Candidate Genes for Cross-Resistance Against DNA-Damaging Drugs1,2

[CANCER RESEARCH 62, 6698–6705, November 15, 2002] Candidate Genes for Cross-Resistance against DNA-damaging Drugs1,2

Rainer Wittig, Michelle Nessling, Rainer D. Will, Jan Mollenhauer, Ru¨diger Salowsky, Ewald Mu¨nstermann, Matthias Schick, Heike Helmbach, Brigitte Gschwendt, Bernhard Korn, Petra Kioschis, Peter Lichter, Dirk Schadendorf, and Annemarie Poustka3 Department of Molecular Genome Analysis [R. W., R. D. W., J. M., R. S., E. M., P. K., A. P.], Department of Molecular Genetics, [M. N., P. L.], and Skin Cancer Unit [H. H., B. G., D. S.], Deutsches Krebsforschungszentrum, and Resource Center for Genome Research [M. S., B. K.], D-69120 Heidelberg, Germany

ABSTRACT has experienced an increasing attention because of its high and still rising incidence, and the poor prognosis associated with this particular Drug resistance of tumor cells leads to major drawbacks in the treat- kind of skin cancer (11). ment of cancer. To identify candidate genes for drug resistance, we In general, the response of malignant melanoma to chemotherapy is compared the expression patterns of the drug-sensitive human malignant melanoma cell line MeWo and three derived sublines with acquired poor, and an improvement of therapeutic protocols is highly desirable. resistance to the DNA-damaging agents cisplatin, etoposide, and fotemus- Several well-characterized drug resistance mechanisms have been tine. Subarray analyses confirmed 57 candidate genes recovered from a studied in regard to their importance for malignant melanoma. Diver- genome-wide scan for differential expression. By specifically addressing gent results have been obtained by these analyses indicating that the cancer genes we retrieved another set of 209 candidates. Exemplary molecular mechanisms underlying drug resistance in melanoma may Northern blot studies indicated qualitative concordance for 110 of 135 be multifaceted and poorly understood (12, 13). (81.4%) data points. Whereas the etoposide-resistant line showed constant The malignant melanoma cell line MeWo has been treated with expression patterns over a period of ϳ2.5 years, the fotemustine- and various cytotoxic compounds, among these the commonly used anti- cisplatin-resistant sublines exhibited considerable variability. Initially cancer drugs cisplatin, fotemustine, and etoposide, to give rise to representing distinct entities, these two sublines finally converged in their sublines with resistance to different concentrations of the respective expression patterns. A total of 110 genes was transiently or permanently deregulated in at least two resistant sublines. Fourteen genes displayed drugs (14). Whereas cisplatin and fotemustine are thought to form differential expression in all three of the sublines. We hypothesize that the DNA adducts, etoposide interferes with topoisomerase II resulting in variations in fotemustine and cisplatin resistance are based on progressive double strand breaks (5, 15–17). Thus, the three drugs have in com- optimization and/or polyclonality. This, in addition to genomic alterations mon the ability to either directly or indirectly lead to DNA damage. investigated by comparative genomic hybridization and evaluation of The resistant cell lines have been characterized in regard to their short-term response genes, can be used as a criterion for the selection of pharmacological properties and genomic alterations (14, 18). Differ- promising candidates. Among these are CYR61, AHCYL1, and MPP1, as ences in gene expression have initially been determined using differ- well as several apoptosis-related genes, in particular STK17A and CRYAB. ential display reverse transcription-PCR and two-dimensional protein As MPP1 and CRYAB are also among the 14 genes differentially expressed gel electrophoresis (19, 20). These efforts have retrieved 15 primary in all three of the drug-resistant sublines, they represent the strongest candidate genes for resistance against the three DNA-damaging drugs, candidates for resistance against DNA-damaging drugs. of which one has been confirmed to date by functional assays (21). The sublines with resistance to DNA-damaging agents have been INTRODUCTION shown to display alterations of pathways involved in the maintenance of DNA integrity (22–25). Drug resistance represents a major problem in cancer therapy. The In specific regard to melanoma therapy cisplatin, but not fotemus- use of in vitro systems has been valuable for the identification and tine or etoposide, presently plays a central role (26). However, the functional verification of several resistance mechanisms. The energy- three MeWo sublines with the highest levels of resistance to cisplatin, dependent drug efflux mediated by members of the ATP binding (MeWoCis1), fotemustine (MeWoFote40), and etoposide (MeWoEto1), cassette transporter protein superfamily (1, 2) is one of the most offer the opportunity to reveal communities in the resistance to important mechanisms for multidrug resistance, and the modulation of DNA-damaging drugs under standardized conditions. Because this is this pathway is in the focus of intense research (3). Other well- of importance for a series of cancer types, we conducted a compre- characterized mechanisms of cellular drug resistance include the de- hensive and detailed analysis of differential gene expression specifi- toxification of drugs via glutathione conjugation (4) and alterations in cally in these three MeWo sublines. the activities or properties of DNA topoisomerases (5). More recently, cellular drug resistance has been associated with alterations in cell death pathways (6, 7), but to date the exact mechanisms remain MATERIALS AND METHODS unclear. Highly parallelized gene expression analyses have been used Cell Culture and Cell Lines. Drug-resistant derivatives of MeWo had to identify additional resistance-associated genes (8–10), because the been selected as described by Kern et al. (14). MeWo cells were grown in presently known mechanisms cannot sufficiently explain the resist- RPMI 1640 supplemented with 10% FCS, the antibiotics penicillin and strep- ance of several tumors. Among these is malignant melanoma, which tomycin (100 units/ml each), and L-glutamine (2 mM). Stable chemoresistant MeWo sublines were grown in supplemented RPMI 1640 with the respective ␮ Received 5/6/02; accepted 9/13/02. drugs: MeWoCis1 with additional cisplatin (1 g/ml), MeWoEto1 with addi- ␮ The costs of publication of this article were defrayed in part by the payment of page tional etoposide (1 g/ml), and MeWoFote40 with additional fotemustine (40 charges. This article must therefore be hereby marked advertisement in accordance with ␮g/ml). In general, cells were grown to 80–90% confluence, and then har- 18 U.S.C. Section 1734 solely to indicate this fact. 1 vested for RNA and DNA isolation. Drug treatment of MeWo cells occurred Supported by the Deutsche Forschungsgemeinschaft Grants Scha 422/7-2 and Scha ϳ 422/7-3 (to D. S.), and the Forschungsfond Mannheim (to D. S.). at 80% confluence for 24 h with the concentrations depicted above. In 2 Supplementary data in this article are available at Cancer Research Online (http:// parallel, MeWo control cells were cultivated for the same time without drugs. cancerres.aacrjournals.org). For the determination of the effects of etoposide treatment on gene expression 3 To whom requests for reprints should be addressed, at Department of Molecular in MeWo , we grew cells with and without etoposide supplementation. For Genome Analysis, Deutsches Krebsforschungszentrum, Im Neuenheimer Feld 280, 69120 Eto1 Heidelberg, Germany. Phone: 49-6221-424742; Fax: 49-6221-423454; E-mail: the array hybridizations, RNA was prepared from at least three different time [email protected]. points (t1-t3) for each of the cell lines. The cell lines were reconstituted from 6698

