Genomics Screen in Transformed Stem Cells Reveals RNASEH2A, PPAP2C, and ADARB1 As Putative Anticancer Drug Targets

249

Genomics screen in transformed stem cells reveals RNASEH2A, PPAP2C, and ADARB1 as putative anticancer drug targets

James M. Flanagan,1 Juan M. Funes,1 PPAP2C decreases cell proliferation by delaying entry Stephen Henderson,1 Laurence Wild,1 into S phase of the cell cycle and is transcriptionally Nessa Carey,2 and Chris Boshoff1 regulated by p53. These in vitro data validate PPAP2C and RNASEH2A as putative cancer targets and endorse 1UCL Cancer Institute, London, United Kingdom and this in silico approach for identifying novel candidates. 2 CellCentric Ltd., Cambridge, United Kingdom [Mol Cancer Ther 2009;8(1):249–60]

Abstract Introduction Since the sequencing of the human genome, recent efforts According to the Food and Drug Administration, the cost of in cancer drug target discovery have focused more on the development for each new anticancer drug ranges from 800 identification of novel functions of known genes and million to 2 billion U.S. dollars (1). It is estimated that only the development of more appropriate tumor models. In the one in five drugs that enter phase I trials will reach U.S. present study, we investigated in vitro transformed regulatory approval; more importantly, half of drugs that human adult mesenchymal stem cells (MSC) to identify enter the expensive phase III stage will not reach authori- novel candidate cancer drug targets by analyzing the zation (1). The reasons for this high attrition rate transcriptional profile of known enzymes compared with are numerous, including issues of on-target and off-target non-transformed MSC. The identified enzymes were com- safety concerns, particularly failure in the biological pared with published cancer gene expression data sets. hypotheses driving the target (2). Much of the recent efforts Surprisingly, the majority of up-regulated enzymes are in target discovery have therefore focused on identification already known cancer drug targets or act within known of novel disease relevant targets with a reduced probability druggable pathways. Only three enzymes (RNASEH2A, of subsequent failure by improving early validation ADARB1, and PPAP2C) are potentially novel targets that employing more suitable cancer models (2). are up-regulated in transformed MSC and expressed in There have been numerous attempts of in silico numerous carcinomas and sarcomas. We confirmed the prediction of cancer drug targets including structural overexpression of RNASEH2A, PPAP2C, and ADARB1 in analyses of all known proteins to define ‘‘the druggable transformed MSC, transformed fibroblasts, and cancer genome,’’ which converges on 2,000 to 3,000 targets that cell lines MCF7, SK-LMS1, MG63, and U2OS. In func- contain protein folds that are amenable to drug-like small tional assays, we show that small interfering RNA chemical compounds (3, 4). Statistical approaches have knockdown of RNASEH2A inhibits anchorage-indepen- also been used to define 250 potentially successful cancer dent growth but does not alter in vitro proliferation of relevant targets of which 60 to 70 are currently being cancer cell lines, normal MSC, or normal fibroblasts. actively investigated (5). However, the druggable genome Knockdown of PPAP2C impaired anchorage-dependent is a ‘‘moving target’’ with protein modeling algorithms in vitro growth of cancer cell lines and impaired the becoming more sophisticated, more hypothetical proteins in vitro growth of primary MSC but not differentiated being characterized allowing better annotations, and a human fibroblasts. We showthat the knockdownof changing perception of what classes of proteins can be targeted (4). The largest categories of current drug targets are G protein-coupled receptors, ion channels, and various classes of enzymes (e.g., protein kinases; ref. 4). Recent advances, including antisense delivery, small inter- Received 7/8/08; revised 9/10/08; accepted 10/3/08. fering RNA (siRNA) knockdown, and small-molecule Grant support: CellCentric (J.M. Flanagan and N. Carey),Medical Research Council UK (L. Wild),and Cancer Research UK (J.M. Flanagan,S. carriers, may extend the range of proteins that could be Henderson,and C. Boshoff). targeted (6, 7). The costs of publication of this article were defrayed in part by the Recent studies have taken a reverse genomics approach payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to to identify potential drug targets by screening short hair- indicate this fact. pin RNA (shRNA) libraries in cancer cell lines and assaying Requests for reprints: James M. Flanagan,Cancer Research UK Viral for genes required for proliferation (8, 9). This approach Oncology Group,University College London Cancer Institute,Paul has revealed a larger number of genes required by normal O’Gorman Building,74 Huntley Street,London,United Kingdom WC1E 6BT. Phone: 44-20-7679-6859; Fax: 44-20-7679-6851. cells for proliferation and substantially less in cancer E-mail: [email protected] cells, reflecting the ability of cancer cells to evade growth- Copyright C 2009 American Association for Cancer Research. inhibitory cues (9). This work defines ‘‘non-oncogene doi:10.1158/1535-7163.MCT-08-0636 addiction’’ genes, which are required by cancer cells to

