. Proteomic approach based on DIGE coupled with MS used in our study proved to be an efficient tool for studying paclitaxel resistance in ovarian cancer cells by identification of proteins implicated in this phenomenon. By PCA and hierarchical clustering analysis, we were able to distinguish between paclitaxel sensitive and the two resistant protein patterns, suggesting that drug resistance could arise from common mechanisms in both natural and induced resistant cell lines. Moreover, results obtained by the use of these powerful techniques, which characterised proteins with high sensitivity, high confidence and high throughput, emphasize the complexity and multifactorial nature of the alterations associated with the development of chemoresistance. In particular we highlighted the importance of some proteins which showed a higher change in expression between paclitaxel sensitive and resistant cell lines and are therefore proposed as candidate biomarkers of chemoresistance in ovarian cancer. These data could have an impact on the development of new therapeutic strategies for ovarian cancer. Acknowledgement This work was partially supported by MIUR (Programma Triennale Ricerca, decreto 1558). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.bbapap.2008.09.017. References [1] Y. Zhu, R. Wu, N. Sangha, C. Yoo, K.R. Cho, K.A. Shedden, H. Katabuchi, D.M. Lubman, Classifications of ovarian cancer tissues by proteomic patterns, Proteomics 6 (21) (2006) 5846–5856. [2] J.P. Gagne, P. Gagne, J.M. Hunter, M.E. Bonicalzi, J.F. Lemay, I. Kelly, C. Le Page, D. Provencher, A.M. Mes-Masson, A. Droit, D. Bourgais, G.G. Poirier, Health and Environment Unit, Laval University Medical Proteome profiling of human epithelial ovarian cancer cell line TOV-112D, Mol. Cell. Biochem. 275 (2005) 25–55. [3] S.A. Cannistra, Cancer of the ovary, N. Engl. J. Med. 9, 351 (24) (2004) 2519–2529.