Downloaded from cancerres.aacrjournals.org on September 25, 2021. © 2002 American Association for Cancer Research. DRUG RESISTANCE IN MELANOMA frozen stocks collected during continuous growth over a period of ϳ2.5 years. imaging plates. In the latter case image acquisition was done using a Fuji For the sensitive MeWo reference, the time point t1 represents the parental FLA3000 phosphorimager and AIDA software. cells from which the drug-resistant sublines were established. The time points CGH. Cells were lysed by incubation in 50 mM Tris, 100 mM EDTA, and t2 and t3 represent cells that were continuously cultured for an additional 23 200 mM NaCl (pH 9) with 1% SDS and 0.5 mg/ml proteinase K at 37°C and 42 months, respectively. For all of the drug-resistant sublines, t1 denotes overnight. Genomic DNA was purified by a standard phenol-chloroform the time point of the definition of drug resistance (14). The nomenclature for extraction procedure and subsequent ethanol precipitation. Genomic DNA was M M additional time points of drug-resistant sublines is as follows: MeWoCis1, t2: 12 reconstituted in 10 m Tris and 1 m EDTA (pH 7.5). Probe preparation, hybridization, image acquisition, and analysis of CGH were performed as months and t3: 31 months; MeWoFote40, t2: 11 months and t3: 30 months; and described previously (28). MeWoEto1, t2: 3 months, t3: 22 months, and t4: 46 months after definition of 4 ϩ drug resistance. In CGH analyses, t1 of MeWoCis1 and MeWoFote40 corresponds to t1 ϩ 8–9 months. RESULTS cDNA Arrays. The genome-wide cDNA array (RZPD p950 UniGene1) was manufactured as described previously (27). The drug resistance subarray Genome-wide Screen for Differentially Expressed Genes in was composed of the 126 candidate genes retrieved from the initial screening Drug-resistant Melanoma Cell Lines. We used UniGene filter ar- of the UniGene array, 52 control genes, and 1143 genes amplified from the rays with ϳ31,500 elements (RZPD p950 UniGene 1; Ref. 27) for a RZPD Onco-library. The entire RZPD Onco-library comprises ϳ2800 cancer genome-wide analysis of differentially expressed genes in drug-resis- candidate genes. For the subarray, cluster-specific oligonucleotides (based on tant MeWo sublines. For the initial selection of drug resistance can- the NCBI UniGene built 90, September 1999) were selected for the amplifi- didate genes, mRNA obtained from three different time points of cation of 300-bp cluster segments that are free of low complexity regions and cultivation under permanent selective pressure (t1 to t3, see “Materials Ј corresponding to 3 termini of mRNA molecules. For the PCR amplification and Methods”) was separately compared with an mRNA pool of three we used pools of plasmid DNA (1 ng/reaction, 96–384 different template different time points (t1 to t3, see “Materials and Methods”)ofthe species per pool) as templates. PCR was performed in 96-well plates for 30 long-term cultivated, drug-sensitive parental MeWo cell line. This cycles (20 s 94°C, 15 s 46°C, and 30 s 72°C). These elements were spotted in a3ϫ 3 pattern on a 7 ϫ 10 cm nylon membrane. All of the PCR products were was intended to select for stable alterations in the expression pattern spotted in duplicate as described before (27) to provide internal controls. For of the drug-resistant sublines and against genes that may be transiently image processing, each block was provided with a guide spot that contains deregulated because of differences in the cell culture conditions. heterologous DNA from the bacterial kanamycin resistance gene. Genes showing distorted ratios of least 2.5 or 1/2.5, respectively, in at Complex cDNA Hybridization. mRNA was isolated using the RNeasy kit least two of three comparisons were selected for additional evaluation. and the Oligotex mRNA kit (Qiagen) according to the manufacturer’s recom- This resulted in an initial number of 139 candidate genes, of which 87 mendations. The mRNA integrity was confirmed by electrophoresis in dena- were retrieved from MeWoEto1. From MeWoCis1 and MeWoFote40, turing 1% agarose gels with formaldehyde, blotting to Hybond Nϩ Nylon respectively, only 23 and 29 candidate genes were recovered. Seven membranes (Amersham-Pharmacia Biotech), and hybridization with radiola- genes were deregulated in two resistances, whereas 3 genes displayed beled dT18V oligonucleotide. The generation of complex cDNA probes was an overlap in all three of the sublines, so that, after elimination of performed as described previously (27). After purification, the specific activity these redundancies, a final set of 126 genes was obtained. of the probes was determined, and volumes of the probes were adjusted to Candidate Gene Evaluation. To additionally confirm and evalu- ascertain equivalent conditions for the array hybridizations. The arrays were ate candidate genes, a drug resistance subarray with a final number of hybridized under the conditions described before (27). UniGene arrays were hybridized at 65°C, whereas the drug resistance subarray was hybridized at 1321 elements was composed. These elements included the 126 63°C to adjust for the smaller average fragment length. candidate genes retrieved from the initial screen, 52 control elements Image Acquisition and Analysis of Array Hybridizations. For RZPD for monitoring the hybridization quality, and 1143 additional elements UniGene 1 arrays, image acquisition, grid alignment, and spot quantification from the RZPD Onco-library. The latter ones were included to spe- was performed as described previously (27). Spot by spot comparison, nor- cifically and more sensitively address known and putative cancer malization, and determination of differential expression was carried out using candidate genes in regard to their differential expression. For the fdiffs, an algorithm developed by Tim Beissbarth (Theoretical Bioinformatics, complex hybridizations, we used the same mRNA populations as in Deutsches Krebsforschungszentrum) based on the matlab package (Math- the initial analyses, i.e., t1 to t3 for each resistant subline. Works, Natick, MA). For the analysis of the drug resistance subarrays we used The subarray analyses indicated consistent changes in the expres- ArrayVision Software (Imaging Research). Here, after spot finding using an sion patterns of all three of the time points investigated for MeWoEto1. automated algorithm, signal intensities were calculated as mean pixel values Stable gene deregulation in MeWo was additionally confirmed by minus a regional background calculated for each 3 ϫ 3 spot group, respec- Eto1 analysis of mRNA of a fourth time point (t4). In contrast, the different tively. For array comparisons, signal intensities were then normalized to the time points of sublines MeWo and MeWo showed substan- mean of all human cDNA containing spots on one filter, and expression ratios Cis1 Fote40 for single spots were calculated. Ratios of 2.5 and 1/2.5 were set as thresholds tial variations. This was confirmed by two to three independent to identify differentially expressed genes from the UniGene arrays. For the hybridization experiments as well as by Northern blot analyses of drug resistance subarray, we used ratios of 1.5 and 1/1.5 as the threshold values selected candidates (Fig. 1). Thus, we concluded that the differences to adapt to the decrease in the average size of the array elements. in the stability of the transcriptomes had led to a differential efficacy Northern Blot Analyses. Fifteen ␮g of total RNA was separated by in recovering candidates from the three sublines. To adjust to this electrophoresis in 1% agarose gels under denaturing conditions in the presence phenomenon, we applied subline-specific criteria to identify primary M of 2.2 formaldehyde, and stained with SYBR Green II (FMC, Rockland, candidates. For MeWoEto1, only genes showing consistent quantita- ME) for visualization under UV light. RNA was transferred to Hybond Nϩ tive changes in all of the data points for three of four analyzed time Nylon membranes (Amersham-Pharmacia Biotech) by capillary transfer over- points were scored as primarily confirmed candidates. For MeWoCis1 night and subsequently immobilized by UV cross-linking. Hybridizations were and MeWoFote40, genes displaying consistent changes in at least five performed with 32P-labeled cDNA probes in 500 mM Na HPO (pH 7.2), 7% 2 4 of six data points for t1 and t2 and in at least three of four data points SDS, and 10 mM EDTA at 65°C overnight. After washing (2 ϫ 10 min in for t3, respectively, were scored as primarily confirmed candidates. 0.5 ϫ SSC, 0.1% SDS at 65°C) membranes were exposed to X-ray films and Using these subline-specific criteria, 57 of the initial 126 genes (45%) were confirmed as being deregulated in at least one of the drug- 4 The abbreviations used are: CGH, comparative genomic hybridization; RZPD, Re- source Center for Genome Research; NCBI, National Center for Biotechnology Informa- resistant sublines. Remarkably, 209 of the 1143 (18%) included tion; HGNC, Human Genome Organisation Gene Nomenclature Committee. Onco-library elements were suggested to be differentially expressed 6699

