Functional Genomics Identifies ABCC3 As a Mediator of Taxane Resistance in HER2-Amplified Breast Cancer

Research Article

Functional Genomics Identifies ABCC3 as a Mediator of Taxane Resistance in HER2-Amplified Breast Cancer Carol O’Brien,1 Guy Cavet,2 Ajay Pandita,1 Xiaolan Hu,2 Lauren Haydu,1 Sankar Mohan,1 Karen Toy,3 Celina Sanchez Rivers,3 Zora Modrusan,3 Lukas C. Amler,1 and Mark R. Lackner1

Departments of 1Oncology Diagnostics, 2Bioinformatics, and 3Molecular Biology, Genentech, Inc., South San Francisco, California

Abstract therapies such as tamoxifen for estrogen receptor (ER)–positive Breast cancer is a heterogeneous disease with distinct cancer (2) and Herceptin for tumors harboring amplification of HER2 molecular subtypes characterized by differential response to the oncogene (3) has had significant effect on patient targeted and chemotherapeutic agents. Enhanced understand- survival; yet, various chemotherapy regimens still form an ing of the genetic alterations characteristic of different important component of breast cancer treatment (4). Although subtypes is needed to pave the way for more personalized chemotherapy is a successful treatment regimen in many cases, an administration of therapeutic agents. We have taken a estimated 50% of patients fail to benefit due to intrinsic or acquired functional genomics approach using a well-characterized multidrug resistance (MDR; ref. 1). MDR refers to the resistance of panel of breast cancer cell lines to identify putative cancer cells to multiple classes of chemotherapeutic drugs that can biomarkers of resistance to antimitotic agents such as be structurally and mechanistically unrelated and is related to the paclitaxel and monomethyl-auristatin-E (MMAE). In vitro overexpression of a variety of proteins that act as ATP-dependent studies revealed a striking difference in sensitivity to these efflux pumps (5). Understanding the molecular alterations that agents between cell lines from different subtypes, with basal- contribute to MDR in breast cancer will be a crucial first step in like cell lines being significantly more sensitive to both enabling the development of diagnostic tests capable of predicting agents than luminal or HER2-amplified cell lines. Genome- resistance to a given therapy and rationally selecting more wide association studies using copy number data from efficacious therapeutic agents. Affymetrix single nucleotide polymorphism arrays identified One means of elucidating the relationship between genomic amplification of the chromosome 17q21 region as being features and therapeutic response is through pharmacogenomics, highly associated with resistance to both paclitaxel and or the study of how inherited and spontaneous genetic variation MMAE. An unbiased approach consisting of RNA interference affects sensitivity to pharmacologic agents (6). Efforts have been and high content analysis was used to show that amplification directed at identifying pharmacogenomic correlates of MDR in and concomitant overexpression of the gene encoding the tumor cell lines, notably in the 60 cancer cell lines used by the ABCC3 drug transporter is responsible for conferring in vitro National Cancer Institute to screen for anticancer activity (the resistance to paclitaxel and MMAE. We also show that NCI-60). At least 100,000 compounds have been screened through amplification of ABCC3 is present in primary breast tumors the NCI-60 over the past 15 years (7), including a recent study that and that it occurs predominantly in HER2-amplified and profiled mRNA expression of the 48 known human ABCtrans- luminal tumors, and we report on development of a specific porters in the NCI-60 and correlated the results with the growth- fluorescence in situ hybridization assay that may have utility inhibitory profiles of 1,429 candidate anticancer drugs. This study as a predictive biomarker of taxane resistance in breast identified several candidate transporters likely to confer resistance cancer. [Cancer Res 2008;68(13):5380–9] to specific agents (8) and provided insights into the general mechanisms of resistance across different tumor types, paving the Introduction way for follow-up studies to investigate mechanisms specific to individual tumor types. A key goal of modern molecular oncology is identifying the Pharmacogenomic studies must also be carried out in the underlying genetic and genomic abnormalities that characterize a context of the emerging picture of breast cancer as a heteroge- given tumor so that the patient can receive targeted and neous disease with distinct molecular subtypes. A number of chemotherapeutic agents likely to provide the most benefit. Breast studies over the past several years have shown that primary breast cancer is the most common form of cancer among women in the tumors may be classified into at least three major subtypes by Western World, with an estimated 1 million new diagnoses and gene expression profiling and that the subtypes have different 400,000 deaths per year worldwide (1). The advent of targeted prognostic outcomes in terms of patient survival (9). Luminal breast cancers are typically ER positive and characterized by coordinate expression of a number of epithelial specific genes, a Note: Supplementary data for this article are available at Cancer Research Online relatively good prognosis, and good response rates to targeted (http://cancerres.aacrjournals.org). C. O’Brien and G. Cavet contributed equally to this work. hormonal therapies. HER2-positive breast cancers are character- Current address for X. Hu: Pfizer Global Research and Development, Groton, ized by high-level gene amplification of the HER2 oncogene, Connecticut. Current address for L. Haydu: Royal Prince Alfred Hospital, Sydney, relatively poor prognosis if untreated, and significant clinical Australia. Requests for reprints: Mark R. Lackner, Genentech, Inc., 1 DNA Way, South San benefit from the HER2-targeting monoclonal antibody trastuzumab Francisco, CA 94080. Phone: 650-467-1846; Fax: 650-225-7571; E-mail: mlackner@ (Herceptin, Genentech; ref. 3). Basal-like breast cancers typically gene.com. I2008 American Association for Cancer Research. lack expression of HER2, ER, and progesterone receptor (PR) and doi:10.1158/0008-5472.CAN-08-0234 hence are sometimes referred to as ‘‘triple-negative’’ tumors (10).

