Cancer Therapy: Preclinical

Gene Expression Profiles Associated with Response to Chemotherapy in Epithelial Ovarian Cancers Amir A. Jazaeri,1Christopher S. Awtrey,4 Gadisetti V.R. Chandramouli,1 Ya o E r i c Chu a n g, 2 Javed Khan,3 Christos Sotiriou,1Olga Aprelikova,1Cindy J. Yee,4 Kristin K. Zorn,6 Michael J. Birrer,6 J. Carl Barrett,1andJeff Boyd4,5

Abstract Purpose: The goal of this study was to determine whether distinct expression profiles are associated with intrinsic and/or acquired chemoresistance in epithelial ovarian carcinoma. Experimental Design: profiles were generated from 21primary chemosensi- tive tumors and 24 primary chemoresistant tumors usingcDNA-based microarrays. Gene expres- sion profiles of both groups of primary tumors were then compared with those of 15 ovarian carcinomas obtained followingplatinum-based chemotherapy (‘‘postchemotherapy’’ tumors). A theme discovery tool was used to identify functional categories of involved in drug resistance. Results: Comparison of primary chemosensitive and chemoresistant tumors revealed differential expression of 85 genes (P < 0.001). Comparison of gene expression profiles of primary chemo- sensitive tumors and postchemotherapy tumors revealed more robust differences with 760 genes differentiatingthe two groups ( P < 0.001). In contrast, only 230 genes were differentially expressed between primary chemoresistant and postchemotherapy groups (P < 0.001). Common to both gene lists were 178 genes representing transcripts differentially expressed between post- chemotherapy tumors and all primary tumors irrespective of intrinsic chemosensitivity.The gene expression profile of postchemotherapy tumors compared with that of primary tumors revealed statistically significant overrepresentation of genes encoding extracellular matrix ^ related . Conclusions: These data show that gene expression profiling can discriminate primary chemo- resistant from primary chemosensitive ovarian cancers. Gene expression profiles were also identi- fied that correlate with states of intrinsic and acquired chemoresistance and that represent targets for future investigation and potential therapeutic interventions.

Platinum-based combination chemotherapy is the standard resistant to this treatment and subsequently show low response first-line treatment for advanced-stage epithelial ovarian carci- rates to other second-line agents (1–3). Early identification of noma. For the f75% of patients diagnosed with advanced- this group of patients could lead to their enrollment in clinical stage disease, 20% to 30% progress on or rapidly become trials or treatment with other experimental therapeutics because standard treatment affords them little benefit. Among initially chemosensitive patients, the vast majority will eventually relapse. Thus, chemoresistance may be present at the outset 1 2 Authors’Affiliations: Laboratory of Biosystems and Cancer and Radiation of treatment (intrinsic resistance) or may develop during Biology Branch, Center for Cancer Research, and 3Pediatric Oncology Branch, National Cancer Institute, Bethesda, Maryland; 4Gynecology and Breast Research treatment (acquired resistance). In practice, ovarian cancers Laboratory, Department of Surgery, and 5Department of Medicine, Memorial are considered ‘‘platinum sensitive’’ if the clinical progression Sloan-KetteringCancer Center, New York,New York; and 6Department of Cell and free interval is >6 months, and evidence suggests that the longer Cancer Biology, Center for Cancer Research, National Cancer Institute, Rockville, this interval, the higher the subsequent response rates to Maryland Received 12/27/04; revised 5/31/05; accepted 6/16/05. additional chemotherapy (4–6). Grant support: NIH grant U01 CA88175 (J. Boyd) and Gynecologic Cancer Understanding the biological mechanisms underlying che- Foundation, National Cancer Institute Gynecologic Oncology Fellowship Program moresistance is of utmost importance for improving the (A.A. Jazaeri and K.K. Zorn). treatment and outcome of ovarian cancer. This topic has been The costs of publication of this article were defrayed in part by the payment of page the subject of intense research, and previous studies on charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. chemoresistance in ovarian cancer have investigated potential Note: Supplementary data for this article are available at Clinical Cancer Research involvement of molecules involved in drug transport, apopto- Online (http://clincancerres.aacrjournals.org/). sis, DNA repair, and detoxification pathways (7–11). Much of Requests for reprints: Jeff Boyd, Department of Surgery, Memorial Sloan- this research has been done using cell culture models and far KetteringCancer Center, Box 201, 1275 York Avenue, New York, NY 10021. Phone: 212-639-8608; Fax: 212-717-3538; E-mail: [email protected]. fewer data are available on the relevance of these studies to, and F 2005 American Association for Cancer Research. biomarkers and potential mechanisms of drug resistance for, doi:10.1158/1078-0432.CCR-04-2682 clinical samples.

Clin Cancer Res 2005;11(17) September 1, 2005 6300 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 23, 2021. © 2005 American Association for Cancer Research. Molecular Profile of Drug Resistance in Ovarian Cancer

The availability of new high-throughput screening techniques associated with both intrinsic and acquired chemoresistance in has allowed for more global investigations of molecular profiles ovarian cancer. The first aim of this investigation was to deter- associated with chemoresistance. In the present study, cDNA mine if intrinsically chemoresistant and chemosensitive tumors microarrays were used to investigate gene expression patterns could be distinguished based on their gene expression profiles.

Ta b l e 1. Clinicopathologic features of primary ovarian cancer cases

c Case no. Age* Histology Stage Grade Chemotherapy DFE Primary chemoresistant R1 49 Clear cell IIIC 3 Carboplatin/paclitaxel 0 R2 6 1 Endometrioid IIIC 3 Carboplatin/paclitaxel 0 R3 6 1 Serous IV 3 Carboplatin/paclitaxel 0 R4 49 Serous IIIC 2 Carboplatin/paclitaxel 0 R5 77 Serous IIIC NAb Carboplatin/paclitaxel 0 R6 49 Serous IIIC 2 Carboplatin/paclitaxel 0 R7 55 Serous IV 3 Carboplatin/paclitaxel 0 R8 56 Serous IIIC 2-3 Carboplatin/paclitaxel 0 R9 6 1 Serous IIIC 3 Carboplatin/paclitaxel 0 R10 73 Serous IV 3 Carboplatin/paclitaxel 0 R11 78 Serous IIIC 3 Carboplatin/paclitaxel 0 R12 7 1 Serous IV 2 Carboplatin/paclitaxel 0 R13 52 Serous IV 2 Carboplatin/paclitaxel 0 R14 64 Serous IIIA 1 Carboplatin/paclitaxel 0 R15 49 Serous IIIC 2 Carboplatin/paclitaxel 0 R16 47 Serous IV 3 Carboplatin/paclitaxel 0 R17 66 Serous IIIC 3 Cisplatin/cyclophosphamide 0 R18 69 Serous IV 2 Cisplatin/cyclophosphamide 0 R19 4 4 Serous IV 2 HD x cisplatin/cyclophosphamide 0 R20 35 Serous IV 2 Cisplatin/cyclophosphamide 0 R21 67 Serous IIIC 2-3 Cisplatin/cyclophosphamide 0 R22 63 Endometrioid IIIC 3 Carboplatin/paclitaxel 0 R23 76 Serous IIIC 3 Carboplatin 0 R24 65 Serous IIIC 2 Carboplatin/paclitaxel 0 Primary chemosensitive S1 5 1 Serous IV 2 Carboplatin/paclitaxel 35 S2 45 Endometrioid IIIC 2-3 HDCarboplatin/paclitaxel 41 S3 65 Mixed IIIC 3 Carboplatin/paclitaxel 13 S4 77 Serous IIIC 2-3 Carboplatin/paclitaxel 3 1 S5 53 Serous IIIC 3 Carboplatin/paclitaxel 32 S6 71 Serous IIIC 3 Carboplatin/paclitaxel 30 S7 55 Serous IIIC 3 Carboplatin/paclitaxel 19 S8 78 Endometrioid IIIC 2-3 Carboplatin/paclitaxel 24 S9 46 Endometrioid IIIC 3 Carboplatin/paclitaxel 2 1 S10 69 Serous IIIC 2 Carboplatin/paclitaxel 14 S11 54 Serous IV 1 Carboplatin/docetaxel 18 S12 53 Serous IIIC 2 Carboplatin/paclitaxel 18 S13 77 Serous IIIB 3 Carboplatin/paclitaxel 2 1 S14 56 Serous IIC 2 Carboplatin/paclitaxel 16 S15 44 Serous IIIC 2 Carboplatin/paclitaxel 16 S16 59 Carcinoma IIIC 3 Carboplatin/paclitaxel 16 S1744Serous IIICNACarboplatin/paclitaxel 13 S18 69 Serous IIIC 2 Carboplatin/paclitaxel 2 1 S19 70 Carcinoma IIIC 3 Carboplatin/paclitaxel 36 S20 74 Serous IIIC 2 Carboplatin/paclitaxel 2 1 S21 57 Serous IIIC 2 Carboplatin/paclitaxel 18

