Mutational Landscape and Antiproliferative Functions of ELF

Published OnlineFirst February 26, 2016; DOI: 10.1158/0008-5472.CAN-14-3816 Cancer Molecular and Cellular Pathobiology Research

Mutational Landscape and Antiproliferative Functions of ELF Transcription Factors in Human Cancer Mizuo Ando1,2, Masahito Kawazu3, Toshihide Ueno1, Daizo Koinuma4, Koji Ando5, Junji Koya6, Keisuke Kataoka6, Takahiko Yasuda1, Hiroyuki Yamaguchi1, Kazutaka Fukumura1, Azusa Yamato1, Manabu Soda1, Eirin Sai3, Yoshihiro Yamashita1, Takahiro Asakage2, Yasushi Miyazaki5, Mineo Kurokawa6, Kohei Miyazono4, Stephen D. Nimer7, Tatsuya Yamasoba2, and Hiroyuki Mano1,8

Abstract

ELF4 (also known as MEF) is a member of the ETS family of ELF1 and ELF2, as in ELF4, were widespread across human transcription factors. An oncogenic role for ELF4 has been cancers, but were almost all mutually exclusive. Moreover, demonstrated in hematopoietic malignancies, but its function chromatin immunoprecipitation coupled with high-throughput in epithelial tumors remains unclear. Here, we show that ELF4 sequencing revealed ELF4-binding sites in genomic regions can function as a tumor suppressor and is somatically inacti- adjacent to genes related to cell-cycle regulation and apoptosis. vated in a wide range of human tumors. We identified a Finally, we provide mechanistic evidence that the antiprolifera- missense mutation affecting the transactivation potential of tive effects of ELF4 were mediated through the induction of ELF4 in oral squamous cell carcinoma cells. Restoration of the HRK, an activator of apoptosis, and DLX3, an inhibitor of cell transactivation activity through introduction of wild-type ELF4 growth. Collectively, our findings reveal a novel subtype of significantly inhibited cell proliferation in vitro and tumor human cancer characterized by inactivating mutations in the xenograft growth. Furthermore, we found that ELF1 and ELF2, ELF subfamily of proteins, and warrant further investigation of closely related transcription factors to ELF4, also exerted anti- the specific settings where ELF restoration may be therapeuti- proliferative effects in multiple cancer cell lines. Mutations in cally beneficial. Cancer Res; 76(7); 1–11. 2016 AACR.

Introduction prostate tumors (3). Forced expression of TMPRSS2-ERG in the context of PI3K pathway activation in prostate epithelial cells ELF4, also known as myeloid Elf-1–like factor (MEF), belongs results in the formation of prostatic intraepithelial neoplasia in to the ETS family of transcription factors that are encoded by 28 mice, demonstrating an oncogenic role for the fusion gene (4). In independent genes in the human genome (1, 2). ETS proteins are addition, fusion of EWSR1 with the ETS family gene FLI1 or implicated in the proliferation, differentiation, and malignant related ones occurs frequently in Ewing's sarcoma (5). Such transformation of various cell lineages. The genes for ERG or fusions are promising therapeutic targets, given that knockdown members of the PEA3 subfamily (ETV1, ETV4, and ETV5) of ETS of EWSR1–FLI1 expression suppresses sarcoma growth. proteins, for instance, are fused to TMPRSS2 in >50% of human ELF4 as well as ELF1 and ELF2 constitute the ELF subfamily of ETS proteins. Elf4-null mice manifest severe defects in the devel- 1Department of Cellular Signaling, Graduate School of Medicine, The opment and function of natural killer cells and NK-T cells (6). University of Tokyo, Tokyo, Japan. 2Department of Otolaryngology Furthermore, the hematopoietic system of such mice harbors an and Head and Neck Surgery, Graduate School of Medicine, The Uni- increased population of quiescent hematopoietic stem cells com- versity of Tokyo, Tokyo, Japan. 3Department of Medical Genomics, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan. pared with wild-type animals, suggesting that ELF4 plays a key 4 Department of Molecular Pathology, Graduate School of Medicine, role in the transition from G0 to G1 phases of the cell cycle (7). The 5 The University of Tokyo, Tokyo, Japan. Department of Hematology, activity of ELF4 is selectively increased in G1 phase and is sup- Nagasaki University Graduate School of Biomedical Sciences, Naga- pressed by the complex of cyclin A and cyclin-dependent kinase 2 saki, Japan. 6Department of Hematology and Oncology, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan. 7Sylvester (CDK2), further linking ELF4 to cell-cycle regulation (8). Comprehensive Cancer Center, University of Miami, Miami, Florida. An oncogenic role for ELF4 has been suggested by several 8 Strategic Basic Research Program, Japan Science and Technology observations. ELF4 is highly expressed in clinical specimens of Agency, Saitama, Japan. leukemia and ovarian cancer (9,10). Elf4 induces Mdm2, and also Note: Supplementary data for this article are available at Cancer Research inhibits Cdkn2a activation, and thereby cooperates with onco- Online (http://cancerres.aacrjournals.org/). genic HRAS(G12V) in malignant transformation (11). Further- Corresponding Author: Hiroyuki Mano, The University of Tokyo, 7-3-1 Hongo, more, cells from Elf4-null mice are impaired in the ability to form Bunkyo-ku, Tokyo 113-0033, Japan. Phone: 81-3-5841-0633; Fax: 81-3-5841- neurospheres, and knockdown of ELF4 was found to inhibit 0634; E-mail: [email protected] neurosphere formation by human primary glioma stem-like cells, doi: 10.1158/0008-5472.CAN-14-3816 suggesting that ELF4 promotes stemness in glioma cells (12). On 2016 American Association for Cancer Research. the other hand, however, one article shows a possibility for the

www.aacrjournals.org OF1

Ando et al.