Cancer Res 2008; 68: (13). July 1, 2008 5380 www.aacrjournals.org

Basal-like breast cancers have a relatively poor prognosis and Materials and Methods currently have not been shown to respond to any targeted therapy (11). It has recently been shown that the subtypes display Cell Lines and Viability Experiments Breast cancer cell lines AU565, BT-474, BT-549, CAMA-1, DU4475, differential response to preoperative chemotherapy regimens (12); HCC1143, HCC1419, HCC1428, HCC2218, HCC70, Hs578T, KPL-1, MCF-7, however, for the most part, the drug resistance mechanisms MDA-MB-231, MDA-MB-361, MDA-MB-435S, MDA-MB-436, MDA-MB-453, underlying these differences have yet to be determined. MDA-MB-468, T-47D, UACC-812, ZR-75-1, and ZR-75-30 were obtained from Recent studies have revealed that large collections of breast the American Type Culture Collection. The cell lines CAL-120, CAL-51, cancer cell lines reflect many of the genetic and genomic changes CAL-85-1, EFM-19, EFM-192A, EVSA-T, and MT-3 were obtained from the characteristic of human breast tumors and hence may serve as a Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH. All model system for a population of molecularly heterogeneous cell lines were maintained in RPMI 1640 or DMEM supplemented with 10% breast cancers (13). For instance, cells may be classified into basal- fetal bovine serum (Sigma), nonessential amino acids, and 2 mmol/L like and luminal subtypes based on gene expression profiling L-glutamine. Although annotated as breast lines, MDA-MB-435S may signatures and they retain most of the high-level amplifications actually be of melanoma origin and MT-3 of colorectal origin based on and deletions that are associated with poor outcome in primary molecular and genetic criteria (19, 20). These findings do not affect the conclusions of this study. For MMAE and paclitaxel IC determination, tumors (13, 14). We have used a panel of 31 breast cancer cell 50 cells were plated in quadruplicate at a density of 3,000 per well in 384-well lines that we have molecularly characterized as a model for plates in normal growth medium and allowed to adhere overnight. pharmacogenomic analysis to identify resistance mechanisms and Paclitaxel (Sigma) or MMAE (Seattle Genetics) were added in 10 concen- subtype differences in response to two antimitotic-based thera- trations based on a 3-fold dilution series (1 Amol/L maximal paclitaxel or peutics, monomethyl-auristatin-E (MMAE) and paclitaxel. MMAE 0.1 Amol/L maximal for MMAE). Cell viability was measured 72 h later is structurally related to dolastatin 10, a pentapeptide natural using the Celltiter-Glo Luminescent Cell Viability Assay (Promega). The product that has been the subject of several human clinical trials concentration of drug resulting in the 50% inhibition of cell viability (IC50) for cancer therapy, and exhibits potent antitumor activities by was calculated from a four-parameter curve analysis (XLfit, IDBS software) inhibiting tubulin polymerization and thus destabilize cellular and was determined from a minimum of three experiments. Cell lines that microtubules (15). Auristatin-monoclonal antibody conjugates did not show 50% reduction in cell viability in response to drug treatment in the majority of experiments conducted were considered to not have reached have been developed with the rationale that targeted delivery of an IC50 by definition and are listed as having an IC50 of >100 nmol/L the drug through specific antigen recognition by the antibody (MMAE) or >1,000 nmol/L (paclitaxel). For ABCC3-overexpressing clones of will lead to enhanced chemotherapeutic efficacy while sparing the EVSA-T cell line that did not achieve IC50, we calculated the half- nontarget-expressing tissues from toxicity (15). Paclitaxel and the maximal effect concentration, or EC50, in GraphPad Prism software related compound docetaxel are anticancer cytotoxic drugs that (GraphPad Software, Inc.). stabilize microtubules and are widely used in the treatment of breast cancer (16). Breast Tumor Samples Primary breast tumors from 145 independent breast cancer patients were Although most in vitro profiling efforts directed at understand- used to make genomic DNA for Agilent Array comparative genomic ing drug resistance to date have focused on gene expression hybridization analysis (aCGH; ref. 21). All the tumors were fresh frozen and analyses, here we sought to identify DNA copy number alterations found to have >70% tumor content, and all were classified as infiltrating that were associated with altered drug sensitivity through analyses ductal carcinoma. ABCC3 FISH studies were conducted on 61 additional of Affymetrix high-density SNP array profiles. Recent studies have independent primary breast tumor samples from the Genentech tumor shown that high-density single nucleotide polymorphism (SNP) bank. arrays, in addition to their intended application in genotyping, can Gene Expression Microarray Studies be used to detect genome-wide DNA copy number changes and Gene expression analysis of breast cancer cell lines was carried out on loss of heterozygosity in human cancers (17). These arrays have RNA extracted from subconfluent cell cultures using Qiagen RNAeasy kits. been shown to have applications in the identification of tumor RNA quality was verified by running samples on an Agilent Bioanalyzer 2100 suppressor and oncogene loci by pinpointing recurrently deleted and samples of sufficient quality were profiled on Affymetrix HGU133- or amplified chromosomal regions (18); we show here that they can Plus_2.0 chips. Preparation of complementary RNA, array hybridizations, be used to identify amplified regions harboring genes that may scanning, and subsequent array image data analysis were done using the modulate the activity of therapeutic drugs. manufacturer’s specified protocol. The major finding of this study is that amplification of a region For overall unsupervised hierarchical clustering analysis of breast cancer of chromosome 17 (17q21) is strongly associated with in vitro cell lines, gene expression data were filtered to remove probe sets that resistance to taxanes and auristatins. The region of amplification showed little variation across the cell lines. Briefly, probes that did not show at least a 5-fold variation across the samples (max/min >10) and an absolute harbors at least 100 genes; therefore, to identify the relevant gene, intensity difference of at least 250 (max-min >250) were excluded from we used an unbiased approach consisting of RNA interference hierarchical clustering analysis. Data preprocessing involved log-trans- (RNAi) and high content analysis. These studies show that forming and median-centering gene expression values, after which average amplification and concomitant overexpression of the ABCC3 gene linkage clustering was carried out using Cluster and TreeView software (22). is most likely responsible for conferring resistance to paclitaxel and MMAE. We also show that this amplicon is present in primary SNP Array and Agilent aCGH Copy Number Studies breast tumors, that it occurs predominantly in HER2-amplified Cell line copy number analysis was carried out on genomic DNA extracted from subconfluent cell cultures using Qiagen DNAeasy kits. For tumors, and we report on development of a specific fluorescence in situ each cell line, 500 ng of genomic DNA were hybridized to Genechip 100 K hybridization (FISH) assay that may have utility as a mapping arrays (Affymetrix, Inc.) according to the manufacturer’s biomarker of taxane response in breast cancer. The results also instructions. These arrays contain probe sets for >116,000 SNP loci derived suggest that genome-wide copy number changes assessed by SNP from all human chromosomes (except the Y chromosome), with a mean arrays are a valuable tool for preclinical biomarker discovery when marker distance of 26 kb (23). SNP calls and signal quantification were used in conjunction with pharmacologic data. obtained with Gene Chip Operating System. Agilent Human Genome 244A www.aacrjournals.org 5381 Cancer Res 2008; 68: (13). July 1, 2008