235

[4] N.I. Marupudi, J.E. Han, K.W. Li, V.M. Renard, B.M. Tyler, H. Brem, Paclitaxel: a review of adverse toxicities and novel delivery strategies, Expert Opin. Drug Saf. 6 (5) (2007) 609–621. [5] G. Ferrandina, G.F. Zannoni, E. Martinelli, A. Paglia, V. Gallotta, S. Mozzetti, G. Scambia, C. Ferlini, Class III beta-tubulin overexpression is a marker of poor clinical outcome in advanced ovarian cancer patients, Clin. Cancer Res. 12 (9) (2006) 2774–2779. [6] T. Fojo, M. Menefee, Mechanisms of multidrug resistance: the potential role of microtubule-stabilizing agents, Ann. Oncol. 18 (Suppl. 5) (2007) v3–8. [7] G. Hütter, P. Sinha, Proteomics for studying cancer cells and the development of chemoresistance, Proteomics 1 (10) (2001) 1233–1248. [8] P. Sinha, S. Kohl, J. Fischer, G. Hütter, M. Kern, E. Köttgen, M. Dietel, H. Lage, M. Schnölzer, D. Schadendorf, Identification of novel proteins associated with the development of chemoresistance in malignant melanoma using two-dimensional electrophoresis, Electrophoresis 21 (14) (2000) 3048–3057. [9] P. Sinha, J. Poland, M. Schnölzer, J.E. Celis, H. Lage, Characterization of the differential protein expression associated with thermoresistance in human gastric carcinoma cell lines, Electrophoresis 22 (14) (2001) 2990–3000. [10] P. Sinha, J. Poland, S. Kohl, M. Schnölzer, H. Helmbach, G. Hütter, H. Lage, D. Schadendorf, Study of the development of chemoresistance in melanoma cell lines using proteome analysis, Electrophoresis 24 (14) (2003) 2386–2404. [11] R. Marouga, S. David, E. Hawkins, The development of the DIGE system: 2D fluorescence difference gel analysis technology, Anal. Bioanal. Chem. 382 (3) (2005) 669–678. [12] M. Unlu, M.E. Morgan, J.S. Minden, Difference gel electrophoresis. A single gel method for detecting changes in protein extracts, Electrophoresis 18 (11) (1997) 2071–2077. [13] A. Kolkman, E.H. Dirksen, M. Slijper, A.J. Heck, Double standards in quantitative proteomics: direct comparative assessment of difference in gel electrophoresis and metabolic stable isotope labelling, Mol. Cell. Proteomics 4 (2005) 255–266. [14] C. Ferlini, G. Raspaglio, S. Mozzetti, M. Distefano, F. Filippetti, E. Martinelli, G. Ferrandina, D. Gallo, F.O. Ranelletti, G. Scambia, Bcl-2 downregulation is a novel mechanism of paclitaxel resistance, Mol. Pharmacol. 64 (2003) 51–58. [15] A. Della Corte, N. Maugeri, A. Pampuch, C. Cerletti, G. de Gaetano, D. Rotilio, Differential analysis by 2-dimensional difference gel electrophoresis (2D-DIGE) of proteins released from human platelets following thrombin activation: identification of lamin A, a novel platelet protein, Platelets 19 (1) (2008) 43–50. [16] A. Shevchenko, M. Wilm, O. Vorm, M. Mann, Mass spectrometric sequencing of proteins silver-stained polyacrylamide gels, Anal. Chem. 68 (5) (1996) 850–858. [17] Gene Ontology Consortium, Gene Ontology Consortium: The Gene Ontology (GO) project, Nucl. Acids Res. 34 (Suppl. 1) (2006) D322–D326. [18] M.F. Burbridge, L. Kraus-Berthier, M. Naze, A. Pierré, G. Atassi, N. Guilbaud, Biological and pharmacological characterisation of three models of human ovarian carcinoma established in nude mice: use of the CA125 tumor marker to predict antitumour activity, Int. J. Oncol. 15 (1999) 1155–1162. [19] Y. Tsuruta, M. Mandai, I. Konishi, H. Kuroda, T. Kusakari, Y. Yura, A.A. Hamid, I. Tamura, M. Kariya, S. Fujii, Combination effect of adenovirus-mediated proapoptotic bax gene transfer with cisplatin o, Eur. J. Cancer 37 (2001) 531–541. [20] N.A. Karp, M. Spencer, H. Lindsay, K. O'Dell, K.S. Lilley, Impact of replicate types on proteomic expression analysis, J. Proteome Res. 4 (5) (2005) 1867–1871. [21] Y. Benjamini, Y.J. Hochberg, Controlling the false discovery rate: a practical and powerful approach to multiple testing, Royal Statist. Soc. B. 57 (1995) 289–300. [22] V. Shridhar, K.C. Bible, J. Staub, R. Avula, Y.K. Lee, K. Kalli, H. Huang, L.C. Hartmann, S.H. Kaufmann, D.I. Smith, Loss of expression of a new member of the DNAJ protein family confers resistance to chemotherapeutic agents used in the treatment of ovarian cancer, Cancer Res. 61 (2001) 4258–4265. [23] Y. Tanaka, K. Fujiwara, H. Tanaka, K. Maehata, I. Kohno, Paclitaxel inhibits expression of heat shock protein 27 in ovarian and uterine cancer cells, Int. J. Gynecol. Cancer 14 (4) (2004) 616–620. [24] D.R. Ciocca, S.K. Calderwood, Heat shock proteins in cancer: diagnostic, prognostic, predictive, and treatment implications, Cell Stress Chaperones 10 (2) (2005) 86–103. [25] J.J. Stewart, J.T. White, X. Yan, S. Collins, C.W. Drescher, N.D. Urban, L. Hood, B. Lin, Proteins associated with cisplatin resistance in ovarian cancer cells identified by quantitative proteomic technology and integrated with mRNA expression levels, Mol. Cell. Proteomics 5 (2006) 433–443. [26] J. Yun, A. Tomida, K. Nagata, T. Tsuruo, Glucose-regulated stresses confer resistance to VP-16 in human cancer cells through a decreased expression of DNA topoisomerase II, Oncol. Res. 7 (12) (1995) 583–590. [27] R. Koomägi, J. Mattern, M. Volm, Glucose-related protein (GRP78) and its relationship to the drug-resistance proteins P170, GST-pi, LRP56 and angiogenesis in non-small cell lung carcinomas, Anticancer Res. 19 (5B) (1999) 4333–4336. [28] V.L. Kinnula, J.D. Crapo, Superoxide dismutases in malignant cells and human tumors, Free Radic. Biol. Med. 36 (6) (2004) 718–744. [29] L. González-Santiago, P. Alfonso, Y. Suárez, A. Núñez, L.F. García-Fernández, E. Alvarez, A. Muñoz, J.I. Casal, Proteomic analysis of the resistance to aplidin in human cancer cells, J. Proteome Res. 6 (4) (2007) 1286–1294. [30] A. Castagna, P. Antonioli, H. Aster, M. Hamdan, S.C. Righetti, P. Perego, F. Zunino, P.G. Righetti, A proteomic approach to cisplatin resistance in the cervix squamous cell carcinoma cell line A431, Proteomics 4 (2004) 3246–3267. [31] K.A. Yoon, Y. Nakamura, H. Arakawa, Identification of ALDH4 as a p53-inducible gene and its protective role in cellular stresses, J. Hum. Genet. 9 (3) (2004) 134–140. [32] J.P. Gagné, C. Ethier, P. Gagné, G. Mercier, M.E. Bonicalzi, A.M. Mes-Masson, A. Droit, E. Winstall, M. Isabelle, G.G. Poirier, Comparative proteome analysis of human epithelial ovarian cancer, Proteome Sci. 5 (2007) 16.

