Oncogene (2011) 30, 1923–1935 & 2011 Macmillan Publishers Limited All rights reserved 0950-9232/11 www.nature.com/onc ORIGINAL ARTICLE A novel isoform of the 8p22 tumor suppressor DLC1 suppresses tumor growth and is frequently silenced in multiple common tumors

JSW Low1,2, Q Tao1,3,4,KMNg4, HK Goh1,2, X-S Shu4, WL Woo1, RF Ambinder1,2,3, G Srivastava5, M Shamay1,3, ATC Chan4, NC Popescu6 and W-S Hsieh1,2,3

1Division of Biomedical Sciences, Johns Hopkins Singapore, Singapore; 2Cancer Science Institute of Singapore, National University of Singapore, Singapore; 3Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins School of Medicine, Baltimore, MD, USA; 4Cancer Laboratory, State Key Laboratory of Oncology in South China, Sir YK Pao Center for Cancer, Department of Clinical Oncology, Hong Kong Cancer Institute and Li Ka Shing Institute of Health Sciences, The Chinese University of Hong Kong, Hong Kong; 5Department of Pathology, University of Hong Kong, Hong Kong and 6Laboratory of Experimental Carcinogenesis, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA

The critical 8p22 tumor suppressor deleted in liver cancer modulating the complex activities of DLC1 during 1(DLC1) is frequently inactivated by aberrant CpG carcinogenesis. methylation and/or genetic deletion and implicated in Oncogene (2011) 30, 1923–1935; doi:10.1038/onc.2010.576; tumorigeneses of multiple tumor types. Here, we report published online 10 January 2011 the identification and characterization of its new isoform, DLC1 isoform 4 (DLC1-i4). This novel isoform encodes Keywords: DLC1; RhoGAP; methylation; tumor sup- an 1125-aa (amino acid) with distinct N-terminus pressor gene; carcinoma; p53 as compared with other known DLC1 isoforms. Similar to other isoforms, DLC1-i4 is expressed ubiquitously in normal tissues and immortalized normal epithelial cells, suggesting a role as a major DLC1 transcript. However, Introduction differential expression of the four DLC1 isoforms is found in tumor cell lines: Isoform 1 (longest) and 3 (short thus Carcinogenesis involves multiple genetic/epigenetic probably nonfunctional) share a promoter and are silenced events including the activation of oncogenes and in almost all cancer and immortalized cell lines, whereas inactivation of tumor suppressor (TSGs) (Fearon isoform 2 and 4 utilize different promoters and are and Vogelstein, 1990; Knudson, 2001). In addition to frequently downregulated. DLC1-i4 is significantly down- genetic mutations, a growing body of evidence has regulated in multiple carcinoma cell lines, including 2/4 shown that TSG inactivation occurs frequently through nasopharyngeal, 8/16 (50%) esophageal, 4/16 (25%) promoter CpG methylation (Jones and Laird, 1999; gastric, 6/9 (67%) breast, 3/4 colorectal, 4/4 cervical and Herman and Baylin, 2003), resulting in TSG silencing 2/8(25%) lung carcinoma cell lines. The functional and subsequent loss of function, thereby contributing DLC1-i4 promoter is within a CpG island and is activated to neoplastic transformation (Jones and Baylin, 2007). by wild-type p53. CpG methylation of the DLC1-i4 Aberrant CpG methylation of TSG-associated CpG promoter is associated with its silencing in tumor cells and islands is a characteristic hallmark of tumor genomic was detected in 38–100% of multiple primary tumors. DNA, and examples of methylation silenced TSGs Treatment with 5-aza-20-deoxycytidine or genetic double include p16, MLH1, VHL, RAR-beta, SOCS1, knockout of DNMT1 and DNMT3B led to demethylation PCDH10, RASAL and so on (Herman and Baylin, of the promoter and reactivation of its expression, 2003; Jones and Baylin, 2007). indicating a predominantly epigenetic mechanism of Deleted in liver cancer 1 (DLC1) gene (also known as silencing. Ectopic expression of DLC1-i4 in silenced ARHGAP7, STARD12, HP and p122-RhoGAP) was tumor cells strongly inhibited their growth and colony isolated by representational difference analysis (Yuan formation. Thus, we identified a new isoform of DLC1 et al., 1998) from a sample of human hepatocellular with tumor suppressive function. The differential expres- carcinoma (HCC), and proposed as a candidate TSG sion of various DLC1 isoforms suggests interplay in because of its frequent deletion in hepatocellular carci- noma tumors and cell lines. Located at 8p22, a site of recurrent deletion in breast, lung and prostate cancers (Yuan et al., 1998), DLC1 encodes a protein of 1091-aa Correspondence: Professor Q Tao, Rm 315, Cancer Center, Depart- ment of Clinical Oncology, PWH, Chinese University of Hong Kong, (amino acid) with extensive homology (86%) to the rat Shatin, Hong Kong. p122-RhoGAP, a GTPase-activating protein (GAP) E-mail: [email protected] or Dr W-S Hsieh, Cancer Science specific for RhoA and Cdc42 that are involved in the Institute of Singapore, National University of Singapore, Centre for regulation of cellular cytoskeleton organization and Life Sciences Level 2, 28 Medical Drive, 117456 Singapore. E-mail: [email protected] other functions (Homma and Emori, 1995; Yuan et al., Received 1 June 2010; revised 15 November 2010; accepted 15 November 1998; Sekimata et al., 1999; Bernards, 2003; Wong et al., 2010; published online 10 January 2011 2003; Durkin et al., 2007b). Subsequent studies identified Novel isoform of DLC1 silenced in tumors JSW Low et al 1924 DLC1 as an epigenetically silenced gene with tumor non-small-cell lung carcinoma cells . In addition, DLC1 suppressive and metastatic inhibitory functions in knockdown in the background of c-myc overexpression breast, liver, colon, lung, stomach and brain cancers promotes the formation of liver tumors in a murine (Kim et al., 2003; Yuan et al., 2003, 2004; Zhou et al., model (Xue et al., 2008). Taken together, these data 2004; Pang et al., 2005). Recently, using suppression suggest that DLC1 function both as a cytoplasmic and substractive hybridization, our group identified DLC1 nuclear tumor suppressor, and highlight the importance as an epigenetically silenced gene in nasopharyngeal of DLC1 in cancer development. (NPC), esophageal and cervical carcinomas (Seng et al., During our study of DLC1, we discovered a new 2007), thus, further supporting the view that DLC1 is a isoform through 50-rapid amplification of complemen- bona fide TSG. tary DNA (cDNA) ends (50-RACE). Here, we report the DLC1 is a member of the human RhoGAP family. characterization of this novel isoform (designated RhoGAP share a conserved 150–200-aa GAP DLC1-isoform 4 (DLC1-i4)), its mRNA expression domain that contains the catalytic activity to convert the and epigenetic alterations in multiple carcinomas. We active GTP-bound Rho proteins to the inactive GDP- also describe the response of the DLC1-i4 promoter to bound state (Bernards, 2003; Moon and Zheng, 2003; p53 and the expression of the two other DLC1 isoforms Tcherkezian and Lamarche-Vane, 2007; Durkin et al., (-i1 and -i3) in human cells. Lastly, we demonstrate that 2007b). Loss of RhoGAP activity leads to aberrant DLC1-i4 and -i1 suppress tumor cell colony formation. activation of GTP-bound Rho proteins, which are involved in the regulation of the cell cycle, adhesion, morphogenesis, polarity and migration; and dysregula- tion of GTP-bound Rho proteins have been implicated Results in tumorigenesis (Martin, 2003; Moon and Zheng, 2003). Members of the DLC family include DLC1, 50-RNA ligase-mediated rapid amplification of cDNA DLC2, and DLC3, and their domain structures share an ends (50 RLM-RACE) identifies a novel isoform of DLC1 N-terminal sterile a motif domain, a serine-rich domain, (DLC1-i4) and expression profiles of all four isoforms a RhoGAP domain and a C-terminal steroidogenic in normal tissues acute regulatory protein-related lipid transfer domain To identify novel isoform(s) of DLC1,50-RLM-RACE (Yuan et al., 1998; Ponting and Aravind, 1999; Ching was performed on NP69 and liver total RNA using et al., 2003; Durkin et al., 2007a). Recently, Yuan et al. antisense primers targeting the DLC1 common 9 (2007) showed that DLC1 harbors a functional bipartite and 12 (Table 1 and Figure 1a). Two PCR bands of nuclear localization signal which works together with 205 bp and 199 bp were obtained and sequenced. the RhoGAP and serine-rich domain to mediate DLC1 Sequence analysis identified two transcriptional start protein nuclear transfer and subsequent in sites only 6 bp apart and overlapped with a previously