*Age at time of surgery. cDisease-free interval followingcompletion of chemotherapy (months). bNot available. xHigh-dose chemotherapy with bone marrow transplant.

www.aacrjournals.org 6301 Clin Cancer Res 2005;11(17) September 1,2005 Downloaded from clincancerres.aacrjournals.org on September 23, 2021. © 2005 American Association for Cancer Research. Cancer Therapy: Preclinical

Ta b l e 2 . Clinicopathologic features of postchemotherapy ovarian cancer cases

Case no. Age* Histology Stage Grade Clinical hx Chemotherapy PC1 57 Serous IIIC 2 NC/IDc Carboplatin/paclitaxel  3 PC2 74 Serous IIIB 3 NC/ID Carboplatin/paclitaxel  3 PC3 59 Serous IIIC 3 NC/ID Carboplatin/paclitaxel  3 PC4 77 Serous IIIC 2 NC/ID Carboplatin/paclitaxel  6 PC5 73 Serous IIIC 2 NC/ID Carboplatin/paclitaxel  6 PC6 68 Undifferentiated IV 3 NC/ID Carboplatin/paclitaxel  6 PC7 22 Serous IIIC 1 NC/ID Carboplatin/paclitaxel  6 PC8 71 Serous IV 3 NC/ID Carboplatin/paclitaxel  6 PC9 71 Serous IIIC 3 NC/ID Carboplatin/paclitaxel  6 PC10 45 Endometrioid IIIC 2 PSLb HDx Carboplatin/paclitaxel PC11 56 Serous IIIC 3 PSL Carboplatin/paclitaxel  6 PC12 58 mixed IIIC 3 PSL Carboplatin/paclitaxel  6 PC13 66 Serous IIIC 3 PSL Carboplatin/paclitaxel  6 PC14 42 Serous IIIC 3 PSL Carboplatin/paclitaxel  6 PC15 28 Serous IIIC 2 PSL Carboplatin/paclitaxel  6

*Age at time of surgery. cNeoadjuvant chemotherapy (as indicated) followed by interval debulkingsurgery. bPositive second-look surgery. xHigh-dose chemotherapy as indicated with bone marrow transplant.

For this part of the investigation, a case was classified as were obtained at the time of surgery, frozen in liquid nitrogen, and intrinsic chemoresistant based on persistent or recurrent disease stored at À80jC until use. Information on treatment and response was within 6 months of initiating first-line platinum-based combi- obtained from patient chart review. Intrinsically chemoresistant tumors nation chemotherapy. Chemosensitive tumors were classified were defined as those associated with persistent or recurrent disease as such based on a complete response to chemotherapy and a within 6 months of the initiation of first-line platinum-based combination chemotherapy. Chemosensitive tumors were defined as platinum-free interval of z13 months. These conservative those with a complete response to chemotherapy and a platinum-free clinical criteria for defining platinum sensitivity and resistance interval of z13 months. The postchemotherapy group is composed of were employed to exclude tumors with intermediate levels of patients who had either surgical debulking following chemotherapy resistance. In the second part of the investigation, gene (i.e., neoadjuvant chemotherapy) or residual tumor at the time of a expression profiles of tumors obtained following chemotherapy second-look procedure. All of these tumor samples were obtained (‘‘postchemotherapy’’ samples) were compared with those of within 6 weeks of the last cycle of chemotherapy. Ovarian tissues from the chemosensitive and chemoresistant primary tumors. The two postmenopausal women obtained at the time of salpingoophor- postchemotherapy group consisted of tumors from nine ectomy for benign indications were used for comparative purposes. patients treated with neoadjuvant chemotherapy who subse- RNA preparation and cDNA microarray analysis. Isolation of RNA quently underwent an interval cytoreductive surgery and from was done using the RNeasy column (Qiagen, Valencia, CA) according six patients who had residual cancer present at the time of to the manufacturer’s instructions. The integrity of RNA was verified by denaturing gel electrophoresis. Total RNA was linearly amplified using a second-look surgery following chemotherapy. modification of the Eberwine procedure (14). Briefly, total RNA was Gene expression profiles of these postchemotherapy tumors reverse transcribed by using a 63-nucleotide synthetic primer contain- were compared with those of our primary (i.e., chemonaive) ing the T7 RNA polymerase binding site [5V-GGCCAGTGAATTGTAA- chemosensitive and chemoresistant tumors. The rationale for TACGACTCACTATAGGGA-GGCGG(T)24-3V]. Second-strand cDNA this approach was 2-fold. First, after each cycle of cytotoxic synthesis (producing double-stranded cDNA) was done with RNase chemotherapy, the ‘‘log kill’’ effect leads to a significant reduction H, Escherichia coli DNA polymerase I, and E. coli DNA ligase in the number of tumor cells that are sensitive to the (Invitrogen, Carlsbad, CA). After cDNA was blunt ended with T4 administered therapy (12, 13). Second, tumor cells that survive DNA polymerase (Invitrogen), purification was accomplished by the treatment are likely to experience changes in gene expression phenol/chloroform/isoamyl alcohol extraction and ammonium ace- that allow them to withstand the selective pressure of the drugs tate/ethanol precipitation. The double-stranded cDNA was then used. Hence, tumor samples obtained shortly following chemo- transcribed using T7 polymerase (T7 Megascript kit, Ambion, Austin, therapy are enriched in resistant clones and are likely to display TX), yielding amplified antisense RNA that was purified using RNeasy mini-columns. Commercially available pooled total RNA from 10 the molecular signature associated with chemoresistance. different human cell lines (Stratagene, La Jolla, CA) was amplified and used as the reference for cDNA microarray experiments. Investigation of gene expression differences between primary chemo- Materials and Methods sensitive and chemoresistant tumors was done using two separate cDNA microarrays to maximize the number of genes screened. The two Tissue specimens. This study was approved by the Memorial Sloan- cDNA microarrays contained 32,448 and 7,585 features each for a Kettering Cancer Center Institutional Review Board. All tumor samples combined total of 40,033 transcripts. The comparison between the