Northern blot analysis using a PDE3A open reading frame sequence as a probe, a 7-kb transcript was detected (data not shown) pointing to the 7-kb PDE3A transcript as the gene identified as differentially regulated in the etoposide- and fotemustine-resistant cell lines. In the case of MPP1 (Fig. 1C), array data matched to the expression of the 2.2-kb variant, which is likely to represent the annotated mRNA (NM_002436). Both recently reported variants of the PDCD6IP gene

(29) displayed a down-regulation in MeWoEto1, which was congruent with the array results (Fig. 1O; Fig. 2F). For EGR1 (Fig. 1H), the expression pattern of the 3.2-kb variant correlated with the array data, and the size of this variant corresponded to the annotated mRNA sequence (X52541). The STMN3 gene (Fig. 1J) had been reported to give rise to a 2.3-kb transcript (NM_015894), which could be detected in Northern blot analysis. The quantitative changes in the levels of this transcript corresponded to the array data. For these 5 genes, other hybridizing bands may represent alternative splice forms, polyadenyl- ation variants, or closely related genes. In the remaining 10 cases, the detected transcript sizes corresponded to the annotated mRNA se- quences in GenBank. Genotype-Transcriptome Correlations. A comparison between the original CGH data (18) and CGH data obtained from the resistant sublines at t3 indicated consistent genomic alterations on eight chromosomes (Fig. 3). The specific genomic alterations allowed verify- ing the integrity of the respective sublines. We determined the cyto- genetic localizations of the candidate genes by screening of NCBI5 (LocusLink; UniGene), RZPD,6 and Genome7 databases. The differential expression of 28 genes could be correlated with gains and losses of the respective chromosomal regions (Table 1). Cluster and Northern Blot Analyses Reveal Convergence of Expression Patterns in Cisplatin and Fotemustine Resistance. On the basis of the 266 differentially expressed genes, we performed cluster analysis (8, 30, 31) using the freely available8 GeneCluster and TreeView software (30) to reveal common patterns of alterations. In