CGH microarrays and Agilent feature extraction software were run previously (30), with some modifications. After an overnight incubation at according to the manufacturer’s instructions and genome-smoothed 56jC, the slides were deparaffinized in three washes of CitroSolv for 5 min analysis DNA copy number (GSA_CN) was calculated based on the each, followed by two washes in alcohol. After air drying, the slides were hybridization intensity (the sum of both allele intensities) for each SNP incubated in a 1 mol/L solution of sodium sulfocyanate for 30 min at 80jC probe with the Affymetrix Chromosome Copy Number Analysis Tool 3.0 and then were treated with pepsin before additional washes in water and a (CNAT 3.0). Copy number data were segmented with the GLAD series of ethanol. Dried slides were then codenatured (76jCfor 6 min) with segmentation algorithm (24). Genome-wide analysis of copy number gains the probe and were hybridized overnight at 37jC(ThermoBrite; Vysis). and losses shown in Supplementary Fig. S2 were determined using the Posthybridization washes and counterstaining were done in a manner Genomic Identification of Significant Targets in Cancer algorithm (25). similar to those previously described. The slides were visualized using an Associations between GSA_CN copy number and drug sensitivity were Olympus BX61 microscope and analyzed using FISHView software (Applied identified with Matlab software (The MathWorks, Inc.) using a version of Spectral Imaging). The copy number analysis and ratio of HER2/ABCC3 to the maxT procedure (26). For each drug, a test statistic was calculated for CEP17 was performed as per the manufacturer’s instructions. each SNP reflecting the difference between log-transformed copy number in sensitive and resistant cell lines. The statistic was calculated as the absolute Functional Validation Experiments value of a standard t statistic (two sample, unequal variance), except that it High-content screening assays were carried out on an Arrayscan VTI was set to zero for those SNPs with <1.75-fold difference in mean copy (Cellomics, Inc.). Cells were transfected in 96-well format using small interfering RNA (siRNA) ‘‘Smartpool’’ oligonucleotides purchased from number between sensitive and resistant classes. Then, the null distribution Dharmacon, Inc., and Oligofectamine transfection reagents. To prioritize of maximum test statistics across all SNPs was estimated in 10,000 random genes for functional studies, two-sided Wilcoxon rank sum tests using the R permutations of the sensitivity labels. The P value for each SNP was programming language4 were done to identify genes with a significant calculated as the fraction of permutations in which the maximum test difference in gene expression in cell lines with >4 copies compared with statistic was greater than or equal to the observed statistic for that SNP. The those with <4 copies of the 17q21.3 amplicon. This analysis combined with resulting P values control the family-wise error rate and take into account availability of reagents for RNA interference experiments led to the selection the number of SNPs tested. of the following 24 genes for functional studies in the EFM-192A cell line: HER2 Copy Number Determination by Quantitative Reverse ABCC3, COL1A1, CROP, EAP, EPN3, FLJ13855, FLJ20920, HOXB7, LOC201191, ITGB3, KIAA0924, KPNB1, LOC400604, LOC81558, MGC11242, MGC15396, Transcription-PCR NDP52, PDK2, PHB, PP1R9B, SLC35B1, SPOP, TOB1 WNT3 Quantitative PCR was done using ABI Prism 7900 Sequence Detection , and . Follow-up System (Applied Biosystems) on genomic DNA prepared as described studies with ABCC3 siRNA were conducted in the additional cell lines above. Quantitative reverse transcription-PCR (qRT-PCR) was done using ZR75-30, MDAMB-453, and HCC-1428. A nontargeting control siRNA that primers CACTGTCTGCACCTTGCTTTG and GCTCTGCAGCTATTGAAA- does not show significant homology to any sequence in the human genome GAACAA for HER2 and AAAGCCGCTCAACTACATGG and TGCTTTGA- was used as a negative control in all RNAi experiments (as described in 5 j ATGCGTCCCAGAG for line-1 repetitive elements. Line-1 is a repetitive technical notes online). After 48 or 72 h incubation at 37 C, cells were fixed element with similar copy numbers per haploid genome between human in 3.7% formaldehyde and permeabilized in 0.1% Triton X-100, followed by normal and neoplastic cells (27). Quantification was based on standard labeling with a 1:500 dilution of anti–phospho-histone H3 (pH3, Upstate) curves from a serial dilution of human normal genomic DNA. The and subsequent 1:250 dilution of Alexa-fluor 488 (Molecular Probes) goat relative target copy number level was also normalized to normal human anti-rabbit secondary antibody. Cells were counterstained with Hoechst- 33258 to allow identification of nuclei and the percentage of cells positive genomic DNA as calibrator. Copy number change of target gene relative for nuclear pH3 immunofluorescence, also known as the mitotic index (31), to the line-1 and the calibrator were determined using the formula À À À À was then quantitated for at least 1,000 cells per well using Cellomics Target E [(CPtarget Cpref)control (CPtarget Cpref)test] as described by Kindich Activation software. All experiments were repeated at least thrice. qRT-PCR and colleagues (28). Conditions for quantitative PCR reaction were as to assess ABCC3 transcript levels after siRNA transfection was performed described in the Invitrogen Platinum SYBR Green qPCR SuperMix-UDG using ABCC3-specific primers (5¶ primer GATTCCAGCCGCTTCAGTT, 3¶ w/ROX package insert. primer CCTGGCTGTGCTCTACACCT). Fluorescence In situ Hybridization Analysis For ABCC3 overexpression experiments, a full-length ABCC3 cDNA Probes. A bacterial artificial chromosome contig comprising of two cloned in the cytomegalovirus (CMV) promoter containing vector pCMV5 overlapping clones, RP11-2605A1 and CTD-3006C13, covering the entire (Invitrogen, Carlsbad, CA) was verified by sequencing the entire coding ABCC3 loci and adjoining areas (based on the UCSS Genome Browser sequence. The construct was transfected into EVSA-T cells and stable clones March 2006 assembly) were used as a probe for the FISH experiments. were selected by growth in 1 mg/mL geneticin (Invitrogen). Overexpression Commercially available probes for HER2/CEP17 (Pathvysion, Vysis/Abbott of ABCC3 in stable clones was confirmed by qRT-PCR on cDNA derived Laboratories) and CEP17 (Vysis/Abbott Laboratories) were also used for the from lines containing pCMV5-ABCC3, pCMV5 vector alone, or the parental FISH experiments. EVSA-T strain. FISH analysis. Cell lines were prepared for cytogenetic analysis by incubation with 0.1 Ag/mL colcemid (Invitrogen) for 30 min, followed by osmotic swelling in KCl (0.075 mol/L) and fixation in methanol/acetic acid Results (3:1), as previously described (29). DNA from the bacterial artificial Molecular characterization of cell lines. Affymetrix gene chromosome clones was extracted by standard methods. The extracted expression profiling was performed on cDNA prepared from total bacterial artificial chromosome DNA was directly labeled with Spectrum mRNA and Affymetrix 100 K SNP array profiling was done on DNA Orange or Spectrum Green (Vysis/Abbott Laboratories) by nick translation from 44 breast cancer cell lines. Unsupervised analysis with the using the Vysis Nick Translation Kit (Vysis/Abbott Laboratories) according 11,000 most differentially expressed genes across the cell line panel to the manufacturer’s instructions. FISH to normal human metaphases was used to classify the cell lines into luminal and basal-like (Abbott Laboratories) confirmed the genomic location of the bacterial subtypes based on gene expression (Supplementary Fig. S1). Cell artificial chromosome clones. Approximately 300 ng of labeled probes were precipitated in excess Human Cot-1 DNA (Invitrogen) and sonicated salmon sperm DNA (Sigma) and resuspended in a 50% formamide, 10% dextran sulfate, and 2Â SSChybridization buffer (Vysis/Abbott Laborato- ries) for the FISH experiments. FISH on cytogenetic preparations and 4 http://www.r-project.org formalin-fixed paraffin-embedded tissue was performed as described 5 http://www.dharmacon.com