Clin Cancer Res 2005;11(17) September 1, 2005 6302 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 23, 2021. © 2005 American Association for Cancer Research. Molecular Profile of Drug Resistance in Ovarian Cancer postchemotherapy samples and the primary tumors was done using the The slides were counterstained with methyl green, dehydrated, and 7,585-feature cDNA microarray. All cDNA microarrays were manufac- mounted. Quantitative scoring was done as the product of two staining tured at the National Cancer Institute. Amplified RNA (4 Ag) was characteristics, the percentage of immunopositive cells and the reverse transcribed and directly labeled using cyanine 5–conjugated intensity of the staining. Slides were examined microscopically at dUTP (tumor RNA) or cyanine 3–conjugated dUTP (pooled reference 20Â power, and five separate areas of each tumor specimen were RNA). Hybridization was done in the presence of 5Â SSC and 25% examined, with f100 cells per area analyzed. Scoring for the number formamide for 14 to 16 hours at 42jC. Slides were washed, dried, and of immunopositive cells was accomplished by assigning 0% to 25% as scanned using an Axon Instruments 4000a laser scanner. A detailed 1, 26% to 50% as 2, 51% to 75% as 3, and 76% to 100% as 4. Intensity protocol for RNA amplification as well as cDNA probe labeling and was scored as 1 to 3. The final score consisted of the product of the hybridization is available at http://nciarray.nci.nih.gov/reference/ immunopositive score (averaged over the five areas) and the intensity (under ‘‘Alternative Methods and Protocols’’). Genepix software score (averaged over the five areas). Statistical analysis was accom- (Molecular Devices, Sunnyvale, CA) was used to analyze the raw data plished using Student’s t test. that were then uploaded to a relational database maintained by the Center for Information Technology at the NIH. Results Data analysis. The log expression ratios for the spots on each array were normalized by subtracting the median log ratio for the same array. Gene expression differences between primary chemosensitive Data were filtered to exclude spots with size <25 Am, intensity less than and chemoresistant tumors. The clinicopathologic features of twice background or <500 units in both red and green channels, and any flagged or missing spots. In addition, any features found to be the samples used in this investigation are presented in Table 1 missing or flagged in >10% of the arrays were not included in the (primary cancers) and Table 2 (postchemotherapy cancers). analysis. The Genepix software assigns intensity levels in arbitrary units Tumors in both chemosensitive and chemoresistant groups with a range between 0 and 65,535. For reference, typical median were predominantly advanced stage, high grade (grade 2 or 3), background on arrays used in this investigation were 250, and the and of serous histology. All patients were treated with platinum- median probe signal was f4,500 (after background subtraction). based chemotherapy. For the purposes of this investigation, all Statistical comparison between tumors groups was done using the ‘‘BRB chemoresistant tumors were from patients with persistent or Array Tools’’ software (http://linus.nci.nih.gov/BRB-ArrayTools.html), recurrent disease present within 6 months of the initiation of consisting of a modified F test with P < 0.001 considered significant. first-line platinum-based combination chemotherapy. The This stringent P was selected in lieu of the Bonferroni correction for median disease-free interval for the chemosensitive samples multiple comparisons, which was deemed excessively restrictive. Gene lists were interrogated using EASE software (15) to identify possible was 21 months (range, 13-41 months). The median age for the overrepresentation of genes belonging to the same biological or chemoresistant group was 61 (range, 35-77), and the median functional class. age for the chemosensitive group was 57 (range, 44-78). Class prediction was done using a compound covariate predictor tool The comparison of gene expression profiles of 21 primary available as part of the BRB Array Tools software. This tool creates a chemosensitive and 24 primary chemoresistant ovarian cancers multivariate predictor for one of two classes to each sample. The genes revealed 85 transcripts with expression levels significantly included in the multivariate predictor were those that were univariately different between the two groups (P < 0.001; Table 3). The significant at the selected significance cutoff of P < 0.0001. The difference in geometric mean expression levels for all of these multivariate predictor is a weighed linear combination of log ratios for transcripts was V2-fold. The ability of the nine most significantly genes that are univariately significant. The weight consists of the differentially expressed genes (P < 0.0001) to predict clinical univariate t statistics for comparing the classes. A leave-one-out response was tested using a leave-one-out prediction model. This approach was then employed to test the ability of the compound covariate predictor to assign chemosensitive or chemoresistant class to analysis revealed an accuracy of 77.8% in correctly classifying each sample. A permutation test was used to assess the significance of refractory and responsive tumors. After 5,000 random permuta- our cross-validated error rate. The random permutations test the null tions, the likelihood of these nine genes differentiating the two hypothesis that there are no systematic differences in gene expression groups with equal or higher accuracy by chance (i.e., the null profiles of the chemoresistant and chemosensitive tumors. This hypothesis) was calculated as P = 0.018. These data showed that, assumption can be tested by randomly permuting labels among the although statistically significant differences at the mRNA level gene expression profiles and determining what proportion of the existed between chemosensitive and chemoresistant primary permuted data sets have a misclassification error rate less than or equal tumors, the magnitude of these differences was modest in to the observed error. This rate serves as the achieved significance level primary tumor samples obtained before chemotherapy. in a test against the null hypothesis. Detailed information about the To determine whether the inclusion of a few tumors of compound covariate predictor and the permutation test for significance is provided by the Biometric Research Branch, National Cancer Institute nonserous histologies (i.e., clear cell or endometrioid) or and is available at http://www.healthsystem.virginia.edu/internet/ comparison of tumors that did not all receive taxane-containing obgyn/supplemental-figure.pdf. chemotherapy regimens biased the results, an unsupervised Immunohistochemical analyses. Immunohistochemical staining was clustering analysis was also done. The resulting dendrogram done on 5 Am sections of formalin-fixed, paraffin-embedded ovarian revealed no segregation based on tumor histology, tumor grade, tumor specimens. After deparaffinization, sections were pretreated with or type of platinum-based chemotherapy received. These steam for 20 minutes in citrate buffer (pH 6.0). Slides were stained with supplementary data (Fig. S1) may be viewed at http:// primary antibodies against Ki-67 (mouse monoclonal clone MIB-1; www.healthsystem.virginia.edu/internet/obgyn/documents/ 1:75 dilution; DAKOCytomation, Carpinteria, CA), proliferating cell Supplemental-figures.pdf. nuclear antigen (PCNA; mouse monoclonal clone PC10; ready-to-use; Semiquantitative immunohistochemical analyses were used DAKOCytomation), and cathepsin D (CTSD; rabbit polyclonal; ready- to-use; DAKOCytomation). Staining was done on a DAKO Autostainer to determine if the gene expression differences between primary using the LSAB2 kit (DAKOCytomation) consisting of biotinylated chemoresistant and chemosensitive ovarian cancers were anti-mouse and anti-rabbit ready-to-use secondary antibodies, strepta- associated with significant expression differences. The vidin-horseradish peroxidase, and the chromogen diaminobenzidine. protein product of the CTSD gene was selected for this purpose

www.aacrjournals.org 6303 Clin Cancer Res 2005;11(17) September 1,2005 Downloaded from clincancerres.aacrjournals.org on September 23, 2021. © 2005 American Association for Cancer Research. Cancer Therapy: Preclinical

