Comparative Oncogenomics Identifies PSMB4 and SHMT2 As Potential Cancer Driver Genes

Published OnlineFirst April 22, 2014; DOI: 10.1158/0008-5472.CAN-13-2683

Cancer Therapeutics, Targets, and Chemical Biology Research

Comparative Oncogenomics Identiﬁes PSMB4 and SHMT2 as Potential Cancer Driver Genes

Genee Y. Lee1, Peter M. Haverty3,LiLi3, Noelyn M. Kljavin2, Richard Bourgon3, James Lee1, Howard Stern4, Zora Modrusan2, Somasekar Seshagiri2, Zemin Zhang3, David Davis1, David Stokoe1, Jeffrey Settleman1, Frederic J. de Sauvage5, and Richard M. Neve1

Abstract Cancer genomes maintain a complex array of somatic alterations required for maintenance and progression of the disease, posing a challenge to identify driver genes among this genetic disorder. Toward this end, we mapped regions of recurrent amplification in a large collection (n ¼ 392) of primary human cancers and selected 620 genes whose expression is elevated in tumors. An RNAi loss-of-function screen targeting these genes across a panel of 32 cancer cell lines identified potential driver genes. Subsequent functional assays identified SHMT2, a key enzyme in the serine/glycine synthesis pathway, as necessary for tumor cell survival but insufficient for transformation. The 26S proteasomal subunit, PSMB4, was identified as the first proteasomal subunit with oncogenic properties promoting cancer cell survival and tumor growth in vivo. Elevated expression of SHMT2 and PSMB4 was found to be associated with poor prognosis in human cancer, supporting the development of molecular therapies targeting these genes or components of their pathways. Cancer Res; 74(11); 1–13. 2014 AACR.

Introduction suggests that individual amplicons contain multiple genes that Genome-wide documentation of tumor gene copy number, collectively alter processes important for tumor growth and gene expression, and somatic mutations has become standard survival (7) and that there may be equivalent cooperativity fi practice, driven by remarkable technologic advances over the between genes in separate but coampli ed genomic regions last decade (1, 2). By elucidating the "omic" landscape of tumor (8, 9). The distinct challenge is to separate passenger genes fi cells, we are only now realizing the vast complexity of the from driver genes (10). In the quest to nd new driver genes diseases that are classed as cancer (3). Assigning function to and, hence, potential therapeutic targets, a number of groups any gene in this morass of observed changes can be a challenge, have used either a reductionist approach by identifying a and years of collective study have begun to hone in on core region of copy-number gain/loss of clinical interest and testing functionalities regulating specific hallmarks of cancer (4). the functionality of each gene within that region (11, 12), or a However, if the long-term goal of diagnostic-driven, individu- holistic approach using RNA interference library screens in a fi alized therapy is to be realized, we need to begin to understand single cell line to nd driver genes (13, 14). how these genetic changes sustain tumor growth, progression, The primary goal of this study was to develop a genome-wide and survival. Oncogenes and tumor suppressors are the clas- oncogenomic screening strategy to identify candidate onco- sical drivers of oncogenesis, but it is still an open question as to genic driver genes. Our approach was to identify recurrent which genes are the drivers in any single tumor when as much regions of copy-number gain and increased expression across as 10% to 15% of the genome can be altered by recurrent copy- multiple solid tumor types, with the hypothesis that these number abnormalities (5). The task of finding "driver genes" is genes represent fundamental processes for sustaining solid further hampered by the fact that the majority of copy-number tumor growth and survival. We selected 620 genes from fi abnormalities are low-level gains, often encompassing large commonly ampli ed regions whose expression is elevated in regions of the genome, which contrasts with high-level, focal tumor versus normal and determined the effect on cell viability copy-number gains as exemplified by HER2 (6). Evidence also by RNAi targeting in 32 cancer cell lines. Here, we report the findings of this study that represents one of the most comprehensive functional oncogenomic screens to date. Authors' Affiliations: Departments of 1Discovery Oncology, 2Molecular Biology, 3Bioinformatics, 4Pathology, and 5Molecular Oncology, Genen- tech Inc., South San Francisco, California Materials and Methods Note: Supplementary data for this article are available at Cancer Research Cell lines Online (http://cancerres.aacrjournals.org/). See Supplementary Table S7 for full list of cell lines, cell line fi Corresponding Author: Richard M. Neve, Genentech, 1 DNA Way, South short tandem repeat (STR) pro les, growth conditions, and San Francisco, CA 94080. Phone: 650-467-4452; Fax: 650-742-5179; transfection conditions used. All cell lines are tested for E-mail: [email protected] mycoplasma, cross contamination, and genetically finger- doi: 10.1158/0008-5472.CAN-13-2683 printed when new stocks are generated to ensure quality and 2014 American Association for Cancer Research. confirm ancestry.

www.aacrjournals.org OF1

Lee et al.

Cell line fingerprinting: SNP fingerprinting. Single- RNA Extraction Kit (Qiagen). DNA was then hybridized to the nucleotide polymorphism (SNP) genotypes are performed each Agilent Human Genome CGH Microarray Kit 244A and geno- time new stocks are expanded for cryopreservation. Cell line mic copy number was calculated by comparing the intensity identity is verified by high-throughput SNP genotyping using ratio between individual tumors and the averaged normal DNA Fluidigm multiplexed assays. SNPs were selected on the basis of content. minor allele frequency and presence on commercial genotyping platforms. SNP profiles are compared with SNP calls from RNAi screen available internal and external data (when available) to deter- Cells were reverse lipid transfected with pools of four mine or confirm ancestry. In cases in which data are unavail- siRNAs; siGENOME siRNAs were used in the primary screen able or cell line ancestry is questionable, DNA or cell lines are and ON-TARGETplus siRNA was used in the secondary screen repurchased to perform profiling to confirm cell line ancestry. and all follow-up studies (Dharmacon). siRNA pools were used SNPs. The following SNPs were used: rs11746396, at 50 nmol/L and singles at 25 nmol/L. Pooled siRNAs targeting rs16928965, rs2172614, rs10050093, rs10828176, rs16888998, PLK1 and TOX transfection reagent were used as positive rs16999576, rs1912640, rs2355988, rs3125842, rs10018359, controls at 25 nmol/L. Lipids used were Dharmafects 1, 2, 3, rs10410468, rs10834627, rs11083145, rs11100847, rs11638893, and 4 (Dharmacon) and Lipofectamine RNAiMAX (Invitrogen). rs12537, rs1956898, rs2069492, rs10740186, rs12486048, See Supplementary Table S7 for transfection conditions for rs13032222, rs1635191, rs17174920, rs2590442, rs2714679, each cell line. siRNA–lipid complexes were formed for 1 hour, rs2928432, rs2999156, rs10461909, rs11180435, rs1784232, then cells were seeded on top of complexes. After 5 days, cell rs3783412, rs10885378, rs1726254, rs2391691, rs3739422, viability was assayed using CellTiter-Glo (Promega) and lumi- rs10108245, rs1425916, rs1325922, rs1709795, rs1934395, nescence readout was performed using Envision plate reader rs2280916, rs2563263, rs10755578, rs1529192, rs2927899, (PerkinElmer). Gene selection criteria: Genes were selected rs2848745, and rs10977980. from the screen by one of three criteria: (i) correlation of RNAi STR profiling. STR profiles are determined for each line response with total RNA expression (expression, value >0.3 using the Promega PowerPlex 16 System. This is performed Pearson correlation); (ii) correlation of RNAi response with once and compared with external STR profiles of cell lines copy-number [comparative genomic hybridization (CGH), val- (when available) to determine cell line ancestry. ue >0.3 Pearson correlation]; (iii) genes that affected cell Loci analyzed. Detection of sixteen loci (fifteen STR loci viability across the panel of cell lines regardless of copy number and Amelogenin for gender identification), including D3S1358, or gene expression (multiple). Genes were classified as a TH01, D21S11, D18S51, Penta E, D5S818, D13S317, D7S820, multiple hit if six or more cell lines had an RNAi response of D16S539, CSF1PO, Penta D, AMEL, vWA, D8S1179, and TPOX. greater than 0.65 (65% cell death).