Table 1 Sequences of primers used in this study PCR Primer name Targeted exona Sequence (50–30) Size Annealing temp (1C) Cycles

50-RLM RACE TJR 12 AGTCCATTTGCCACTGATGG — 55 35 E10R2 9 TAAAGCTGTGCATACTGGGG E10R 9 CCGTAGCCAATCACAAGCTT

RT–PCR DLC1-i4 V4F 6 AACACTAGAGACAGACGGCT 294 bp 55 35 TJR 12 AGTCCATTTGCCACTGATGG DLC1-i1 NewF 4 ACTCCAGTAGCCAATTCTGG 342 bp 55 35 TJR 12 AGTCCATTTGCCACTGATGG DLC1-i2 V2F 8 GGACACCATGATCCTAACAC 289 bp 55 35 TJR 12 AGTCCATTTGCCACTGATGG DLC1-i3 NewF 4 ACTCCAGTAGCCAATTCTGG 226 bp 55 35 V3R 7 CGGCCTAGGTGATGTTTTCT GAPDH GAPDH55 — ATCTCTGCCCCCCTGCTGA 304 bp 60 25 GAPDH33 — GATGACCTTGCCCACAGCCT

MSP for the DLC1-i4 promoter DLC1-i4 M1 — TAGGCGATAGTTTGCGGTC 143 bp 58 40 M2 — AAAAAAACTCGCAAAAAACGCG U1 — GGTAGGTGATAGTTTGTGGTT 147 bp 58 40 U2 — AAAAAAAAACTCACAAAAAACACA

BGS for the DLC1-i4 promoter DLC1-i4 BGS1 — GAAAGTTAAAGATAAGGTTATTTG 564 bp 58 40 BGS2 — CCAAATAACATCCAAAACTCTAA

Abbreviations: DLC, deleted in liver cancer; MSP, methylation-specific PCR; RT–PCR, reverse transcription PCR. aExon number according to Figure 1a.

Oncogene Novel isoform of DLC1 silenced in tumors JSW Low et al 1925

Figure 1 mRNA expression, and protein structure of the various DLC1 isoforms and the CpG-rich DLC1-i4 promoter. (a) Exon structure of the human DLC1 gene (not drawn to scale). The DLC1 gene contains internal promoters and expresses multiple isoforms. Boxes represent exons and are numbered to correspond to coding exons of the various DLC1 isoforms with the 50-most exon denoted as exon 1 (exon 1 of isoforms 1 and 3). Closed boxes indicate exons common in all DLC1 isoforms except isoform 3. Bent arrows indicate the transcriptional start sites of the various DLC1 isoforms. Black, dotted, gray and dashed lines indicate the splicing structure for isoform 1, 2, 3 and 4, respectively. Primer locations are shown by short black arrows. (b) The CpG island promoter of DLC1-i4. Sequences of the DLC1-i4 promoter, first exon and the locations of the BGS and MSP primers are shown. The full first exon sequences as defined in GenBank are shown in caps. Transcriptional start sites as determined by 50-RLM RACE in liver and the SV40 T-antigen immortalized nasopharyngeal epithelial cell line NP69 are marked with arrowheads. The translational start site (ATG) is boxed and individual CpG dinucleotides are shown in bold caps. MSP primer locations (DLC1-i4-M1/M2, U1/U2) are double underlined, whereas BGS primer locations (DLC1-i4-BGS1/2) are underlined with thick lines. (c) DLC1 isoform protein structures (not drawn to scale). The domain organization of all the DLC1 isoforms are shown with the sterile a motif, serine-rich domain, NLS, RhoGAP and START domain highlighted. The putative mitochondrial targeting sequence of DLC1-i4 is highlighted with a gray box. Name and amino- acid lengths of each isoform are shown on the left and right respectively. Curved bars denote coding exons. (d) Expression profile of all DLC1 isoforms in normal human adult and fetal tissues as assessed by semi-quantitative RT–PCR. The house-keeping gene GAPDH was used as a control. Underlined tissues were examined for DLC1-i4 silencing in their corresponding tumor cell lines. S. muscle, skeletal muscle. deposited partial cDNA sequence (AK025544) in NCBI DA853751, DA409883, DA403097, DA401187, DA231722, database (Figure 1b). Subsequent BLAST searches DA328255, DA334650, BP287999, BP285601, CN388450, identified 15 expressed sequence tags (Genbank accession: BX474714, BX474696, BP238467, CN388449 and

Oncogene Novel isoform of DLC1 silenced in tumors JSW Low et al 1926 BP287477) with their transcriptional start sites close to of 1087 bp spanning the promoter, exon 1 and part of the two established by us. intron1 was found (Takai and Jones, 2002): GC content, We then cloned and sequenced the full-length 3,378 bp 57%; observed/expected CpG ratio, 0.737; with a high open reading frame (ORF) of DLC1-i4 from human concentration of 55 CpG sites in a 564-bp region liver RNA and identified a 1,125-aa open reading (Figures 2a and 4a). Potential transcription factor (TF)- frame precisely matched with DLC1-i1 and -i2 except binding sites predicted by three TF search programs for a different short N-terminal region (Figure 1c and (TFSEARCH, MotifSearch and MatInspector) suggest Supplementary Figure 1). Domain analysis showed the that the DLC1-i4 promoter might be regulated by TP53, presence of domains typical of all the DLC members E2F, STAT and heat-shock factor (Figure 2a). (Bernards, 2003; Ching et al., 2003; Durkin et al., 2007a) Next, we evaluated whether the putative DLC1-i4 (Figure 1c). In addition, a putative mitochondrial tar- promoter was transcriptionally functional by cloning geting sequence was identified within the first 30 aa of three fragments covering the promoter separately into a DLC1-i4 using five online bioinformatics program, promoter-less luciferase construct and transfected cell- iPSORT, MitoProt II, Predotar, Mitopred and TargetP lines with/without DLC1-i4 expression (HEK293, CNE2 1.1 server (Claros and Vincens, 1996; Bannai et al., 2002; and HK1). All three constructs could drive transcrip- Small et al., 2004; Guda et al., 2004a, b; Emanuelsson tion, with the shortest fragment (À480/ þ 140) showing et al., 2007). This sequence is coded by the unique first the highest activity and the longest (À1840/ þ 140) exon of DLC1-i4 and is absent in other isoforms showing the lowest activity (Figure 2b), showing that (Figure 1c and Supplementary Figure 1). the DLC1-i4 promoter is functional. Genbank and Ensembl database searches querying Bioinformatics analysis predicts that the DLC1-i4 the predicted aa sequence encoded by the new first exon promoter harbors five putative p53-binding sites found both full-length and partial mRNA transcripts (Figure 2a). To see whether p53 regulates this promoter, coding for DLC1-i4 orthologs in mouse, rabbit, Rhesus the promoter construct (pGL2-DLC1i4-PF1/R(À1840/ macaque, cattle and elephant (Supplementary Figure 2). þ 140)) with or without a p53 expression vector was This cDNA was thus designated DLC1-i4 in accordance transfected into HEK293, CNE2 and the p53-null with the nomenclature used by GenBank. Due to a HCT116/p53KO cells, and luciferase activity was assayed. sequence variation in the previous cDNA sequence As shown in Figure 2c, p53 induced upregulation of AK025544, in silico translation of this cDNA would luciferase activity ranging from 3- to 38-fold, with the yield a truncated protein of 804-aa up to the end of the highest activity observed in HCT116/p53KO cells RhoGAP domain (Supplementary Figure 1). With most (Figure 2c). Co-transfection of DLC1-i4 promoter con- of the available expression data on DLC1 isoform 2 struct with mutant p53 expression plasmids (G245C and only, we investigated the expression of all DLC1 R248W) into HCT116-p53KO, showed that only wild- isoforms including DLC1-i4 in human normal adult type p53 could upregulate DLC1-i4 and control p21 and fetal tissues by reverse transcription PCR (RT– promoter activities (10-fold and 2.5-fold, respectively) PCR) using exon-specific primers to distinguish each (Figure 2d). As the p53 mutants are defective in DNA- isoform. All isoforms were detected in all normal tissues binding (Zhou et al., 1999; Bullock and Fersht, 2001), this examined, with isoforms 2 and 4 showing stronger finding suggests that the ability of wild-type p53 to expression, and isoform 3 the weakest (Figure 1d). upregulate DLC1-i4 promoter activity is attributed to its DNA binding ability. In addition, both gene promoters The functional DLC1-i4 promoter is upregulated by p53 showed upregulation of activity in response to increasing and stress p53 concentration in a statistically significant (Po0.001– We used bioinformatic screening to analyze B3Kb 0.0015), dose-dependent manner up to 50 ng (Figure 2e) upstream of the putative DLC1-i4 promoter. A CpG-island before dropping off at higher p53 concentrations.