Ta b l e 3 . Genes differentially expressed between primary chemoresistant and primary chemosensitive ovarian cancers (P < 0.001)

c IMAGE clone* UniGene Gene Description Fold difference P

366971 Hs.156346 TOP2A Topoisomerase (DNA) IIa170-kDa 1.81 0.0008 810263 Hs.335798 RHPN2 Rhophilin, Rho GTPase-bindingprotein 2 1.70 0.00006 435303 Hs.397426 KIAA4146 KIAA1416 protein 1.58 0.0006 753464 Hs.33540 LOC389677 Similar to RIKEN cDNA 3000004N20 1.57 0.00004 742581 Hs.42173 C6orf107 6 open readingframe107 1.57 0.00002 450653 Hs.84063 BCL2L11 BCL2-like11(apoptosis facilitator) 1.57 0.0009 129345 Hs.173946 PAPD1 PAP-associated domain containing1 1.57 0.00004 589967 Hs.301431 ZNF71 protein 71 (Cos26) 1.56 0.0002 725223 Hs.119563 PSME4 (prosome, macropain) activator subunit 4 1.55 0.0008 452963 Hs.321390 CUGBP1 CUG triplet repeat, RNA-bindingprotein1 1.54 0.0003 71622 Hs.496511 PRKCI Protein kinase C, i 1. 5 2 0.0 0 07 757383 Hs.59236 ANKRD27 Ankyrin repeat domain 27 (VPS9 domain) 1.52 0.0001 220658 Hs.440394 MSH2 MutS homologue 2, colon cancer, nonpolyposis type1 (E. coli) 1.49 0.0003 1470530 Hs.435788 NCOA6 Nuclear coactivator 6 1.47 0.0004 261567 Hs.25812 NBS1 Nijmegen breakage syndrome1 (nibrin) 1.47 0.00006 592928 Hs.173001 KIAA1221 KIAA1221protein 1.47 0.00003 839048 Hs.156682 IGSF4 Immunoglobulin superfamily, member 4 1.47 0.0003 773203 Hs.43627 SOX12 SRY (sex-determiningregionY) box12 1.42 0.0009 452423 Hs.326392 SOS1 Son of sevenless homologue1 (Drosophila) 1.42 0.00006 306568 Hs.415997 COL6A1 Collagen, typeVI, a1 1.42 0.00002 855563 Hs.306251 ERBB3 v-erb-b2 erythroblastic leukemia viral oncogene homologue 3 1.42 0.0009 128159 Hs.170472 TPR Translocated promoter region (to activated METoncogene) 1. 4 1 0.0 0 0 2 343490 Hs.381189 CBX3 Chromobox homologue 3 (HP1chomologue, Drosophila) 1.41 0.0004 51737 Hs.437224 RBBP8 Retinoblastoma-bindingprotein 8 1.38 0.00003 247240 Hs.8258 C19orf13 Chromosome 19 open readingframe 13 1.38 0.0005 950667 Hs.36761 HRASLS HRAS-like suppressor 1. 3 7 0.0 0 0 1 264117 Hs.343475 CTSD CathepsinD(lysosomalaspartylprotease) 1.35 0.0006 741790 Hs.7942 AFTIPHILIN Aftiphilin protein 1.34 0.0004 490414 Hs.386198 EML4 Echinoderm microtubule-associated protein-like 4 1.34 0.001 230235 Hs.215766 GTPBP4 GTP-bindingprotein 4 1.34 0.0001 853066 Hs.5719 CNAP1 Chromosome condensation-related SMC-associated protein1 1.33 0.0002 950092 Hs.405144 SFRS3 Splicingfactor, arginine/serine ^ rich 3 1. 3 2 0.0 0 0 2 859857 Hs.442787 ZNF148 Zinc finger protein148 (pHZ-52) 1. 3 2 0.0 0 07 447569 Hs.282901 RNPC2 RNA-bindingregion(RNP1, RRM) containing2 1. 3 2 0.0 0 0 1 344942 Hs.109299 PPFIA3 Protein tyrosine phosphatase, interactingprotein (liprin), a3 1.32 0.0009 767419 Hs.274351 ZDHHC9 Zinc finger, DHHC domain-containing 9 1. 3 2 0.0 0 0 6 882434 Hs.362996 KIAA0779 KIAA0779 protein 1.31 0.00005 283329 Hs.252451 SEMA3A Semaphorin 3A 1.31 0.0007 359184 Hs.439523 PRKR Protein kinase, IFN-inducible 1.30 0.0006 180156 Hs.123654 PCF11 Pre-mRNA cleavage complex II protein Pcf11 1.30 0.0002 180156 Hs.128959 EST 1.30 0.0002

(Continued on the followingpage )

because of the availability of a commercial antibody and prior markers showed significantly higher expression in the chemo- studies implicating this protease in the pathogenesis of several sensitive primary tumors. These results suggest that the higher cancers, including ovarian and breast (16, 17). Using a expression of CTSD in the chemosensitive tumors correlates semiquantitative immunohistochemical scoring system, che- with a higher proliferative state that may in turn render them mosensitive samples displayed significantly higher CTSD more sensitive to cytotoxic chemotherapy. Notably, CTSD protein expression than the chemoresistant samples (Fig. 1A). expression was also significantly higher in the primary chemo- Because CTSD expression correlates with high proliferation sensitive group compared with the postchemotherapy group states in several tumors (17–19), the expression of Ki-67 and (two-tailed t test; P = 0.0008), consistent with the hypothesis PCNA was also analyzed in this set of tumors using that the postchemotherapy tumors are enriched for chemo- semiquantitative immunohistochemistry (Fig. 1B and C). Both resistant clones.

Clin Cancer Res 2005;11(17) September 1, 2005 6304 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 23, 2021. © 2005 American Association for Cancer Research. Molecular Profile of Drug Resistance in Ovarian Cancer

Ta b l e 3 . Genes differentially expressed between primary chemoresistant and primary chemosensitive ovarian cancers (P < 0.001) (Cont’d)