Cloning Gene expression analysis Full-length open reading frame clones were obtained from Patient tissue samples with the appropriate Institutional Invitrogen and Genentech. C-terminus FLAG tags were added by Review Board approval and patient-informed consent were PCR then tagged proteins were subcloned into pLPCX retroviral obtained from commercial sources (Supplementary Table S1). expression vector (Clontech). NIH/3T3 and MCF10A stable cell The human tissue samples used in the study were deidentified lines were generated by infection with retroviral particles expres- (double-coded) before their use, and, hence, the study using sing FLAG-tagged constructs and selection with puromycin. See these samples is not considered human subject research under Supplementary Table S7 for a full list of clones and primers used. the US Department of Human and Health Services regulations and related guidance (45 CFR Part 46). All tumor tissues were Gene selection for RNAi screen subject to pathology review. Tumor DNA and RNA were An independent set of tumor expression data (source, Gene- extracted using the Qiagen AllPrep DNA/RNA Kit (Qiagen). Logic, Inc.; ref. 15) was used to identify genes upregulated in Total RNA was harvested from cells using the RNeasy Kit tumor versus normal samples (breast, 24 normal, 91 tumors; with on-column DNase digestion (Qiagen). RNA for validation lung, 80 normal, 105 tumors; ovary, 101 normal, 82 tumors; and of siRNA function was acquired 3 days after transfection unless prostate, 27 normal, 70 tumors). Percentile analysis for differ- otherwise noted. RNA was quantified using a Nanodrop spec- ential gene expression (16), a statistical tool that compares both trophotometer, amplified with TaqMan One-Step RT-PCR the magnitude and the variability between two sample groups, Master Mix, assayed with TaqMan gene expression assays, was used to identify differential gene expression between normal and then analyzed using a 7900HT Fast Real-Time PCR System and cancer samples in the GeneLogic (15) database. Genes found (Applied Biosystems). See Supplementary Table S7 for full list within an amplicon that showed a 1.5-fold increase in cancer of gene expression assays used. Normalization control assays samples versus normal were selected for the screen. were HPRT1 (human) and ACTB (mouse).

Tumor samples, DNA preparation, and copy-number Immunoblotting arrays Protein was harvested from cells with RIPA buffer, passed Tumor samples had a tumor percentage of >80% as assessed through a syringe, and cleared by centrifugation. Protein for by an expert pathologist. DNA was extracted from frozen validation of siRNA function was acquired 3 days after trans- tissues and cell lines by a standard protocol using the DNA/ fection unless otherwise noted. Protein was quantiﬁed using

OF2 Cancer Res; 74(11) June 1, 2014 Cancer Research

Oncogenomic Target Discovery

www.aacrjournals.org Cancer Res; 74(11) June 1, 2014 OF3

Lee et al.

Table 1. Summary of GISTIC analysis of copy-number aberrations for each tumor type

Gain Loss

Tumor type No. of peaks Genes OIC Genes No. of peaks Genes UIC Genes Breast (HER2) 27 250 69 25 327 193 Breast (TN) 22 372 103 21 375 237 Breast (HR) 20 308 85 18 180 131 Ovarian 26 486 130 25 330 233 Melanoma 36 1,019 197 32 119 65 NSCLC 26 636 145 24 274 187

Total 3,071 729 1,605 1,046

NOTE: Number of signiﬁcant peaks (as determined by GISTIC), total gene, and miRNA count within those regions are listed for both copy-number gain and loss. Genes identiﬁed as overexpressed in cancer (OIC) and underexpressed in cancer (UIC) compared with normal controls are also listed. Abbreviations: HER2, HER2-positive; HR, hormone receptor–positive; TN, triple-negative.

bicinchoninic acid protein assay (Pierce). Protein was sepa- Cell-cycle analysis rated on 4% to 12% Bis-Tris gels (Invitrogen), transferred to For steady-state assay, NIH/3T3 cells were cultured for 48 nitrocellulose membranes, blocked with 5% bovine serum hours and labeled with 10 mmol/L bromodeoxyuridine albumin or milk in TBST for 30 minutes, then blotted with (BrdUrd) for 30 minutes. Cells were then detached and stained primary antibody overnight at 4C. See Supplementary Table using the BrdUrd Flow Kit (BD Biosciences). Flow cytometry S7 for list of primary antibodies used. Membranes were then data were acquired with FACSCalibur (BD Biosciences) and washed and incubated with appropriate horseradish peroxi- analyzed with FlowJo software. For cell synchronization assay dase–conjugated secondary antibodies for 1 hour, washed and (double thymidine block), NIH/3T3 cells were cultured for detected with SuperSignal West Femto Chemiluminescent 48 hours then blocked with 2 mmol/L thymidine (Sigma) Substrate (Pierce). Luminescence signal was acquired with for 15 hours. Cells were released into full media for 10 hours FluorChem Q (Alpha Innotech). and reblocked with 2 mmol/L thymidine for 15 hours. Cells were then released and pulsed at indicated time points for 3D assays 30 minutes with 10 mmol/L BrdUrd in the presence or absence Matrigel drip cultures. MCF10A cells were cultured in of 100 ng/mL nocodazole (Sigma). Cells were then analyzed for three-dimensional (3D) culture as previously described (17). Of cell-cycle status as described above. note, 125,000 cells were seeded in a 6-well plate on top of a layer of polymerized Matrigel (BD Biosciences), overlaid with a Gene ontology analysis solution of 5% Matrigel, and cultured for 21 days. Gene sets were classified by Gene Ontology using the Soft agar assay. NIH/3T3 cells were cultured in soft agar Database for Annotation, Visualization and Integrated Discov- colony formation assay with a base layer of 1.2% agarose and an ery (DAVID) pathway analysis online tool and standard para- assay layer of 0.4% agarose (BD Biosciences). Of note, 4,000 cells meters (18). were seeded per well in a 12-well plate. Plates were scanned and analyzed after 21 days using Gelcount scanner and software Cell-based assays (Oxford Optronics). Apoptosis was assayed using Caspase-Glo 3/7 Assay (Pro- mega). Proteasome activity was assayed using Proteasome-Glo Xenograft studies Chymotrypsin-Like Cell-Based Assays (Promega). One million NIH/3T3 cells were implanted subcutaneously in the right flank of NCr nude mice (Taconic), with 5 mice per Kaplan–Meier analysis group. Tumor dimensions were measured by caliper. At end of Data were generated with KMPlotter (http://kmplot.com/ assay, tumors were harvested and flash-frozen in liquid nitro- analysis/) using a median cutoff for gene expression. Pro- gen. RNA was extracted from tumors using a TissueLyser II and besets used for analysis: 202244_at (PSMB4); 214095_at RNeasy extraction (Qiagen). (SHMT2).