Figure 2 Functional localization of the DLC1-i4 promoter. (a) Structure of the DLC1-i4 50-promoter region (Not drawn to scale). The enlarged diagram of the 50-promoter region of DLC1-i4 is shown. Horizontal line represents the genomic sequence. Gray boxes with numbers below indicate exons. The transcriptional start site is labeled with an arrow. Putative binding sites of TP53, E2F, STATx and HSFs predicted by three different bioinformatics programs are labeled. Thick black bar shows the location of the CpG island. (b) Localization of the functional DLC1-i4 promoter using promoter luciferase assays. Promoter activities of three constructs containing different regions of the putative DLC1-i4 promoter relative to the promoter-less control vector using Luciferase assays are presented. Data shown are from three independent Luciferase assays (±s.d) in three cell lines (HEK293, CNE2 and HK1). Every one of the three different constructs of the DLC1-i4 promoter can drive the transcription of the target gene in all cell lines tested. (c) DLC1-i4 promoter is p53-responsive. The schematic diagram shows five putative p53-binding sites (BS) predicted by bioinformatics on the DLC1-i4 promoter. DLC1-i4 promoter construct were transfected into HEK293, CNE2 or HCT116/p53KO cell lines with (black bars) or without (shaded bars) TP53 expression construct, and assayed for luciferase activity after 48 h. Results of luciferase assays with respect to control pGL2 plasmid (open bars) in triplicates are shown in the histograms. The respective promoter constructs are shown on the left. Values on all histograms shown are the mean of three independent assays (±s.d). (d) Wild-type TP53, but not mutant TP53 is able to transactivate DLC1-i4 promoter activity. Wild-type or mutant p53 containing expression vectors were cotransfected with either DLC1-i4 promoter construct (black bars) or pGL2-p21P (open bars) into the p53-null HCT116/p53KO cell line for 48 h before luciferase activity was measured. Histograms show DLC1-i4 promoter activity upregulated by wild-type, but not mutant p53. The p21 promoter construct was used in parallel as a control for p53 transactivation. (e) TP53 upregulates DLC1-i4 promoter activity in a dose-dependent manner. Upregulation of DLC1-i4 promoter activity was observed with increasing amount of wild-type p53-expression vector. Increasing amount of the wild-type p53-expression vector (1 ng to 1 mg) was cotransfected together with either DLC1-i4 promoter construct (black bars) or pGL2-p21P (open bars) into HCT116/p53KO cell line. A two-tailed paired Student’s t-test was performed to determine the statistical differences between the increasing p53 concentrations. Enh, enhancer.

Oncogene Novel isoform of DLC1 silenced in tumors JSW Low et al 1927 We also transfected HCT116 or HCT116/p53KO cells also silenced in tumors. The expression of DLC1-i4, with either the DLC1-i4 or the p21-promoter construct together with -i1, -i2 and -i3, were examined by semi- for 48 h before irradiation with 10 J/m2 of UV light and quantitative RT–PCR in a panel of carcinoma cell lines, assayed for luciferase activity after 2–6 h incubation. and we correlated DLC1-i4 expression with its promoter Upregulation of either DLC1-i4 or p21 promoter CpG methylation using methylation-specific PCR activities was observed (Supplementary Figure 3), indi- (MSP) and bisulfite genomic sequencing (BGS). Results cating that the DLC1-i4 promoter is stress-responsive. showed weak or no expression of DLC1-i4 transcripts in multiple carcinoma cell lines, including 2/4 NPC, 8/ Promoter CpG methylation silences DLC1-i4 expression 16(50%) esophageal, 4/16(25%) gastric, 6/9(67%) in multiple carcinoma cell lines breast, 3/4 colorectal, 4/4 cervical, 2/8(25%) lung As the functional DLC1-i4 promoter contains a CpG carcinomas, as well as the various other cell lines tested island, we assessed whether like DLC1-i2, DLC1-i4 is (Figure 3 and Supplementary Figure 4). Meanwhile,

Oncogene Novel isoform of DLC1 silenced in tumors JSW Low et al 1928

Figure 3 Expression analysis of all DLC1 isoforms and DLC1-i4 promoter methylation in multiple tumor cell lines and normal control. NPC, esophageal, gastric (GsCa), breast, colon, cervical, lung and cell lines; immortalized normal cell lines (NP69, HEK293, RHEK1, NE1 and NE3), normal epithelial cell lines (HMEC and HMEpC) and normal placenta tissue as controls. M, methylated; U, unmethylated. Fragment size indicated on right of panels.

Table 2 Frequencies of DLC1-i4 methylation in tumors and normal tissues studied Origin Cell line Tissue

Tumor Nasopharyngeal carcinoma 100% (4/4) 98% (46 þ 2 weak/49) Esophageal carcinoma 38% (6/16) 47% (14/30) Gastric carcinoma 12.5% (2/16) 82% (9/11) Breast carcinoma 22% (2/9) 77% (8 þ 2 weak/13) Colorectal carcinoma 50% (2/4) 100% (11/11) Cervical carcinoma 75% (3/4) Lung carcinoma 13% (1/8) Hepatocellular carcinoma (HCC) 0% (0/12) 38% (14/37) Renal cell carcinoma 0% (0/9) Prostate carcinoma 33% (1/3) Ovarian carcinoma 0% (0/2) Burkitt lymphoma 100% (5/5) Nasal NK/T-cell lymphoma 100% (2/2) Hodgkin lymphoma 83% (5/6) Immortalized normal epithelial cell lines NP69, 293-HEK, RHEK1, NE1, NE3 0% (0/5) Normal epithelial cell lines HMEC and HMEpC 0% (0/2) Normal tissues Normal nasopharynx tissues 33% (3 weak/9) Normal PBMCs 0% (0/5)

strong methylated alleles were consistently detected in 55 CpG sites in the DLC1-i4 promoter (Figure 4a) cell lines without DLC1-i4 expression by MSP (Table 2, validated the MSP data (Figure 4b). Collectively these Figure 3 and Supplementary Figure 4). In contrast, results show that aberrant promoter CpG methylation is DLC1-i4 was readily detected in all normal (HMEC associated with DLC1-i4 silencing, and like DLC1-i2, is and HMEpC) and normal immortalized cell lines a frequent event in multiple tumors. (NP69, HEK293, RHEK1, NE1 and NE3) tested, with DLC1-i1 and -i3 share a common promoter and their no methylated promoter alleles detected (Table 2 and expression were silenced in almost all the tumor cell lines Figure 3). High-resolution BGS analysis performed for tested (Figure 3). Analysis on their shared CpG poor