c c IMAGE clone* clone* UniGene UniGene Gene Description Description FoldFold difference difference PP 740707 Hs.269902 KIAA0494 KIAA0494 protein 1.28 0.0006 884892 Hs.7838 MKRN1 Makorin, ringfinger protein, 1 1.27 0.00002 826135 Hs.367811 STK38 Serine/threonine kinase 38 1.26 0.0007 838359 Hs.77890 GUCY1B3 Guanylate cyclase1,soluble, b3 1.26 0.00008 878174 Hs.388164 FADS2 Fatty acid desaturase 2 1.26 0.0006 837864 Hs.94262 DD5 Progestin-induced protein 1.26 0.0008 753914 Hs.512235 ITPR2 Inositol1,4,5-triphosphate receptor, type 2 1.26 0.0006 825197 Hs.131168 SEP1 Strand-exchange protein1 1.25 0.0006 815772 Hs.369284 C20orf6 Chromosome 20 open readingframe 6 1.24 0.0009 754654 Hs.118964 p66a P66 a 1.23 0.0002 384018 Hs.411300 WBP4 WW domain-bindingprotein 4 (formin-bindingprotein 21) 1.23 0.00002 320834 Hs.386404 UBE4B Ubiquitination factor E4B (UFD2 homologue, yeast) 1.22 0.0007 73596 Hs.35086 USP1 Ubiquitin-specific protease1 1.19 0.0003 743727 Hs.389638 FLJ41501 Clone BRTHA2006975 0.86 0.0006 277134 Hs.93836 CIP98 CASK-interactingprotein CIP98 0.85 0.0009 433289 Hs.117331 TREML1 Triggering receptor expressed on myeloid cells-like1 0.84 0.0009 811999 Hs.6455 RUVBL2 RuvB-like 2 (E. coli) 0.83 0.0003 1292893 Hs.125785 LOC149018 Hypothetical LOC149018, mRNA 0.83 0.0009 813735 Hs.410314 PCDH16 Protocadherin16 dachsous-like (Drosophila) 0.81 0.0006 814271 Hs.18885 CGI-116 CGI-116 protein 0.79 0.0004 399563 Hs.120332 EST 0.78 0.0004 814129 Hs.283716 MSCP Mitochondrial solute carrier protein 0.77 0.00002 490484 Hs.356349 ZNF145 Zinc finger protein145 (Kruppel-like, expressed in PML) 0.77 0.0007 47043 Hs.154138 CHI3L2 Chitinase3^like2 0.73 0.001 470035 Hs.14060 PROK1 Prokineticin 1 0.71 0.0007 2571195 Hs.406683 RPS15 Ribosomal protein S15 0.71 0.0003 303109 Hs.123464 P2RY5 Purinergic receptor P2Y,G-protein coupled, 5 0.70 0.0005 431231 Hs.381870 EFEMP2 Epidermal growth factor ^ containing fibulin-like ECM protein 2 0.70 0.0006 280950 Hs.422340 SRI Sorcin 0.66 0.0001 453183 Hs.301302 SCAM-1 Vinexin b (SH3-containingadaptor molecule-1) 0.66 0.0004 490556 Hs.26518 TM4SF7 Transmembrane 4 superfamily member 7 0.63 0.0001 502499 Hs.75835 PMM1 Phosphomannomutase1 0.61 0.0001 344720 Hs.81994 GYPC Glycophorin C (Gerbich blood group) 0.60 0.0003 2559389 Hs.150833 C4A Complement component 4A 0.58 0.0008 2784073 Hs.322431 NEUROD2 Neurogenic differentiation 2 0.54 0.0005 898305 Hs.439671 NBL1 Neuroblastoma, suppression of tumorigenicity1 0.54 0.00002 756372 Hs.37682 RARRES2 responder (tazarotene-induced) 2 0.52 0.0003 1474174 Hs.367877 MMP2 Matrix metalloproteinase-2 (72-kDa type IV collagenase) 0.50 0.00004 462939 Hs.526933 EST 0.48 0.001 823851 Hs.469463 AEBP1 AE-bindingprotein1 0.41 0.0007

*IMAGE Integrated MolecularAnalysis of Genomes and their Expression Consortium clone (http://madb.nci.nih.gov/CR_query.shtml). cFold difference in geometric means of chemosensitive tumors (numerator) compared with chemoresistant tumors (denominator).

Gene expression differences between primary and postchemo- differentially expressed between postchemotherapy samples therapy tumors. The overview of these comparisons is depicted and primary tumors irrespective of intrinsic chemoresistance. in Fig. 2A. This analysis revealed that 759 genes differentiated The magnitude and direction of change in the expression levels the postchemotherapy samples from the primary chemo- of these 178 genes were very similar in the two comparisons sensitive tumors (P < 0.001). In contrast, only 229 genes were (postchemotherapy versus primary chemoresistant and post- differentially expressed between postchemotherapy and primary chemotherapy versus primary chemosensitive; Fig. 2B). A chemoresistant groups (P < 0.001), suggesting smaller differ- partial list of these genes and their expression levels in primary ences between the molecular profiles of the latter two groups. A tumors and postchemotherapy samples is presented in Fig. 3A. comparison of the two gene lists revealed 178 genes that were Notably, genes encoding several oxidizing enzymes, including common to both and thus represented those genes that were ADH1B, ADH1C, and ALDH2, showed higher expression in the

www.aacrjournals.org 6305 Clin Cancer Res 2005;11(17) September 1,2005 Downloaded from clincancerres.aacrjournals.org on September 23, 2021. © 2005 American Association for Cancer Research. Cancer Therapy: Preclinical postchemotherapy samples. Also showing higher expression in and 3C). When the expression levels of these 51 differentially postchemotherapy samples were DOC1, CAV1, DUSP1, expressed genes in normal postmenopausal ovary were graphed ITM2A, DCN, and , all of which have been reported to along with the tumor expression levels, an interesting and be down-regulated in ovarian cancer when compared with unexpected pattern emerged. The expression profile of the normal ovary (20–22). In contrast, TOP1, TOP2A, and ZWINT postchemotherapy samples resembled that of normal ovarian have been reported to be overexpressed in ovarian cancer (23, tissue to a much greater degree than that of the primary 24) but were lower expressed in the postchemotherapy samples chemosensitive tumors. These data are consistent with a model compared with the primary tumors. in which the postchemotherapy samples show a partial ‘‘return Only 51 genes uniquely differentiated the primary chemo- to normal’’ or ‘‘low proliferative state’’ molecular expression resistant samples from the postchemotherapy group. In profile with respect to this set of genes. One notable exception contrast, 581 genes uniquely differentiated the primary chemo- to the overall similarities between postchemotherapy and sensitive samples from the postchemotherapy group (Fig. 2A). normal ovarian gene expression pattern was in the expression In addition, the latter comparison resulted in higher magnitude of CYR61. This gene has been implicated in angiogenesis (29) differences in expression as indicated by 51 of the 581 genes and chemoresistance (30) and was expressed at a significantly showing z2-fold difference in their geometric mean expression. higher level in postchemotherapy samples compared with both The comparison of the primary chemoresistant group with the primary chemosensitive and normal ovarian samples (Fig. 3A). postchemotherapy samples revealed 13 genes that showed at In addition, several of the genes higher expressed in the least a 1.5-fold difference in mean expression levels (Fig. 3D). postchemotherapy group were noted to be components of the Furthermore, only one gene, SPP1 (also known as osteopon- extracellular matrix (ECM) or involved in its remodeling. This tin), had a >2-fold change in expression and was higher impression was more formally investigated by using EASE expressed in the primary chemoresistant group. This gene has software (15) to analyze the biological categories within this been implicated previously in ovarian cancer (25) and has been gene list. This analysis confirmed the statistically significant proposed to represent a diagnostic biomarker for ovarian (P < 0.05 after Bonferroni correction) overrepresentation of cancer (26). Interestingly, the postchemotherapy samples genes involved in ECM among the genes differentiating the expressed higher levels of CDKN1C (p57KIP2) and ADAMTS1, postchemotherapy and primary chemosensitive tumors. One both of which have been shown to function as negative hypothesis that may be derived from this observation is that regulators of proliferation (27, 28). However, these levels were stromal-epithelial interactions or the ECM per se may be still far lower than those observed in normal postmenopausal involved in acquired chemoresistance in ovarian cancer. ovarian samples (Fig. 3D). In the comparison of the postchemotherapy tumors with the Discussion primary chemosensitive samples, 41 genes showed at least 2-fold higher expression and 10 genes showed at least a 2-fold These data suggest that gene expression patterns in primary lower expression in the postchemotherapy tumors (Fig. 3B (pretreatment) ovarian cancers can discriminate intrinsically

Fig. 1. Semiquantitative immunohistochemical analyses of CTSD, Ki-67, and PCNA expression. Representative photomicrographs of primary chemosensitive (A) and primary chemoresistant (B) ovarian cancers immunostained for CTSD at Â40 magnification. Results of immunohistochemical scoringfor CTSD (C), Ki-67 (D), and PCNA (E). Columns , mean score; bars,SD.