accordance with the previous results, t1 to t4 of MeWoEto1 located

within one cluster (Fig. 2, columns 17–27). MeWoCis1t1 (Fig. 2,

columns 14–16) was related to MeWoEto1 according to the clustering,

whereas MeWoFote40t1 and t2 (Fig. 2, columns 01–06) represented a

Fig. 1. Verification of differential expression of selected drug resistance candidate distinct entity. However, MeWoCis1t2/t3 (Fig. 2, columns 09–13) and genes by Northern blot analysis. Radiolabeled cDNA fragments specific for the candidate MeWo t3 (Fig. 2, columns 07–08) deviated from their original genes depicted in rows A–O (nomenclature corresponds to official HGNC gene symbols) Fote40 were hybridized to Northern blots containing total RNA. The RNAs were taken from patterns and appeared to finally converge in regard to their expression different time points (t1 to t3) of cultivation of the resistant sublines and untreated MeWo. patterns. We verified these results by Northern blot analysis of se- ϩ Ϫ MeWoEto1 time point t4 was either cultivated in presence ( ) or absence ( ) of etoposide. lected candidates. Twelve of 15 genes showed consistent quantitative In addition, RNA from 24 h drug-treated MeWo cells (Lane 8:Cϩ, cisplatin; Lane 12: Eϩ, etoposide; Lane 18:Fϩ, fotemustine) and the respective untreated control (Lane 4) changes over all MeWoEto1 time points analyzed (for example was included in the analyses. The panels denoted as 28S show 28S rRNA bands of the ID133864, APOD, CRYAB, and AHCYL1 in Fig. 1) Six of the 15 RNA gels before blotting to demonstrate the RNA integrity and equal loading. Arrows at the right mark differently sized bands obtained for five of the genes. The respective sizes genes displayed convergence in their expression levels at the later in kb are specified at the right of the panels. time points when comparing MeWoCis1 and MeWoFote40 (see ID133864, APOD, CRYAB, PEPP2, PLAB, IFITM3 in Fig. 1). compared with the sensitive parental MeWo cell line. Thus, a total of CRYAB, one of the genes represented and up-regulated in both groups, 266 candidate genes resulted from these efforts (for the primary data, has been related to the repression of apoptosis (32). see supplementary data2). We selected a subset of 15 genes for Evaluation of Short-Term Response and Cross-Resistance Can- additional verification by Northern blot analyses. The expression of didate Genes. To distinguish between stably deregulated genes and these candidates was analyzed for the three drug-resistant sublines at genes deregulated because of the presence of the drugs, we conducted time points t1 to t3 so that a total of 135 data points was investigated. a series of additional experiments. At first, we included RNA samples For 110 of the 135 data points (81.4%) we obtained qualitatively of the sensitive parental MeWo cell line treated for 24 h with cisplatin concordant results (Fig. 1; Fig 2F), confirming the applicability of the and fotemustine in the Northern blot analyses. Secondly, we addition- subline-specific criteria that were set previously. ally carried out Northern blot and subarray analyses with 24-h etopo- For 5 of the 15 genes analyzed, we obtained more than one side-induced MeWo cells as well as with MeWoEto1 omitting the hybridizing band (Fig. 1, A, C, H, J, O). The clone ID133864 mapped etoposide from the culture medium. The latter subarray analyses to two UniGene clusters, Hs.24305 and Hs.191045, and therefore was represented by two elements on the drug resistance subarray (Fig. 1A). 5 Internet address: http://www.ncbi.nlm.nih.gov/. 6 Internet address: http://www.rzpd.de/. By reverse transcription-PCR analyses, we identified the sequence of 7 Internet address: http://www.gdb.org/. ID133864 as part of the 3Ј-untranslated region of the PDE3A gene. In 8 Internet address: http://rana.lbl.gov/EisenSoftware.htm. 6700

Fig. 2. Cluster analysis. Different array elements (2 per gene) are displayed in rows. Pairwise comparisons between a sensitive reference probe and different resistant or drug-treated cell populations are depicted in columns. Up-regulation of expression compared with the reference is depicted by nuances of red, down-regulation by nuances of green color. Black color reports equal expression levels, and gray color marks fields for which the data were not taken into consideration for cluster analyses because of divergent expression ratios within duplicate spots (Ͼ50% of the larger value) or immeasurable expression levels. For the raw data please refer to the supplementary material. Gene nomenclature corresponds to HGNC, NCBI LocusLink IDs (LL), and the annotated chromosomal localizations are listed. The dendrograms at the top of the two panels base on the cluster analysis of the entire set of 266 genes. The degree of similarity between the analyzed samples corresponds to vertical branch lengths of the evolutionary tree, which is generated as a graphical result of pairwise similarity assessment. The different time points (t1 to t3) of the resistant cell lines MeWoFote40 (columns 01–08), MeWoCis1 (columns 09–16), and MeWoEto1 [columns 17–27, t4 cultivated in the presence (ϩ) and absence (Ϫ) of etoposide] were compared with the sensitive reference. Columns 28–30 depict the results of the comparison of 24 h etoposide-treated MeWo cells to the untreated control. A–D, expression patterns of cross-resistance candidate genes. A, genes deregulated in MeWoCis1 and MeWoFote40; B, genes deregulated in MeWoCis1 and MeWoEto1; C, genes deregulated in MeWoEto1 and MeWoFote40; D, genes deregulated in all three drug resistant sublines. E, expression patterns of deregulated genes with relationship to apoptosis. F, expression patterns of the differentially expressed genes that have been verified by Northern blot analyses shown in Fig. 1.

indicated no major influence of etoposide on the overall expression course of the short-term response to all three of the DNA-damaging pattern in MeWoEto1, as suggested by clustering (Fig. 2, columns drugs. The down-regulation of all 3 of the genes became manifested 17–24). In contrast, the patterns observed for short-term treated in the long-term resistant cell line MeWoEto1, and 2 of these (CYR61 MeWo substantially differed from both untreated MeWo and drug- and the 4.5-kb transcript of STMN3) were also transiently down- resistant MeWoEto1 (Fig. 2, columns 28–30). These data suggested regulated at one of the time points in MeWoCis1. The transcriptional that genes deregulated because of the short-term response to the repression of the 3 genes in MeWoEto1 was maintained even in the presence of the drug are represented only to a minor extent in the absence of etoposide (Fig. 1). Moreover, 2 of these short-term re- candidate panel. Basically, these data were confirmed by the Northern sponse genes mapped close to a region at chromosome 1 identified as blot analyses. Twelve of the 15 genes showing quantitative differ- genomic loss in MeWoEto1 by CGH (Table 1 Fig. 3). CYR61 was ences in the resistant sublines were not responsive to short-term drug located at 1p31-p22 (1p22.3 according to GeneCards), whereas treatment. Remarkably, however, 3 of the 15 genes (AHCYL1, CYR61, AHCYL1 had been mapped to 1p12. and the 4.5-kb variant of STMN3) displayed a down-regulation in the To identify potential cross-resistance candidates, we compared the 6701