Oncogene Novel isoform of DLC1 silenced in tumors JSW Low et al 1929

Figure 4 DLC1-i4 is silenced by DNA methylation in carcinoma cell lines and demethylation by pharmacological or genetic approach could induce DLC1-i4 reexpression. (a) Schematic diagram of the location of 59 CpG dinucleotides along an 800 bp fragment encompassing exon 1 on the 50-promoter region of DLC1-i4. Vertical line represents one CpG site. Locations of CpG sites 1–55 examined by high-resolution methylation analysis are demarcated by the horizontal bracket. Arrow denotes the transcriptional start site and box represents the first exon of DLC1-i4. Gray block arrows indicates location of MSP primers. (b) High resolution mapping of individual CpG sites on the DLC1-i4 promoter in tumor and normal cell lines as assessed by BGS. A 564 bp fragment of the core promoter containing 55 CpG sites was analyzed using dideoxynucleotide sequencing. CpG site is shown on top with numbered ballooned arrows. Gray block arrows denote MSP primers location. Methylation status of individual CpG site is expressed as percentage methylation calculated from 5–12 sequenced colonies. Number of clones sequenced, MSP and RT–PCR expression results are shown on the right. (c) Demethylation by Aza restores DLC1-i4 expression in Aza-treated NPC cell lines C666-1 and HK1 (day 3 (72 h) or day 6 (144 h)) with concomitant demethylation of its promoter. (d) Reexpression of DLC1-i4 can only be mediated by Aza alone (72 h) or combined with TSA (96 h). (e) Genetic double knockout of both DNMT1 and DNMT3B strongly induces DLC1-i4 promoter demethylation and mRNA expression. (f) High-resolution methylation mapping of 55 CpG sites in a 564 bp region of the DLC1-i4 promoter by BGS confirmed the genetic demethylation in HCT116-DKO. Horizontal line represents the DLC1-i4 promoter region with short vertical lines representing each CpG site analyzed. Percentage methylation was established as total percentage of methylated cytosines from 5–12 randomly sequenced colonies. Numbers of clones sequenced, MSP and RT–PCR expression results are shown on the right. The locations of MSP primers are indicated by thick gray arrows. M, methylated; U, unmethylated. þ , expressed; À, silenced.

promoter through promoter luciferase assay showed However, unlike DLC1-i2, DLC1-i4 is not expressed that it is a transcriptionally functional promoter in normal peripheral blood mononuclear cells (PBMCs) (Supplementary Figure 5A and 5B). Lack of TFs to with no methylated alleles detected, but is expressed in drive the expression of DLC1-i1, -i3 and -i4 was normal lymph node (Figure 1d), indicating that mecha- unlikely, as all promoter constructs were active in nisms other than DNA methylation are involved in HK1, a cell line not expressing any DLC1 isoforms silencing of DLC1-i4 promoter in normal PBMCs (Figure 3). Further work is required to elucidate the (Supplementary Figure 4B). This raised the possibility mechanism of the silencing of DLC1-i1 and -i3. that different DLC1 isoforms might have functional

Oncogene Novel isoform of DLC1 silenced in tumors JSW Low et al 1930

Figure 5 Promoter methylation of DLC1-i4 in multiple primary carcinomas. Representative MSP data showing methylation of DLC1-i4 in multiple primary tumors but not in normal nasopharynx and matched tissues. (a) Normal nasopharynx tissues as controls and primary NPC tumors from Asians. (b) Expression of DLC1-i4 and MSP in a subset of primary NPC tumors. (c) Methylation in multiple primary carcinomas. (d) DLC1-i4 expression in matched breast primary tumor and normal tissue pair. M, methylated; U, unmethylated; N, paired surgical marginal tissues; T, tumor tissues.

specialization in different subsets of cells. Interestingly, Promoter methylation and expression of DLC1-i4 DLC1-i4 is not expressed in any of the lymphoma cell in multiple primary carcinomas lines tested (Supplementary Figure 4B), correlating with DLC1-i4 methylation was next examined in multiple methylation of the DLC1-i4 promoter. primary tumors, including NPC, esophageal, gastric, hepatocellular, breast and colorectal carcinoma. We first examined DLC1-i4 methylation in nine normal naso- Pharmacologic or genetic demethylation restores pharynx tissues and found weak methylation in only DLC1-i4 expression three (33%) (Figure 5a and Table 2), meanwhile, To determine whether CpG methylation is directly methylation was detected in 48/49 (98%, with two implicated in the silencing of DLC1-i4, two cell weakly methylated) endemic NPC tumors from Asian lines (C666-1 and HK1) with methylated and silenced Chinese (Figure 5a and Table 2), and in 14/30 (47%) DLC1-i4 were treated with DNA methyltransferase esophageal carcinomas but less frequently in the paired 0 inhibitor 5-aza-2 -deoxycytidine (Aza). DLC1-i4 ex- non-tumor tissues (Figure 5c and Table 2). In addition, pression was restored after Aza treatment, although at DLC1-i4 methylation was detected in 9/11 (82%) different levels, with a concomitant increase in un- primary gastric, 10/13 (77%, with two weakly methy- methylated promoter allele detected by MSP (Figure 4c). lated) breast, 14/37 (38%) HCC and 11/11 (100%) The same phenomenon was observed when HCT116, colorectal carcinomas (Figure 5c and Table 2). RT–PCR EC109 and HK1 were treated with Aza alone, or and MSP analysis of primary NPC showed down- combination with a inhibitor regulation of DLC1-i4 expression in 5/9 specimens but trichostatin A (TSA) (Figure 4d). TSA treatment alone weak correlation of DLC1-i4 expression with methyla- is insufficient to induce DLC1-i4 expression, implying a tion of DLC1-i4 promoter (Figure 5b). RT–PCR critical role for DNA methylation. Moreover, for analysis of matched breast primary tumor and adjacent HCT116 with genetic double knockout of both DNMT1 normal tissue showed DLC1-i4 downregulation in the and DNMT3B genes (HCT116-DKO cell line) (Rhee tumor as compared with the matched normal tissue et al., 2002), robust induction of DLC1-i4 RNA was (Figure 5d). These results indicate that DLC1-i4 accompanied by increased unmethylated DLC1-i4 al- methylation occurs frequently in tumors and affects leles (Figure 4e). Weak DLC1-i4 expression with some DLC1-i4 expression. demethylated alleles were also observed in DNMT1- knockout HCT116 cells (HCT116-1KO), but knocking- out DNMT3B seems to have little effect on DLC1-i4 DLC1-i4 suppresses tumor cell colony formation reexpression, even though weak demethylated alleles were To examine the effects of DLC1-i4 on tumor cell also detected. Subsequent high-resolution BGS methy- growth, we transfected a DLC1-i4 expression construct lation analysis confirmed the substantial demethylation of into HCT116, HK1 and KYSE510 cell lines, all of which DLC1-i4 promoter in HCT116-DKO cells, indicating that have methylated and silenced endogenous DLC1-i4, and the DLC1-i4 promoter methylation is cooperatively the clonogenicity of transfected cells was assessed by maintained by DNMT1 and DNMT3B (Figure 4f). monolayer culture. Significant reduction in colony