Clin Cancer Res 2005;11(17) September 1, 2005 6306 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 23, 2021. © 2005 American Association for Cancer Research. Molecular Profile of Drug Resistance in Ovarian Cancer

to be an underestimation for three reasons that are not mutually exclusive. First, according to the Goldie-Coldman hypothesis, chemoresistance is believed to result from a clonal selection process driven by the acquisition of drug resistance mutations (31). Thus, in primary tumors, only a small per- centage of the cells are likely to possess a chemoresistant phe- notype and the associated molecular changes. This results in a ‘‘dilution’’ of the observed gene expression differences when such tumors are compared with chemosensitive cancers. Second, the comparison of gene expression profiles in primary tumors evaluates only intrinsic chemoresistance. It is likely that a substantial component of clinical chemoresistance is biolog- ically acquired and is therefore only manifested following exposure to chemotherapeutic agents. In support of this hypothesis, the greatest differences in gene expression were observed between postchemotherapy samples and primary chemosensitive tumors. Finally, as is the case with most inves- tigations involving cDNA microarrays, the primary expression data are in the form of logarithmic intensity ratios. Secondary data, such as average expression levels for genes within a group, are derived by calculating the geometric rather than the arith- metic means of logarithmic intensity ratios, resulting in smaller values and smaller apparent differences. The higher expression of CTSD in chemosensitive tumors, as determined by immunohistochemistry, shows that small differ- ences in geometric mean expression as determined by cDNA microarrays may be associated with substantially greater differ- ences in protein expression. Although one previous report failed to show prognostic value of CTSD expression in ovarian cancer (32), most other studies show low CTSD expression to be an adverse prognostic indicator in ovarian cancer Fig. 2. Gene expression differences between postchemotherapy and primary (16, 33, 34). In view of our findings, the prognostic value of chemosensitive or primary chemoresistant tumors. A, an overview of the number of CTSD may be related to its higher expression in intrinsically differentially expressed genes (P < 0.001), with the top circle representing chemosensitive tumors. Consistent with this hypothesis, CTSD postchemotherapy versus primary chemosensitive samples, the bottom circle representingpostchemotherapy versus primary chemoresistant samples, and the has been implicated in -depenedent apoptosis following overlap region representing genes differentially expressed between DNA damage induced by drugs and g-irradiation (35). Further postchemotherapy samples and primary tumors irrespective of intrinsic investigations are needed to evaluate a possible causal chemosensitivity. B, changes in the magnitude and direction of the 178 genes that discriminated the postchemotherapy samples from all primary tumors (P < 0.001). relationship between these observations and to better define For each gene, the fold difference between postchemotherapy and primary the relationship between CTSD and chemosensitivity. The chemoresistant tumors (Xaxis) is plotted against the fold difference between postchemotherapy and primary chemosensitive tumors (Yaxis).Values less than 1.0 chemosensitive tumors also showed higher expression of the reflect higher expression in the primary tumors. r, Pearson correlation coefficient. proliferative markers PCNA and Ki-67. In agreement with this observation, a previous morphologic study found a highly chemoresistant from chemosensitive tumors. An accurate significant correlation between proliferation and mitotic assessment of predictors of response to chemotherapy is best indices and the presence of apoptotic bodies in primary ovarian accomplished through a prospective clinical trial involving cancers (36). Thus, higher rate of proliferation may contribute adequate numbers of patients. However, conducting such a trial to the chemosensitive nature of these tumors by predisposing requires the identification of potential molecular predictors of them to undergo apoptosis following chemotherapy. response in ovarian cancer. The data derived from screening This investigation also revealed substantially different gene >40,000 transcripts provide candidate targets for such an expression between primary ovarian cancers and tumor evaluation. samples obtained following chemotherapy. The postchemo- Although highly statistically significant, the magnitude of the therapy tumors are difficult to classify, based on customary mean expression differences between chemosensitive and clinical criteria, as either chemoresistant (cancer progression on chemoresistant groups were modest for the discriminating gene chemotherapy or recurrence within 6 months of completing set. The observed magnitude of expression differences is likely chemotherapy) or chemosensitive (disease-free interval of at —! Fig. 3. Specific genes with statistically significant differential expression between postchemotherapy and primary tumors (P < 0.001).The correspondingexpression values from normal postmenopausal ovaries are shown for comparison. A, genes that were differentially expressed between postchemotherapy tumors and both groups of primary tumors.The top 50 genes are shown; the full list may be found in the Supplementary Data. B, genes that were differentially higher expressed by z2-fold (n =41)inthe postchemotherapy compared with primary chemosensitive tumors. Genes encodingECM-related proteins are shown in bold type. C, genes that were differentially higher expressed by z2-fold (n = 10) in primary chemosensitive compared with postchemotherapy tumors. D, genes differentially expressed between the postchemotherapy tumors and the primary chemoresistant tumors.

www.aacrjournals.org 6307 Clin Cancer Res 2005;11(17) September 1,2005 Downloaded from clincancerres.aacrjournals.org on September 23, 2021. © 2005 American Association for Cancer Research. Cancer Therapy: Preclinical

Clin Cancer Res 2005;11(17) September 1, 2005 6308 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 23, 2021. © 2005 American Association for Cancer Research. Molecular Profile of Drug Resistance in Ovarian Cancer