Fig. 3. CGH. Regions with chromosomal alterations are marked by colored bars (MeWo, gray; MeWoEto1, green; MeWoCis1, red; MeWoFote40, blue) beneath the particular chromosome ideograms. The CGH ratio profiles at the different time points are arranged at the right of the respective ideograms and are averaged from data of n metaphase cells. For the time point definitions refer to “Materials and Methods.” In the profiles, the central black line represents the balanced value 1. The adjacent lines represent the thresholds for under representation (0.75, red line) and overrepresentation (1.25, green line). Additional lines are arranged in 0.25 ratio value intervals. The chromosomal region 1p32-p36 (marked by black dots) as well as the gray shaded areas of tandem repetitive DNA clusters were ex- cluded from the evaluation, because no representative flu- orescence intensities could be measured in these regions because of suppression with Cot-1 DNA (28).

data obtained for the three different drug-resistant sublines from the changes were persistent in MeWoEto1, these genes again displayed the subarray analyses. This revealed substantial overlaps in the differen- lower levels of the parental MeWo cells in MeWoCis1t2/t3 (Fig. 2B). tial expression of genes during acquired drug resistance. In total, 110 Secondly, several stress-inducible genes, e.g., PLAB, HSPA5, differentially expressed genes showed overlaps in their deregulation HSPA9B, DDIT3, and GADD45A, were often down-regulated in the

(Fig. 2, A–D). Fourteen of these were found to be deregulated in all drug-resistant sublines, to the major part in MeWoEto1 (Figs. 1 and 2). three of the drug-resistant cell lines (Fig. 2D). Remarkably, 12 genes belonging to apoptosis-related pathways were Arrangement According to Common Pathways Points to Fre- commonly found to be deregulated, as for example CRYAB, DFFA, quent Deregulation of Apoptosis-related Genes. Subsets of the PDCD6IP, or SH3BP5 (Fig. 1; Fig. 2E). deregulated genes could be identified to belong to common pathways. For example, 6 genes, G1P3, ISG15, IFI27, IFITM1, IFITM3, and DISCUSSION PRKR, represent a group of IFN-inducible genes. These genes were highly overexpressed in MeWoEto1 and in MeWoCis1t1, which is in Drug resistance is considered to represent a complex pleiotropic agreement with the data obtained for a paclitaxel-resistant subline of response (5, 34–36). In vitro systems can reflect fundamental mech- the breast cancer cell line MCF-7 (33). However, whereas these anisms of acquired drug resistance under simplified but standardized 6702

Table 1 Genes with correlation between alterations at the genomic and transcriptional level The corresponding CGH profiles for affected regions are depicted in Fig. 3. Expression data for all deregulated genes are disposed in the supplementary material. Ϫ, genomic losses; ϩ, genomic gains; o, not changed in drug resistant sublines. Genomic alteration Correlating differential gene expression