Oncogene Novel isoform of DLC1 silenced in tumors JSW Low et al 1931

Figure 6 Ectopic expression of DLC1-i4 suppresses clonogenicity of carcinoma cells. (a) Colorectal carcinoma (HCT116) and esophageal carcinoma (KYSE510) cell lines were transfected with pcDNA3.1( þ ) vector alone, pcDNA3.1( þ )DLC1-i4 or pcDNA3.1( þ )TP53 and selected with G418. DLC1-i4 and TP53 greatly inhibited clonogenicity of tumor cells. Control cells without vector would not survive G418 selection. (b) Quantitative analyses of colony numbers of HCT116, KYSE510 and NPC cell line HK1. Number of G418-resistant colonies in vector-only transfected cell line were set as 100%. Values are the mean±s.e. from three independent experiments. Asterisk indicates statistically significant difference (**Po0.001). numbers were observed in all cell lines tested (down to them—the prototype DLC1-isoform 2. From our B24–44% of vector control, Po0.001) (Figures 6a and b), previous work on DLC1 in carcinoma (Seng et al., similar to that by TP53 (data not shown). Thus, DLC1- 2007) we identified a fourth isoform of DLC1. In this i4 is a functional tumor suppressor having growth report, we provide evidence that this novel isoform 4 inhibitory activities in tumor cells. As DLC1-i1 shares (DLC1-i4) codes from an alternative promoter a 1125-aa structural domains with DLC1-i4 and DLC1-i2, we protein, which differs from DLC1-isoforms 1 and 2 at tested DLC1-i1 in the same assay and showed that the extreme N-terminal because of the use of an DLC1-i1 is also a functional tumor suppressor (Supple- alternative first exon, but nevertheless share the rest of mentary Figure 5C). the exons common to both isoforms and thus have the same protein domain structure. In addition, the protein sequence of DLC1-i4 was found to contain a mitochon- drial targeting sequence in its first 30 aa coded for by the Discussion new first exon, a signal missing from the rest of the DLC1 isoforms. The possible mitochondria localization DLC1 is a well-defined bona fide tumor suppressor, of DLC1-i4 however awaits further confirmation. With frequently silenced in multiple tumors (Yuan et al., the variability at the N-terminal region of DLC1-i1, -i2 1998, 2003; Ng et al., 2000; Goodison et al., 2005; Seng and -i4, it may be inferred that the conserved C-terminal et al., 2007; Xue et al., 2008). This gene is transcribed region domains are critical in mediating specific protein– from two different promoters, resulting in transcripts protein interactions. encoding three insofar known isoforms. To date, Similar to DLC1-i2, the DLC1-i4 promoter is virtually all studies have focused on only one of hypermethylated in multiple tumor types including

Oncogene Novel isoform of DLC1 silenced in tumors JSW Low et al 1932 NPC, esophageal, gastric, breast, colorectal, cervical this promoter is not a CpG island, and could not be and HCC cell lines and primary tumors, which is activated by Aza treatment (unpublished observation). associated with transcriptional silencing. Methylation- Taken together, the data suggest that different DLC1 mediated silencing could be reversed pharmacologically isoforms may have different functional specialization in or by genetic double knockout of DNMT1 and cellular homeostasis and oncogenesis, and whether they DNMT3B. In addition, ectopic-DLC1-i4 expression in are functionally redundant or not still needs further carcinoma cells shows tumor suppressive properties. In exploration. contrast, no methylation or silencing was detected in p53 is an important tumor suppressor, mutated in any of the immortalized normal cell line controls tested, B50% of all cancers (Efeyan and Serrano, 2007). indicating that DLC1-i4 methylation probably occurs DLC1-i4 promoter harbors at least five putative p53 late in carcinogenesis. The correlation between promoter binding sites; and our data showed that it is upregulated methylation and in primary tumor will by wild-type p53, requiring the DNA-binding ability of require further characterization in tumors other than p53. UV treatments also moderately upregulated this NPC, as the presence of large proportions of non- promoter, similar to the p21 promoter, indicating that malignant stromal cells in NPC tumors makes clear the DLC1-i4 promoter is p53-regulated. delineation of gene expression difficult. In summary, we identified a new isoform of DLC1, The use of alternative promoters and exon skipping to transcribed from an alternative promoter. This novel generate transcripts with different 50-ends appear to be isoform, designated DLC1-i4, was found to have tumor common among the DLC family of RhoGAPs. DLC2,a inhibitory properties, silenced epigenetically in multiple gene homologous to DLC1 at 13q12.3, yields four carcinomas and regulated by the p53 tumor suppressor. different transcripts to generate four protein products of Although several important issues remain to be ad- differing lengths (Leung et al., 2005). DLC3 at Xq13 was dressed, such as the functional overlap of each found to yield three transcripts generated by the same individual DLC1 isoforms, the frequent deregulation means (Durkin et al., 2007a). Similar type of regulation of DLC1-i4 in multiple tumors indicates that it may be (alternative promoters and exon variability) are also an important tumor suppressor. found in other TSGs, for example, TP53, APC and BRCA1 (Horii et al., 1993; Xu et al., 1995; Bourdon et al., 2005), and is associated with the generation of Materials and methods proteins with different functional properties. Our data and others show that DLC1 encode four major isoforms Cell lines, tumor and normal tissue samples from three alternative promoters. Interestingly, it was Tumor cell lines used in this study are described elsewhere 0 recently verified by 3 -RACE in mouse that a 6.2 Kb (Ying et al., 2005b) include NPC, esophageal, gastric, breast, transcript found in the Vega database is indeed colorectal, cervical, lung, hepatocellular, renal, prostate, expressed and probably codes for a 127 kDa protein ovarian carcinomas, Burkitt lymphoma, nasal NK/T-cell with a sequence nearly identical to DLC1-i4 (Sabbir lymphoma and Hodgkin lymphomas (Supplementary Figures). et al., 2010). The complexity and expression of different Epithelial cell lines (HMEC and HMEpC) and immortalized DLC1 transcripts may be a way of differential regula- epithelial cell lines (NP69, RHEK1, HEK293, NE1 and NE3) tion to cater for different functions of each isoform. were used as normal controls. Cell lines were routinely As the tissue expression pattern of DLC1-i1 and -i3 maintained in RPMI or DMEM, or Keratinocyte SFM Medium (for NP69 only) (Tsao et al., 2002; Lo et al., 2006). transcripts is controversial, we addressed this question Total RNA and DNA were extracted from cell pellets using using isoform-specific primers and RT–PCR to check TRI Reagent (Molecular Research Centre Inc., Cincinnati, their expression in a panel of normal adult and fetal OH, USA) (Tao et al., 2002). Normal PBMCs from healthy tissues. Our results show that all DLC1 isoforms are individuals were collected as previously described (Cheung expressed at different levels and they have an over- et al., 1993). Human normal adult and fetal tissue RNA lapping ubiquitous expression pattern in all normal samples were purchased commercially (Ying et al., 2005a; Seng tissues tested, except that only DLC-i2 is expressed in et al., 2007). RNA from matched human breast primary PBMCs. Our work on the common promoter of DLC-i1 tumor and normal tissue pair was purchased from Biochain and -i3 demonstrated that it is a functional promoter, (Hayward, CA, USA). DNA extracted from primary tumors and the full-length open reading frame of DLC1-i1 does are described elsewhere (Tao et al., 1998, 1999; Qiu et al., 2004; Ying et al., 2005a, b; Seng et al., 2007; Cui et al., 2008). have tumor suppressor function. DLC1-i3 is unlikely to have tumor suppressor activity as it does not harbor the canonical DLC1 domains and so was not tested. The 50 RLM-RACE speculation that DLC1-i3 is a regulator of other DLC1 50-RLM-RACE was carried out with RNA from NP69 cell line isoforms remains to be verified. Our findings are and liver RNA (BD Clontech, Palo Alto, CA, USA), using the corroborated by a recently published study in which FirstChoice RLM-RACE kit (Ambion Inc., Austin, TX, USA) DLC1b (-DLC1-i1) and DLC1a (DLC1-i2) but not according to the manufacturer’s protocol. Reverse transcrip- tion was carried out with random hexamers and nested PCR DLC1g (DLC1-i3) were shown to suppress stress fiber used to amplify the 50 end of DLC1 transcripts using kit formation and HCC cell growth (Ko et al., 2010). It is provided sense primers and antisense primers specific to both noteworthy that both isoforms are almost totally DLC1-i1 and -i2 (Primer sequences available in Table 1). silenced in all tumors and immortalized normal cell Nested PCR amplification was carried out in a 25 ml volume lines tested, but the silencing mechanism is unclear as consisting of 1 ml of cDNA, 0.6 U of AmpliTaq Gold (PE