Figure 1. Recurrent gene copy-number changes in breast, ovarian, lung, melanoma, and prostate primary tumors. A, frequencies of gene copy gain (red) and loss (green) are plotted as a function of genomic location for each tumor types. Positive values indicate frequencies of samples showing copy-number increases [log2 (copy number) > 0.3], and negative values indicate frequencies of samples showing copy-number decreases [log2 (copy number) < 20.3]. Vertical solid lines indicate boundaries between chromosomes. Vertical dotted lines indicate positions of centromeres. Sample number is indicatedtothe right of each graph. B, analysis of signiﬁcant regions of copy-number gain by GISTIC. Q-scores for gene copy-number gains are overlaid on a single plot for breast, ovarian, lung, and melanoma samples. C, averaged Q values across all four tumor types identify most signiﬁcant regions for the entire dataset.

OF4 Cancer Res; 74(11) June 1, 2014 Cancer Research

Oncogenomic Target Discovery

Results (GISTIC) algorithm (20) was applied to calculate significant DNA copy-number analysis of human tumors regions of copy-number change and identify the genes within Anonymous tumor specimens were collected and appraised each region. Table 1 contains the summary of this analysis; for tumor content and pathology by expert pathologists. The gene lists are supplied in Supplementary Table S2. collection consisted of 161 breast tumors (HER2-positive, n ¼ 53; hormone receptor–positive, n ¼ 54; and triple-negative, n ¼ Common regions of gene amplification and deletion 54), 51 ovarian tumors (serous, n ¼ 37; papillary serous n ¼ 14), across tumor types 52 lung tumors (adenocarcinoma, n ¼ 5; squamous cell car- Our data provided a direct comparison among the gen- cinoma, n ¼ 47), 51 melanomas, and 57 prostate tumors omes of five solid tumor types analyzed on a single platform (Supplementary Table S1). DNA copy number was measured and it was evident that the frequency of genomic aberrations using Agilent arrays (see Materials and Methods). Gain and found in breast, ovarian, lung, and melanoma tumors had Loss Analysis of DNA (R package from Bioconductor; ref. 19) strikingly similar architecture in several chromosomal was applied to infer the segmented copy numbers for each regions (Fig. 1). In contrast, the prostate dataset shared few sample normalized to two copies. Figure 1A compares the common features with the other tumor types, showing frequency of recurrent abnormalities across all five indications. greater frequency of copy-number loss overall and few The genomic identification of significant targets in cancer regions of significant copy-number gain. To select genes for

A B 150 Breast, Melanoma, Lung, Ovary, and Prostate Genes 100 Amplified regions overexpressed in 50 (GISTIC) cancer vs. normal 0

620 Genes −50 Primary siGenome RNAi Screen − 32 cell lines 100 Percentage of inhibition Percentage −150 620 Genes Expression Copy number Multiple correlation correlation hits 150 Copy number Expression Multiple

100

105 Genes OTP RNAi screen 50 32 cell lines

Secondary Expression Copy number Multiple correlation correlation hits of inhibition Percentage −50 105 Genes

C 5 D E 150 C3ORF62 40,000 ERBB2 EIF4A3 4 30,000 3 100 20,000 2 10,000 50 Copy number Copy 1

mRNA expression 0 0 −20 0 20 40 60 80 −40 −20 0 20 40 60 80 of inhibition Percentage 0 Percentage of inhibition Percentage of inhibition Cell line

Figure 2. RNAi screen design and results. A, schematic outline of the screening approach used in this study. B, summary of all data for primary RNAi screen (top) and secondary RNAi screen (bottom). Details of genes and cell lines can be found in Supplementary Tables S3 and S7, respectively. For each graph, individual targeted genes and controls are plotted on the horizontal axis, and cell growth is represented as the percentage of control plotted on the vertical axis. There are 32 data points for each gene or control, representing a data point for each cell line in the screen. Bottom, genes are ordered by the criteria by which the genes were selected for the secondary screen (indicated above the data points). Examples of how genes were selected for further analysis are shown for individual genes; C, correlation with copy number, C3ORF62 (Pearson correlation ¼ 0.53); D, correlation with expression, ERBB2 (Pearson correlation ¼ 0.52); and F, multiple hit, EIF4A3 [genes were classiﬁed as a multiple hit if six or more cell lines had an RNAi response of greater than 0.65 (65% cell death), indicated by dotted line], respectively. See Materials and Methods for detailed selection criteria.

www.aacrjournals.org Cancer Res; 74(11) June 1, 2014 OF5

Lee et al.