sented in the comparison between postchemotherapy and primary tumors even after rigorous statistical correction for multiple comparisons. Such a finding is consistent with a recent study of ovarian cancer chemoresistance in vitro, where a group of several ECM components was found to be up- regulated in platinum-resistant cells (37). These genes included DCN and COL6A3, two of the ECM-related genes that were found to be higher expressed in the postchemo- therapy group in this study. Another member of this group found to be overexpressed in postchemotherapy tumors was SPARC (also known as osteo- nectin), a matricellular glycoprotein involved in angiogenesis, cell adhesion, and ECM turnover (38). This gene is also up- Fig. 4. Diagrammatic depiction of chemoresistance in primary and regulated following chemotherapy in breast cancer (39) and postchemotherapy samples. Smaller differences in gene expression observed has antiproliferative and tumor suppressor function in ovarian between primary chemosensitive and chemoresistant tumors are likely the result of low relative abundance of intrinsically chemoresistant clones as predicted by the (40, 41) and breast (42) cancer cells. Furthermore, SPARC Goldie-Coleman hypothesis. Chemotherapy results in a reduction in the number of stimulates matrix metalloproteinase-2 expression in other chemosensitive and an enrichment of chemoresistant clones in the tissues (43, 44) and may account for the observed higher postchemotherapy samples. In addition, chemotherapy is likely to induce additional genetic changes contributing to acquired chemoresistance.The combination of expression of matrix metalloproteinase-2 in the postchemo- these effects is likely to be responsible for the robust differences in gene expression therapy tumors. Another related ECM gene, SPARCL1 (also observed between primary and postchemotherapy samples. known as hevin, SC1, and MAST-9), was significantly higher expressed in the postchemotherapy samples compared with least 12 months) as they fit neither clinical criterion. They all both groups of primary tumors, is down-regulated in several eventually became resistant to chemotherapy. Our assertion is cancers, and has a negative effect on cell proliferation (45). that these tumor samples represent a state of enrichment in The higher expression of these and other antiproliferative chemoresistant clones, as these are tumors that have survived genes (e.g., KLF4 and CAV1; refs. 21, 46) in the postchemo- three to six cycles of chemotherapy. A reasonable hypothesis is therapy samples, as well as similarities in the gene expression that the gene expression profile of these postchemotherapy profiles of the postchemotherapy tumors and normal post- samples is likely to include molecular changes associated with menopausal ovaries, support the concept that a decreased acquired chemoresistance, as these samples were obtained proliferative state may be involved in the development of within a few weeks of completing three to six cycles of acquired chemoresistance. Furthermore, given that the primary chemotherapy. Consistent with this hypothesis, fewer and chemoresistant tumors exhibited significantly lower Ki-67, smaller magnitude gene expression differences were observed PCNA, and CTSD protein expression compared with the between postchemotherapy and primary chemoresistant sam- chemosensitive samples, decreased proliferation may also be ples. However, the data also suggest that intrinsic and acquired a contributing feature to intrinsic chemoresistance. Senescent or chemoresistance are likely to manifest through nonoverlapping slow-growing cells may be more tolerant of cytotoxic chemo- molecular pathways (Fig. 4). This was evident by the lack of therapy, thus allowing more time for selection of advantageous significant overlap (some genes are part of both lists) between mutations and development of resistant clones. Consistent with the gene list differentiating primary chemosensitive and such a hypothesis, the majority of postchemotherapy samples chemoresistant groups and the list differentiating each group exhibit a decreased mitotic index compared with the preche- from the postchemotherapy samples. In comparing the primary motherapy sample of the same tumor (47). tumors with postchemotherapy samples, three separate lists of This investigation represents an initial effort to discover differentially expressed genes were generated. Two lists identi- potentially important molecular mediators of intrinsic and fied genes that uniquely differentiated primary chemoresistant acquired chemoresistance in ovarian cancer in vivo. It is critical and chemosensitive tumors from the postchemotherapy to further investigate the molecular basis for chemoresistance samples and one that included genes that discriminated the prospectively in a large cohort with prechemotherapy and latter from both former groups. All three lists contain genes that postchemotherapy sampling from the same patients. This have been implicated previously in tumorigenesis and provide investigation provides preliminary data that may inform such targets for prospective investigations of acquired chemoresist- future studies and perhaps clinical protocols. ance in ovarian cancers. Finally, certain patterns of expression deserve special consi- deration. First, several ECM-related genes revealed differential Acknowledgments expression between postchemotherapy and primary tumors, We thank Drs. Edison T. Liu and William J. Hoskins for their vital roles in with all being higher expressed the postchemotherapy tumors. obtainingthe Director’s Challengegrant from the National Cancer Institute that This functional category of genes was significantly overrepre- made this work possible.

References 1. Chi DS, Sabbatini P. Advanced ovarian cancer. Curr recurrent ovarian cancer: evidence-based decisions. chemotherapy: current standards of care. Br J Cancer Treat Options Oncol 2000;1:139 ^ 46. Curr Opin Oncol 2002;14:519 ^ 27. 2003;89:S3 ^ 8. 2. Salom E, Almeida Z, Mirhashemi R. Management of 3. McGuire WP, Markman M. Primary ovarian cancer 4. Markman M, Hoskins W. Responses to salvage

www.aacrjournals.org 6309 Clin Cancer Res 2005;11(17) September 1,2005 Downloaded from clincancerres.aacrjournals.org on September 23, 2021. © 2005 American Association for Cancer Research. Cancer Therapy: Preclinical