Oncogene Novel isoform of DLC1 silenced in tumors JSW Low et al 1933 Biosystems, Foster City, CA, USA), 0.6 mM of each primer, using Fugene 6 (Roche Diagnostics, Mannheim, Germany). and 2 mM MgCl2. The amplification cycle involve an initial Transfected cells were grown for 48 h before luciferase assay ‘hot start’ at 95 1C for 10 min, followed by 35 cycles of using the Dual Luciferase Reporter Assay System (Promega). amplifications (94 1C, 30 s; 55 1C, 30 s; 72 1C, 2 min) with a final Three independent assays were conducted and the mean±s.d. extension step of 72 1C for 10 min. Nested PCR reaction was values were calculated. diluted 10 Â with sterile water and 1 ml of the diluted reaction To assess the effect of p53 on DLC1-i4 promoter activity, was used for subsequent nested PCR with the same conditions. concentrations of the p53 expression vector ranging from 1 to PCR products were analyzed on 1.8% agarose gels. Two bands 500 ng (pcDNA3.1( þ )TP53) or two mutant p53 expression of 205 bp and 199 bp from Liver and NP69, respectively, were constructs (G245C and R248W) (gifts from Bert Vogelstein) excised, purified and sequenced with the ABI PRISM BigDye were cotransfected with 2 mg of pGL2-DLC1i4-PF1/R(À1840/ v1.1 Kit (PE Biosystems) with the inner primer E10R. þ 140) and 100 ng pRL-SV40 into respective cells lines. For comparative purposes, the p21 promoter construct pGL2– Bioinformatics analysis p21P was used as a positive control for p53-dependent Sequences obtained were analyzed using BLASTN (Altschul transcriptional activation. For UV treatment, cells after et al., 1990) program (www.ncbi.nlm.nih.gov/blast). Gene transfection are allowed to grow for 48 h and the growth 2 structure and orthologs were obtained from GenBank medium removed before irradiation with 10 J/m UV in a (www.ncbi.nlm.nih.gov/entrez/query.fcgi?CMD ¼ search&DB Stratalinker UV Crosslinker 1800 (Stratagene, La Jolla, CA, ¼ gene) and Ensembl databases (http://www.ensembl.org) USA). After UV irradiation, growth media were replaced and (Maglott et al., 2007; Flicek et al., 2010). Alignment of incubated for a further 2–6 h before luciferase assay. aa sequences and phylogenetic tree generation were carried out using the ClustalW2 program (http://www.ebi.ac.uk/ Construction of expression plasmids Tools/clustalw2/index.html) (Larkin et al., 2007) and shaded The full-length cDNA of DLC1-i4 and -i1 with a Hemagglu- using BioEdit 7.0.1 (www.mbio.ncsu.edu/BioEdit/bioedit.html) tinin tag at its N-terminus and its ATG ‘start’ codon within a (Hall, 1999). Potential TF-binding sites on DLC1-i4 promoter Kozak consensus sequence (Kozak, 1986, 1987), and the were predicted using TFSEARCH (www.cbrc.jp/research/db/ common ‘stop’ codon present in exon 21 (Figure 1a) was PCR TFSEARCH.html) (Yutaka Akiyama, 1995; Heinemeyer et al., cloned from human liver-tissue RNA (BD Clontech) using a 1998), MotifSearch (www.motif.genome.jp/) (Heinemeyer high-fidelity DNA polymerase, Phusion . The cDNA was then et al., 1999) and MatInspector (www.genomatix.de) (Cartharius restriction enzyme digested, cloned into the expression vector et al., 2005). CpG island was predicted with CpG Island pcDNA3.1( þ ) (Invitrogen, Carlsbad, CA, USA) and se- Searcher (www.uscnorris.com/cpgislands2/cpg.aspx) (Takai and quenced. pcDNA3.1( þ )TP53 was constructed by subcloning Jones, 2002, 2003) using default settings. N-terminal signal the full-length wild-type TP53 from plasmid pC53-SN (gift peptide analysis were done using iPSORT (http://hc.ims. from Dr Bert Vogelstein) into pcDNA3.1( þ ). The two mutant u-tokyo.ac.jp/iPSORT/) (Bannai et al.,2002),MitoProtII p53 expression constructs (G245C & R248W) were also gifts (http://ihg2.helmholtz-muenchen.de/ihg/mitoprot.html) (Claros from Dr Bert Vogelstein. and Vincens, 1996), Predotar (http://urgi.versailles.inra.fr/ predotar/predotar.html) (Small et al., 2004), Mitopred (http:// bioapps.rit.albany.edu/MITOPRED/) (Guda et al., 2004a, b) Bisulfite treatment and promoter methylation analysis and TargetP 1.1 (http://www.cbs.dtu.dk/services/TargetP/) Bisulfite modification of DNA and the subsequent MSP and (Emanuelsson et al., 2007). BGS analyses were previously described (Tao et al., 1999, 2002). MSP and BGS primer sets are listed in Table 1. All primer sets were previously tested for not amplifying any unbisulfited DNA. Semi-quantitative RT–PCR For BGS, the PCR products were TA-cloned into the pCR2.1- RT–PCR was done as previously described (Tao et al., 1998, TOPO vector (Invitrogen) and 5–12 colonies were randomly 2002), with primers in Table 1, together with house-keeping chosen and sequenced. All PCR reactions were carried out using gene GAPDH as a control. RT–PCR was performed for 35 the AmpliTaq-Gold DNA polymerase (Applied Biosystems). cycles for all genes except GAPDH (25 cycles), using the GeneAmp RNA PCR system (Applied Biosystems). Aza and TSA treatment Freshly seeded cells (C666-1 and HK1) at a concentration of Gene reporter assays 1 Â 105 cells per ml in T-75 flasks were grown overnight with From normal placenta DNA (Sigma-Aldrich Corporation, fresh medium containing Aza (Sigma-Aldrich Corporation) at St Louis, MO, USA), three regions of the putative DLC1-i4 a final concentration of 50 mM (Qiu et al., 2004; Ying et al., and -i1 promoter were PCR amplified using a high-fidelity 2005b). Cells were treated continuously for 72 or 144 h, with DNA polymerase, Phusion (Finnzymes, Espoo, Finland) and the Aza-containing medium changed every 24 h before being ligated to the promoter-less pGL2-Enhancer vector (Promega, harvested for DNA and RNA extraction. For TSA treatment, Madison, WI, USA) to create pGL2-DLC1i4-PF1/R(À1840/ cells (HCT116, EC109 and HK1) were treated as above for þ 140), pGL2-DLC1i4-PF2/R(À960/ þ 140), pGL2-DLC1i4- 72 h using 10 mM of Aza and subsequently for another 24 h with PF3/R(À480/ þ 140); pGL2-DLC1i1-F1/R(À1272/ þ 286), pGL2- 100 ng/ml TSA. DLC1i1-F2/R(À656/ þ 286) and pGL2-DLC1i1-F3/R(À376/ þ 286). The p21 promoter plasmid (pGL2–p21P) was a kind gift from Drs Satya Narayan and Bert Vogelstein (University Colony formation assays of Florida, Gainesville, FL, USA; Johns Hopkins, Baltimore, HCT116, HK1 and EC109 cells (3 Â 105 cells per ml) were MD, USA). Plasmids used for transfection were prepared and plated in a 12-well plate and allowed to grow for 24 h before purified using the Endofree Plasmid Maxi Kit from Qiagen being transfected with 2 mg of expression plasmids or empty (Qiagen GmbH, Germany). Promoter activities were tested vector using Fugene 6 (Roche Diagnostics) according to the by co-transfection of cell lines (HEK293, CNE2 and HK1) in manufacturer’s protocol. At 48 h post-transfection, transfected 12-well plates with 2 mg of promoter construct and 100 ng of cells were trypsinized and diluted into 6-well plates with G-418 Renilla Luciferase plasmid pRL-SV40 as an internal control (500 mg/ml) selection for 2–3 weeks. Surviving colonies (with