Mol Cancer Ther 2009;8(1). January 2009

proliferate but are not required for normal cell proliferation quantitative reverse transcription-PCR (qRT-PCR) valida- (10). tion of the data set has been done previously, and further The cancer stem cell hypothesis suggests that tumors details and quality-control measures can be found in the derive from a subpopulation of multipotent stem cells, ArrayExpress database4 with the accession no. E-MEXP- which continually renew and sustain the malignant growth 563 (14). (11). The tumor stem cells are potentially more resistant to The carcinoma expression data were downloaded from cytotoxic therapies and may be molecularly distinct from the expO public repository5 by way of the National Center the bulk of the tumor cells (12). Most cell culture models of for Biotechnology Information Gene Expression Omnibus carcinogenesis have used differentiated cells or well- database.6 We selected 146 samples hybridized to Affy- established cancer cell lines to investigate the properties metrix hgu133plus2 genechips, covering 11 common types of oncogenes, tumor suppressor genes, or to identify novel of carcinoma (breast, colon, endometrium, kidney clear- cancer drug targets (13). However, if the cancer stem cell cell carcinoma, kidney papillary carcinoma, lung adeno- hypothesis is correct, then the most appropriate cells in carcinoma, ovary, prostate, rectum, thyroid, and bladder) which to investigate the genes involved in carcinogenesis with no fewer than 10 of each type. Our mesenchymal may be stem cells. To this end, we have generated tumor expression data represent 19 different subtypes, transformed human adult mesenchymal stem cells (MSC) including alveolar rhabdomyosarcoma, chondroblas- using a stepwise introduction of five genetic hits including toma, chondrosarcoma, chordoma, chondromyxoid hTERT, inactivation of p53 and pRb, introduction of small fibroma, dedifferentiated chondrosarcoma, embryonal T antigen to inactivate PP2A leading to stabilization of rhabdomyosarcoma, Ewing’s sarcoma, fibromatosis, c-Myc, and finally introduction of H-Ras (14). We have lipoma, leiomyosarcoma, myxoid liposarcoma, synovial shown that MSC require the same number of genetic hits sarcoma, malignant peripheral nerve sheath tumor, (five) as differentiated cells to become fully transformed neurofibroma, osteosarcoma, pleiomorphic sarcoma, (tumor formation in athymic mice). schwannoma, and well-differentiated liposarcoma (15). We hypothesize that the enzymes specifically up- Raw CEL file data were processed and normalized to regulated in the transformed MSC model will also be produce expression profiles using the ‘‘rma’’ algorithm. up-regulated in various sarcomas and carcinomas and Cancer versus normal tissue control gene expression data may provide novel targets for cancer drug development. were identified employing the Oncomine Web resource7 In this study, we have taken a forward genomics using the differential expression function with a P cutoff approach for drug target discovery by mining gene of <0.0001(Supplementary Table). 8 Oncomine uses a data expression data from numerous cancer microarray data analysis method, whereby all data are log-transformed, sets. Due to the current interest in enzymes for drug median-centered per array, and SD-normalized to 1per targets, we have focused here on identifying novel array. All tumor array data are described in Supplemen- enzymes as candidate cancer drug targets. This trans- tary Tables S2 to S4.8 formed primary stem cell model also provides an Hierarchical clustering of MSC transformation gene excellent resource for in vitro validation of any identified expression data was done using DNA-Chip Analyzer therapeutic targets or to test novel drugs in the early (dChip).9 Using a distance metric of 1-correlation and a stages of stepwise oncogenic transformation. centroid linkage method, significant clusters were identified with a P cutoff of <0.001. Further significance of Materials and Methods clusters was done using the R package ‘‘pvclust,’’ which P Enzyme Data Set computes an approximate unbiased value for clusters using a multiscale bootstrap resampling method (16). We identified all known enzymes in the human genome n n using the Expasy Enzyme database3 August 21, 2007 pvclust was done with both = 1,000 and = 10,000 bootstrap with similar results (Supplementary Fig. S1).8 release, which categorizes human enzymes into 207 oxi- P doreductases (EC 1), 319 transferases (EC 2), 300 hydrolases Significant clusters were identified with < 0.05. qRT-PCR (EC 3), 63 lyases (EC 4), 33 isomerases (EC 5), and 73 ligases f qRT-PCR was done on an Eppendorf Mastercycler using (EC 6). These 1,000 enzymes are represented by 2,256 ¶ probe sets from the Affymetrix hgu133plus2 genechips. For the following primer pairs: PPAP2C_F (5 -ctgtatgtgcaggcac- gact-3¶) and PPAP2C_R (5¶-aaaggccaccaggaagaact-3¶), RNA- this study, we have not included candidate enzymes that ¶ ¶ do not have an EC number. SEH2A_F (5 -ccagaccatcctggagaaag-3 ) and RNASEH2A_R Gene Expression Microarray Data We have used our transformed MSC cancer model Gene Expression Microarray data as a primary screen for up- 4 ArrayExpress (http://www.ebi.ac.uk/microarray-as/aer/). regulation of enzymes during transformation using a 5 expO repository (https://expo.intgen.org/geo/home.do). cutoff of 2-fold up-regulation (P < 0.05; ref. 14). Extensive 6 National Center for Biotechnology Information Gene Expression Omnibus (ftp://ftp.ncbi.nlm.nih.gov/pub/geo/DATA/SeriesMatrix/GSE2109). 7 Oncomine (http://www.oncomine.org). 8 Supplementary material for this article is available at Molecular Cancer Therapeutics Online (http://mct.aacrjournals.org/). 3 Expasy (http://www.expasy.ch/enzyme/). 9 dchip (www.dchip.org).

Mol Cancer Ther 2009;8(1). January 2009

(5¶-tgagtccctcctgattctcg-3¶), ADARB1_F (5¶-gtccgactcgaccat- containing 10% human serum (Stem Cell Technologies) and gactt-3¶) and ADARB1_R (5¶-tgccagagacaggaggattt-3¶), and 1ng/mL basic fibroblast growth factor (R&D Systems). The GAPDH_F (5¶-ggagtcaacggatttggtcgta-3¶) and GAPDH_R primary human dermal fibroblasts (HF 0), immortalized (5¶-ggcaacaatatccactttaccagagt-3¶). The reaction mix human dermal fibroblasts (HF 1), transformed human contained 1 SYBR Green master mix (Applied Biosystems) dermal fibroblasts (HF 5), breast cancer cell line MCF7, and 0.5 mmol/L of each forward and reverse primers in a leiomyosarcoma cell line SK-LMS1, and osteosarcoma cell volume of 30 AL. PCR cycling consisted of 95jC for 10 min lines MG63 and U2OS were cultured in DMEM containing and then 40 cycles of 95jC for 30 s, 60jC for 60 s followed by 10% fetal bovine serum (Life Technologies/Invitrogen). a melt-curve analysis. Fold change in expression was siRNA/shRNA Knockdown and In vitro Assays DDC calculated by t normalized to GAPDH for each sample Knockdown was done by transfection with a pool of four and normalizing each of the cell lines to the average of the siRNA oligonucleotides (100 nmol/L final concentration; control cells: parental MSC (MSC 0), primary human Dharmacon) for each gene: RNASEH2A, ADARB1, and fibroblasts (HF 0), or transformed MSC (MSC 5). PPAP2C or ON-TARGETplus Nontargeting Control. Four Cell Culture individual siRNA oligonucleotides were also used for Primary MSC and each line containing a different set of PPAP2C knockdown. Transfection was done using Oligo- oncogenes (see Fig. 1) were grown in Mesencult medium fectamine reagent (Invitrogen) and the manufacturer’s

Figure 1. Gene expression microarray data for 45 enzymes in the MSC cancer model. Three replicate gene expression microarrays data for each of the five transforming steps (hits) is compared with the parental MSC (see ref. 14 for more details). The parental MSC was immortalized with hTERT (MSC 1hit); then,pRb and p53 were inactivated by addition of the human papillomavirus proteins E7 (MSC 2 hits) and E6 (MSC 3 hits) followed by inactivation of PP2A and stabilization of c-Myc by the addition of SV40 small T antigen (MSC 4 hits); finally,constitutively active H-Ras was expressed (MSC 5 hits). Enzymes that showed statistically significant >2-fold increase in expression in MSC 5 hits compared to parental MSC were selected and hierarchical clustering was done using dChip as described in Materials and Methods. Positive control enzymes (c) described in Fig. 2 and Supplementary Fig. S2 and novel enzymes (b) described in Fig. 2 are marked.