Cancer Res 2008; 68: (13). July 1, 2008 5382 www.aacrjournals.org

Downloaded from cancerres.aacrjournals.org on September 27, 2021. © 2008 American Association for Cancer Research. ABCC3 Amplification and Taxane Resistance lines classified as luminal expressed high levels of ERa and many colleagues,6 but a finding relevant to this study is that amplification of the target genes regulated by ER, including GATA3, HNF3A, at 17q21.3 is common in HER2-amplified and luminal cell lines but IGF-IR, and XBP1. Cell lines classified as basal-like expressed high not in basal-like cell lines (Supplementary Fig. S2). levels of some or all of the well-described basal markers vimentin, In vitro sensitivity to antimitotic drugs. We screened 31 caveolin, MFGE8, and the basal cytokeratins such as KRT5 breast cancer cell lines for in vitro sensitivity to paclitaxel and (Supplementary Fig. S1; ref. 32). Because amplification of the MMAE as assessed by the IC50 value for each compound in a HER2 oncogene clearly defines a separate disease subtype that is standard luciferin-based viability assay (Supplementary Table S1; not apparent from overall gene expression classification in cell Supplementary Fig. S3). Notably, there was significant correlation lines (13), we determined HER2 copy number by qRT-PCR on between the relative sensitivity to each agent across the panel of genomic DNA and normalization to line 1 repetitive elements for cell lines (Spearman rank-order correlation coefficient, rs = 0.55). all cell lines (Supplementary Table S1). The molecular subtype in In addition, Fig. 1 shows that cell lines with the basal-like gene Fig. 1 is a classification derived from both the overall gene expression signature had lower average IC50 values and were more expression results as well as the HER2 copy number analysis. Our sensitive to each agent than luminal or HER2-amplified cell lines as findings agree with previous reports (13) and suggest that this determined by Kruskal-Wallis rank sum test (P = 0.002 for MMAE, collection of breast cancer cell lines reflects to some degree the P = 0.005 for paclitaxel). major transcriptional distinctions that define breast cancer Identification of genomic alterations that correlate with subtypes and to some extent are representative as models of in vitro sensitivity. We next sought to identify regions of subtypes as luminal, basal-like, and HER2-amplified tumors. chromosomal gain or loss that correlated with sensitivity to Genome-wide patterns of copy number gain and loss in the cell paclitaxel or MMAE through supervised analysis of SNP array line panel show that the breast cancer cell lines harbor most of copy number data. First, cell lines were classified into either the major copy number alterations (e.g., MYC, CCND1, HER2 gain sensitive (IC50 <10 nmol/L) or resistant (MMAE IC50 >100 nmol/L, p16, PTEN loss) that have been described in tumors (Supplemen- paclitaxel IC50 >1,000 nmol/L) groups based on the sensitivity data. tary Fig. S2). Subtype-specific differences exist and are shown in We then used the maxT algorithm (26) to analyze data from Supplementary Fig. S2 as well as discussed in detail by Hu and f115,000 SNPs and identify individual SNPs where the mean copy number differed between sensitive and resistant classes with genome-wide significance. In the case of paclitaxel, a group of SNPs on chromosome 17 starting at chromosome position 44,303,217 and ending at position 44,724,301 (17q21.21 to 17q21.23) showed statistically significant copy number differences between sensitive and resistant classes (P for rs2411377 = 0.04). The same group of markers also showed significant association between copy number and MMAE sensitivity (P for rs2411377 = 0.05). Figure 2 shows the relationship between paclitaxel sensitivity and genomic DNA copy number in this part of chromosome 17 and that a significant number of cell lines (8 of 14) that showed resistance to paclitaxel have at least four gene copies within the region. Identification of candidate genes in the interval. The chromosomal region from 17q21.31 to 17q21.33 encodes f100 expressed transcripts according to the University of California Santa Cruz Genome Browser.7 Based on the principle that functionally relevant genes in regions of amplification should exhibit a concomitant increase in mRNA expression, we filtered this list down to 24 genes that showed significant overexpression upon amplification and had reagents available to conduct RNAi studies. An example of significantly higher expression of the candidate gene ABCC3 in amplified cell lines compared with nonamplified cell lines is shown in Supplementary Fig. S4. These 24 genes were selected for subsequent functional analysis to identify the locus responsible for conferring resistance to taxanes and auristatins. Functional validation of ABCC3 by RNA interference. We used an RNA interference strategy to identify the gene responsible for mediating resistance to taxanes and auristatins in amplified cell lines. The assay utilized made use of the fact that treatment of cells with paclitaxel or MMAE results in a block of cell cycle progression at the M phase that can be assayed by the presence of the mitotic Figure 1. Cell lines of the basal-like subtype show greater average sensitivity to paclitaxel and MMAE than luminal or HER2-amplified cell lines. A, in vitro response of breast cancer cell lines to MMAE. B, in vitro response of breast cancer cell lines to paclitaxel.On the X axis, cell lines were classified into major molecular subtypes of breast cancer as described in the text.The Y axis 6 X. Hu, C. O’Brien, C.D. Honchell, T. Wu, J. Chant, M. Lackner, G. Cavet. Genetic indicates the in vitro IC50 value or concentration of drug that resulted in 50% alterations and oncogenic pathways associated with breast cancer subtypes, inhibition of cell viability. Horizontal lines, mean sensitivity to each agent for cell submitted. lines of a given subtype. 7 http://genome.ucsc.edu www.aacrjournals.org 5383 Cancer Res 2008; 68: (13). July 1, 2008