Table 2. Genes passing selection criteria after two rounds of RNAi screens

Gene Chr Cytoband Description Analysis ELOVL1 1 1p34.2 CGI-88 PROTEIN CGH C30RF62 3 3p21.31 CHROMOSOME 3 OPEN READING FRAME 62 CGH ENO3 17 17pter-p11 ENOLASE 1, (ALPHA) CGH ERBB2 17 17q11.2-q12j17q21 V-ERB-B2 ERYTHROBLASTIC LEUKEMIA VIRAL EXPRESSION ONCOGENE HOMOLOG 2 ACTN4 19 19q13 ACTININ, ALPHA 4 EXPRESSION MYC 8 8q24.21 V-MYC MYELOCYTOMATOSIS VIRAL ONCOGENE EXPRESSION HOMOLOG (AVIAN) EMILIN1 2 2p23.3-p23.2 ELASTIN MICROFIBRIL INTERFACER 1 EXPRESSION WARS 14 14q32.31 INTERFERON-INDUCED PROTEIN 53 EXPRESSION EIF2S2 2 20pter-q12 EUKARYOTIC TRANSLATION INITIATION FACTOR 2, EXPRESSION SUBUNIT 2 BETA, 38KDA BOP1 8 8q24.3 BLOCK OF PROLIFERATION 1 EXPRESSION SHMT2 12 12q12-q14 SERINE HYDROXYMETHYLTRANSFERASE 2 EXPRESSION (MITOCHONDRIAL) EIF6 20 20q12 EUKARYOTIC TRAHSLATION INITIATION FACTOR 6 EXPRESSION EDEM2 20 20q11.22 CHROMOSOME 20 OPEN READING FRAME 31 EXPRESSION ALDOA 16 16p11.2 ALDOLASE A, FRUCTOSE-BISPHOSPHATE MULTIPLE EIF3S8 16 16p11.2 EUKARYOTIC TRANSLATION INITIATION FACTOR 3. MULTIPLE SUBUNIT 8, 110KDA EIF4A3 17 17q25.3 DEAD (ASP-GLU-ALA-ASP) BOX POLYPEPTIDE 48 MULTIPLE PHF5A 22 22q13.2 PHD FINGER PROTEIN 5A MULTIPLE PSMA6 14 14q13 PROTEASOME (PROSOME, MACROPAIN) SUBUNIT, MULTIPLE ALPHA TYPE, 6 PSMB4 1 1q21 PROTEASOME (PROSOME, MACROPAIN) SUBUNIT, MULTIPLE BETA TYPE, 4 PSMD11 17 17q11.2 PROTEASOME (PROSOME, MACROPAIN) MULTIPLE 26S SUBUNIT, NON-ATPASE, 11 PSMD2 3 3q27.1 PROTEASOME (PROSOME, MACROPAIN) 26S MULTIPLE SUBUNIT. NON-ATPASE, 2 PSMD3 17 17q12-q21.1 PROTEASOME (PROSOME, MACROPAIN) MULTIPLE 26S SUBUNIT, NON-ATPASE, 3 PSMD8 19 19q13.2 PROTEASOME (PROSOME, MACROPAIN) MULTIPLE 26S SUBUNIT. NON-ATPASE, 8 SNRPD2 19 19q13.2 SMALL NUCLEAR RIBONUCLEOPROTEIN D2 MULTIPLE POLYPEPTIDE 16.5KDA THOC4 17 17q25.3 THO COMPLEX 4 MULTIPLE

NOTE: Analysis indicates the criteria by which the genes were selected; correlation of phenotype with copy number (CGH), correlation of phenotype with mRNA level (expression), and effect on multiple lines (multiple). Genes highlighted in bold were selected for follow-up functional studies. Cell line copy number and gene mRNA baseline levels are shown in Supplementary Table S5.

an RNAi-based screen to identify oncogenic drivers, we analysis (18) of these genes showed enrichment for several focused on genomic regions that exhibited recurrent ampli- cancer-related functions such as regulation of DNA, protein, fication across the breast, ovarian, lung, and melanoma and glucose metabolism (Supplementary Table S4). datasets. Figure 1B shows an overlay of the GISTIC Q-scores for these four tumor types and Fig. 1C shows peaks derived Loss-of-function RNAi screen to identify oncogenic from a cross-analysis of these significant peaks. Genes driver genes within these 86 peaks (Supplementary Table S2) were fil- To identify amplified and overexpressed genes required for tered by gene expression levels (see Materials and Methods) initiation and/or maintenance of tumor cell proliferation, we to select genes within amplicons that are overexpressed in performed an RNAi loss-of-function screen targeting 620 genes cancer compared with normal tissues, resulting in a final list across a panel of 32 tumor cell lines representing the four solid of 620 genes (Supplementary Table S3). Gene Ontology tumor types used to generate the gene list as outlined in Fig. 2A.

OF6 Cancer Res; 74(11) June 1, 2014 Cancer Research

Oncogenomic Target Discovery

A B ACTN4 ALDOA ENO3 SHMT2 WARS EMILIN1 EDEM2 ELOVL1 Parental EIF4A3 PSMB4 PSMA6 EIF2S2 EIF6 260 kD

160 kD 110 kD 80 kD

60 kD FLAG Parental PSMB4 SHMT2 Hras.G12V 50 kD

600 250 200 30 kD 400 150 20 kD 100 200 50 Total colonies 0 0 β-Actin 42 kD Average diameter ( μ m)

Parental PSMB4 SHMT2 Parental PSMB4 SHMT2

HRAS.G12V HRAS.G12V C

pLPCX PSMB4 SHMT2 HRAS. G12V

D 2,500 Parental PSMB4

) 2,000 3 SHMT2 HRAS. G12V 1,500

1,000

500 Tumor volume (mm volume Tumor

0 0 10 20 30 Day

Figure 3. Effect of PSMB4 and SHMT2 on anchorage-independent growth and cell cycle. A, Western blot analysis of FLAG-tagged target genes stably expressed in NIH/3T3 cells. Dotted line, lanes placed next to each other from different regions of the same gel. B, soft agar assays to assess effect of target genes on anchorage-independent growth. Total number of colonies and average colony diameter are shown in bar graphs below; scale bar, 1 mm. C, effect of target gene expression on MCF10A acinar morphogenesis. Selected target genes were expressed in the nontumorigenic breast epithelial cell line to assay gross acinar morphogenesis. Oncogenic HRAS.G12V is shown as a positive control at a 4-day time point; scale bar, 100 mm. D, effect of PSMB4 and SHMT2 on tumor growth in vivo. Nude mice bearing NIH/3T3 cell lines stably expressing PSMB4, SHMT2, and HRAS.G12V (positive control) were monitored for xenograft growth. The mean tumor volume (þ SEM) for each group (n ¼ 5) is shown.

Each of the 86 ampliﬁed regions identiﬁed by GISTIC was response with copy number (CGH; Fig. 2C), (ii) correlation of represented by at least three cell lines. Two rounds of RNAi RNAi response with total RNA level (expression; Fig. 2D), and screening were conducted, the results of which are summa- (iii) genes that affect multiple lines (multiple; Fig. 2E) inde- rized in Fig. 2B; Supplementary Table S3. Three independent pendent of copy number or expression correlations (Supple- criteria were used to select genes; (i) correlation of RNAi mentary Table S3). Using these selection criteria, 105 genes

www.aacrjournals.org Cancer Res; 74(11) June 1, 2014 OF7

Lee et al.

A 0.0006 <0.0001 <0.0001 <0.0001 <0.0001 0.004 <0.0001 <0.0001 0.02 0.07 12 12 10 10 8 8 6 6

RPKM (SHMT2) 4 RPKM (PSMB4) 2 4 2

Log 2 Log 2

0 0 N C N C N C N C N C N C N C N C N C N C Breast Lung Ovary Skin Prostate Breast Lung Ovary Skin Prostate