Mol Cancer Ther 2009;8(1). January 2009

Table 1. Gene list of up-regulated enzymes in transformed MSC cancer model

HGNC HGNC name GeneRifs* PubMed Pathway/function PharmGKB symbol cites* drugs/drug pathwaysc

RNASEH2A RNase H2, subunit A 0 11 RNA degradation None PPAP2C Phosphatidic acid phosphatase type 2C 2 13 Glycerolipid None synthesis ADARB1Adenosine deaminase, RNA-specific, B1 10 58 RNA pre-mRNA None editing PRSS3 Protease, serine, 3 (mesotrypsin) 8 47 Trypsin inhibitor None degradation CTSH Cathepsin H 6 117 Lysosomal None degradation MTAP Methylthioadenosine phosphorylase 12 108 Polyamine None metabolism CPE Carboxypeptidase E 4 117 Neuropeptide None cleavage DHFR Dihydrofolate reductase 22 145 Purine Methotrexate, biosynthesis 5-fluorouracil, antimetabolite, antineoplastic agents MMP3 Matrix metalloproteinase 3 95 257 Extracellular Pravastatin matrix degradation MMP1 Matrix metalloproteinase 1 148 311 Extracellular Doxorubicin matrix degradation PTGS2 Prostaglandin-endoperoxide 549 616 Prostaglandin Aspirin, celecoxib, synthase 2 (cyclooxygenase-2) biosynthesis clomipramine, glucocorticoids, prostaglandins, rofecoxib, statin, valdecoxib

NOTE: For full table of 44 enzymes, see Supplementary Table S1. *The number of GeneRifs or PubMed citations related to each enzyme (as of September 2007). cIdentification of known drugs or drug pathways that target each gene. bNumber of drug programs targeting each gene. xIdentification of patents identifying these genes specifically as cancer drug targets. kIdentified as a potential target for small molecules based on protein structure from Russ and Lampel (4). {Bioinformatic comparison with successful drug targets and ranking (percentile) from Mayburd et al. (5).

protocol. Additional knockdown of RNASEH2A was was assessed by soft-agarose assays as described previous- done by transfection of shRNA vectors for RNASEH2A, ly (14). For cell cycle analysis, MCF7 cells were transfected nonsilencing control, and vector control from the Open with siRNA control pool or PPAP2C siRNA and grown for Biosystems whole-genome pGIPZ shRNA library using 48 h followed by serum starvation for 24 h to synchronize FuGene reagent (Roche) and the manufacturer’s protocol. cells in G1 phase. Cells were analyzed at 8 and 24 h post- Transfection efficiency of shRNA vectors was estimated at serum release to assess S-phase progression. Cells were f30% by monitoring green fluorescent protein from the harvested and fixed in 70% ethanol at 0jC, and the nuclear vector. Following transfection, the cells were grown for DNA was stained using a solution of propidium iodide 48 h before harvesting for qRT-PCR or in vitro assays. Cell (50 mg/mL), RNase A (1mg/mL), and Triton X-100 proliferation and viability was determined using the (0.02%) in PBS. Cell suspensions were analyzed on a CellTiter 96 Aqueous One solution (Promega). Cells were FACSCalibur (Becton Dickinson) using CellQuest and plated in triplicate at 1,000 cells per well of a 96-well plate. Modfit data analysis software. The CellTiter solution was added at both 0 and 48 h time Western Blot points using 30 AL/well and further incubated for 2 h at Antibodies used in this study include sc-126 (DO-1) for 37jC, and the absorbance was read at 490 nm to quantify p53(SantaCruzBiotechnology)andAb-1foractin the formazan product. Anchorage-independent growth (Oncogene).

Mol Cancer Ther 2009;8(1). January 2009

Table 1. Gene list of up-regulated enzymes in transformed MSC cancer model (Cont’d)

Therapeutic Thomson Patentsx Druggable Successful Mayburd target database c Pharmab targetk anticancer target{ rank{

None None No No Yes 0.9845 None None No No No 0.8114

None None No No No 0.3103

None None No Yes No 0.8400

None None No Yes No 0.7801

Research target None WO 2007/097648 A1Yes No

Research target None No Yes No 0.4929

Lamotrigine; 45 WO 2007/132146 A1 Yes Yes 0.9782 malarone; pyrimethamine; pemetrexed; proguanil; trimetrexate Research target 9 US 2005/250789 A1, US 5473100 Yes Yes 0.3412