Oncogene Novel isoform of DLC1 silenced in tumors JSW Low et al 1934 450 cells per colony) were counted and analyzed after staining Acknowledgements with Gentian Violet. The experiment was repeated thrice independently. This project was supported by an A*STAR grant (Johns Hopkins Singapore, QT/WS/RA), a Chinese University of Accession number Hong Kong grant (QT) and a grant from Singapore National The sequences of the RLM-RACE products of DLC1 from Research Foundation and the Ministry of Education under the liver and NP69 have been deposited into GenBank (Accession Research Center of Excellence Program (WS/RA). We thank no: EU159199 and EU159200, respectively). Drs Bert Vogelstein, George Tsao, (Dolly Huang), Michael Obster and Guiyuan Li for some cell lines; and DSMZ (German Collection of Microorganisms & Cell Cultures) for Conflict of interest the KYSE cell lines (Shimada et al., Cancer 69: 277–284, 1992). We also thank Drs Wenling Han, Thomas Putti, Luke Tan for The authors declare no conflict of interest. some tumor samples.

References

Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. (1990). Basic Guda C, Fahy E, Subramaniam S. (2004a). MITOPRED: a genome- local alignment search tool. J Mol Biol 215: 403–410. scale method for prediction of nucleus-encoded mitochondrial Bannai H, Tamada Y, Maruyama O, Nakai K, Miyano S. (2002). proteins. Bioinformatics 20: 1785–1794. Extensive feature detection of N-terminal protein sorting signals. Guda C, Guda P, Fahy E, Subramaniam S. (2004b). MITOPRED: a Bioinformatics 18: 298–305. web server for the prediction of mitochondrial proteins. Nucl Acids Bernards A. (2003). GAPs galore! A survey of putative Ras Res 32: W372–W374. superfamily GTPase activating proteins in man and Drosophila. Hall TA. (1999). Bioedit: A user friendly biological sequence alignment Biochim Biophys Acta 1603: 47–82. editor and analysis program for Windows 95/98/NT. Nucl Acids Bourdon JC, Fernandes K, Murray-Zmijewski F, Liu G, Diot A, Symp Ser 41: 95–98. Xirodimas DP et al. (2005). p53 isoforms can regulate p53 Heinemeyer T, Chen X, Karas H, Kel AE, Kel OV, Liebich I et al. (1999). transcriptional activity. Genes Dev 19: 2122–2137. Expanding the TRANSFAC database towards an expert system of Bullock AN, Fersht AR. (2001). Rescuing the function of mutant p53. regulatory molecular mechanisms. Nucleic Acids Res 27: 318–322. Nat Rev Cancer 1: 68–76. Heinemeyer T, Wingender E, Reuter I, Hermjakob H, Kel AE, Kel OV Cartharius K, Frech K, Grote K, Klocke B, Haltmeier M, Klingenhoff et al. (1998). Databases on transcriptional regulation: TRANSFAC, A et al. (2005). MatInspector and beyond: promoter analysis TRRD and COMPEL. Nucleic Acids Res 26: 362–367. based on transcription factor binding sites. Bioinformatics 21: Herman JG, Baylin SB. (2003). in cancer in association 2933–2942. with promoter hypermethylation. N Engl J Med 349: 2042–2054. Cheung WY, Chan AC, Loke SL, Srivastava G, Pittaluga S, Lim LY Homma Y, Emori Y. (1995). A dual functional signal mediator et al. (1993). Latent sites of Epstein-Barr virus infection. Am J Clin showing RhoGAP and -delta stimulating activities. Pathol 100: 502–506. EMBO J 14: 286–291. Ching YP, Wong CM, Chan SF, Leung TH, Ng DC, Jin DY et al. Horii A, Nakatsuru S, Ichii S, Nagase H, Nakamura Y. (1993). (2003). Deleted in liver cancer (DLC) 2 encodes a RhoGAP protein Multiple forms of the APC gene transcripts and their tissue-specific with growth suppressor function and is underexpressed in hepato- expression. Hum Mol Genet 2: 283–287. cellular carcinoma. J Biol Chem 278: 10824–10830. Jones PA, Baylin SB. (2007). The epigenomics of cancer. Cell 128: 683–692. Claros MG, Vincens P. (1996). Computational method to predict Jones PA, Laird PW. (1999). Cancer epigenetics comes of age. Nat mitochondrially imported proteins and their targeting sequences. Genet 21: 163–167. European Journal of Biochemistry 241: 779–786. Kim TY, Jong HS, Song SH, Dimtchev A, Jeong SJ, Lee JW et al. (2003). Cui Y, Ying Y, van Hasselt A, Ng KM, Yu J, Zhang Q et al. (2008). Transcriptional silencing of the DLC-1 by OPCML is a broad tumor suppressor for multiple carcinomas and epigenetic mechanism in gastric cancer cells. Oncogene 22: 3943–3951. lymphomas with frequently epigenetic inactivation. PLoS ONE Knudson AG. (2001). Two genetic hits (more or less) to cancer. Nat 3: e2990. Rev Cancer 1: 157–162. Durkin ME, Ullmannova V, Guan M, Popescu NC. (2007a). Deleted Ko FC, Yeung YS, Wong CM, Chan LK, Poon RT, Ng IO et al. in liver cancer 3 (DLC-3), a novel Rho GTPase-activating protein, is (2010). Deleted in liver cancer 1 isoforms are distinctly expressed downregulated in cancer and inhibits tumor cell growth. Oncogene in human tissues, functionally different and under differential 26: 4580–4589. transcriptional regulation in hepatocellular carcinoma. Liver Durkin ME, Yuan BZ, Zhou X, Drazen BZ, Douglas RL, International 30: 139–148. Thorgeirsson SS et al. (2007b). DLC-1:a Rho GTPase-activating Kozak M. (1986). Point mutations define a sequence flanking the AUG protein and tumour suppressor. Journal of Cellular and Molecular initiator codon that modulates translation by eukaryotic ribosomes. Medicine 11: 1185–1207. Cell 44: 283–292. Efeyan A, Serrano M. (2007). p53: guardian of the genome and Kozak M. (1987). An analysis of 50-noncoding sequences from 699 policeman of the oncogenes. Cell Cycle 6: 1006–1010. vertebrate messenger RNAs. Nucleic Acids Res 15: 8125–8148. Emanuelsson O, Brunak S, von Heijne G, Nielsen H. (2007). Locating Larkin MA, Blackshields G, Brown NP, Chenna R, McGettigan PA, proteins in the cell using TargetP, SignalP and related tools. Nat McWilliam H et al. (2007). Clustal W and Clustal  version 2.0. Protocols 2: 953–971. Bioinformatics 23: 2947–2948. Fearon ER, Vogelstein B. (1990). A genetic model for colorectal Leung TH, Ching YP, Yam JW, Wong CM, Yau TO, Jin DY et al. tumorigenesis. Cell 61: 759–767. (2005). Deleted in liver cancer 2 (DLC2) suppresses cell transforma- Flicek P, Aken BL, Ballester B, Beal K, Bragin E, Brent S et al. (2010). tion by means of inhibition of RhoA activity. Proc Natl Acad Sci Ensembl’s 10th year. Nucl Acids Res 38: D557–D562. USA 102: 15207–15212. Goodison S, Yuan J, Sloan D, Kim R, Li C, Popescu NC et al. (2005). Lo AK, Lo KW, Tsao SW, Wong HL, Hui JW, To KF et al. (2006). The RhoGAP protein DLC-1 functions as a metastasis suppressor Epstein-Barr virus infection alters cellular signal cascades in human in breast cancer cells. Cancer Res 65: 6042–6053. nasopharyngeal epithelial cells. Neoplasia 8: 173–180.