growth-suppressive role of ELF4 in a human lung cancer cell line resulted in the identification of 12 missense mutations including A549 (13). ELF4(L211M) (20). Given that L211 maps within the conserved We now describe an L211M amino acid substitution in the ETS ETS domain of ELF4 (Fig. 1A), we chose this nonsynonymous domain of ELF4 that is present in the human squamous cell mutation for further analysis. A homozygous V382I mutation of carcinoma cell line T3M-1 Cl-10 (14). Forced expression of wild- ELF4 was also identified in the genome of the human breast type ELF4 resulted in marked inhibition of cell proliferation not cancer cell line MDA-MB-231 (data not shown). only in this cell line but also in a human fibrosarcoma cell line Further searching for nonsynonymous mutations of ELF4 in HT1080 (15) and a human endometrial carcinoma cell line the Catalogue of Somatic Mutations in Cancer (COSMIC; http:// HEC59 (16) as well as in the nontransformed human mammary cancer.sanger.ac.uk/cancergenome/projects/cosmic), Cancer Cell epithelial cell line MCF10A (17). We further demonstrate that Line Encyclopedia (CCLE; http://www.broadinstitute.org/ccle/ somatic mutations of ELF1, ELF2, and ELF4 are present and home), International Cancer Genome Consortium (ICGC; mutually exclusive in a wide range of human tumors, and that http://icgc.org), and The Cancer Genome Atlas (TCGA; https:// the antiproliferative effect of ELF4 is mediated through the induc- tcga-data.nci.nih.gov/tcga/tcgaHome2.jsp) databases revealed tion of HRK and DLX3 expression. >40 mutations widely distributed throughout the protein structure of ELF4 (Fig. 1A). In addition, Totoki and colleagues (21) fi BCORL1–ELF4 Materials and Methods identi ed a fusion gene in hepatocellular carcinoma. Cell lines We examined the transactivation potential of ELF4(L211M), T3M-1 Cl-10 or HEC59 was obtained from RIKEN Cell Bank or ELF4(V382I), and BCORL1-ELF4 in comparison with that of the Japanese Collection of Research Bioresources Cell Bank, respec- wild-type protein (Fig. 1B). The activity of a reporter construct tively, and human embryonic kidney 293T (HEK293T), HT1080, containing the ETS-binding regions within the IL3 promoter MDA-MB-231, KATO-III, A549 and MCF10A were obtained (APET; ref. 19) was markedly increased by expression of wild- from ATCC. MCF10A was maintained in DMEM-F12 supplemen- type ELF4. Whereas the V382I substitution did not substantially ted with 5% horse serum (Biowest), recombinant human EGF affect the transactivation activity of ELF4, the activities of both (20 ng/mL; Peprotech), bovine insulin (10 mg/mL; Sigma- ELF4(L211M) and BCORL1-ELF4 were significantly attenuated Aldrich), hydrocortisone (0.5 mg/mL; Sigma-Aldrich), and cholera compared with that of the wild-type protein. The observed tran- toxin (100 ng/mL; Sigma-Aldrich). All other cells were main- scriptional activity was dependent on the ETS-binding regions of tained in DMEM-F12 supplemented with 10% FBS and 2 mmol/L the IL3 promoter, given that substitution of nucleotides within the L-glutamine (Invitrogen). ETS-binding sites (the mAPET reporter) abolished all activity. Furthermore, expression level of endogenous as well as artificially Functional analysis of ELF4 and other proteins introduced ELF4 mRNA was determined by reverse transcription We inserted an internal ribosomal entry site (ires) and (RT) and real-time PCR (PCR) analysis. As demonstrated in enhanced green fluorescent protein (EGFP) cDNA (Clontech) Supplementary Fig. S1, almost 20 times larger amounts of ELF4 into pMXS (18) to yield the pMXS-ires-EGFP plasmid, which mRNA were expressed in the transfected cells compared with the allows simultaneous expression of inserted cDNAs together with control, suggesting that competitive effects of endogenous ELF4 EGFP. For reporter assays of ELF4 activity, HEK293T cells were are negligible. transfected with pMXS- or pMXS-ires-EGFP-based expression We next examined whether such loss of function of ELF4 plasmids for various cDNAs, the firefly luciferase–based APET might contribute to carcinogenesis. Whereas introduction of reporter plasmid (19), and the pGL-TK plasmid for Renilla lucif- ELF4(L211M) or BCORL1-ELF4 had no marked effect on the erase (Promega). As a negative control for ELF4 activity, cells were proliferation of T3M-1 Cl-10 cells, expression of wild-type ELF4 transfected with the mAPET plasmid (19) instead of APET. Quan- or ELF4(V382I) greatly inhibited cell growth (Fig. 1C), sugges- titative data are presented as means SD and were analyzed as tive of a tumor-suppressive role for ELF4 in oral squamous cell indicated. A P value of <0.05 was considered statistically carcinoma. Expression level of ELF4 mediated by retrovirus significant. transduction was >40 times higher than that of endogenous Infectious retroviruses were generated by transfection of ELF4, again suggesting a negligible effect of endogenous ELF4 HEK293T cells with pMXS-ires-EGFP-based expression vectors protein (Supplementary Fig. S2). The antiproliferative activity together with pGP and pAmpho packaging plasmids (Takara of wild-type ELF4 was also apparent in HT1080 cells (Fig. 1C), Bio). After infection of T3M-1 Cl-10 or other cell lines with the which do not harbor any nonsynonymous mutations within resulting retroviruses, EGFP-positive cells were monitored by flow ELF1, ELF2, or ELF4 (22). In this cell line, however, ELF4 cytometry. Cell-cycle profiles were examined with the use of an (L211M) had a moderate antiproliferative effect, likely reflect- APC BrdU Flow Kit (BD Biosciences), and the enzymatic activities ing its residual transactivation activity (Fig. 1B). Unexpectedly, of caspase-3 and 7 were measured with the use of a Caspase-Glo similar effects of ELF4 and its mutants were also apparent in 3/7 Assay (Promega). Other experimental methods are described nontransformed MCF10A cells (Fig. 1C), suggesting that the in detail in Supplementary Methods. antiproliferative function of ELF4 might be operative in a wider range of cells than previously appreciated. The same findings were also confirmed in an inducible Results system. A doxycycline-inducible expression plasmid of FLAG- Tumor-suppressive role of ELF4 tagged ELF4 or ELF4(L211M) was constructed, and stably T3M-1 Cl-10 is a cell line of squamous cell carcinoma of the oral transduced into T3M-1 Cl-10 or HEC59 that carries ELF4 cavity established from 33-year-old male (14). Our previous RNA- (Q341fs). We first confirmed that the addition of a FLAG sequence analysis of the cells with a next-generation sequencer epitope tag to the COOH-terminus of ELF4 did not affect its