Research target 8 US 2005/250789 A1, US 5473100 Yes Yes 0.8068

Celecoxib; 124 EP1239879 B1, EP1654262 B1 Yes Yes 0.9582 rofecoxib; valdecoxib

Other Online Resources primary MSC, and also in various cancers. Such enzymes We screened for novelty of targets using Entrez Gene10 could potentially be considered as new targets for cancer and PubMed11 and identification of known drug targets drug development. We identified 56 probe sets, represent- from PharmGKB,12 Thomson Pharma,13 and Therapeutic ing 44 enzymes (one or two probe sets per gene), which are Target Database.14 Patent information was screened using significantly up-regulated >2-fold (false discovery rate- esp@cenet15 and Patent Lens.16 corrected P < 0.05) in the transformed MSC (5 hits, MSC 5) compared with the primary MSC (0 hits, MSC 0; Fig. 1). Results Hierarchical clustering analysis of each of these probe sets P Expression of Enzymes in the Transformed MSC identifies four significant ( < 0.05) clusters that stratify Model with the genetic alterations during stepwise transformation (Supplementary Fig. S1).8 The first cluster includes 4 We analyzed the expression of all known enzymes in the enzymes (UPP1, XDH, MMP3, and CPE) up-regulated only human genome, with the aim of identifying novel enzymes after the fifth oncogenic hit, correlating with activation of that are up-regulated in transformed MSC, and not in the Ras pathway (constitutively active H-Ras) and the fully transformed state. The second major cluster contains 22 enzymes (from GCDH to RNMT) significantly increased after the inactivation of PP2A and stabilization of c-Myc 10 Entrez Gene (http://www.ncbi.nlm.nih.gov/sites/entrez?db=gene). (fourth hit), and a third cluster contains 17 enzymes (from 11 PubMed (http://www.ncbi.nlm.nih.gov/sites/entrez?db=pubmed). 12 PharmGKB (http://www.pharmgkb.org/index.jsp). GPD2 to MTAP), which are significantly up-regulated after 13 Thomson Pharma (http://www.thomson-pharma.com). the inactivation of pRb. The fourth cluster contains 4 14 Therapeutic Target Database (http://xin.cz3.nus.edu.sg/group/cjttd/ttd.asp). 15 esp@cenet (http://gb.espacenet.com/). enzymes (PTGS2, OAS1, NP, and PRSS3) and displays an 16 Patent Lens (http://www.patentlens.net/patentlens/structured.cgi). interesting profile being up-regulated after introduction of

Mol Cancer Ther 2009;8(1). January 2009

hTERT (first hit) or p53 (third hit) suppression but then Identification of Novel Cancer Targets down-regulated after the inactivation of PP2A (fourth hit) Novelty of candidate enzymes was investigated using and again increasing on H-Ras activation. We have shown several measures (Table 1). GeneRifs or ‘‘Gene Reference previously that these clusters are representative of genes into Function’’ consist of phrases taken from publications known to be transcriptionally regulated by each oncogenic describing a function of the gene or association with hit (Supplementary Fig. S68 in ref. 14). Furthermore, as an particular diseases or cancers. The number of PubMed example, we show that gene set enrichment analysis citations for particular genes also serves as a measure of identifies a significant overrepresentation of known Myc how well characterized it is, and a systematic search of targets (P = 0.009262) in the genes significantly up- the literature on PubMed was used to identify any known regulated or down-regulated after the fourth hit (17). associations with cancers. Finally, PharmGKB, Therapeu- Identification of Known CancerTargets tic Target Database, Thomson Pharma, and Patent data- The majority of significantly up-regulated enzymes (32 of bases indicate whether the gene is already a target of a 44, 73%) in the transformed MSC either are already well- known drug or acts within a known drug pathway. known cancer drug targets, act within known drug path- Specifically, if the gene is listed on any of the Web sites ways, or are current research targets previously identified as a drug target, then we have considered that as a as up-regulated in various malignancies (Table 1; Supple- known target. Additionally, if the gene is within a very mentary Table S1).8 This is confirmed statistically in well described pathway, such as the methotrexate comparison with the percentage of known target enzymes pathway (Supplementary Fig. S4),8 then we have also in those that do not change expression in this model (12% considered this a known target. Taken together, these are known targets) or are significantly down-regulated measures have identified 3 enzymes (RNASEH2A, (11% are known targets; P = 8.425e-08, m2 test). This can be ADARB1, and PPAP2C), which are expressed in various considered as a positive control for the identification of malignancies, up-regulated in some cancers compared candidate drug targets in this model and suggests that with normal tissues, and are candidate novel cancer there may be few truly novel up-regulated enzymes in this targets (Table 1; Fig. 2). model system left to be identified as cancer drug targets. RNASEH2A is the catalytic subunit of the RNase H2 We have identified four examples of well-characterized complex, which is the major source of RNase H activity cancer drug targets that are also identified as significantly in mammalian cells (22), and is up-regulated in MSC up-regulated in our transformed MSC cells: PTGS2 (cyclo- after the inactivation of pRb and also in fully trans- oxygenase-2), MMP1and MMP3, and DHFR. Analysis of formed MSC. It is also up-regulated in a variety of the expression of these candidates in the transformed MSC different cancer types including bladder, brain (glioblas- and in various cancer gene expression data sets reveals toma multiforme, oligodendroglioma, and oligoastrocy- novel insights into the regulation of these genes and other toma), breast, head, and neck squamous cell carcinomas enzymes with similar expression profiles (Fig. 1) and also as well as leukemias (T- and B-cell acute lymphoblastic identifies novel cancers in which these genes are overex- leukemia and acute myeloid leukemia), melanomas, and pressed (Fig. 2A; Supplementary Fig. S2).8 As an example, seminomas (Fig. 2B). It is expressed in all of the we show that hTERT alone is capable of up-regulating carcinomas and sarcomas investigated with the highest transcription of PTGS2 in MSC with a 1.8-fold increase (P = expression observed in breast carcinomas and dediffer- 0.039) compared with parental MSC. Furthermore, we entiated chondrosarcomas, respectively. RNASEH2A was observe increased expression of PTGS2 in seminoma, also one of the top 2% of genes (ranked 0.9845) identified chondromyxoid fibroma, and malignant peripheral nerve by Mayburd et al. that showed significant correlation sheath tumor, which have not been reported previously. with the profiles of all known successful cancer drug Structural Analysis of Druggable Enzymes targets (5). A structural analysis of all known genes has been done PPAP2C is a member of the phosphatidic acid phospha- previously to identify the ‘‘druggable genome,’’ that is, tase family of enzymes (also known as lipid phosphate proteins that contain structural elements amenable to phosphatases) that regulate the dephosphorylation of small-molecule inhibition (3, 4). A comparison of the up- lipid phosphates (23). Compared with matched normal regulated enzymes in the MSC model and the druggable tissues, PPAP2C is significantly overexpressed in ovarian proteins identified 16 enzymes that are already targeted by carcinoma similarly to other family members (ref. 24; specific drugs (ODC1, MME, NP, ALDH1A3, ADCY3, Fig. 2C). However, our analysis has shown that it is also RRM1, ASNS, PPAT, UMPS, SOAT1, XDH, DHFR, DPYD, overexpressed in bladder, lung, and prostate cancers MMP3, MMP1, and PTGS2). In addition, we have also compared with normal tissues. In the cancer gene expres- identified 4 enzymes that have been characterized as sion data, we detect expression of PPAP2C in all druggable and are up-regulated in numerous cancers, carcinomas and sarcomas and it appears to be particularly although no drugs have yet been developed against these highly expressed in breast carcinoma, dedifferentiated targets (MTAP, PRSS3, CPE, and CTSH; Supplementary chondrosarcoma, fibromatosis, synovial sarcoma, and Fig. S3).8 These 4 enzymes have been implicated previously neurofibroma. PPAP2C is also significantly overexpressed in cancer and thus cannot be considered novel candidates in high-grade breast tumors compared with low-grade (18–21). tumors (Supplementary Fig. S5).8