B Figure 4. Expression of SHMT2 and <0.0001 <0.0001 0.0002 <0.0001 <0.0001 0.19 12 12 PSMB4 in normal and cancer 10 10 tissues. A, expression of SHMT2 is shown for breast, lung, ovary, skin, 8 8 and prostate tissues from 6 6 GeneLogic (left) and TCGA (right) RPKM (PSMB4) RPKM (SHMT2) 4 2 4 data sources. B, expression of 2 PSMB4 2 Log 2 is shown for breast, lung, Log ovary, skin, and prostate tissues 0 0 from GeneLogic (left) and TCGA N C N C N C N C N C N C N C N C N C N C (right) data sources; N, normal; C, Breast Lung Ovary Skin Prostate Breast Lung Ovary Skin Prostate cancer; numbers above normal/ cancer pairings represent P values C of a nonparametric t test (Mann– HR, 1.53 (1.35–1.73) HR, 1.56 (1.37–1.77) Whitney test). C, Kaplan–Meier Log-rank P = 5.4e–12 Log-rank P = 3.9e–11 survival plots showing the prognostic effect on RFS in breast cancer for SHMT2 (left) and PSMB4 (right). D, Kaplan–Meier survival plots showing the Probability prognostic effect on OS in lung cancer for SHMT2 (left) and Low SHMT2 Low PSMB4 PSMB4 (right). Data were High SHMT2 High PSMB4 generated with KMPlotter (25) 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 using a median cutoff for gene 0 5 10 15 20 0 5 10 15 20 expression. Time (y) Time (y) D HR, 1.4 (1.2–1.63) HR, 0.93 (0.8–1.08) Log-rank P = 1.4e–05 Log-rank P = 0.33 Probability

Low SHMT2 Low PSMB4 High SHMT2 High PSMB4 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 0 50 100 150 200 0 50 100 150 200 Time (mo) Time (mo)

passed and moved to second round screening, of which 25 were copy number. Overall correlation of phenotype with expression validated (Table 2). This list was significantly enriched for levels yielded more hits than that with copy number, perhaps genes with functions associated with the proteasome, spliceo- due to the noted lack of linearity between copy number/mRNA some, DNA replication, cell cycle, and metabolism (Supple- levels/protein levels for the majority of genes (21, 22). mentary Table S4). ERBB2 and MYC were among the genes meeting the validation criteria, confirming the screen was able Experimental and functional validation of RNAi screen to identify amplified/overexpressed driver oncogenes, candidates although these were both selected on the basis of a correlation Fifteen of the 25 hits from the RNAi screen were selected between phenotype and total RNA levels, not phenotype and for further validation based on function and/or potential

OF8 Cancer Res; 74(11) June 1, 2014 Cancer Research

Oncogenomic Target Discovery

www.aacrjournals.org Cancer Res; 74(11) June 1, 2014 OF9

Lee et al.

druggability (Table 2; see Materials and Methods). Protein Relevance of SHMT2 and PSMB4 in human cancer levels of these genes were assessed in the cell line panel by SHMT2 and PSMB4 were primarily selected for our screen as Western blot analysis (Supplementary Fig. S1). Of the eight amplified in lung and ovarian cancer, respectively. To inves- candidate genes for which specific antibodies were found, only tigate the broader clinical relevance, we compared expression three, ACTN4, ERBB2, and EMILIN1, showed correlation of SHMT2 and PSMB4 in cancer and normal tissue in two between mRNA (detected by Affymetrix expression arrays or published datasets (15, 24). For the five primary indications TaqMan) and protein (Supplementary Table S5). RNAi-medi- surveyed in this study, SHMT2 expression is significantly ated knockdown was confirmed for 13 of the 15 target genes. increased in cancer samples compared with normal tissue in Levels of two genes, ENO3 and WARS, were unaffected by RNAi all five indications in both datasets (Fig. 4A, left). Normal and dropped from further evaluation (Supplementary Fig. S2). samples were not available for ovary and skin in The Cancer To assess the transforming activity of the remaining 13 genes, Genome Atlas (TCGA) RNAseq dataset (24); however, Agilent each was cloned and expressed in NIH/3T3 cells to perform expression array data from the same group contained three soft agar colony formation assays (Fig. 3A). Of these, ELOVL1 normal ovary samples and showed an increase in expression in could not be expressed, PSMA6 expression was only detected cancer (P ¼ 0.0226; Supplementary Table S6). In comparison, by TaqMan, and all other genes expressed at or above phys- the cytosolic form of serine hydroxymethyltransferase, SHMT1, iologic levels. Two of the 13 genes (PSMB4 and SHMT2) was found overexpressed in only four tumor types, and showed promoted colony formation in addition to the HRAS.G12V- decreased expression in several others. PSMB4 showed signif- positive control (Fig. 3B). Overall proliferation rates of PSMB4- icant upregulation in breast, lung, ovarian, and skin tumors and SHMT2-expressing cells were higher than parental cells (Fig. 4A and B, right). Overexpression of the three proteasome (Supplementary Fig. S3A). A more detailed analysis of cell-cycle b-ring catalytic subunits was also observed (Supplementary effects of these genes in both steady-state and after release Table S6). PSMB5 overexpression followed a similar pattern to from synchronization revealed that the expression of PSMB4, whereas levels of PSMB6 and 7 were unrelated and less PSMB4, but not SHMT2, resulted in aberrations in the frequently elevated in cancer. Increases in expression of cell-cycle profile. In the steady-state condition, the cell-cycle SHMT2 and PSMB4 were observed in a wide range of other profile of SHMT2-expressing cells was indistinguishable tumor types supporting the involvement of these genes in the from that of control cells, whereas PSMB4-expressing cells etiology of multiple cancers (Supplementary Table S6). fi – – spent a signi cantlylowerproportionoftimeinG0 G1 Kaplan Meier analysis (25) indicated high expression of phase in favor of more time in both S and G2–M phases SHMT2 is associated with worse relapse-free survival (RFS) (Supplementary Fig. S3B). Synchronization of the cells using and distant metastasis-free survival and overall survival (OS) in a double thymidine block and release into media with or breast cancer, and increased time to first progression and without nocodozole revealed that PSMB4-expressing cells decreased OS in lung cancer (Fig. 4C and D, left; Supplementary have a similar cell-cycle profile as HRAS.G12V-expressing Table S6). Increased PSMB4 expression was associated with cells, with accelerated transition through the G2–M phase worse RFS in breast cancer and decreased OS in ovarian cancer and bypass of the G2 DNA damage and mitotic spindle (Fig. 4C, right; Supplementary Table S6). checkpoints (Supplementary Fig. S3C and S3D; ref. 23). PSMB4 and SHMT2 were expressed in MCF10A cells to Suppression of PSMB4 inhibits proteasomal activity and assess effects on epithelial cell morphogenesis in 3D cul- processing of b-ring catalytic subunits tures. These cells exhibited no morphologic differences Inhibition of proteasome activity is an established antican- compared with control (Fig. 3C), in contrast with the highly cer therapeutic strategy (26, 27). Although PSMB4 does not invasive phenotype of the HRAS.G12V control, suggesting contain intrinsic hydrolytic activity, it is rate-limiting for 20S these genes have minimal effect on polarity, migration, and proteasome assembly (28). Cell viability and protease activity invasion. Knockdown efficiency for the individual RNAis was assessed after PSMB4 knockdown revealing a time-depen- within pools was confirmed by TaqMan and Western blot dent decrease in both cell number and protease activity (Fig. analysis for PSMB4 and SHMT2 (Supplementary Fig. S4A). In 5A, top). Cells treated with the proteasome inhibitor, MG132, vivo assessment of oncogenic properties of PSMB4 and showed a similar effect in the cells tested (Fig. 5A, bottom). SHMT2demonstratedthatPSMB4butnotSHMT2expres- Expression of PSMB5,-6 and -7 mRNA increased over time in sion was sufficienttopromoteNIH/3T3xenografttumor response to PSMB4 knockdown and MG132 treatment (Fig. 5B) growth (Fig. 3D). Expression of each transgene in the tumors and an accumulation of polyubiquinated (K48-linked) protein was confirmed using human-specificTaqManprobeson products were observed (Fig. 5C). This did not result in an mRNA extracted from the tumors (Supplementary Fig. S4B). obvious increase of the total protein levels of PSMB5, -6 and -7,