OF2 Cancer Res; 76(7) April 1, 2016 Cancer Research

ELF Proteins Function as a Tumor Suppressor

Figure 1. Antiproliferative activity of ELF4. A, nonsynonymous mutations of ELF4. The positions of missense, nonsense, and frameshift mutations of human ELF4 identified from database searches are shown on the schematic protein structure. The positions of L211M and V382I mutations detected in T3M-1 Cl-10 and MDA-MB-231 cells, respectively, as well as that of the point of fusion to BCORL1 are also indicated. A, acidic domain; AID, AML1-interacting domain; ST, serine- and threonine-rich domain; and P, proline-rich domain. B, luciferase reporter activity for HEK293T cells transfected with pMXS-based expression plasmids for ELF4, ELF4(L211M), ELF4(V382I), or BCORL1-ELF4 (or with a mock plasmid) together with the APET reporter plasmid (or its mutant form mAPET) and the pGL-TK plasmid for Renilla luciferase. Data represent fireflyluciferaseactivitynormalizedbyRenilla luciferase – – activity and are means þ SD from three independent experiments; , P ¼ 6.52 10 3; †, P ¼ 5.23 10 3 versus the corresponding value for wild-type ELF4 (Student t test). C, growth curves for T3M-1 Cl-10, HT1080, or MCF10A cells infected with recombinant retroviruses generated from pMXS-ires- EGFP–based expression plasmids for ELF4, ELF4(L211M), ELF4(V382I), or BCORL1-ELF4 (or from a mock plasmid). The EGFP–positive fraction was monitored by flow cytometry at the indicated times after infection. Data, means SD from three independent experiments. Proportion of EGFP-positive cells at day 15 was compared between the cells expressing the wild-type ELF4 and those expressing ELF4(L211M) by the Student t test; – – – , P ¼ 4.19 10 3; †, P ¼ 3.51 10 3; , P ¼ 2.05 10 2. transactivation activity (Supplementary Fig. S3), and doxycy- We further evaluated how expression of ELF4 disturbs cline-dependent expression of the proteins was verified by cell growth. Cell-cycle analysis of T3M-1 Cl-10 cells revealed immunoblot analyses (Supplementary Fig. S4). As shown in that forced expression of either wild-type or V382I mutant Supplementary Fig. S5, induction of the wild-type ELF4 sup- form of ELF4 resulted in a significant increase in the pro- pressed the growth of both cell lines, but such effect was portion of cells in G0–G1 and a concomitant decrease in the significantly attenuated with the ELF4(L211M) mutant. In S-phase fraction (Supplementary Fig. S7). Such effects contrast, knockdown of ELF4 message did not affect the growth were not observed with ELF4(L211M) or BCORL1-ELF4. The profile of MCF10A cells harboring the wild-type ELF4 (Sup- activities of caspase-3/7 were also increased in these cells plementary Fig. S6). by forced expression of ELF4 or ELF4(V382I), but not by

www.aacrjournals.org Cancer Res; 76(7) April 1, 2016 OF3

Ando et al.

(center). In contrast, apoptosis was only focally observed in HEC59 cells expressing ELF4(L211M).