Region MeWoEto1 MeWoFote40 MeWoCis1 MeWoEto1 MeWoFote40 MeWoCis1 7p Ϫ ooRALA, STK17A, ARL4, CDC10, HIBADH, GARS 12p ϩϪϪPDE3A MGST1, HSPD1, PEPP2, LDHB, PEPP2, KRAS2, KRAS2, DKFZP564O1863 DKFZP564O1863 1p Ϫ ooGTF2B, CYR61 3p Ϫ ooRAF1, OSBPL10, SH3BP5, TOP2B 9p o ϪϪ NFIB, TYRP1, FANCG, SMARCA2 15q o ϩϩ COX5A COX5A Xq o ϩ o ELF4, CETN2, HPRT1, MPP1 experimental conditions. When investigating primary tumor cells, being resistant to exposure to the respective drugs, the two sublines parameters introducing variations are clonal heterogeneity of the showed considerable variations in their expression patterns over the tumor cells, variable genetic backgrounds of donors, contamination time course. However, we speculate that one can take several advan- with other cell types, and cross-talk with other nontumor structures. tages out of this phenomenon. Although extrapolation to the in vivo situation is linked with certain At first, both cluster analyses and Northern blot studies suggest limitations, because additional factors have to be taken into consid- distinct changes in the gene expression profiles of MeWoCis1 and eration, the limited sources of external variations in in vitro systems MeWoFote40 at early time points but convergence at later time points. provide important entry points to the mechanisms of drug resistance We hypothesize that this convergence reflects progressive optimiza- (1–5). We selected a well-defined in vitro system (14) to comprehen- tion of the drug resistance response. Moreover, convergence because sively analyze differential gene expression and genomic alterations of optimization would be in line with the similar mode of action of during acquired resistance to DNA-damaging drugs. It is remarkable cisplatin and fotemustine (15–17). Indeed, earlier studies support the that, independent of the perspective, deregulation of apoptosis-related view that MeWoCis1 and MeWoFote40 exert cross-resistance against genes appears to play a role. cisplatin and fotemustine (14). Thus, converging genes identify as Starting from a genome-wide scan for differentially expressed strong candidates for drug resistance. PEPP2 and CRYAB, for exam- genes, we identified 57 genes confirmed by subarray analyses. Inclu- ple, are noteworthy. PEPP2, which has been suggested to take part in sion of genes of the Onco-library to more specifically address known phosphoinositide-mediated signaling (37), is successively silenced and putative cancer genes retrieved an additional set of 209 candi- over the time course. In contrast, CRYAB, an antagonist of caspase dates. The recovery of the latter genes presumably bases on the 3-mediated apoptosis (32), is consecutively up-regulated. redefinition of the parameters for the subarray analyses, which was Secondly, however, progressive optimization cannot completely necessary to adapt for the decrease of the mean element length, and explain the variations in the expression pattern of MeWoFote40. When for the fluctuations in MeWoCis1 and MeWoFote40. Because Northern overlaying the expression levels of the genes PLAB, SH3BP5, blot analyses suggested an 81.4% concordance with the subarray PDE3A, and CRYAB in t2 and t3 (Fig. 1), the combinatorial pattern analyses for the genes tested, we hypothesize that a major subset of agrees with the one obtained for t1. Thus, most probably, a split of the the 266 genes is truly differentially expressed in the drug-resistant cell population has taken place. Consequently, MeWoFote40 was poly- sublines. Twelve of these genes are involved in the modulation of clonal at t1. Possibly, differential reconstitution from the frozen stocks apoptosis pathways. Three of these originated from the 57 confirmed has led to shifts in the otherwise stable polyclonal population. Under UniGene clones, whereas 9 derived from the 209 genes recovered this premise, genes with differential behavior in t2 and t3 would from the Onco-library. This suggests that by using genes from the participate in alternative pathways of fotemustine resistance. A very Onco-library for the subarray no bias toward the isolation of apopto- suggestive example is that t3 shows an activation of the apoptosis sis-related genes was introduced. antagonist CRYAB, whereas it is absent from t2. Strikingly, however, The genes recovered may basically subdivide into two categories t2 shows a repression of the proapoptotic genes CRADD (38), RIPK1 with differential importance for drug resistance itself. The first cate- (39), DFFA (40), and DAPK3 (41), which, in fact, could be able gory comprises genes that could directly be related to the phenotypes. to compensate for the lack of the antagonistic action of CRYAB The second category may represent genes of which the deregulation is (Fig. 2E). rather caused by genomic rearrangements or coactivation, or core- We determined correlations between chromosomal aberrations and pression because of deregulation of transcription factors. In fact, at up- or down-regulation for 28 genes. This is particularly helpful for least 35 of 266 deregulated genes have known or putative functions in candidate evaluation. STK17A, which has a stimulatory effect on the regulation of transcription. To date, a clear distinction between apoptosis (42), is down-regulated in MeWoEto1 and locates to chro- genes belonging to these two groups is not possible, because this mosome 7p that has undergone a loss in this subline. The perfect would require assays for high throughput screening. However, based agreement between genomic, expression, and functional data points to on the design of our experiments and the definition of subgroups, we this gene as a considerably strong candidate. MPP1 at Xq28 is to date can encircle the most promising candidates. poorly characterized at the functional level. Its up-regulation in ϳ Primarily, the comparison of the expression patterns over 2.5 MeWoFote40 attributes to the amplification of the respective genomic years may point to certain limitations, even within this well-defined in locus, which was additionally confirmed by Southern blot analyses vitro system of acquired drug resistance. The constant deregulation of (data not shown). In MeWoEto1 and MeWoCis1 a distinct mechanism some genes as well as the consistency of the genomic alterations must be responsible for the MPP1 up-regulation, because neither confirmed the integrity of MeWoCis1 and MeWoFote40. Although CGH nor additional Southern blot studies (data not shown) suggested 6703