Figure 5. Effects of PSMB4 loss on proteasome activity and protein expression of the proteasome catalytic b subunits. A, siRNA knockdown of PSMB4 in NCI-H1299 and ES-2 cells over a 72-hour time course measuring cell viability (top left) and chymotrypsin-like proteasome activity normalized to cell number (top right). Effects of the proteasome inhibitor, MG132, on cell viability and chymotrypsin-like proteasome activity are shown at the bottom.B, measurement of RNA levels of the proteasome b-ring subunits PSMB4, PSMB5, PSMB6, and PSMB7 in cells treated with PSMB4 RNAi for the stated times or 1 mmol/L MG132 for 24 hours. C, protein levels (Western blots) of the proteasome b-ring subunits PSMB4, PSMB5, PSMB6, and PSMB7 in cell lines transfected with PSMB4 RNAi for the stated times or 1 mmol/L MG132 for 24 hours. D, comparison of cell line sensitivity with 10 nmol/L bortezomib and PSMB4 RNAi in a panel of cell lines. RNAi knockdown of PSMB4 was conﬁrmed in Supplementary Fig. S5B.

OF10 Cancer Res; 74(11) June 1, 2014 Cancer Research

Oncogenomic Target Discovery

but a decrease in the processed (lower molecular weight, glucose and amino acid metabolism has been elucidated, higher mobility) forms with a concomitant accumulation of indicating that serine is an allosteric regulator of pyruvate the precursor forms of these subunits (Fig. 5C) was observed. kinase M2 (35, 36). Our data show that cell lines with These data indicate loss of PSMB4 disrupts the formation of high SHMT2 require the gene for survival, but exogenous the 20S proteasome, because assembly of 20S half-mers expression of the gene does not alter the acinar development requires PSMB4 and propeptide removal precedes maturation of MCF10A cells grown in Matrigel (in contrast with of the 20S proteasome (29). This effect seems to be specificto PHGDH; 32), whereas it is sufficienttopromoteanchor- loss of PSMB4, because knockdown of PSMA6 had little or no age-independent growth in colony formation assays but not effect on the protein levels of the core b-ring catalytic subunits to drive growth of tumors in vivo.Furthermore,SHMT2is (Supplementary Fig. S5A). To compare the effects of PSMB4 found overexpressed in many tumors and is associated with loss with proteasome inhibitor activity, we selected five cell worse clinical outcome in breast and lung cancer. Therefore, lines with a range of sensitivities to bortezomib (PS-341, our findings contribute to an evolving paradigm that serine/ Velcade) and compared the effect on cell viability of bortezo- glycine metabolism is of critical importance to the devel- mib against PSMB4 knockdown. Although one line (NCI- opment and or maintenance of cancer. H1838) was exceptionally resistant to both bortezomib treat- Our studies also provided the first evidence for the require- ment and loss of PSMB4, other lines tested showed variable ment of a proteasomal subunit, PSMB4, as a potential driver sensitivities to both treatments with little correlation between oncogene in multiple tumor types. PSMB4 and five other the two results (Fig. 5D). A similar result was observed with 20 proteasome subunits were found to be essential for the survival of the lines used in the RNAi screen (Supplementary Fig. S5C). of a broad range of tumor types (breast, lung, skin, and ovary; Together, these data support a central role for PSMB4 in Supplementary Table S3). PSMB4 has also been identified as a maintaining a functional 20S proteasome and suggest that gene required for the survival of human glioblastoma cells (37), targeting PSMB4 may offer an alternative therapeutic option to although there is little evidence for elevated levels in brain existing proteasome inhibitors. cancer versus normal brain in the datasets we studied. Of the two proteasome subunits that underwent functional screening, PSMB4 was the only one that promoted both anchorage- Discussion independent growth and tumorigenesis and, to the authors' In this report, we used genomics-based selection and a loss- knowledge, is the first proteasomal subunit shown to possess of-function screen to identify genes required for the survival oncogenic properties. and growth of human cancer cells. Two rounds of RNAi screens The two FDA-approved proteasome inhibitors, bortezomib reduced our initial pool of 620 candidates to 25 genes, 15 of and carfilzomib, both target the core proteolytic subunits which were put through a stringent series of follow-up screens PSMB5, PSMB6, and PSMB7 (38). The noncatalytic subunit to assess oncogenic potential. Known oncogenes ERBB2 and PSMB4 represents a novel potential target as it plays an MYC were identified using this approach, confirming the important role in regulating the assembly of the proteasome validity of this screen. A number of shared biologic functions (28, 29), and, thus, by inhibiting its function and proteasome were enriched in the 25 candidate genes; SHMT2 and ALDOA assembly one could potentially prevent the catalytic activity of have glycolytic/metabolic functions; PHF5A, THOC4, and all three proteolytic subunits. There is precedence that such an SNRPD2 are involved in mRNA processing; and 40% (6/15) of approach is feasible, as CRBN has been shown to directly the genes are components of the 26S proteasome (four from interact with PSMB4, negatively regulating proteasome activity the 19S regulatory particle and one each from the 20S a- and in vivo, possibly through interfering with the assembly of the b-rings). We used stringent criteria to identify potential tar- 20S proteasome (39). Because proteasomal assembly requires gets, and, in addition to SHMT2 and PSMB4, our data revealed the 15-amino acid C-terminal tail of PSMB4 to intercalate into a number of genes on which cancer cells from multiple tis- a groove between the PSMB6 and PSMB7 subunits, interfering sues rely for continued proliferation and that warrant further with this interaction offers a potential therapeutic opportunity investigation (Table 2). (40). Despite PSMB4 being amplified and overexpressed in On the basis of our results, we established that the mito- cancer (a prerequisite for being screened in this project), tumor chondrial serine hydroxymethyltransferase gene (SHMT2) is cell sensitivity to loss of PSMB4 did not correlate with gene required for cancer cell survival, but is not sufficient to copy number or gene expression. Evidence suggests that promote tumorigenesis. SHMT enzymes catalyze the conver- regulation of proteins in the proteasomal complex seems to sion of serine to glycine by catalyzing the transfer of the occur at the protein level, rather than at the copy number/ b-carbon of serine to tetrahydrofolate (THF) generating mRNA level (21). Paradoxically, in vitro and patient-derived 5,10-methylene-THF and glycine (30, 31). Phosphoglycerate data suggest that elevated expression of certain proteasome dehydrogenase (PHGDH), another enzyme in the pathway of subunits may contribute to bortezomib resistance (41, 42). generating glycine from glucose, was recently shown to be Although it is plausible that increased copy number and amplified in melanoma (32) and breast cancer (33) and the next expression of PSMB4 is required in certain tumors to maintain enzyme in this pathway, phosphoserine aminotransferase, is PSMB4 transcript and/or protein levels, the exact mechanism overexpressed in colorectal cancers, associated with chemore- of PSMB4 deregulation remains to be evaluated and it is likely a sistance (34), and required for serine pathway flux in breast complex mechanism tied to the intrinsic turnover rate of the cancer (33). More recently, a delicate interdependency between proteasome.

www.aacrjournals.org Cancer Res; 74(11) June 1, 2014 OF11

Lee et al.