Mol Cancer Ther 2009;8(1). January 2009

Figure 2. Cancer gene expression microarray data for PTGS2 (A) and novel cancer drug targets (B-D). Gene expression microarray data obtained from the National Center for Biotechnology Information Gene Expression Omnibus for carcinomas and sarcoma gene expression data (15) are presented as box plots of log2 expression values. The normal versus tumor gene expression data were obtained from Oncomine and is presented as log-transformed,median- centered per array,and SD-normalized to 1 per array. Only tumor types with significant overexpression compared with normal tissue ( P < 0.0001) are presented. Asterisk, novel cancers in which detectable levels of expression (log2 expression >5 in the carcinomas or sarcomas) or overexpression (in normal versus tumor comparisons) are detected for PTGS2. The full list of sarcoma abbreviations is listed in Supplementary Tables.

Mol Cancer Ther 2009;8(1). January 2009

bone tumor, compared with all other sarcomas. The regulation of ADARB1has not been described. We detect significant 5-fold increase in expression of ADARB1in the gene expression microarray data (Fig. 1) and by qRT-PCR (data not shown) following the addition of small T antigen, which stabilizes the c-Myc oncogene, suggesting that Myc may be involved in transcriptional regulation of this enzyme. Validation of Novel CancerTargets We have confirmed the increased expression of RNA- SEH2A (4.8-fold; P = 0.012), ADARB1 (6.3-fold; P = 0.0002), and PAPP2C (7.7-fold; P = 0.0014) by qRT-PCR in the transformed MSC (MSC 5 hits) compared with the parental cells (Fig. 3A). We have also sought to confirm the overexpression of these enzymes in transformed human fibroblasts, which may be more representative of carcinomas. These primary human fibroblasts were transformed with the same five oncogenic hits as were used in the MSC and showed similar signs of transformation including the ability to form colonies in soft agarose and tumors in athymic mice (14). In the transformed fibroblasts, significant overexpression was detected for both RNASEH2A (2.0-fold; P = 0.0058) and PPAP2C (6.2-fold; P = 0.00065) compared with the primary fibroblasts. However, ADARB1 was not significantly different to the parental fibroblasts (Fig. 3B). Expression was also confirmed in each of the cancer cell lines tested with similar expression levels to the transformed MSC and significantly lower expression in the two normal cell lines (Fig. 3C). We have investigated the role of RNASEH2A, PPAP2C, and ADARB1in proliferation and transformation by siRNA knockdown in the transformed MSC, transformed human Figure 3. Validation of overexpression of novel candidate enzymes. fibroblasts, and the cancer cell lines MCF7 (breast cancer), qRT-PCR was done for ADARB1,RNASEH2A,and PPAP2C in the SK-LMS1(leiomyosarcoma), MG63 (osteosarcoma), and transformed MSC (MSC 5) compared with the parental MSC (MSC 0) (A) and in human dermal fibroblasts (HF 0) compared with transformed U2OS (osteosarcoma). We compared the cell viability human dermal fibroblasts (HF 5) (B). Fold change in expression was results in cancer cells with the immortalized normal cell DD calculated in three replicates by C t normalized to GAPDH for each lines MSC 1(MSC + hTERT) and HF 1(human fibroblasts + sample and normalizing each of the cell lines to the average of the parental cells. Student’s t test was done for statistical analysis. C, expression of hTERT; Fig. 4). Knockdown for each of the genes was RNASEH2A,PPAP2C,and ADARB1 in cancer cell lines was calculated in confirmed by qRT-PCR analysis of mRNA levels in each of DD 8 three replicates by C t normalized to GAPDH for each sample and the cell lines tested (Supplementary Fig. S6). siRNA normalizing each of the cell lines to the average of the transformed MSC knockdown of RNASEH2A or ADARB1significantly for each of the genes separately. altered the cell viability of one osteosarcoma cell line, U2OS; however, neither gene altered the proliferation of Adenosine deaminase is a key degradative enzyme of any of the other cell lines tested. Knockdown of PPAP2C purine metabolism and is crucial for DNA replication and significantly decreased the cell viability of all cancer cell several purine nucleoside analogues have been developed lines tested, except the leiomyosarcoma cell line SK-LMS1. to specifically target adenosine deaminase, including In the normal immortalized MSC and fibroblasts, inhibition pentostatin and cladribine (Thomson Pharma Web site10). of RNASEH2A and ADARB1did not alter cell viability; There are, however, nine other members of the adenosine however, PPAP2C knockdown altered the viability of the deaminase family in the human genome, many of which MSC but not the differentiated fibroblasts (Fig. 4A). The are RNA-specific including ADARB1, one of the candidates sequence specificity of PPAP2C knockdown by the pooled identified by our analysis. ADARB1shows significant up- siRNA oligonucleotides was confirmed by knocking down regulation in lymphomas and seminomas in comparison expression with the four oligonucleotides separately in the with normal tissues and is expressed consistently in transformed fibroblasts compared with the normal fibro- most carcinomas and sarcomas that we have investigated blasts (Fig. 4B). (Fig. 2D). Of note is the significantly increased expression We have shown previously that transformed MSC form in chondromyxoid fibromas (CMF), a benign cartilaginous colonies rapidly in soft agarose similarly to many cancer