Partially redundant functions of ELF1 and ELF2 relative to ELF4 Given that the protein structures of ELF1 and ELF2 are similar to that of ELF4, we compared the effects of these ELF subfamily members as well as ELF3 and ELF5 (both of which belong to the ESE subfamily of ETS proteins) on IL3 promoter activity (Fig. 3A). Both ELF1 and ELF2 increased reporter activity, albeit to a lesser extent than did ELF4. Whereas ELF3 had no effect, ELF5 also induced a small increase in the reporter activity. These results, thus, suggested that the ELF subfamily proteins have partially redundant roles in transcriptional regulation. We therefore next examined whether the ELF subfamily proteins might also share a tumor-suppressive function. Forced expression of ELF2 inhibited the proliferation of T3M-1 Cl-10 cells to the same marked extent as did ELF4 (P ¼ 9.82 10 3 for mock vs. ELF2, Student t test; Fig. 3B). Both ELF1 and ELF3 had a moderate antiproliferative effect, whereas ELF5 had no such effects. ELF2 also markedly inhibited the proliferation of HT1080 and MCF10A cells, but the effects of ELF1, ELF3, and ELF5 varied in these cell lines. The differences in the growth retardation potential of these ETS proteins among the cell lines did not appear to be due to differences in the expression level of the corresponding endogenous proteins, given that the mRNA levels for each protein were similar among the three cell lines (with the exception Figure 2. of that for ELF4 in MCF10A; Supplementary Fig. S8). In vivo tumor-suppressive potential of ELF4. A, HEC59 cells were stably Because our results suggested that the members of the ELF transduced with a doxycycline-inducible expression plasmid for ELF4 or subfamily may share a tumor-suppressive role, we screened the ELF4(L211M), or the empty vector (Mock). One million of each cells were s.c. injected into NOG mice. Sixteen days later, oral administration COSMIC, CCLE, ICGC, and TCGA databases for cancer specimens of doxycycline (200 mg/L) was started. Tumor size was measured at and cell lines that harbor nonsynonymous mutations in the entire days 16, 18, 20, 23, and 26, and shown as means þ SD from eight protein of ELF1, ELF2, or ELF4. Such mutations were found to be independent experiments. Tumor size at day 26 was compared with the widespread among human cancers (Fig. 3C). Furthermore, muta- Student t test. B, two days after doxycycline administration, tumor tions of these three proteins were almost completely mutually section was subjected to hematoxylin and eosin (H&E) staining exclusive, with the exception of a lung adenocarcinoma specimen (left), immunohistostaining with antibodies to FLAG tag (middle), or TUNEL assay (right); scale bars, 1 mm. positive for ELF1 and ELF2 mutations and a specimen of adenocarcinoma of the large intestine positive for mutations in ELF1, ELF2, and ELF4. that of ELF4(L211M) or BCORL1-ELF4 (Supplementary We next examined whether nonsynonymous mutations within Table S1). the ETS domain of these three proteins affect their transactivation To confirm the tumor-suppressive function of ELF4 in vivo,we potential. Most such amino acid substitutions reduced the ability of chose HEC59 cells infected with a recombinant retrovirus for the proteins to activate the IL3 promoter, albeit to various extents doxycycline-inducible FLAG-tagged ELF4 or ELF4(L211M) for the (Fig. 4A). Moreover, whereas forced expression of ELF1 moderately NOG (NOD/Shi-scid, IL-2Rg-null) mouse tumorigenicity assay, inhibited the growth of T3M-1 Cl-10 cells, that of the loss-of- because, among the cell lines we used, only HEC59 was function mutants P225A or G265R had no inhibitory effect (Fig. tumorigenic in the mice. Each transduced cells were subcuta- 4B). Likewise, loss-of-function mutants of ELF2 (E140 and neously injected into mice, and the expression of ELF4 or its W238) did not affect the growth of T3M-1 Cl-10 cells, whereas mutant was induced with administration of doxycycline at 16 the wild-type protein had a marked antiproliferative action. Similar days after the injection. While tumor size of the mock-trans- results were obtained for expression of these various mutant duced HEC59 cells were increased during the observation proteins in HT1080 or MCF10A cells, whereas the growth-suppres- period, growth of HEC59 expressing the wild-type ELF4 was sive effect of wild-type ELF1 was cell line dependent (Fig. 4B). significantly retarded, again confirming a tumor-suppressive With regard to the ELF4 mutants, the growth of T3M-1 Cl-10 function of ELF4 (Fig. 2A). Such effect was less profound for cells was significantly suppressed by forced expression of ELF4 ELF4(L211M), however. (E282K) [P ¼ 1.10 10 2 for mock vs. ELF4(E282K), Student t At 2 days after doxycycline administration, tumors were also test; Fig. 4C], the transactivation activity of which was similar to pathologically examined. As shown in the left of Fig. 2B, expres- that of the wild-type protein (Fig. 4A). Other ELF4 mutants, sion of ELF4 induced massive apoptosis of HEC59 cells, which however, had lost such growth-suppressive ability to an extent was confirmed by terminal deoxynucleotidyl-transferase–medi- that approximately paralleled their loss of transactivation activity. ated dUTP nick end labeling (TUNEL) assay (right). Accordingly, Furthermore, we tested whether ELF4 mutants act on the wild- expression of ELF4 was confirmed in marginal zone of the tumor type protein in a dominant-negative manner. As demonstrated in

OF4 Cancer Res; 76(7) April 1, 2016 Cancer Research

ELF Proteins Function as a Tumor Suppressor

Figure 3. Redundant tumor-suppressive roles of ELF subfamily members. A, APET reporter activity for HEK293T cells transfected with pMXS-based expression plasmids for ELF1, ELF2, ELF3, ELF4, or ELF5. Data represent firefly luciferase activity normalized by Renilla luciferase activity and are means SD for three independent experiments. B, growth curves for T3M-1 Cl-10, HT1080, or MCF10A cells infected with recombinant retroviruses generated from pMXS-ires-EGFP– based expression plasmids for ELF1, ELF2, ELF3, ELF4, or ELF5. The EGFP- positive fraction was monitored by flow cytometry at the indicated times after infection. Data, means SD from three independent experiments. C, nonsynonymous mutations of ELF1, ELF2, or ELF4 identified in human cancer specimens and cell lines by database screening. The specimen carrying ELF4(L211M), ELF4(V382I), or BCORL1-ELF4 is indicated at the bottom.