Downloaded from cancerres.aacrjournals.org on September 25, 2021. © 2002 American Association for Cancer Research. DRUG RESISTANCE IN MELANOMA the presence of an amplification at the gene locus. Thus, this phe- 6. Los, M., Herr, I., Friesen, C., Fulda, S., Schulze-Osthoff, K., and Debatin, K. M. nomenon argues for a role in multidrug resistance and against being a Cross-resistance of CD95- and drug-induced apoptosis as a consequence of deficient activation of caspases (ICE/Ced-3 proteases). Blood, 90: 3118–3129, 1997. bystander. Correlation with genomic data also allows resolving ap- 7. Makin, G., and Hickman, J. A. Apoptosis and cancer chemotherapy. Cell Tissue Res., parent contradictions between functional data and the direction of 301: 143–152, 2000. 8. Scherf, U., Ross, D. T., Waltham, M., Smith, L. H., Lee, J. K., Tanabe, L., Kohn, deregulation. For example, RAF1, an antagonist of apoptosis (43), is K. W., Reinhold, W. C., Myers, T. G., Andrews, D. T., Scudiero, D. A., Eisen, M. B., down-regulated in MeWoEto1. Obviously, this is based on the loss of Sausville, E. A., Pommier, Y., Botstein, D., Brown, P. O., and Weinstein, J. N. A gene the respective chromosomal region, i.e., 3p. Thus, its down-regulation expression database for the molecular pharmacology of cancer. Nat. Genet., 24: 236–244, 2000. is likely a gene dosage effect attributable to the genomic alterations. 9. Kudoh, K., Ramanna, M., Ravatn, R., Elkahloun, A. G., Bittner, M. L., Meltzer, P. S., In association with data obtained from the experiments on the Trent, J. M., Dalton, W. S., and Chin, K-V. Monitoring the expression profiles of short-term responses, the genomic alterations provide additional val- doxorubicin-induced and doxorubicin-resistant cancer cells by cDNA microarray. Cancer Res., 60: 4161–4166, 2000. uable information. For example, the genes AHCYL1 and CYR61 reveal 10. Perou, C. M., Sorlie, T., Eisen, M. B., van de Rijn, M., Jeffrey, S. S., Rees, C. A., down-regulation in MeWo already during the short-term response to Pollack, J. R., Ross, D. T., Johnsen, H., Akslen, L. A., Fluge, O., Pergamenschikov, drug exposure. Apparently, this down-regulation became manifested A., Williams, C., Zhu, S. X., Lonning, P. E., Borresen-Dale, A. L., Brown, P. O., and Botstein, D. Molecular portraits of human breast tumours. Nature (Lond.), 406: in MeWoEto1, independently of the presence of etoposide in growth 747–752, 2000. medium. This is possibly caused by a loss of the respective chromo- 11. Hall, H. I., Miller, D. R., Rogers, J. D., and Bewerse, B. Update on the incidence and somal region at 1p. mortality from melanoma in the United States. J. Am. Acad. Dermatol., 40: 35–42, 1999. The utilization of a homogenous source offered the opportunity to 12. Serrone, L., and Hersey, P. The chemoresistance of human malignant melanoma: an identify genes commonly deregulated in more than one drug resist- update. Melanoma Res., 9: 51–58, 1999. ance phenotype basically independent of external variations such as 13. Helmbach, H., Rossmann, E., Kern, M. A., and Schadendorf, D. Drug-resistance in human melanoma. Int. J. Cancer, 93: 617–622, 2001. different genetic backgrounds or tumor architecture. According to 14. Kern, M. A., Helmbach, H., Artuc, M., Karmann, M., Jurgovsky, K., and Schaden- subarray analyses, a total number of 110 genes matches this criterion. dorf, D. Human melanoma cell lines selected in vitro displaying various levels of drug The most promising candidates are 14 genes either up- or down- resistance against cisplatin, fotemustine, vindesine or etoposide: Modulation of proto- oncogene expression. Anticancer Res., 17: 4359–4370, 1997. regulated in each of the three drug resistance phenotypes (Fig. 2D), 15. Lage, H., and Dietel, M. Involvement of the DNA mismatch repair system in among these the 4 IFN-inducible genes G1P3, ISG15, IFITM1, and antineoplastic drug resistance. J. Cancer Res. Clin. Oncol., 125: 156–165, 1999. 16. Fink, D., Aebi, S., and Howell, S. B. The role of DNA mismatch repair in drug IFITM3, as well as MPP1 and UP, which has also been shown to be resistance. Clin. Cancer Res., 4: 1–6, 1998. down-regulated in gastric cancer cells with cross-resistance to cispla- 17. Coultas, L., and Strasser, A. The molecular control of DNA damage-induced cell tin and doxorubicin (44), and again the apoptosis antagonist CRYAB. death. Apoptosis, 5: 491–507, 2000. 18. Nessling, M., Kern, M. A., Schadendorf, D., and Lichter, P. Association of genomic In specific regard to chemotherapy in melanoma, the consistent up- imbalances with resistance to therapeutic drugs in human melanoma cell lines. regulation of the four IFN-inducible genes may be noteworthy. Sim- Cytogenet. Cell Genet., 87: 286–290, 1999. ilar observations were made in the acquired resistance of the breast 19. Sinha, P., Kohl, S., Fischer, J., Hutter, G., Kern, M., Kottgen, E., Dietel, M., Lage, H., Schnolzer, M., and Schadendorf, D. Identification of novel proteins associated with cancer cell line MCF-7 against the antimicrotubule agent paclitaxel. the development of chemoresistance in malignant melanoma using two-dimensional However, the resistant MCF-7 cells exerted no cross-resistance electrophoresis. Electrophoresis, 21: 3048–3057, 2000. against the DNA-damaging drug doxorubicin (33). Nevertheless, it 20. Grottke, C., Mantwill, K., Dietel, M., Schadendorf, D., and Lage, H. Identification of differentially expressed genes in human melanoma cells with acquired resistance to seems worthwhile to conduct additional studies on these 4 IFN- various antineoplastic drugs. Int. J. Cancer, 88: 535–546, 2000. inducible genes, because IFN therapy is presently evaluated for its 21. Lage, H., Helmbach, H., Grottke, C., Dietel, M., and Schadendorf, D. DFNA5 usefulness in melanoma treatment (26). (ICERE-1) contributes to acquired etoposide resistance in melanoma cells. FEBS Lett., 494: 54–59, 2001. In conclusion, we present a set of 266 genes deregulated in mela- 22. Runger, T. M., Emmert, S., Schadendorf, D., Diem, C., Epe, B., and Hellfritsch, D. noma cell lines resistant to etoposide, fotemustine, and cisplatin. Of Alterations of DNA repair in melanoma cell lines resistant to cisplatin, fotemustine or these 110 are deregulated in more than one and 14 are differentially etoposide. J. Investig. Dermatol., 114: 34–39, 2000. 23. Lage, H., Christmann, M., Kern, M. A., Dietel, M., Pick, M., Kaina, B., and expressed in all three of the drug resistant phenotypes. Among the Schadendorf, D. Expression of DNA repair proteins hMSH2, hMSH6, hMLH1. latter are MPP1 and CRYAB, which identify as excellent candidates. O6-methylguanine DNA methyltransferase and N-methylpurine-DNA glycosylase in melanoma cells with acquired drug resistance. Int. J. Cancer, 80: 744–750, 1999. 24. Lage, H., Helmbach, H., Dietel, M., and Schadendorf, D. Modulation of topoisomerase II activity and expression in melanoma cells with acquired drug resistance. Br. J. ACKNOWLEDGMENTS Cancer, 82: 488–491, 2000. 25. Christmann, M., Pick, M., Lage, H., Schadendorf, D., and Kaina, B. Acquired We thank the Bioinformatics Section (Oliver Heil, Lars Ebert, Daniel resistance of melanoma cells to the antineoplastic agent fotemustine is caused by Bongartz, and Frank Schwarz) of the RZPD in Heidelberg for excellent reactivation of the DNA repair gene MGMT. Int. J. Cancer, 92: 123–129, 2001. computational support. We also thank Christian Maercker, Daniel Mertens, 26. Retsas, S. Adjuvant therapy of malignant melanoma: Is there a choice? Crit. Rev. and Jan Tuckermann for additional computational support and critical discus- Oncol. Hematol., 40: 187–193, 2001. 27. Boer, J. M., Huber, W. K., Sueltmann, H., Wilmer, F., von Heidebreck, A., Haas, S., sions, Holger Sueltmann for critical reading of the manuscript, Sara Burmester Korn, B., Gunawan, B., Vente, A., Fuezesi, L., Vingron, M., and Poustka, A. and Ute Ernst for sequencing, and Effi Rees and Markus Schuster for technical Identification and classification of differentially expressed genes in renal cell carci- support in composition of the drug resistance subarray. noma by expression profiling on a global human 31, 500 element cDNA array. Genome Res., 11: 1861–1870, 2001. 28. Lichter, P., Bentz, M., and Joos, S. Detection of chromosomal aberrations by means REFERENCES of molecular cytogenetics: painting of chromosomes and chromosomal subregions and comparative genomic hybridization. Methods Enzymol., 254: 334–359, 1995. 1. Chen, C. J., Chin, J. E., Ueda, K., Clark, D. P., Pastan, I., Gottesman, M. M., and 29. Vito, P., Pellegrini, L., Guiet, C., and D’Adamio, L. Cloning of AIP1, a novel protein Roninson, I. B. Internal duplication and homology with bacterial transport proteins in that associates with the apoptosis-linked gene ALG-2 in a Ca2ϩ-dependent reaction. the mdr1 (P-glycoprotein) gene from multidrug-resistant human cells. Cell, 47: J. Biol. Chem., 274: 1533–1540, 1999. 381–389, 1986. 30. Eisen, M. B., Spellman, P. T., Brown, P. O., and Botstein, D. Cluster analysis and 2. Cole, S. P., Bhardwaj, G., Gerlach, J. H., Mackie, J. E., Grant, C. E., Almquist, K. C., display of genome-wide expression patterns. Proc. Natl. Acad. Sci. USA, 95: 14863– Stewart, A. J., Kurz, E. U., Duncan, A. M., and Deeley, R. G. Overexpression of a 14868, 1998. transporter gene in a multidrug-resistant human lung cancer cell line. Science (Wash. 31. Ross, D. T., Scherf, U., Eisen, M. B., Perou, C. M., Rees, C., Spellman, P., Iyer, V., DC), 258: 1650–1654, 1992. Jeffrey, S. S., Van de Rijn, M., Waltham, M., Pergamenschikov, A., Lee, J. C., 3. Persidis, A. Cancer multidrug resistance. Nat. Biotechnol., 17: 94–95, 1999. Lashkari, D., Shalon, D., Myers, T. G., Weinstein, J. N., Botstein, D., and Brown, 4. Zhang, K., Mack, P., and Wong, K. P. Glutathione-related mechanisms in cellular P. O. Systematic variation in gene expression patterns in human cancer cell lines. Nat. resistance to anticancer drugs. Int. J. Oncol., 12: 871–882, 1998. Genet., 24: 227–235, 2000. 5. Larsen, A. K., and Skladanowski, A. Cellular resistance to topoisomerase-targeted 32. Kamradt, M. C., Chen, F., and Cryns, V. L. The small heat shock protein ␣ drugs: from drug uptake to cell death. Biochim. Biophys. Acta, 1400: 257–274, 1998. B-crystallin negatively regulates cytochrome c- and caspase-8-dependent activation 6704