Mol Cancer Ther 2009;8(1). January 2009

cell lines (14). The transient siRNA or shRNA knockdown Discussion of RNASEH2A in transformed MSC significantly reduced In this study, we have sought to identify novel cancer drug the number of colonies formed in soft agarose (Fig. 4C targets by investigating a transformed MSC model. This and D). This finding was also observed in transformed model represents the oncogenic alterations that occur in the fibroblasts HF 5, MCF7, SK-LMS1, and MG63. The transformation process and the transcriptional changes osteosarcoma cell line U2OS and the immortalized MSC observed are specific to each oncogenic hit in a homoge- and fibroblasts do not form colonies in soft agarose. This nous cell type. We use this model in addition to studying is the first evidence that RNASEH2A may play a role in tumor tissue compared with normal tissue, which represent transformation. Although ADARB1knockdown sup- the molecular alterations that have occurred late in the pressed soft-agarose colony formation in transformed tumor development (at the time of diagnosis and surgical MSC, we did not observe a similar result in any other resection of the tumor) and in a heterogeneous cell cell line. population. In rat fibroblasts, the homologous gene, Ppap2c, caused a The approach that we have taken in this study provides premature entry into S phase when overexpressed and an excellent resource for investigating the regulation and delayed S-phase entry when knocked down (25). We have expression of established successful drug targets, as we confirmed this in the human breast cancer cell line, MCF7, have shown that many of these enzymes are up-regulated which expresses the highest level of PPAP2C and showed a in this model. We have elaborated on four of these targets, high degree of inhibition of in vitro proliferation. We show PTGS2, DHFR, MMP1, and MMP3, which we observe are that PPAP2C siRNA knockdown in MCF7 delays S-phase up-regulated in numerous cancer types and serve as a progression and does not induce apoptosis (Fig. 5A). benchmarkforthenoveltargetsthatwehavealso Furthermore, we show that PPAP2C is up-regulated in identified. Despite the cardiovascular side effects of one the transformed MSC model after the addition of human inhibitor (Vioxx), PTGS2 inhibitors (coxibs) are currently in papillomavirus E6 (which suppresses p53 function) and is clinical trials for various malignancies including colorectal, also up-regulated following shRNA knockdown of p53 or non-small cell lung, breast, prostate, and pancreatic cancers overexpression of a dominant-negative form of p53 (Fig. (26–28). PTGS2 was shown previously to be overexpressed 5B-D). These data are consistent with a role for p53 in the in chordoma and chondrosarcoma; however, the use of transcriptional regulation of PPAP2C. coxibs for sarcomas has yet to be explored (29, 30). Our

Figure 4. Functional consequence of knockdown of RNASEH2A,PPAP2C,and ADARB1. A, transfection of pooled siRNA,individual siRNA oligonucleotides,or shRNA vectors was done to knockdown RNASEH2A,PPAP2C,and ADARB1 in the transformed MSC (MSC 5),transformed human fibroblasts (HF 5),breast cancer cell line MCF7,leiomyosarcoma cell line SK-LMS1,osteosarcoma cell lines MG63 and U2OS,and immortalized MSC (MSC 1) and human fibroblasts (HF 1). Cell viability was assessed by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay. Data represent the percentage growth compared with untreated cells calculated as a ratio of 48/0 h and an average of three replicate wells. B, transfection of individual siRNA oligonucleotides to PPAP2C was used to knockdown PPAP2C in transformed human fibroblasts (HF 5) and human fibroblasts (HF 1). C, growth of colonies in soft agarose was done as described previously (14). The number of colonies was quantitated by counting colonies in three fields from three replicate experiments. D, confirmation of the specificity of the knockdown with an additional shRNA vector. Student’s t test was done for statistical analysis. Bars, SE from triplicate experiments.

Mol Cancer Ther 2009;8(1). January 2009

Figure 5. PPAP2C is involved in S-phase progression and is regulated by p53. A, transfection of siRNA oligonucleotides was done to knockdown PPAP2C in the breast cancer cell line MCF7. Cells were grown for 48 h subsequent to transfection,serum starved for 24 h to synchronize the cells in G 1 phase (data not shown),and released for 24 h before fixing,staining with propidium iodide,and analysis by flow cytometry. Cell cycle phase percenta ges were calculated by the Modfit software package. B, expression of PPAP2C was measured by qRT-PCR in the transformed MSC model showing up- regulation subsequent to the addition of human papillomavirus E6. C, Western blot analysis showing knockdown of p53 by E6 in the MSC model (MSC2- E6) compared with MSC2-E7 or by siRNA knockdown of p53 in MSC2-E7 compared with vector control or overexpression of a dominant-negative form of p53 (p53-175H). D, expression of PPAP2C was measured by qRT-PCR in the cells with abrogated p53.

analysis identified several novel cancers in which PTGS2 appears to be in the development of specific active site- appears to be overexpressed including the difficult-to-treat directed inhibitors against specific MMPs (32). Although sarcomas, chondromyxoid fibroma, and malignant periph- our analysis has confirmed the overexpression of these eral nerve sheath tumor. These data provide evidence that genes in many cancers, we have also identified several coxibs could be worth investigating in these tumor types novel tumors that express MMP1(chondroblastoma) and and suggest novel indications for these drugs. Methotrex- MMP3 (chordoma and seminoma), suggesting that the full ate is one of the first clinically useful broad-range range of malignancies that the drugs against these targets antimetabolic anticancer drugs that targets DHFR (31). It could be used for is yet to be explored (Supplementary is a drug that has been used for numerous cancer types and Fig. S2).8 Similarly to these known targets, the novel it is encouraging to see that the MSC model includes not enzymes that we have identified, RNASH2A, PPAP2C, only up-regulation of DHFR but also many other enzymes and ADARB1, are overexpressed in numerous cancer types involved in the de novo purine biosynthesis pathway and targeting these genes could have potentially broad (Supplementary Fig. S6).8 The MMP gene family including therapeutic options. MMP1and MMP3 are also recognized as pharmaceutical The lysophosphatidic acid signaling cascade may be a targets (32). The challenge ahead for this class of drugs novel target for therapy in ovarian cancer (24). Indeed,