Supplementary Fig. S9, ELF4(L211M) slightly enhances the 17,127 ELF4-binding peaks throughout the genome with an FDR reporter activation of the wild-type in a dose-dependent way, of <0.1. These peaks included intergenic regions (46.16%), pro- reflecting the residual transcriptional factor activity of the L211M moter regions (2.03%), coding exons (13.33%), and introns mutant as shown Fig. 4A. On the other hand, expression of the (32.81%; Fig. 5A). Promoters, 50 noncoding exons, coding exons, activity-null mutant (F239L and R267Q) did not affect the 30 noncoding exons, and noncoding RNA sequences were signif- reporter activity. Therefore, these loss-of-function mutations are icantly overrepresented among the ELF4-binding sites relative to unlikely to generate dominant-negative proteins. regions randomly selected from the human genome (n ¼ 100,000; – – – c2 test, P ¼ 4.07 10 42, 6.49 10 3,0,3.85 10 25, and 3.12 – Genome-wide screening for ELF4-binding sites 10 132, respectively). At this FDR cutoff, no single peaks were To explore further the tumor-suppressive function of ELF4, we detected for the specific binding of ELF4(L211M), consistent with searched the human genome for ELF4-binding sites with the use the notion that the L211M substitution is a loss-of-function of chromatin immunoprecipitation coupled with sequencing mutation with regard to the transactivation activity of ELF4. (ChIP-seq) with a next-generation sequencer. We expressed We verified some of the ELF4-specific peaks by ChIP coupled ELF4-FLAG in T3M-1 Cl-10 cells, and ELF4 and its bound DNA with real-time PCR analysis (Supplementary Fig. S10). fragments were immunoprecipitated with antibodies to FLAG for De novo prediction of consensus sequences among the ELF4- DNA sequencing. More than 22 million sequence reads were binding sites picked up motif #15, which contains GGAA, the obtained for cells expressing ELF4 or ELF4(L211M) as well as for representative binding sequence for ETS proteins (1), thus vali- mock-infected cells. dating our analysis (Fig. 5B). Other predicted motifs include #11 Comparison of the high-quality reads between the ELF4-expres- as a potential binding site for ETS proteins and STAT3, as well as sing and mock-infected cells resulted in the identification of #13 as a binding site for SOX10 and SOX2.

www.aacrjournals.org Cancer Res; 76(7) April 1, 2016 OF5

Ando et al.

Figure 4. Loss-of-function mutations in the ELF subfamily members. A, APET reporter activity for HEK293T cells expressing ELF1, ELF2, or ELF4 or their ETS domain mutants. Data represent firefly luciferase activity normalized by Renilla luciferase activity and are means SD for three independent experiments. B, growth curves for T3M-1 Cl-10, HT1080, or MCF10A cells infected with recombinant retroviruses generated from pMXS-ires-EGFP– based expression plasmids for ELF1 or ELF2 or their ETS domain mutants. The EGFP-positive fraction was monitored by flow cytometry at the indicated times after infection. Data, means SD from three independent experiments. C, growth curves for T3M-1 Cl-10 cells infected with recombinant retroviruses generated from pMXS-ires-EGFP– based expression plasmids for ELF4 or its ETS domain mutants. The EGFP- positive fraction was monitored by flow cytometry at the indicated times after infection. Data, means SD from three independent experiments.

Binding peaks for ELF4 were mapped within the region span- The expression profilesrevealedthat90and5geneswere ning positions 1,000 and þ50 bp relative to the transcription markedly induced (>8-fold) or suppressed (<1/8-fold), start site for 204 independent protein-coding genes. The "biolog- respectively, in the cells expressing wild-type ELF4 compared ical process" terms of the Gene Ontology (GO) classification for with those expressing ELF4(L211M) (Supplementary Tables these genes were significantly enriched in those related to a S3 and S4). Given that forced expression of ELF4 in T3M-1 Cl- proapoptotic function (Supplementary Table S2), suggestive of 10 cells suppressed the G0–G1 to S phase transition (Supple- a direct role for ELF4 in cell death. mentary Fig. S7) as well as increased caspase-3/7 activity (Supplementary Table S1), we focused on DLX3 and HRK DLX3 and HRK mediate the tumor-suppressive function of among the ELF4-induced genes. DLX3 encodes a homeobox ELF4 transcription factor, and its forced expression in the basal cell To elucidate how ELF4 exerts its antiproliferative activity, layer of skin inhibits proliferation of the basal cells (24). HRK we compared gene-expression profiles between T3M-1 Cl-10 encodes a protein (Harakiri) that binds to and markedly cells expressing ELF4 or ELF4(L211M) with the use of DNA inhibits the activity of the antiapoptotic proteins BCL2 and microarray analysis. Gene set enrichment analysis (GSEA; BCL2L1 (25). ref. 23) of the two datasets resulted in the isolation of 344 Real-time RT-PCR analysis confirmed that DLX3 and HRK and 169 gene sets for transcription factor targets and biologic expression was significantly induced by wild-type ELF4, but processes, respectively, with a q value of <0.25. These gene sets not by ELF4(L211M), in T3M-1 Cl-10 cells (Fig. 6B). Forced included "ETS motif," "growth," "G1 phase," and "apoptosis" expression of DLX3 or HRK also inhibited the proliferation of (Fig. 6A), again implicating ELF4 in cell-cycle regulation and these cells to an extent similar to that apparent with ELF4 (Fig. apoptosis. 6C). Likewise, induction of DLX or HRK affected the G0–G1 to

OF6 Cancer Res; 76(7) April 1, 2016 Cancer Research

ELF Proteins Function as a Tumor Suppressor

Figure 5. ELF4-binding sites in the human genome. A, the proportion of mapped regions in the human genome for ChIP-seq reads obtained with immunoprecipitates prepared with antibodies to FLAG specifically from T3M-1 Cl- 10 cells expressing ELF4-FLAG is shown on the left. Mean values of mapped regions for randomly selected 100-bp fragments of the human genome (n ¼ 100,000) are shown on the right (simulation). B, significantly enriched consensus sequences among the identified ELF4-binding regions are shown in both forward and reverse directions.