Overall, our approach has led to the discovery of potential Disclosure of Potential Conflicts of Interest therapeutic targets, but has also underscored the complexity R.M. Neve is a scientist, has received commercial research support, and HER2 has ownership interest (including patents) from Genentech. S. Seshagiri is a of cancer biology. is the paradigm for a gene driven by scientist in Roche/Genentech. F.J. de Sauvage has ownership interest (includ- amplification of its locus, generating high RNA and protein ing patents) in Roche. No potential conflicts of interest were disclosed by the expression, resulting in oncogenic activity. Although our other authors. screen was predicated on this hypothesis, very few of the targets were selected on the basis of copy number, highlighting Authors' Contributions Conception and design: G.Y. Lee, H. Stern, D. Stokoe, F.J. de Sauvage, R.M. Neve the biologic complexity evident in a cancer cell. A number of Development of methodology: G.Y. Lee, L. Li, J. Lee, D. Davis, D. Stokoe, studies have indicated that there is ever decreasing linearity in R.M. Neve Acquisition of data (provided animals, acquired and managed patients, the relationship of gene copy number to RNA to protein (43) provided facilities, etc.): J. Lee, Z. Modrusan, S. Seshagiri, R.M. Neve due to a number of regulatory mechanisms such as methyl- Analysis and interpretation of data (e.g., statistical analysis, biostatistics, ation, RNA editing and stability, and protein translation and computational analysis): G.Y. Lee, P.M. Haverty, L. Li, N.M. Kljavin, R. Bourgon, Z. Zhang, J. Settleman, R.M. Neve ubiquitination. Amplicons may also contain more than one Writing, review, and/or revision of the manuscript: G.Y. Lee, L. Li, H. Stern, potential driver gene that cooperates in tumor etiology and D. Stokoe, J. Settleman, R.M. Neve maintenance (44, 45), whereas many genes within any given Administrative, technical, or material support (i.e., reporting or orga- nizing data, constructing databases): G.Y. Lee, L. Li, R.M. Neve amplicon are likely to be passengers and do not contribute to Study supervision: D. Davis, J. Settleman, R.M. Neve the survival/homeostasis for a given cancer cell. Therefore, expanding the strategy used in this study by selecting genes Acknowledgments based on multiple biologic parameters (copy number, muta- The authors thank Mamie Yu and Suresh Selvaraj for maintaining and providing cell lines for this study. The authors also thank Don Kirkpatrick for tion, RNA, epigenetics, and protein expression) may improve technical assistance and insightful discussions. our ability to identify bona fide cancer targets. As technologies The costs of publication of this article were defrayed in part by the payment of advance to generate accurate high-density datasets, the use of page charges. This article must therefore be hereby marked advertisement in high-content screening approaches to assess multiple hall- accordance with 18 U.S.C. Section 1734 solely to indicate this fact. marks of cancer in relevant cancer models will greatly improve Received September 19, 2013; revised February 18, 2014; accepted March 7, our ability to identify relevant targets (46, 47). 2014; published OnlineFirst April 22, 2014.

References 1. Stratton MR. Exploring the genomes of cancer cells: progress and 14. Yang J, Fan J, Li Y, Li F, Chen P, Fan Y, et al. Genome-wide RNAi promise. Science 2011;331:1553–8. screening identifies genes inhibiting the migration of glioblastoma 2. Chin L, Gray JW. Translating insights from the cancer genome into cells. PLoS ONE 2013;8:e61915. clinical practice. Nature 2008;452:553–63. 15. GeneLogic. Available from: http://www.genelogic.com/. 3. Collisson EA, Cho RJ, Gray JW. What are we learning from the cancer 16. Li L, Chaudhuri A, Chant J, Tang Z. PADGE: analysis of heterogeneous genome? Nat Rev Clin Oncol 2012;9:621–30. patterns of differential gene expression. Physiol Genomics 2007;32: 4. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. 154–9. Cell 2011;144:646–74. 17. Lee GY, Kenny PA, Lee EH, Bissell MJ. Three-dimensional culture 5. Korkola J, Gray JW. Breast cancer genomes–form and function. Curr models of normal and malignant breast epithelial cells. Nat Methods Opin Genet Dev 2010;20:4–14. 2007;4:359–65. 6. Leary RJ, Lin JC, Cummins J, Boca S, Wood LD, Parsons DW, et al. 18. Huang da W, Sherman BT, Lempicki RA. Systematic and integrative Integrated analysis of homozygous deletions, focal amplifications, and analysis of large gene lists using DAVID bioinformatics resources. Nat sequence alterations in breast and colorectal cancers. Proc Natl Acad Protoc 2009;4:44–57. Sci U S A 2008;105:16224–9. 19. Olshen AB, Venkatraman ES, Lucito R, Wigler M. Circular binary 7. Chin K, Devries S, Fridlyand J, Spellman PT, Roydasgupta R, Kuo WL, segmentation for the analysis of array-based DNA copy number data. et al. Genomic and transcriptional aberrations linked to breast cancer Biostatistics 2004;5:557–72. pathophysiologies. Cancer Cell 2006;10:529–41. 20. Beroukhim R, Getz G, Nghiemphu L, Barretina J, Hsueh T, Linhart D, 8. Brown LA, Johnson K, Leung S, Bismar TA, Benitez J, Foulkes WD, et al. Assessing the significance of chromosomal aberrations in can- et al. Coamplification of CCND1 and EMSY is associated with an cer: methodology and application to glioma. Proc Natl Acad Sci U S A adverse outcome in ER-positive tamoxifen-treated breast cancers. 2007;104:20007–12. Breast Cancer Res Treat 2010;121:347–54. 21. Geiger T, Cox J, Mann M. Proteomic changes resulting from gene copy 9. Kwek SS, Roy R, Zhou H, Climent J, Martinez-Climent JA, Fridlyand J, number variations in cancer cells. PLoS Genet 2010;6:e1001090. et al. Coamplified genes at 8p12 and 11q13 in breast tumors cooperate 22. Neve RM, Chin K, Fridlyand J, Yeh J, Baehner FL, Fevr T, et al. A with two major pathways in oncogenesis. Oncogene 2009;28: collection of breast cancer cell lines for the study of functionally distinct 1892–903. cancer subtypes. Cancer Cell 2006;10:515–27. 10. Albertson DG, Collins C, McCormick F, Gray JW. Chromosome aber- 23. Knauf JA, Ouyang B, Knudsen ES, Fukasawa K, Babcock G, Fagin JA.