Figure 2. In vitro resistance to paclitaxel is associated with amplification of the chromosome 17q21 region.Ideogram ( left) indicates relative area of detail shown in heat map.Heat map indicates copy number (GSA_CN algorithm) with green indicating losses and red indicating gains according to the scale shown.Cell l ines are shown in order of increasing paclitaxel sensitivity from left to right.The classifications ( sensitive, intermediate, resistant) used for supervised analysis of SNP array data and identification of biomarkers of resistance are indicated at the top of the figure. marker pH3 (33). Phosphorylation at Ser10 of histone H3 is tightly the mitotic index in nonamplified cell lines HCC-1428 and MDA- correlated with chromosome condensation during the M phase, MB-453 (Fig. 3C and D). siRNA knockdown of ABCC3 at the and the percentage of cells that are positive for pH3 staining, or transcript level was confirmed by qRT-PCR (Supplementary mitotic index, can be determined through an immunofluorescence Fig. S5). Similar results were obtained with MMAE (data not assay. We reasoned that cellular knockdown of the gene-mediating shown). ABCC3/MRP3 (henceforth referred to as ABCC3) is a resistance should increase sensitivity of cell lines harboring the member of the MDR-associated protein (MRP) subfamily of ATP- amplification to paclitaxel and MMAE and hence result in an dependent drug efflux pumps (34). accumulation of arrested cells and a higher mitotic index relative Overexpression of ABCC3 causes in vitro MDR. Given that to control-treated cells at a given drug concentration. Higher ABCC3 knockdown increased sensitivity to paclitaxel and MMAE, mitotic index correlates with reduction in viability and prolifera- we tested whether ABCC3 overexpression in taxane- and MMAE- tion determined by other assays8 but is a more specific readout of sensitive cells with low levels of ABCC3 would render cells resistant the antimitotic effects of these drugs. We found that RNAi of 23 of to these agents. EVSA-T cells were selected as a model to generate the 24 candidate genes did not reproducibly result in accumulation ABCC3-overexpressing lines because they do not show ABCC3 of arrested cells and increased mitotic index in EFM-192A cells amplification and express low levels of ABCC3 transcripts. Three (data not shown), but that RNAi of ABCC3/MRP3 resulted in a 2- to independently derived lines were confirmed to overexpress ABCC3 3-fold increase in mitotic index relative to control treatment with a transcripts (Supplementary Fig. S5) and screened for in vitro nontargeting control siRNA in the cell lines EFM-192A and ZR75-30 sensitivity to paclitaxel and MMAE using an ATP-based lumines- (Fig. 3A and B). In contrast, ABCC3 RNAi did not appreciably alter cence assay. All three overexpressing cell lines were at least 20-fold less sensitive to paclitaxel and MMAE based on EC50 values and also showed markedly less inhibition of cell growth compared with a vector-alone control stable cell line in an ATP-based lumines- 8 C. O’Brien and M.R. Lackner, unpublished observations. cence assay (Fig. 4).

Cancer Res 2008; 68: (13). July 1, 2008 5384 www.aacrjournals.org

Amplification of ABCC3 occurs in breast tumors. Analysis of with single or multiple integration into various chromosomes the region of chromosome 17 encompassing HER2 and ABCC3 while maintaining single copies of HER2 and ABCC3 on in the cell line 100K SNP array data suggested that the ABCC3 chromosome 17 (Fig. 6A and B). Cell lines predicted to be diploid amplicon was most commonly associated with the HER2-amplified for ABCC3 based on SNP array analysis were confirmed to be subtypes and was not seen in the cell lines classified as luminal or diploid based on FISH analysis with CEP17 and ABCC3 (data not basal-like (Fig. 5, top two panels). shown). FISH analysis of the 61 HER2-positive primary tumors To ensure that ABCC3 amplification was not a cell line–specific that were screened for ABCC3 amplification confirmed that phenomenon, we also characterized copy number data at the elevated copy number at ABCC3 is common in HER2-positive ABCC3 locus using Agilent aCGH arrays on DNA from 145 primary breast tumors (Fig. 6C and D; Supplementary Table S2). High-level breast tumors. These tumor samples were also classified into gene amplification (>2.2 ratio of ABCC3/CEP17; see Fig. 6C) was luminal, basal-like, and HER2 subtypes using a predictor based on seen in 25% of the tumors, whereas an additional 11% of the expression levels of ER, PR, and HER2 as described by Hu and tumors showed moderate increases for ABCC3 (3–7 copies of colleagues.6 ABCC3 copy number gains (>3.5 copies) are present in ABCC3, see Fig. 6D) for ABCC3. Interestingly, a number of tumors 25% of HER2-amplified and 11% of luminal tumors but were not show evidence of heterogeneity and exhibit cells with both very present in basal-like tumors (Fig. 5, bottom two panels). high-level amplification of ABCC3 alongside cells with diploid copy Finally, to confirm the cytogenetic basis of the apparent copy number of ABCC3. number gains observed by SNP and aCGH arrays, we developed a FISH assay using a bacterial artificial chromosome clone (see Materials and Methods) spanning the ABCC3 locus and performed Discussion FISH analysis on select cell lines and 61 primary tumors that had Molecular classification of breast cancers into subtypes with been classified as overexpressing HER2 based on the HerceptTest shared features and similar prognostic outcomes provides a (immunohistochemistry assay, reviewed in ref. 35; data not shown). framework to begin efforts to individualize cancer therapy. We The FISH results from cell lines corroborated the data obtained have shown here that a large collection of breast cancer cell lines from the SNP array and qPCR analyses. As depicted in Fig. 6A, reflects many of the genetic and genomic alterations characteristic cell line EFM-192A predicted from SNP arrays to have elevated of human breast tumors and that subtypes show clear differences ABBC3 copy number indeed exhibited a high-level amplification of in response to antimitotic agents, with the basal-like subtype ABCC3, which is manifested as homogeneously staining regions being the most sensitive. We present functional evidence that one

Figure 3. RNAi of ABCC3 increases sensitivity of cell lines with 17q21 amplification to paclitaxel.Graphs represent mitotic index in response to paclitaxel t reatment for four different cell lines after ABCC3 or control siRNA treatment. A and B, ZR-75-30 and EFM-192A cells have amplification and overexpression of ABCC3 and display enhanced sensitivity (increased mitotic index) after knockdown. C and D, HCC-1428 and MDA-MB-453 have low copy number and expression and do not show increased sensitivity upon ABCC3 knockdown.