236

M. Di Michele et al. / Biochimica et Biophysica Acta 1794 (2009) 225–236

[33] S. Mori-Iwamoto, Y. Kuramitsu, S. Ryozawa, K. Mikuria, M. Fujimoto, S. Maehara, Y. Maehara, K. Okita, K. Nakamura, I. Sakaida, Proteomics finding heat shock protein 27 as a biomarker for resistance of pancreatic cancer cells to gemcitabine, Int. J. Oncol. 31 (6) (2007) 1345–1350. [34] Y. Ye, H.H. Meyer, T.A. Rapport, The AAA ATPase Cdc48/p97 and its partners transport proteins from the ER into the cytosol, Nature 414 (2001) 652–656. [35] N.W. Bays, R.Y. Hampton, Cdc48-Ufd1-Npl4: stuck in the middle with Ub, Curr. Biol. 12 (2002) 366–371. [36] K.J. Choi, Y.J. Piao, M.J. Lim, J.H. Kim, J. Ha, W. Choe, S.S. Kim, Overexpressed cyclophilin A in cancer cells renders resistance to hypoxia- and cisplatin-induced cell death, Cancer Res. 67 (8) (2007) 3654–3662. [37] Y. Shi, Y. Han, X. Wang, Y. Zhao, X. Ning, B. Xiao, D. Fan, MGr1-Ag is associated with multidrug-resistant phenotype of gastric cancer cells, Gastric Cancer 5 (2002) 154–159. [38] M. Prokocimera, A. Margalita, Y. Gruenbauma, The nuclear lamina and its proposed roles in tumorigenesis: projection on the hematologic malignancies and future targeted, J. Struc. Biol. 155 (2006) 351–360. [39] A. Johnsson, I. Zeelenberg, Y. Min, J. Hilinski, C. Berry, S.B. Howell, G. Los, Identification of genes differentially expressed in association with acquired cisplatin resistance, Br. J. Cancer 83 (8) (2000) 1047–1054. [40] R. Balachandran, M.J. Welsh, B.W. Day, Altered levels and regulation of stathmin in paclitaxel-resistant ovarian cancer cells, Oncogene 22 (55) (2003) 8924–8930. [41] E. Alli, J.M. Yang, J.M. Ford, W.N. Hait, Reversal of stathmin-mediated resistance to paclitaxel and vinblastine in human breast carcinoma cells, Mol. Pharmacol. 71 (5) (2007) 1233–1240. [42] S. Shimizu, Y. Matsuoka, Y. Shinohara, Y. Yoneda, Y. Tsujimoto, Essential role of voltage-dependent anion channel in various forms of apoptosis in mammalian cells, J. Cell. Biol. 152 (2) (2001) 237–250. [43] Y. Tsujimoto, S. Shimizu, The voltage-dependent anion channel: an essential player in apoptosis, Biochimie 84 (2–3) (2002) 187–193. [44] E. Kyu-Ho Han, S.K. Tahir, S.P. Cherian, N. Collins, S.-C. Ng, Modulation of paclitaxel resistance by annexin IV in human cancer cell lines, Br. J. Cancer 83 (1) (2000) 83–88. [45] N.M. Verrills, S.T. Po'uha, M.L. Liu, T.Y. Liaw, M.R. Larsen, M.T. Ivery, G.M. Marshall, P.W. Gunning, M. Kavallaris, Alterations in gamma-actin and tubulin-targeted