S phase transition in the cell cycle (Supplementary Fig. S7) pressor and is somatically inactivated in a wide array of human and caspase-3/7 activity (Supplementary Table S1). tumors. The wide distribution of nonsynonymous mutation sites To examine whether DLX3 and HRK are essential mediators of within the ELF4 protein structure further suggests that loss of the antiproliferative effect of ELF4, we prepared shRNA constructs function of ELF4 contributes to carcinogenesis. to deplete the corresponding mRNAs (sh#5 and sh#8 for DLX3, Whereas our ﬁndings provide evidence for a tumor-suppressive and sh#7 and sh#9 for HRK; Supplementary Fig. S11). T3M-1 Cl-10 function of ELF4, they are not necessarily inconsistent with cells stably expressing a control shRNA, sh#5, or sh#8 were then previous results indicating an oncogenic function of ELF4. ELF4 infected with retroviruses for wild-type ELF4 or ELF4(L211M). The has, thus, been found to be essential for neurosphere formation by suppression of cell growth by ELF4 was markedly attenuated by glioma cells, implicating the protein in the maintenance of expression of sh#5 or sh#8 (Fig. 7A), showing that DLX3 is an stemness (12). Such stemness of cancer is frequently coincident effector for ELF4. Similarly, the antiproliferative activity of wild- with a quiescent state or slow cell cycling (26). Our observation type ELF4 was partially abrogated in cells expressing sh#7 or sh#9 that ELF4 expression suppresses the transition from G0–G1 to S (Fig. 7B), indicating that HRK is also an effector for ELF4. The phase of the cell cycle is, thus, consistent with the proposed role of suppression of the antiproliferative effect of ELF4 was most pro- ELF4 in glioma stem-like cells. nounced in T3M-1 Cl-10 cells stably expressing both sh#8 (target- In the T3M-1 Cl-10 cell line, we found that the major tumor- ing DLX3) and sh#9 (targeting HRK) simultaneously (Fig. 7C). suppressive pathways downstream of ELF4 involve HRK and To exclude the possibility of off-target effects of the shRNAs, DLX3. HRK is a member of the BH3-only protein family and a we prepared DLX3 cDNAs resistant to sh#5 or sh#8 as well as potent inducer of apoptosis (27). HRK expression is frequently HRK cDNAs resistant to sh#7 or sh#9. As expected, the growth silenced by promoter methylation in multiple cancers, and of T3M-1 Cl-10 cells expressing DLX3 shRNA was markedly such silencing is associated with a low apoptotic index in suppressed by introduction of the corresponding shRNA- clinical specimens (28). We have now revealed that HRK is an resistant DLX3 cDNA (Fig. 7D, left). Similarly, the introduction effector for ELF4. Given that many BH3 mimetics are under of shRNA-resistant HRK cDNA inhibited the proliferation of development for clinical use, human cancers with dysfunc- cells expressing the corresponding HRK shRNA (right). Together, tional ELF4 are potential targets for such agents. Mutations in these results thus conﬁrmed that both DLX3 and HRK are essent- DLX3 are linked to tricho-dento-osseous syndrome, an auto- ial mediators of the tumor-suppressive activity of ELF4. somal-dominant disorder characterized by ectoderm dysplasia (29). DLX3 inhibits the growth of basal cells of the skin (24), and a high level of DLX3 expression predisposes tumor cells Discussion to the induction of apoptosis (30). Furthermore, similar to We have found that ELF4 (and other members of the ELF HRK, DLX3 is frequently silenced by promoter methylation subfamily of ETS proteins) functions primarily as a tumor sup- in childhood acute lymphoblastic leukemia harboring the

www.aacrjournals.org Cancer Res; 76(7) April 1, 2016 OF7

Ando et al.

Figure 6. ELF4 induces expression of HRK and DLX3. A, comparison of gene-expression proﬁles between T3M-1 Cl-10 cells expressing wild-type ELF4 or ELF4(L211M) by GSEA led to the isolation of "ETS motif," "growth," "G1 phase," and "apoptosis" gene sets. B, RT and real-time PCR analysis of DLX3, HRK,andGAPDH transcript abundance in T3M-1 Cl-10 cells expressing wild-type ELF4 or ELF4(L211M). Data are expressed relative to the corresponding value for mock-infected cells and are means SD for three independent experiments; , P ¼ 1.03 10–3; †, P ¼ 6.00 10–3 (Student t test). C, growth curves for T3M-1 Cl-10 cells infected with recombinant retroviruses generated from pMXS-ires-EGFP–based expression plasmids for ELF4, DLX3, or HRK. The EGFP-positive fraction was monitored by ﬂow cytometry at the indicated times after infection. Data, means SD from three independent experiments; , P ¼ 1.03 10–3; †, P ¼ 6.00 10–3 (Student t test).

MLL–AF4 fusion gene (31), suggestive of an antiproliferative stream of the last exon of HRK with an FDR of 0.069, we did not function for DLX3. detect any enhancer activity of this region in luciferase reporter Our ChIP-seq data did not reveal the speciﬁc binding of ELF4 to assays (data not shown). The regulation of HRK and DLX3 by ELF4 the DLX3 locus (Supplementary Fig. S12). Whereas a putative is, thus, likely to be indirect and dependent on intermediary binding region for ELF4 is present approximately 500 bp down- proteins. Given this complexity, it is not surprising that the