of caspase-3 by inhibiting its autoproteolytic maturation. J. Biol. Chem., 276: 16059– 39. Hsu, H., Huang, J., Shu, H. B., Baichwal, V., and Goeddel, D. V. TNF-dependent 16063, 2001. recruitment of the protein kinase RIP to the TNF receptor-1 signaling complex. 33. Luker, K. E., Pica, C. M., Schreiber, R. D., and Piwnica-Worms, D. Overexpression Immunity, 4: 387–396, 1996. of IRF9 confers resistance to antimicrotubule agents in breast cancer cells. Cancer 40. Liu, X., Zou, H., Slaughter, C., and Wang, X. DFF, a heterodimeric protein that Res., 61: 6540–6547, 2001. functions downstream of caspase-3 to trigger DNA fragmentation during apoptosis. 34. Tew, K. D., O’Brien, M., Laing, N. M., and Shen, H. Coordinate changes in Cell, 89: 175–184, 1997. expression of protective genes in drug-resistant cells. Chem. Biol. Interact. 111–112: 41. Kawai, T., Matsumoto, M., Takeda, K., Sanjo, H., and Akira, S. ZIP kinase, a novel 199–211, 1998. serine/threonine kinase which mediates apoptosis. Mol. Cell. Biol., 18: 1642–1651, 35. Akiyama, S., Chen, Z. S., Sumizawa, T., and Furukawa, T. Resistance to cisplatin. 1998. Anticancer Drug Des., 14: 143–151, 1999. 42. Sanjo, H., Kawai, T., and Akira, S. DRAKs, novel serine/threonine kinases related to 36. Kartalou, M., and Essigmann, J. M. Mechanisms of resistance to cisplatin. Mutat. Res., 478: 23–43, 2001. death-associated protein kinase that trigger apoptosis. J. Biol. Chem., 273: 29066– 37. Dowler, S., Currie, R. A., Campbell, D. G., Deak, M., Kular, G., Downes, C. P., and 29071, 1998. Alessi, D. R. Identification of pleckstrin-homology-domain-containing proteins with 43. Majewski, M., Nieborowska-Skorska, M., Salomoni, P., Slupianek, A., Reiss, K., novel phosphoinositide-binding specificities. Biochem. J., 351: 19–31, 2000. Trotta, R., Calabretta, B., and Skorski, T. Activation of mitochondrial Raf-1 is 38. Ahmad, M., Srinivasula, S. M., Wang, L., Talanian, R. V., Litwack, G., Fernandes- involved in the antiapoptotic effects of Akt. Cancer Res., 59: 2815–2819, 1999. Alnemri, T., and Alnemri, E. S. CRADD, a novel human apoptotic adaptor molecule 44. Chung, Y. M., Park, S., Park, J. K., Kim, Y., Kang, Y., and Yoo, Y. D. Establishment for caspase-2, and FasL/tumor necrosis factor receptor-interacting protein RIP. Can- and characterization of 5-fluorouracil-resistant gastric cancer cells. Cancer Lett., 159: cer Res., 57: 615–619, 1997. 95–101, 2000.

6705

Rainer Wittig, Michelle Nessling, Rainer D. Will, et al.

Cancer Res 2002;62:6698-6705.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/62/22/6698

Cited articles This article cites 43 articles, 13 of which you can access for free at: http://cancerres.aacrjournals.org/content/62/22/6698.full#ref-list-1

Citing articles This article has been cited by 8 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/62/22/6698.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/62/22/6698. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.