chemotherapy in ovarian cancer: a critical need for 20. Mok SC, WongKK, Chan RK, et al. Molecular clon- Kristensen GB. The significance of metastasis-related precise definitions of the treated population [editori- ingof differentially expressed genes in human epitheli- factors cathepsin-D and nm23 in advanced ovarian al]. J Clin Oncol 1992;10:513^ 4. al ovarian cancer. Gynecol Oncol 1994;52:247 ^ 52. cancer. Ann Oncol 1999;10:1335 ^ 41. 5. Alberts DS. Treatment of refractory and recurrent 21.Wiechen K, Diatchenko L, Agoulnik A, et al. Caveo- 35. WuGS,SaftigP,PetersC,El-DeiryWS.Potential ovarian cancer. Semin Oncol 1999;26:8 ^ 14. lin-1is down-regulated in human ovarian carcinoma role for cathepsin D in p53-dependent tumor sup- 6. Spriggs D. Optimal sequencing in the treatment of and acts as a candidate tumor suppressor gene. Am J pression and chemosensitivity. Oncogene 1998;16: recurrent ovarian cancer. Gynecol Oncol 2003;90: Pathol 2001;159:1635^43. 2177 ^ 83. S39 ^ 44. 22. Manzano RG, Montuenga LM, Dayton M, et al. 36. Brustmann H. Apoptotic bodies as a morphological 7. Johnson SW, Ozols RF, HamiltonTC. Mechanisms of CL100 expression is down-regulated in advanced feature in serous ovarian carcinoma: correlation with drugresistance in ovarian cancer. Cancer 1993;71: epithelial ovarian cancer and its re-expression nuclear grade, Ki-67 and mitotic indices. Pathol Res 644^9. decreases its malignant potential. Oncogene 2002; Pract 2002;198:85 ^ 90. 8. Aebi S, Kurdihaidar B, Gordon R, et al. Loss of DNA 21:4435^7. 37. Sherman-Baust CA,Weeraratna AT, Rangel LB, et al. mismatch repair in acquired resistance to cisplatin. 23. van der Zee AG, Hollema H, de JongS, et al. P-gly- Remodelingof the extracellular matrix throughoverex- Cancer Res 1996;56:3087 ^ 90. coprotein expression and DNA topoisomerase I and II pression of collagen VI contributes to cisplatin resis- 9. Coukos G, Rubin SC. Chemotherapy resistance in activity in benign tumors of the ovary and in malignant tance in ovarian cancer cells. Cancer Cell 2003;3: ovarian cancer: new molecular perspectives. Obstet tumors of the ovary, before and after platinum/cyclo- 377 ^ 86. Gynecol 1998;91:783 ^ 92. phosphamide chemotherapy. Cancer Res 1991;51: 38. BradshawAD, Sage EH. SPARC, a matricellular pro- 10. Vasey PA. Resistance to chemotherapy in advanced 5915^20. tein that functions in cellular differentiation and tissue ovarian cancer: mechanisms and current strategies. Br 24. Jazaeri AA,YeeCJ, Sotiriou C, Brantley KR, Boyd J, response to injury. J Clin Invest 2001;107:1047 ^ 54. J Cancer 2003;89:S23 ^ 8. Liu ET. Gene expression profiles of BRCA1-linked, 39. Korn EL, McShane LM,Troendle JF, Rosenwald A, 11. Agarwal R, Kaye SB. Ovarian cancer: strategies for BRCA2-linked, and sporadic ovarian cancers. J Natl Simon R. Identifyingpre-post chemotherapy differen- overcomingresistance to chemotherapy. Nat Rev Cancer Inst 2002;94:990 ^ 1000. ces in gene expression in breast tumours: a statistical Cancer 2003;3:502^ 16. 25. Tiniakos DG,YuH, Liapis H. Osteopontin expression method appropriate for this aim. BrJCancer 2002;83: 12. Skipper HE. Kinetics of mammary tumor cell in ovarian carcinomas and tumors of low malignant 1093^ 6. growth and implications for therapy. Cancer 1971; potential(LMP).HumPathol1998;29:1250^4. 40. Mok SC, Chan WY,WongKK, Muto MG, Berkowitz 28:1479 ^ 99. 26. Kim JH, Skates SJ, UedeT, et al. Osteopontin as a RS. SPARC, an extracellular matrix protein with tumor- 13. Norton L, Simon R, Brereton HD, Bogden AE. potential diagnostic biomarker for ovarian cancer. suppressingactivity in human ovarian epithelial cells. Predictingthe course of Gompertzian growth. Nature J Am Med Assoc 2002;287:1671 ^ 9. Oncogene1996;12:1895^901. 1976;264:542^5. 27. ZhangP, Liegeois NJ,WongC, et al. Altered cell dif- 41. Yiu GK, Chan WY, NgSW, et al. SPARC (secreted 14. Van Gelder RN, von Zastrow ME, Yool A, Dement ferentiation and proliferation in mice lackingp57KIP2 protein acidic and rich in cysteine) induces apopto- WC, BarchasJD, EberwineJH. Amplified RNA synthe- indicates a role in Beckwith-Wiedemann syndrome. sis in ovarian cancer cells. Am J Pathol 2001;159: sized from limited quantities of heterogeneous cDNA. Nature 1997;387:151 ^ 8. 609 ^ 22. Proc Natl Acad Sci U S A1990;87:1663 ^ 7. 28. LuqueA, Carpizo DR, Iruela-Arispe ML. ADAMTS1/ 42. Dhanesuan N, Sharp JA, Blick T, Price JT, 15. HosackDA,DennisG,ShermanBT,LaneHC, METH1inhibits endothelial cell proliferation by direct Thompson EW. Doxycycline-inducible expression of Lempicki RA. Identifyingbiological themes within lists bindingand sequestration of VEGF165. J Biol Chem SPARC//BM40 in MDA-MB-231 human of genes with EASE. Genome Biol 2003;4:R70. 2003;278:23656^65. breast cancer cells results in growth inhibition. Breast 16. Losch A, Schindl M, Kohlberger P,et al. Cathepsin D 29. Babic AM, Kireeva ML, Kolesnikova TV, Lau LF. Cancer ResTreat 2002;75:73 ^ 85. in ovarian cancer: prognostic value and correlation CYR61, a product of a growth factor-inducible im- 43. Gilles C, Bassuk JA, Pulyaeva H, Sage EH, Foidart with p53 expression and microvessel density.Gynecol mediate early gene, promotes angiogenesis and tu- JM,Thompson EW. SPARC/osteonectin induces ma- Oncol 2004;92:545 ^ 52. mor growth. Proc Natl Acad Sci U S A 1998;95: trix metalloproteinase 2 activation in human breast 6355^60. 17. Ioachim E,Tsanou E, Briasoulis E, et al. Clinicopatho- cancer cell lines. Cancer Res 1998;58:5529 ^ 36. logical study of the expression of hsp27, pS2, cathep- 30.WittigR, NesslingM,Will RD, et al. Candidate genes 44. FujitaT, Shiba H, Sakata M, Uchida Y, Nakamura S, sin D and metallothionein in primary invasive breast for cross-resistance against DNA-damaging drugs. Kurihara H. SPARC stimulates the synthesis of OPG/ cancer. Breast 2003;12:111^ 9. Cancer Res 2002;62:6698 ^ 705. OCIF, MMP-2 and DNA in human periodontal ligament 18. Glondu M, Liaudet-Coopman E, Derocq D, Platet N, 31. Goldie JH, Coldman AJ. A mathematic model for cells. J Oral Pathol Med 2002;31:345 ^ 52. Rochefort H, Garcia M. Down-regulation of cathepsin- relatingthe drugsensitivity of tumors to their sponta- 45. Claeskens A, Ongenae N, Neefs JM, et al. Hevin is D expression by antisense gene transfer inhibits tumor neous mutation rate. Cancer Treat Rep 1979;63: down-regulated in many cancers andis a negative reg- growth and experimental lung metastasis of human 172 7 ^ 3 3 . ulator of cell growth and proliferation. Br J Cancer breast cancer cells. Oncogene 2002;21:5127 ^ 34. 32. FerrandinaG,ScambiaG,FagottiA,etal.Immunor- 2000;82:1123^ 30. 19. Ioachim E, Charchanti A, Stavropoulos N, adiometric and immunohistochemical analysis of ca- 46. Chen X, Whitney EM, Gao SY,YangVW. Transcrip- Athanassiou E, Bafa M, Agnantis NJ. Expression of thepsin D in ovarian cancer: lack of association with tional profilingof Kruppel-like factor 4 reveals a cathepsin Din urothelial carcinoma of the urinary blad- clinical outcome. Br J Cancer 1998;78:1645 ^ 52. function in cell cycle regulation and epithelial dif- der: an immunohistochemical study includingcorrela- 33. Scambia G, Panici PB, Ferrandina G, et al. Clinical ferentiation. J Mol Biol 2003;326:665 ^ 77. tions with extracellular matrix components, CD44, significance of cathepsin D in primary ovarian cancer. 47. McCluggage WG, Lyness RW, Atkinson RJ, et al. p53, Rb, c-erbB-2 and the proliferation indices. Anti- EurJ Cancer 1994;30A:935 ^ 40. Morphological effects of chemotherapy on ovarian cancer Res 2002;22:3383 ^ 8. 34. Baekelandt M, Holm R, Trope CG, Nesland JM, carcinoma. J Clin Pathol 2002;55:27 ^ 31.

Clin Cancer Res 2005;11(17) September 1, 2005 6310 www.aacrjournals.org Downloaded from clincancerres.aacrjournals.org on September 23, 2021. © 2005 American Association for Cancer Research. Gene Expression Profiles Associated with Response to Chemotherapy in Epithelial Ovarian Cancers

Amir A. Jazaeri, Christopher S. Awtrey, Gadisetti V.R. Chandramouli, et al.

Clin Cancer Res 2005;11:6300-6310.

Updated version Access the most recent version of this article at: http://clincancerres.aacrjournals.org/content/11/17/6300

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2006/01/10/11.17.6300.DC1

Cited articles This article cites 44 articles, 9 of which you can access for free at: http://clincancerres.aacrjournals.org/content/11/17/6300.full#ref-list-1

Citing articles This article has been cited by 27 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/11/17/6300.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://clincancerres.aacrjournals.org/content/11/17/6300. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.

Downloaded from clincancerres.aacrjournals.org on September 23, 2021. © 2005 American Association for Cancer Research.