OF8 Cancer Res; 76(7) April 1, 2016 Cancer Research

ELF Proteins Function as a Tumor Suppressor

Figure 7. HRK and DLX3 mediate tumor-suppressive function of ELF4. A, growth curves for T3M-1 Cl-10 cells that had been infected with lentiviruses containing a puromycin resistance gene and encoding either a control shRNA (Ctrl-sh) or shRNAs targeted to DLX3 (sh#5 or sh#8), selected with puromycin, and then infected with retroviruses encoding EGFP and either ELF4 or ELF4(L211M). The EGFP-positive fraction was monitored by flow cytometry at the indicated times after retrovirus infection. Data, means SD from three independent experiments. B, growth curves for T3M-1 Cl-10 cells that had been infected with lentiviruses containing a puromycin resistance gene and encoding either a control shRNA (Ctrl-sh) or shRNAs targeted to HRK (sh#7 or sh#9), selected with puromycin, and then infected with retroviruses encoding EGFP and either ELF4 or ELF4(L211M). The EGFP-positive fraction was monitored by flow cytometry at the indicated times after retrovirus infection. Data, means SD from three independent experiments. C, growth curves for T3M-1 Cl-10 cells that had been infected simultaneously with a lentivirus containing a puromycin resistance gene and encoding an shRNA targeting DLX3 (sh#8) as well as with a lentivirus containing a neomycin resistance gene and encoding an shRNA targeting HRK (sh#9), selected with puromycin and neomycin, and then infected with retroviruses encoding EGFP and either ELF4 or ELF4(L211M). The EGFP-positive fraction was monitored by flow cytometry at the indicated times after retrovirus infection. Data, means SD from three independent experiments. D, growth curves for T3M-1 Cl-10 cells stably expressing shRNAs targeting DLX3 (sh#5 or sh#8, left) or HRK (sh#7 or sh#9, right) determined after infection with retroviruses encoding EGFP and containing corresponding shRNA-resistant cDNAs for DLX3 or HRK. Data, means SD from three independent experiments. www.aacrjournals.org Cancer Res; 76(7) April 1, 2016 OF9

Ando et al.

tumor-suppressive function of ELF4 is highly dependent on cell Authors' Contributions context. Conception and design: M. Ando, T. Yamasoba, H. Mano Sashida and colleagues (11) revealed that MDM2 is a Development of methodology: M. Ando, M. Kawazu, D. Koinuma, J. Koya direct target of ELF4. We thus compared the expression level Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Ando, M. Kawazu, J. Koya, K. Kataoka, of HRK, DLX3,orMDM2 in the TCGA RNA-seq dataset of K. Fukumura, A. Yamato, M. Soda, Y. Yamashita, T. Asakage, Y. Miyazaki 520 cases with head and neck squamous cell carcinoma with Analysis and interpretation of data (e.g., statistical analysis, biostatistics, or without nonsynonymous ELF4 mutations. As shown in computational analysis): M. Ando, T. Ueno, S.D. Nimer, H. Mano Supplementary Fig. S13, expression level of these three genes Writing, review, and/or revision of the manuscript: M. Ando, M. Kawazu, was not signiﬁcantly affected by ELF4 mutations, whereas a T. Asakage, Y. Miyazaki, K. Miyazono, T. Yamasoba, H. Mano tendency of DLX3 induction and MDM2 suppression was Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Ando, J. Koya, T. Yasuda, H. Yamaguchi, observed in the tumors carrying wild-type ELF4.Thus,reg- E. Sai, M. Kurokawa, T. Yamasoba ulation of the expression of these three genes is highly cell Study supervision: T. Yamasoba, H. Mano context dependent. Indeed, ELF4-mediated induction of DLX3 or HRK in cancer is cell type dependent (Supplemen- tary Fig. S14). Grant Support Collectively, our data indicate that the antitumor activity of This study was supported in part by a grant for Third-Term Comprehen- ELF4 (and of the other ELF subfamily members) is likely more sive Control Research for Cancer and by a grant for Research on Develop- widespread among human tissues than previously appreciated, ment of New Drugs from the Ministry of Health, Labor, and Welfare of with tumors arising from the loss of function of ELF4 (or of ELF1 Japan. The costs of publication of this article were defrayed in part by the payment of or ELF2) likely constituting a distinct subtype across multiple page charges. This article must therefore be hereby marked advertisement in organs. accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Disclosure of Potential Conﬂicts of Interest Received January 16, 2015; revised November 28, 2015; accepted December No potential conﬂicts of interest were disclosed. 21, 2015; published OnlineFirst February 26, 2016.