rations in solid tumors. Nat Genet 2003;34:369–76. Oncogenic RAS induces accelerated transition through G2–M and 11. Gelsi-Boyer V, Orsetti B, Cervera N, Finetti P, Sircoulomb F, Rouge C, promotes defects in the G2 DNA damage and mitotic spindle check- et al. Comprehensive profiling of 8p11–12 amplification in breast points. J Biol Chem 2006;281:3800–9. cancer. Mol Cancer Res 2005;3:655–67. 24. TCGA consortium. Available from: http://cancergenome.nih.gov/. 12. Kao J, Pollack JR. RNA interference–based functional dissection of the 25. Gyorffy B, Lanczky A, Eklund AC, Denkert C, Budczies J, Li Q, et al. An 17q12 amplicon in breast cancer reveals contribution of coamplified online survival analysis tool to rapidly assess the effect of 22,277 genes genes. Genes Chromosomes Cancer 2006;45:761–9. on breast cancer prognosis using microarray data of 1,809 patients. 13. Sawey ET, Chanrion M, Cai C, Wu G, Zhang J, Zender L, et al. Breast Cancer Res Treat 2010;123:725–31. Identification of a therapeutic strategy targeting amplified FGF19 in 26. Ruschak AM, Slassi M, Kay LE, Schimmer AD. Novel proteasome liver cancer by Oncogenomic screening. Cancer Cell 2011;19: inhibitors to overcome bortezomib resistance. J Natl Cancer Inst 347–58. 2011;103:1007–17.

OF12 Cancer Res; 74(11) June 1, 2014 Cancer Research

Oncogenomic Target Discovery

27. Hoeller D, Dikic I. Targeting the ubiquitin system in cancer therapy. 38. Pellom ST Jr, Shanker A. Development of proteasome inhibitors as Nature 2009;458:438–44. therapeutic drugs. J Clin Cell Immunol 2012;S5:5. 28. Murata S, Yashiroda H, Tanaka K. Molecular mechanisms of protea- 39. Lee KM, Lee J, Park CS. Cereblon inhibits proteasome activity by some assembly. Nat Rev 2009;10:104–15. binding to the 20S core proteasome subunit beta type 4. Biochem 29. Hirano Y, Kaneko T, Okamoto K, Bai M, Yashiroda H, Furuyama K, et al. Biophys Res Commun 2012;427:618–22. Dissecting beta-ring assembly pathway of the mammalian 20S pro- 40. Li X, Kusmierczyk AR, Wong P, Emili A, Hochstrasser M. Beta-subunit teasome. EMBO J 2008;27:2204–13. appendages promote 20S proteasome assembly by overcoming an 30. Hebbring SJ, Chai Y, Ji Y, Abo RP, Jenkins GD, Fridley B, et al. Serine Ump1-dependent checkpoint. EMBO J 2007;26:2339–49. hydroxymethyltransferase 1 and 2: gene sequence variation and 41. Oerlemans R, Franke NE, Assaraf YG, Cloos J, van Zantwijk I, Berkers functional genomic characterization. J Neurochem 2012;120:881–90. CR, et al. Molecular basis of bortezomib resistance: proteasome 31. Kalhan SC, Hanson RW. Resurgence of serine: an often neglected but subunit beta5 (PSMB5) gene mutation and overexpression of PSMB5 indispensable amino Acid. J Biol Chem 2012;287:19786–91. protein. Blood 2008;112:2489–99. 32. Locasale JW, Grassian AR, Melman T, Lyssiotis CA, Mattaini KR, Bass 42. Shuqing L, Jianmin Y, Chongmei H, Hui C, Wang J. Upregulated AJ, et al. Phosphoglycerate dehydrogenase diverts glycolytic flux and expression of the PSMB5 gene may contribute to drug resistance in contributes to oncogenesis. Nat Genet 2011;43:869–74. patient with multiple myeloma when treated with bortezomib-based 33. Possemato R, Marks KM, Shaul YD, Pacold ME, Kim D, Birsoy K, et al. regimen. Exp Hematol 2011;39:1117–8. Functional genomics reveal that the serine synthesis pathway is 43. Raeder MB, Birkeland E, Trovik J, Krakstad C, Shehata S, Schumacher essential in breast cancer. Nature 2011;476:346–50. S, et al. Integrated genomic analysis of the 8q24 amplification in 34. Vie N, Copois V, Bascoul-Mollevi C, Denis V, Bec N, Robert B, et al. endometrial cancers identifies ATAD2 as essential to MYC-dependent Overexpression of phosphoserine aminotransferase PSAT1 stimulates cancers. PLoS ONE 2013;8:e54873. cell growth and increases chemoresistance of colon cancer cells. Mol 44. Wu J, Liu S, Liu G, Dombkowski A, Abrams J, Martin-Trevino R, et al. Cancer 2008;7:14. Identification and functional analysis of 9p24 amplified genes in human 35. Ye J, Mancuso A, Tong X, Ward PS, Fan J, Rabinowitz JD, et al. breast cancer. Oncogene 2012;31:333–41. Pyruvate kinase M2 promotes de novo serine synthesis to sustain 45. Yang ZQ, Streicher KL, Ray ME, Abrams J, Ethier SP. Multiple inter- mTORC1 activity and cell proliferation. Proc Natl Acad Sci U S A acting oncogenes on the 8p11-p12 amplicon in human breast cancer. 2012;109:6904–9. Cancer Res 2006;66:11632–43. 36. Chaneton B, Hillmann P, Zheng L, Martin AC, Maddocks OD, Chok- 46. Silverman EK, Loscalzo J. Network medicine approaches to the kathukalam A, et al. Serine is a natural ligand and allosteric activator of genetics of complex diseases. Discov Med 2012;14:143–52. pyruvate kinase M2. Nature 2012;491:458–62. 47. Vidal M, Chan DW, Gerstein M, Mann M, Omenn GS, Tagle D, et al. 37. Thaker NG, Zhang F, McDonald PR, Shun TY, Lewen MD, Pollack IF, The human proteome - a scientific opportunity for transforming et al. Identification of survival genes in human glioblastoma cells by diagnostics, therapeutics, and healthcare. Clin Proteomics 2012; small interfering RNA screening. Mol Pharmacol 2009;76:1246–55. 9:6.

www.aacrjournals.org Cancer Res; 74(11) June 1, 2014 OF13

Comparative Oncogenomics Identifies PSMB4 and SHMT2 as Potential Cancer Driver Genes

Genee Y. Lee, Peter M. Haverty, Li Li, et al.

Cancer Res Published OnlineFirst April 22, 2014.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-13-2683

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2014/04/22/0008-5472.CAN-13-2683.DC1

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/early/2014/05/14/0008-5472.CAN-13-2683. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.