www.aacrjournals.org 5385 Cancer Res 2008; 68: (13). July 1, 2008

Figure 4. Stable overexpression of ABCC3 results in in vitro resistance to paclitaxel and MMAE.Stable cell lines derived from single cell clones that overexpress ABCC3 from the CMV promoter or a control line with empty vector were assayed for growth-inhibitory effects of paclitaxel (A) or MMAE (B).Treatment of ABCC3-overexpressing clones did not result in 50% reduction of cell viability in a 3 d assay so the fold change in sensitivity was assessed by calculating the concentration that resulted in half-maximal response, or EC50.Specifically, the EC 50 for the vector control treated with paclitaxel was 0.2 nmol/L, whereas the EC50 values for the ABCC3-expressing lines were 5, 10, and 80 nmol/L, respectively.The EC 50 for the vector control–treated with MMAE was 0.05 nmol/L, whereas the EC50 values for the ABCC3-expressing lines were 1.5, 12, and 90nmol/L, respectively. The vector control cell line was clonally derived from EVSA-T and unlike the parental cell line exhibited an IC50 in response to paclitaxel and MMAE in the experiment shown.

mechanism for this differential response is amplification of the overexpression does not confer resistance to doxorubicin in long- ABCC3 drug efflux pump in a subset of luminal and HER2- term assays despite other published reports of functional studies amplified cell lines but not in basal-like cell lines. Future functional suggesting a role for ABCC3 in transporting this agent (36). Taken genomic efforts in this cell line panel should help identify targeted together, these findings suggests that either the ability of ABCC3 agents likely to specifically show efficacy in one or more of the to transport paclitaxel is context or cell type dependent, or that subtypes, a particularly pressing need for the basal-like or triple- differences in assay format or duration account for the apparent negative class of tumors. The detailed understanding of the role for ABCC3 in transporting paclitaxel in breast cancer molecular alterations in the cell lines also raises the likelihood of suggested by our studies. identifying biomarkers of efficacy that are not only correlated with An unusual aspect of our findings is that resistance may be response but also linked to the biology of the target pathway being derived from an amplification event present in the primary tumor modulated. independent of prior selection due to drug treatment because a ABCC3 overexpression has been implicated in acquired MDR number of the samples profiled are mostly either primary tumors in cancer cell lines in previous studies. For instance, Liu and or cell lines derived from such tumors or pleural effusions collected colleagues report 459-fold overexpression of ABCC3 relative to the at the time of diagnosis. Therefore, it may be postulated that parental in a cell line, MCF-7/AdVp3000, that was derived by Chr17q21.31-21.33 amplification is selected as part of tumor selection for growth in the presence of doxorubicin (36). In evolution/homeostasis and that ABCC3 amplification is coampli- addition, it has recently been shown that treatment of carcinoma fied as a passenger gene that confers serendipitous chemotherapy cell lines with vincristine results in significant up-regulation of resistance. As such, ABCC3 amplification may represent a ABCC2 and ABCC3 transcripts in these cells (37). The closely mechanism of intrinsic resistance present in the tumor due to related pumps ABCC2 (MRP2) and ABCC10 (MRP7) have both amplification of adjacent sequences that are presumably being been shown to confer paclitaxel resistance when overexpressed selected for during the process of tumorigenesis. One possible (37, 38), and moreover ABCC2 has been shown to be an important candidate for such selection of Chr17q21.31-21.33 amplification is determinant of paclitaxel pharmacokinetics in vivo in mouse the nearby gene Myst2, which encodes a histone acetyltransferase models (39). Paclitaxel has not been previously shown to be a that has been shown by Hu and colleagues6 to result in substrate for ABCC3 and indeed studies of ectopic overexpression transformation in soft agar when overexpressed in cell lines that of ABCC3 in Madin-Darby canine kidney cells or NIH-3T3 cells do not normally have amplification or overexpression of Myst2. have failed to show increased resistance to paclitaxel (40, 41). Other examples of amplification of ABCtransporters have been Notably, one of these studies also found (40) that ABCC3 reported, but usually in the context of the tumor-evolving

Cancer Res 2008; 68: (13). July 1, 2008 5386 www.aacrjournals.org

Downloaded from cancerres.aacrjournals.org on September 27, 2021. © 2008 American Association for Cancer Research. ABCC3 Amplification and Taxane Resistance resistance. For instance, resistant variants of the T-cell leukemia subtypes such as HER2-positive breast cancer may harbor cell line CCRF-CEM selected by growth in increasing levels of the unexpected heterogeneity that may affect response to therapeutics, anthracycline epirubicin were shown to have acquired amplifica- in this case amplification of the ABCC3 transporter in 25% to 35% tion of the ABCC1/MRP1 gene (42). of primary HER2-amplified breast tumors. All of these results taken Other studies have also addressed resistance mechanisms to together highlight the fact that breast cancer is a heterogeneous paclitaxel in breast cancer and have hinted at a multiplicity of disease that can evade chemotherapy through multiple mecha- mechanisms. For instance, a recent report suggested that in nisms and highlight the need for panels of biomarkers that can be primary cell cultures derived from breast carcinomas, expression of used to predict response in individual tumors. PR was negatively correlated with response to paclitaxel in the The results presented here also illustrate the utility of high- adjuvant setting (43). This finding is consistent with the density SNP arrays for biomarker discovery when used in observation here that basal-like breast cancer cell lines are more conjunction with pharmacologic data in cell lines. A key advantage sensitive to paclitaxel on average than luminal or HER2-amplified of this approach is that gene amplification events are relatively lines. Another potential mechanism that has been identified is stable and can ultimately be assayed by FISH on archival samples paclitaxel-triggered phosphorylation of the protein caveolin-1, from clinical trials. FISH assays are already part of routine clinical which has been shown to sensitize cells to apoptosis by regulating practice in the diagnosis of HER2-positive MBC(47) A recent cell cycle progression and activation of the apoptotic signaling retrospective analysis of 1,500 women with node-positive breast molecules p53 and p21 (44). Because caveolin-1 expression is cancer showed that expression or amplification HER2 in a breast highest in basal-like breast cancers (45), this is another potential cancer is associated with enhanced clinical benefit from the mechanism to explain the differential antimitotic response addition of paclitaxel after adjuvant treatment with doxorubicin observed across subtypes. Finally, gene expression studies on compared with patients with HER2-negative, ER-positive, and biopsies obtained before adjuvant chemotherapy followed by node-positive breast cancer (48). However, a significant fraction functional studies in vitro have suggested that low expression of of women with HER2-positive tumors failed to show a survival the microtubule-binding protein tau is a potential biomarker for benefit in this study, suggesting the possibility that paclitaxel pathologic response to paclitaxel-containing regimens (46). The resistance mechanisms are also present in a proportion of HER2- results described in this article affirm that even well-defined positive breast tumors. Utilizing a FISH assay to stratify patients by