References 1. Lacorazza HD, Nimer SD. The emerging role of the myeloid Elf-1 like 13. Seki Y, Suico MA, Uto A, Hisatsune A, Shuto T, Isohama Y, et al. The ETS transcription factor in hematopoiesis. Blood Cells Mol Dis 2003;31: transcription factor MEF is a candidate tumor suppressor gene on the X 342–50. chromosome. Cancer Res 2002;62:6579–86. 2. Hollenhorst PC, McIntosh LP, Graves BJ. Genomic and biochemical 14. Okabe T, Sato N, Kondo Y, Asano S, Ohsawa N, Kosaka K, et al. insights into the speciﬁcity of ETS transcription factors. Annu Rev Biochem Establishment and characterization of a human cancer cell line that 2011;80:437–71. produces human colony-stimulating factor. Cancer Res 1978;38: 3. Kumar-Sinha C, Tomlins SA, Chinnaiyan AM. Recurrent gene fusions in 3910–7. prostate cancer. Nat Rev Cancer 2008;8:497–511. 15. Rasheed S, Nelson-Rees WA, Toth EM, Arnstein P, Gardner MB. Charac- 4. King JC, Xu J, Wongvipat J, Hieronymus H, Carver BS, Leung DH, et al. terization of a newly derived human sarcoma cell line (HT-1080). Cancer Cooperativity of TMPRSS2-ERG with PI3-kinase pathway activation in 1974;33:1027–33. prostate oncogenesis. Nat Genet 2009;41:524–6. 16. Kuramoto H, Nishida M, Morisawa T, Hamano M, Hata H, Kato Y, et al. 5. Lessnick SL, Ladanyi M. Molecular pathogenesis of Ewing sarcoma: new Establishment and characterization of human endometrial cancer cell therapeutic and transcriptional targets. Annu Rev Pathol 2012;7: lines. Ann N Y Acad Sci 1991;622:402–21. 145–59. 17. Soule HD, Maloney TM, Wolman SR, Peterson WDJr, Brenz R, McGrath 6. Lacorazza HD, Miyazaki Y, Di Cristofano A, Deblasio A, Hedvat C, Zhang J, CM, et al. Isolation and characterization of a spontaneously immor- et al. The ETS protein MEF plays a critical role in perforin gene expression talized human breast epithelial cell line, MCF-10. Cancer Res 1990; and the development of natural killer and NK-T cells. Immunity 2002; 50:6075–86. 17:437–49. 18. Onishi M, Kinoshita S, Morikawa Y, Shibuya A, Phillips J, Lanier LL, et al. 7. Lacorazza HD, Yamada T, Liu Y, Miyata Y, Sivina M, Nunes J, et al. The Applications of retrovirus-mediated expression cloning. Exp Hematol transcription factor MEF/ELF4 regulates the quiescence of primitive 1996;24:324–9. hematopoietic cells. Cancer Cell 2006;9:175–87. 19. Ando K, Tsushima H, Matsuo E, Horio K, Tominaga-Sato S, Imanishi D, 8. Miyazaki Y, Boccuni P, Mao S, Zhang J, Erdjument-Bromage H, Tempst P, et al. Mutations in the nucleolar phosphoprotein, nucleophosmin, pro- et al. Cyclin A-dependent phosphorylation of the ETS-related protein, MEF, mote the expression of the oncogenic transcription factor MEF/ELF4 in restricts its activity to the G1 phase of the cell cycle. J Biol Chem leukemia cells and potentiates transformation. J Biol Chem 2013;288: 2001;276:40528–36. 9457–67. 9. Fukushima T, Miyazaki Y, Tsushima H, Tsutsumi C, Taguchi J, Yoshida S, 20. Ando M, Kawazu M, Ueno T, Fukumura K, Yamato A, Soda M, et al. Cancer- et al. The level of MEF but not ELF-1 correlates with FAB subtype of acute associated missense mutations of caspase-8 activate nuclear factor-kappaB myeloid leukemia and is low in good prognosis cases. Leuk Res 2003;27: signaling. Cancer Sci 2013;104:1002–8. 387–92. 21. Totoki Y, Tatsuno K, Yamamoto S, Arai Y, Hosoda F, Ishikawa S, et al. High- 10. Yao JJ, Liu Y, Lacorazza HD, Soslow RA, Scandura JM, Nimer SD, et al. resolution characterization of a hepatocellular carcinoma genome. Nat Tumor promoting properties of the ETS protein MEF in ovarian cancer. Genet 2011;43:464–9. Oncogene 2007;26:4032–7. 22. Kawazu M, Ueno T, Kontani K, Ogita Y, Ando M, Fukumura K, et al. 11. Sashida G, Liu Y, Elf S, Miyata Y, Ohyashiki K, Izumi M, et al. ELF4/MEF Transforming mutations of RAC guanosine triphosphatases in human activates MDM2 expression and blocks oncogene-induced p16 activation cancers. Proc Natl Acad Sci U S A 2013;110:3029–34. to promote transformation. Mol Cell Biol 2009;29:3687–99. 23. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, 12. Bazzoli E, Pulvirenti T, Oberstadt MC, Perna F, Wee B, Schultz N, et al. MEF et al. Gene set enrichment analysis: a knowledge-based approach for Promotes Stemness in the Pathogenesis of Gliomas. Cell Stem Cell interpreting genome-wide expression proﬁles. Proc Natl Acad Sci U S A 2012;11:836–44. 2005;102:15545–50.

OF10 Cancer Res; 76(7) April 1, 2016 Cancer Research

ELF Proteins Function as a Tumor Suppressor

24. Morasso MI, Markova NG, Sargent TD. Regulation of epidermal differ- 28. Nakamura M, Shimada K, Konishi N. The role of HRK gene in human entiation by a Distal-less homeodomain gene. J Cell Biol 1996;135: cancer. Oncogene 2008;27Suppl 1:S105–S113. 1879–87. 29. Price JA, Bowden DW, Wright JT, Pettenati MJ, Hart TC. Identiﬁcation of a 25. Inohara N, Ding L, Chen S, Nunez G. harakiri, a novel regulator of cell mutation in DLX3 associated with tricho-dento-osseous (TDO) syndrome. death, encodes a protein that activates apoptosis and interacts selectively Hum Mol Genet 1998;7:563–9. with survival-promoting proteins Bcl-2 and Bcl-X(L). EMBO J 1997;16: 30. Ferrari N, Paleari L, Palmisano GL, Tammaro P, Levi G, Albini A, et al. 1686–94. Induction of apoptosis by fenretinide in tumor cell lines correlates with 26. Moore N, Lyle S. Quiescent, slow-cycling stem cell populations in cancer: a DLX2, DLX3, and DLX4 gene expression. Oncol Rep 2003;10:973–7. review of the evidence and discussion of signiﬁcance. J Oncol 2011;2011: 31. Campo Dell'Orto M, Banelli B, Giarin E, Accordi B, Trentin L, Romani M, e396076. et al. Downregulation of DLX3 expression in MLL-AF4 childhood lym- 27. Willis SN, Adams JM. Life in the balance: how BH3-only proteins induce phoblastic leukemias is mediated by promoter region hypermethylation. apoptosis. Curr Opin Cell Biol 2005;17:617–25. Oncol Rep 2007;18:417–23.

www.aacrjournals.org Cancer Res; 76(7) April 1, 2016 OF11

Mutational Landscape and Antiproliferative Functions of ELF Transcription Factors in Human Cancer

Mizuo Ando, Masahito Kawazu, Toshihide Ueno, et al.

Cancer Res Published OnlineFirst February 26, 2016.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-14-3816

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2021/03/04/0008-5472.CAN-14-3816.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/2016/03/22/0008-5472.CAN-14-3816. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.