Mol Cancer Ther 2009;8(1). January 2009

targeting the lipid signaling cascade is an emerging In conclusion, the in silico approach investigating all therapeutic strategy for cancer, inflammation, and meta- known enzymes in a transformed stem cell model is a bolic diseases (33). Furthermore, phosphatidic acid phos- useful approach that could equally be applied to investi- phatase enzymes are involved in receptor-activated signal gate all G protein-coupled receptors, ion channel genes, or transduction by the phospholipase D enzyme family, which putative enzymes. In addition to identifying novel candi- are key requirements for Ras-mediated transformation (34). date cancer targets, this genetic approach can also improve Our data suggest that overexpression of PPAP2C, observed current drug therapy in humans by identifying novel in numerous human cancers, may be a requirement for cancers in which known targets are up-regulated and increased cell proliferation. It is, therefore, not unexpected provide further insight into the oncogenic regulation of that its regulation is controlled in part by p53, which is these known targets. We have identified three enzymes, known to suppress many other cell cycle regulating genes RNASEH2A, PPAP2C, and ADARB1, which, to our such as chk1, cdc20, cdc25A, and cdc25c (35). Our data also knowledge, are novel putative candidate cancer drug show that the normal MSC are inhibited by the knockdown targets. We have confirmed the overexpression in cancer of PPAP2C, whereas the differentiated cells are not affected cell lines and have shown that inhibition of two of these despite the similar expression levels. We suggest that this is candidates, RNASEH2A and PPAP2C, leads to a reduction due to a common feature shared by both the cancer cell in the in vitro proliferation or transformation of cancer cells. lines and the normal stem cells, which could be the ability This suggests that our analytical screen, combining both for self-renewal or an increased susceptibility to aberrant our in vitro model and public gene expression data, may cell cycle signals. This particular enzyme family could be have enriched for genes that alter the proliferation of cancer attractive targets as they are expressed on the plasma cells. However, the extent to which this analysis has membrane, with the active site facing the extracellular enriched for true drug targets can only be determined matrix potentially allowing recognition by therapeutic with additional screening of more genes compared with a antibodies. variety of genes that were not hits in this screen. Both of The RNase H2 complex is a heterotrimer composed of the these identified genes could be classified as ‘‘non-oncogene catalytic subunit RNASEH2A and the noncatalytic subunits addiction’’ genes that are required for continued cancer cell RNASEH2B and RNASEH2C, which together are able to proliferation but not essential for normal differentiated cell recognize and cleave a single ribonucleotide embedded in a proliferation. This category of genes is the most desirable DNA-DNA complex (36). This complex also mediates the for targeting for cancer therapy (10). However, validation excision of single ribonucleotides from DNA:RNA of these genes as true cancer targets can only be claimed duplexes such as in the removal of the lagging-strand once a therapeutic agent is developed and shown to be Okazaki fragment RNA primers during DNA replication clinically effective by acting against the target to which it (37). Thus, a plausible biological hypothesis for this gene as was designed (43). a candidate drug target is likely to involve its role in DNA Disclosure of Potential Conflicts of Interest replication. It is now apparent that the role of RNA in the N. Carey: CellCentric grant support,CellCentric employee. J. Flanagan: cell includes regulation of gene expression via RNA CellCentric-sponsored post-doctoral fellow. No other potential conflicts interference pathways, heterochromatin formation, and of interest were disclosed. targeting DNA methylation and histone modifications (38, References 39). RNA regulating enzymes, therefore, could represent a 1. DiMasi JA,Grabowski HG. Economics of new oncology drug novel class of enzymes that could be considered as development. J Clin Oncol 2007;25:209 – 16. potential cancer drug targets. Inactivating mutations in 2. Lindsay MA. Target discovery. Nat Rev 2003;2:831 – 8. the three RNase H2 subunits have recently been implicated 3. Hopkins AL,Groom CR. The druggable genome. Nat Rev 2002;1: in the autosomal recessive neurologic disorder Aicardi- 727 – 30. Goutie`res syndrome (36). If RNASEH2A is involved in 4. Russ AP,Lampel S. The druggable genome: an update. Drug Discov transformation as suggested by our data, it would be Today 2005;10:1607 – 10. interesting to further investigate whether heterozygous 5. Mayburd AL,Golovchikova I,Mulshine JL. Successful anti-cancer drug targets able to pass FDA review demonstrate the identifiable signature carriers of these Aicardi-Goutie`res syndrome mutations are distinct from the signatures of random genes and initially proposed to a certain extent protected against cancer. targets. Bioinformatics 2008;24:389 – 95. ADARB1is an RNA-specific adenosine deaminase, 6. Okuyama M,Laman H,Kingsbury SR,et al. Small-molecule mimics of an a-helix for efficient transport of proteins into cells. Nat Methods 2007; which encodes an enzyme that is responsible for pre- 4:153 – 9. mRNA editing of the glutamate receptor subunit B by 7. Pai SI,Lin YY,Macaes B,Meneshian A,Hung CF,Wu TC. Prospects of site-specific deamination of adenosines (40). Although we RNA interference therapy for cancer. Gene Ther 2006;13:464 – 77. detect no adverse effects of ADARB1knockdown in the 8. Silva JM,Marran K,Parker JS,et al. Profiling essential genes in cancer cell lines tested, the role of RNA editing may play a human mammary cells by multiplex RNAi screening. Science 2008;319: 617 – 20. role in brain tumor formation (41, 42). This has been 9. Schlabach MR,Luo J,Solimini NL,et al. Cancer proliferation gene described as an epigenetic mechanism and may also help discovery through functional genomics. Science 2008;319:620 – 4. or hinder miRNA functions by editing to interfere with 10. Solimini NL,Luo J,Elledge SJ. Non-oncogene addiction and the stress pri-mRNA processing, altering target sites, or editing of phenotype of cancer cells. Cell 2007;130:986 – 8. miRNA sequences themselves (42). 11. Bjerkvig R,Tysnes BB,Aboody KS,Najbauer J,Terzis AJ. Opinion.