Figure 5. Presence of HER2 and ABCC3 amplification determined by SNP array (cell lines) or array CGH (primary tumors) with copy number gains (red) and losses (green) represented by heat map.The ideogram to the left indicates the location of the areas of detail in the heat map.Each column represents a separate DNA sample and each row an independent SNP (cell lines) or probe (tumor samples). Arrows, positions of HER2 and ABCC3.Samples were classified into three major molecular subtypes of breast cancer as described in the text. www.aacrjournals.org 5387 Cancer Res 2008; 68: (13). July 1, 2008

Figure 6. FISH analysis of ABCC3 amplification in cell lines and tumors. A, FISH probes for ABCC3 (green)and HER2 (red) show high-level amplification of both loci in the well-established breast cancer cell line EFM-192A. Multiple homogenously staining regions are found integrated in various chromosomes and HER2 and ABCC3 seem to colocalize (yellow staining).The bright green signal (arrows) indicates the CEP17 centromeric probe. B, HER2 and ABCC3 sequences are highly amplified and are present in a 1:1 ratio in this cell line. C and D, ABCC3 amplification in two independent primary breast tumors.FISH probes for ABCC3 ( red)and CEP17 (green) show (C) amplification in tight clusters, indicative of very high level amplification in this tumor. The tumor sample in D shows evidence of multiple copies of ABCC3 and polysomy for chromosome 17 as indicated by multiple copies of CEP17 in each nucleus.

ABCC3 amplification status in this setting would allow determi- Disclosure of Potential Conflicts of Interest nation as to whether ABCC3 amplification correlates with lack of G. Cavet, A. Pandita, X. Hu, S. Mohan, K. Toy, C. Sanchez Rivers, Z. Modrusan, L.C. clinical benefit from taxane-containing regimens and hence Amler, and M.R. Lackner are employees of Genentech, Inc. The other authors disclosed whether ABCC3 amplification could serve as a biomarker to no potential conflicts of interest. identify patients who should receive other chemotherapeutic agents in addition to taxanes. Acknowledgments Future studies with this large panel of well-characterized cell Received 1/18/2008; revised 3/18/2008; accepted 4/29/2008. lines should help to identify additional genomic alterations that The costs of publication of this article were defrayed in part by the payment of page predict response to chemotherapeutics as well as novel targeted charges. This article must therefore be hereby marked advertisement in accordance agents and lead to the development of companion diagnostic tests with 18 U.S.C. Section 1734 solely to indicate this fact. We thank Elaine Storm, Richard Neve, Jane Fridlyand for helpful comments on the that can be validated in clinical studies and ultimately to match manuscript, as well as Fred de Sauvage and the Genentech Cancer Genome Project for patients with appropriate therapies. support.

References 5. Szakacs G, Paterson JK, Ludwig JA, Booth-Genthe C, observation of breast tumor subtypes in independent Gottesman MM. Targeting multidrug resistance in gene expression data sets. Proc Natl Acad Sci U S A 1. O’Driscoll L, Clynes M. Biomarkers and multiple drug cancer. Nat Rev 2006;5:219–34. 2003;100:8418–23. resistance in breast cancer. Curr Cancer Drug Targets 6. Katsnelson A. Cautious welcome for FDA pharmaco- 10. Finnegan TJ, Carey LA. Gene-expression analysis and 2006;6:365–84. genomics guidance. Nat Biotechnol 2005;23:510. the basal-like breast cancer subtype. Future Oncol 2007; 2. Cigler T, Goss PE. Breast cancer adjuvant endocrine 7. Lee JK, Havaleshko DM, Cho H, et al. A strategy for 3:55–63. therapy. Cancer J 2007;13:148–55. predicting the chemosensitivity of human cancers and its 11. Kurt M, Harputluoglu H, Dede DS, Gullu IH, Altundag 3. Slamon DJ, Leyland-Jones B, Shak S, et al. Use of application to drug discovery. Proc Natl Acad Sci U S A K. Potential molecular targeted therapies in the chemotherapy plus a monoclonal antibody against 2007;104:13086–91. management of the basal-like subtype of breast cancer. HER2 for metastatic breast cancer that overexpresses 8. Szakacs G, Annereau JP, Lababidi S, et al. Predict- Breast 2007;16:111–2. HER2. N Engl J Med 2001;344:783–92. ing drug sensitivity and resistance: profiling ABC 12. Rouzier R, Perou CM, Symmans WF, et al. Breast 4. Guarneri V, Conte PF. The curability of breast cancer transporter genes in cancer cells. Cancer Cell 2004;6: cancer molecular subtypes respond differently to and the treatment of advanced disease. Eur J Nucl Med 129–37. preoperative chemotherapy. Clin Cancer Res 2005;11: Mol Imaging 2004;31 Suppl 1:S149–61. 9. Sorlie T, Tibshirani R, Parker J, et al. Repeated 5678–85.