[46]

[47]

[48]

[49]

[50] [51]

[52]

[53]

[54] [55] [56]

[57]

drug resistance in childhood leukemia, J. Natl. Cancer Inst. 98 (19) (2006) 1363–1374. H. Kajiyama, K. Shibata, M. Terauchi, M. Yamashita, K. Ino, A. Nawa, F. Kikkawa, Chemoresistance to paclitaxel induces epithelial–mesenchymal transition and enhances metastatic potential for epithelial ovarian carcinoma cells, Int. J. Oncol. 31 (2) (2007) 277–283. S. Peñuelas, V. Noé, C.J. Ciudad, Modulation of IMPDH2, survivin, topoisomerase I and vimentin increases sensitivity to methotrexate in HT29 human colon cancer cells, FEBS J. 272 (3) (2005) 696–710. S. Chen, X. Liao, J. Pan, K. Yang, C. Hua, Y. Wen, In vitro study of the effect of nm23H1 on metastasis ability and chemo-sensitivity of Acc-M cell lines, Sichuan Da Xue Xue Bao Yi Xue Ban 34 (4) (2003) 628–630. Y. Liu, H. Liu, B. Han, J.-T. Zhang, Identification of 14-3-3S as a contributor to drug resistance in human breast cancer cells using functional proteomic analysis, Cancer Res. 66 (6) (2006) 3248–3255. T. Fukumori, H.O. Kanayama, A. Raz, The role of galectin-3 in cancer drug resistance, Drug Resist Updat. 10 (3) (2007) 101–108. S. Califice, V. Castronovo, M. Bracke, F. van den Brule, Dual activities of galectin-3 in human prostate cancer: tumor suppression of nuclear galectin-3 vs. tumor promotion of cytoplasmic galectin-3, Oncogene 23 (2004) 7527–7536. M. Kavallaris, D.Y. Kuo, C.A. Burkhart, D.L. Regl, M.D. Norris, M. Haber, SB. Horwitz, Taxol-resistant epithelial ovarian tumors are associated with altered expression of specific beta-tubulin isotypes, J. Clin. Invest. 100 (5) (1997) 1282–1293. M. Schmidt, G. Schler, P. Gruensfelder, F. Hoppe, Differential gene expression in a paclitaxel-resistant clone of a head and neck cancer cell, Eur. Arch. Otorhinolaryngol. 263 (2) (2006) 127–134. Y. Sun, E3 ubiquitin ligases as cancer targets and biomarkers, Neoplasia 8 (8) (2006) 645–654. A. Ciechanover, K. Iwai, The ubiquitin system: from basic mechanisms to the patient bed, IUBMB Life 56 (4) (2004) 193–201. X.D. Yan, L.Y. Pan, Y. Yuan, J.H. Lang, N. Mao, Identification of platinum-resistance associated proteins through proteomic analysis of human ovarian cancer cells and their platinum-resistant sublines, J Proteome Res. 6 (2) (2007) 772–780. L. Smith, K.J. Welham, M.B. Watson, P.J. Drew, M.J. Lind, L. Cawkwell, The proteomic analysis of cisplatin resistance in breast cancer cells, Oncol Res. 16 (11) (2007) 497–506.