Mol Cancer Ther 2009;8(1). January 2009

The origin of the cancer stem cell: current controversies and new insights. 27. Gasparini G,Longo R,Sarmiento R,Morabito A. Inhibitors of cyclo- Nat Rev Cancer 2005;5:899 – 904. oxygenase 2: a new class of anticancer agents? Lancet Oncol 2003;4: 12. Alison MR,Murphy G,Leedham S. Stem cells and cancer: a deadly 605 – 15. mix. Cell Tissue Res 2008;331:109 – 24. 28. McGettigan P,Henry D. Cardiovascular risk and inhibition of cyclo- 13. Hahn WC,Counter CM,Lundberg AS,Beijersbergen RL,Brooks MW, oxygenase: a systematic review of the observational studies of selective Weinberg RA. Creation of human tumour cells with defined genetic and nonselective inhibitors of cyclooxygenase 2. JAMA 2006;296: elements. Nature 1999;400:464 – 8. 1633 – 44. 14. Funes JM,Quintero M,Henderson S,et al. Transformation of human 29. Fasig JH,Dupont WD,Olson SJ,Lafleur BJ,Cates JM. Steroid mesenchymal stem cells increases their dependency on oxidative hormone receptor and COX-2 expression in chordoma. Am J Clin Pathol phosphorylation for energy production. Proc Natl Acad Sci U S A 2007; 2007;128:375 – 81. 104:6223 – 8. 30. Endo M,Matsumura T,Yamaguchi T,et al. Cyclooxygenase-2 15. Henderson SR,Guiliano D,Presneau N,et al. A molecular map of overexpression associated with a poor prognosis in chondrosarcomas. mesenchymal tumors. Genome Biol 2005;6:R76. Hum Pathol 2006;37:471 – 6. 16. Suzuki R,Shimodaira H. Pvclust: an R package for assessing the 31. Huennekens FM. The methotrexate story: a paradigm for develop- uncertainty in hierarchical clustering. Bioinformatics 2006;22:1540 – 2. ment of cancer chemotherapeutic agents. Adv Enzyme Regul 1994;34: 397 – 419. 17. Zeller KI,Zhao X,Lee CW,et al. Global mapping of c-Myc binding sites and target gene networks in human B cells. Proc Natl Acad Sci U S A 32. Pirard B. Insight into the structural determinants for selective 2006;103:17834 – 9. inhibition of matrix metalloproteinases. Drug Discov Today 2007;12: 640 – 6. 18. Bataille F,Rogler G,Modes K,et al. Strong expression of methylthioadenosine phosphorylase (MTAP) in human colon carcinoma 33. Wymann MP,Schneiter R. Lipid signalling in disease. Nat Rev Mol Cell cells is regulated by TCF1/h-catenin. Lab Invest 2005;85:124 – 36. Biol 2008;9:162 – 76. 19. Martinez I,Wang J,Hobson KF,Ferris RL,Khan SA. Identifica- 34. Buchanan FG,McReynolds M,Couvillon A,et al. Requirement of tion of differentially expressed genes in HPV-positive and HPV- phospholipase D1 activity in H-RasV12-induced transformation. Proc Natl negative oropharyngeal squamous cell carcinomas. Eur J Cancer Acad Sci U S A 2005;102:1638 – 42. 2007;43:415 – 32. 35. Riley T,Sontag E,Chen P,Levine A. Transcriptional control of human 20. Arolas JL,Vendrell J,Aviles FX,Fricker LD. Metallocarboxypepti- p53-regulated genes. Nat Rev Mol Cell Biol 2008;9:402 – 12. dases: emerging drug targets in biomedicine. Curr Pharm Des 2007;13: 36. Crow YJ,Leitch A,Hayward BE,et al. Mutations in genes encoding 349 – 66. ribonuclease H2 subunits cause Aicardi-Goutieres syndrome and mimic 21. Schweiger A,Christensen IJ,Nielsen HJ,Sorensen S,Brunner N,Kos congenital viral brain infection. Nat Genet 2006;38:910 – 6. J. Serum cathepsin H as a potential prognostic marker in patients with 37. Qiu J,Qian Y,Frank P,Wintersberger U,Shen B. Saccharomyces colorectal cancer. Int J Biol Markers 2004;19:289 – 94. cerevisiae RNase H(35) functions in RNA primer removal during lagging- 22. Frank P,Braunshofer-Reiter C,Wintersberger U,Grimm R,Busen W. strand DNA synthesis,most efficiently in cooperation with Rad27 Cloning of the cDNA encoding the large subunit of human RNase HI,a nuclease. Mol Cell Biol 1999;19:8361 – 71. homologue of the prokaryotic RNase HII. Proc Natl Acad Sci U S A 1998; 38. Chapman EJ,Carrington JC. Specialization and evolution of endog- 95:12872 – 7. enous small RNA pathways. Nat Rev Genet 2007;8:884 – 96. Brindley DN. Lipid phosphate phosphatases and related proteins: 23. 39. Yin H,Lin H. An epigenetic activation role of Piwi and a Piwi- signaling functions in development,cell division,and cancer. J Cell associated piRNA in Drosophila melanogaster. Nature 2007;450: Biochem 2004;92:900 – 12. 304 – 8. 24. Tanyi JL,Morris AJ,Wolf JK,et al. The human lipid phosphate 40. Schoft VK,Schopoff S,Jantsch MF. Regulation of glutamate phosphatase-3 decreases the growth,survival,and tumorigenesis of receptor B pre-mRNA splicing by RNA editing. Nucleic Acids Res 2007; ovarian cancer cells: validation of the lysophosphatidic acid signaling 35:3723 – 32. cascade as a target for therapy in ovarian cancer. Cancer Res 2003;63: 1073 – 82. 41. Maas S,Patt S,Schrey M,Rich A. Underediting of glutamate receptor GluR-B mRNA in malignant gliomas. Proc Natl Acad Sci U S A 2001;98: 25. Morris KE,Schang LM,Brindley DN. Lipid phosphate phosphatase-2 14687 – 92. activity regulates S-phase entry of the cell cycle in Rat2 fibroblasts. J Biol Chem 2006;281:9297 – 306. 42. Paz N,Levanon EY,Amariglio N,et al. Altered adenosine-to-inosine RNA editing in human cancer. Genome Res 2007;17:1586 – 95. 26. Harris RE,Alshafie GA,Abou-Issa H,Seibert K. Chemoprevention of breast cancer in rats by celecoxib,a cyclooxygenase 2 inhibitor. Cancer 43. Benson JD,Chen YN,Cornell-Kennon SA,et al. Validating cancer Res 2000;60:2101 – 3. drug targets. Nature 2006;441:451 – 6.

Mol Cancer Ther 2009;8(1). January 2009

Downloaded from mct.aacrjournals.org on September 26, 2021. © 2009 American Association for Cancer Research. Genomics screen in transformed stem cells reveals RNASEH2A, PPAP2C, and ADARB1 as putative anticancer drug targets

James M. Flanagan, Juan M. Funes, Stephen Henderson, et al.

Mol Cancer Ther 2009;8:249-260.

Updated version Access the most recent version of this article at: http://mct.aacrjournals.org/content/8/1/249

Supplementary Access the most recent supplemental material at: Material http://mct.aacrjournals.org/content/suppl/2008/12/29/8.1.249.DC1

Cited articles This article cites 43 articles, 13 of which you can access for free at: http://mct.aacrjournals.org/content/8/1/249.full#ref-list-1

Citing articles This article has been cited by 5 HighWire-hosted articles. Access the articles at: http://mct.aacrjournals.org/content/8/1/249.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://mct.aacrjournals.org/content/8/1/249. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.