Cancer Res 2008; 68: (13). July 1, 2008 5388 www.aacrjournals.org

13. Neve RM, Chin K, Fridlyand J, et al. A collection of 26. Westphal P, Young S. Resampling-based multiple 38. Huisman MT, Chhatta AA, van Tellingen O, breast cancer cell lines for the study of functionally testing: examples and methods for P-value adjustment. Beijnen JH, Schinkel AH. MRP2 (ABCC2) transports distinct cancer subtypes. Cancer Cell 2006;10:515–27. John Wiley & Sons; 1993. taxanes and confers paclitaxel resistance and both 14. Chin K, DeVries S, Fridlyand J, et al. Genomic and 27. Fanning TG, Singer MF. LINE-1: a mammalian processes are stimulated by probenecid. Int J Cancer transcriptional aberrations linked to breast cancer transposable element. Biochim Biophys Acta 1987;910: 2005;116:824–9. pathophysiologies. Cancer Cell 2006;10:529–41. 203–12. 39. Lagas JS, Vlaming ML, van Tellingen O, et al. 15. Doronina SO, Toki BE, Torgov MY, et al. Development 28. Kindich R, Florl AR, Jung V, et al. Application of a Multidrug resistance protein 2 is an important deter- of potent monoclonal antibody auristatin conjugates for modified real-time PCR technique for relative gene copy minant of paclitaxel pharmacokinetics. Clin Cancer Res cancer therapy. Nat Biotechnol 2003;21:778–84. number quantification to the determination of the 2006;12:6125–32. 16. Crown J, O’Leary M. The taxanes: an update. Lancet relationship between NKX3.1 loss and MYCgain in 40. Kool M, van der Linden M, de Haas M, et al. 2000;355:1176–8. prostate cancer. Clin Chem 2005;51:649–52. MRP3, an organic anion transporter able to transport 17. Zhao X, Li C, Paez JG, et al. An integrated view of 29. Bayani J, Zielenska M, Marrano P, et al. Molecular anti-cancer drugs. Proc Natl Acad Sci U S A 1999;96: copy number and allelic alterations in the cancer cytogenetic analysis of medulloblastomas and supra- 6914–9. genome using single nucleotide polymorphism arrays. tentorial primitive neuroectodermal tumors by using 41. Zeng H, Bain LJ, Belinsky MG, Kruh GD. Expression Cancer Res 2004;64:3060–71. conventional banding, comparative genomic hybridiza- of multidrug resistance protein-3 (multispecific organic 18. GarrawayLA,WidlundHR,RubinMA,etal. tion, and spectral karyotyping. J Neurosurg 2000;93: anion transporter-D) in human embryonic kidney 293 Integrative genomic analyses identify MITF as a lineage 437–48. cells confers resistance to anticancer agents. Cancer Res survival oncogene amplified in malignant melanoma. 30. Pandita A, Aldape KD, Zadeh G, Guha A, James CD. 1999;59:5964–7. Nature 2005;436:117–22. Contrasting in vivo and in vitro fates of glioblastoma cell 42. Mahjoubi F, Hill RJ, Peters GB. Chromosome 19. Gorringe KL, Chin SF, Pharoah P, et al. Evidence that subpopulations with amplified EGFR. Genes Chromo- microdissection identifies genomic amplifications asso- both genetic instability and selection contribute to the somes Cancer 2004;39:29–36. ciated with drug resistance in a leukemia cell line: an accumulation of chromosome alterations in cancer. 31. Grove LE, Ghosh RN. Quantitative characterization approach to understanding drug resistance in cancer. Carcinogenesis 2005;26:923–30. of mitosis-blocked tetraploid cells using high content Chromosome Res 2006;14:263–76. 20. Rae JM, Ramus SJ, Waltham M, et al. Common analysis. Assay Drug Dev Technol 2006;4:421–42. 43. Schmidt M, Bremer E, Hasenclever D, et al. Role of origins of MDA-MB-435 cells from various sources with 32. Livasy CA, Karaca G, Nanda R, et al. Phenotypic the progesterone receptor for paclitaxel resistance in those shown to have melanoma properties. Clin Exp evaluation of the basal-like subtype of invasive breast primary breast cancer. Br J Cancer 2007;96:241–7. Metastasis 2004;21:543–52. carcinoma. Mod Pathol 2006;19:264–71. 44. Shajahan AN, Wang A, Decker M, Minshall RD, Liu 21. Barrett MT, Scheffer A, Ben-Dor A, et al. Comparative 33. Gasparri F, Cappella P, Galvani A. Multiparametric MC, Clarke R. Caveolin-1 tyrosine phosphorylation genomic hybridization using oligonucleotide microar- cell cycle analysis by automated microscopy. J Biomol enhances paclitaxel-mediated cytotoxicity. J Biol Chem rays and total genomic DNA. Proc Natl Acad Sci U S A Screen 2006;11:586–98. 2006;282:5934–43. 2004;101:17765–70. 34. Borst P, Evers R, Kool M, Wijnholds J. A family of 45. Pinilla SM, Honrado E, Hardisson D, Benitez J, 22. Eisen MB, Spellman PT, Brown PO, Botstein D. drug transporters: the multidrug resistance-associated Palacios J. Caveolin-1 expression is associated with a Cluster analysis and display of genome-wide expression proteins. J Natl Cancer Inst 2000;92:1295–302. basal-like phenotype in sporadic and hereditary breast patterns. Proc Natl Acad Sci U S A 1998;95:14863–8. 35. Carlsson J, Nordgren H, Sjostrom J, et al. HER2 cancer. Breast Cancer Res Treat 2006;99:85–90. 23. Matsuzaki H, Dong S, Loi H, et al. Genotyping over expression in breast cancer primary tumours and 46. Rouzier R, Rajan R, Wagner P, et al. Microtubule- 100,000 SNPs on a pair of oligonucleotide arrays. Nat corresponding metastases. Original data and literature associated protein tau: a marker of paclitaxel sensitivity Methods 2004;1:109–11. review. Br J Cancer 2004;90:2344–8. in breast cancer. Proc Natl Acad Sci U S A 2005;102: 24. Hupe P, Stransky N, Thiery JP, Radvanyi F, Barillot E. 36. Liu Y, Peng H, Zhang JT. Expression profiling of ABC 8315–20. Analysis of array CGH data: from signal ratio to gain and transporters in a drug-resistant breast cancer cell line 47. Lambros MB, Natrajan R, Reis-Filho JS. Chromogenic loss of DNA regions. Bioinformatics 2004;20:3413–22. using AmpArray. Mol Pharmacol 2005;68:430–8. and fluorescent in situ hybridization in breast cancer. 25. Beroukhim R, Getz G, Nghiemphu L, et al. Assessing 37. Huang R, Murry DJ, Kolwankar D, Hall SD, Foster DR. Hum Pathol 2007;38:1105–22. the significance of chromosomal aberrations in cancer: Vincristine transcriptional regulation of efflux drug 48. Hayes DF, Thor AD, Dressler LG, et al. HER2 and Methodology and application to glioma. Proc Natl Acad transporters in carcinoma cell lines. Biochem Pharma- response to paclitaxel in node-positive breast cancer. N Sci U S A 2007;104:20007–12. col 2006;71:1695–704. Engl J Med 2007;357:1496–506.

www.aacrjournals.org 5389 Cancer Res 2008; 68: (13). July 1, 2008

Correction: ABCC3 Amplification and Taxane Resistance

In the article on ABCC3 amplification and taxane resistance in the July 1, 2008 issue of Cancer Research (1), there is an error in Fig. 2. The corrected figure appears below.

Figure 2.

1. O’Brien C, Cavet G, Pandita A, Hu X, Haydu L, Mohan S, Toy K, Rivers CS, Modrusan Z, Amler LC, Lackner MR. Functional genomics identifies ABCC3 as a mediator of taxane resistance in HER2-amplified breast cancer. Cancer Res 2008;68:5380–9.

I2008 American Association for Cancer Research. doi:10.1158/0008-5472.CAN-68-21-COR2 www.aacrjournals.org 9105 Cancer Res 2008; 68: (21). November 1, 2008 Functional Genomics Identifies ABCC3 as a Mediator of Taxane Resistance in HER2-Amplified Breast Cancer

Carol O'Brien, Guy Cavet, Ajay Pandita, et al.

Cancer Res 2008;68:5380-5389.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/68/13/5380

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2008/06/25/68.13.5380.DC1

Cited articles This article cites 46 articles, 13 of which you can access for free at: http://cancerres.aacrjournals.org/content/68/13/5380.full#ref-list-1

Citing articles This article has been cited by 16 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/68/13/5380.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/68/13/5380. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.