Oncogene Novel isoform of DLC1 silenced in tumors JSW Low et al 1935 Maglott D, Ostell J, Pruitt KD, Tatusova T. (2007). gene: gene- latent promoter C in iatrogenic B cell lymphoproliferative disease. centered information at NCBI. Nucleic Acids Res 35: D26–D31. Application of PCR-based analysis. Am J Pathol 155: 619–625. Martin GS. (2003). Cell signaling and cancer. Cancer Cell 4: 167–174. Tcherkezian J, Lamarche-Vane N. (2007). Current knowledge of the Moon SY, Zheng Y. (2003). Rho GTPase-activating proteins in cell large RhoGAP family of proteins. Biol Cell 99: 67–86. regulation. Trends Cell Biol 13: 13–22. Tsao SW, Wang X, Liu Y, Cheung YC, Feng H, Zheng Z et al. (2002). Ng IO, Liang ZD, Cao L, Lee TK. (2000). DLC-1 is deleted in primary Establishment of two immortalized nasopharyngeal epithelial cell hepatocellular carcinoma and exerts inhibitory effects on the lines using SV40 large T and HPV16E6/E7 viral oncogenes. Biochim proliferation of hepatoma cell lines with deleted DLC-1. Cancer Biophys Acta 1590: 150–158. Res 60: 6581–6584. Wong CM, Lee JM, Ching YP, Jin DY, Ng IO. (2003). Genetic and Pang JC, Chang Q, Chung YF, Teo JG, Poon WS, Zhou LF et al. epigenetic alterations of DLC-1 gene in hepatocellular carcinoma. (2005). Epigenetic inactivation of DLC-1 in supratentorial primitive Cancer Res 63: 7646–7651. neuroectodermal tumor. Hum Pathol 36: 36–43. Xu CF, Brown MA, Chambers JA, Griffiths B, Nicolai H, Solomon E. Ponting CP, Aravind L. (1999). START: a lipid-binding domain (1995). Distinct transcription start sites generate two forms of in StAR, HD-ZIP and signalling proteins. Trends Biochem Sci 24: BRCA1 mRNA. Hum Mol Genet 4: 2259–2264. 130–132. Xue W, Krasnitz A, Lucito R, Sordella R, VanAelst L, Cordon-Cardo Qiu GH, Tan LK, Loh KS, Lim CY, Srivastava G, Tsai ST et al. C et al. (2008). The ITGB2 immunomodulatory gene (CD18), (2004). The candidate tumor suppressor gene BLU, located at the enterocolitis, and Hirschsprung’s disease. Genes & Development 22: commonly deleted region 3p21.3, is an E2F-regulated, 1439–1444. stress-responsive gene and inactivated by both epigenetic and Ying J, Li H, Seng TJ, Langford C, Srivastava G, Tsao SW et al. genetic mechanisms in nasopharyngeal carcinoma. Oncogene 23: (2005a). Functional epigenetics identifies a protocadherin PCDH10 4793–4806. as a candidate tumor suppressor for nasopharyngeal, esophageal Rhee I, Bachman KE, Park BH, Jair KW, Yen RW, Schuebel KE et al. and multiple other carcinomas with frequent methylation. Oncogene (2002). DNMT1 and DNMT3b cooperate to silence genes in human 25: 1070–1080. cancer cells. Nature 416: 552–556. Ying J, Srivastava G, Hsieh WS, Gao Z, Murray P, Liao SK et al. Sabbir MG, Wigle N, Loewen S, Gu Y, Buse C, Hicks GG et al. (2005b). The stress-responsive gene GADD45G is a functional (2010). Identification and characterization of Dlc1 isoforms in the tumor suppressor, with its response to environmental stresses mouse and study of the biological function of a single gene trapped frequently disrupted epigenetically in multiple tumors. Clin Cancer isoform. BMC Biology 8: 17. Res 11: 6442–6449. Sekimata M, Kabuyama Y, Emori Y, Homma Y. (1999). Morpho- Yuan BZ, Durkin ME, Popescu NC. (2003). Promoter hypermethyla- logical changes and detachment of adherent cells induced by p122, a tion of DLC-1, a candidate tumor suppressor gene, in several GTPase-activating protein for Rho. J Biol Chem 274: 17757–17762. common human cancers. Cancer Genet Cytogenet 140: 113–117. Seng TJ, Low JSW, Li H, Cui Y, Goh HK, Wong MLY et al. (2007). Yuan BZ, Jefferson AM, Baldwin KT, Thorgeirsson SS, Popescu NC, The major 8p22 tumor suppressor DLC1 is frequently silenced by Reynolds SH. (2004). DLC-1 operates as a tumor suppressor gene in methylation in both endemic and sporadic nasopharyngeal, human non-small cell lung carcinomas. Oncogene 23: 1405–1411. esophageal, and cervical carcinomas, and inhibits tumor cell colony Yuan BZ, Jefferson AM, Millecchia L, Popescu NC, Reynolds SH. formation. Oncogene 26: 934–944. (2007). Morphological changes and nuclear translocation of DLC1 Small IF, Peeters NF, Legeai FF, Lurin C. (2004). Predotar: a tool for tumor suppressor protein precede apoptosis in human non-small cell rapidly screening proteomes for N-terminal targeting sequences. lung carcinoma cells. Exp Cell Res 313: 3868–3880. PROTEOMICS 4: 1581–1590. Yuan BZ, Miller MJ, Keck CL, Zimonjic DB, Thorgeirsson SS, Takai D, Jones PA. (2002). Comprehensive analysis of CpG islands Popescu NC. (1998). Cloning, characterization, and chromosomal in human 21 and 22. Proc Natl Acad Sci USA 99: localization of a gene frequently deleted in human liver cancer 3740–3745. (DLC-1) homologous to rat RhoGAP. Cancer Res 58: 2196–2199. Takai D, Jones PA. (2003). The CpG island searcher: a new WWW Yutaka Akiyama, TFSEARCH: searching transcription factor binding resource. In Silico Biol 3: 235–240. sites, 1995 Ref Type: Unpublished Work. Tao Q, Huang H, Geiman TM, Lim CY, Fu L, Qiu GH et al. (2002). Zhou X, Thorgeirsson SS, Popescu NC. (2004). Restoration of DLC-1 Defective de novo methylation of viral and cellular DNA sequences gene expression induces apoptosis and inhibits both cell growth and in ICF syndrome cells. Hum Mol Genet 11: 2091–2102. tumorigenicity in human hepatocellular carcinoma cells. Oncogene Tao Q, Robertson KD, Manns A, Hildesheim A, Ambinder RF. 23: 1308–1313. (1998). Epstein-Barr virus (EBV) in endemic Burkitt’s lymphoma: Zhou X, Wang XW, Xu L, Hagiwara K, Nagashima M, Wolkowicz R molecular analysis of primary tumor tissue. Blood 91: 1373–1381. et al. (1999). COOH-terminal domain of p53 modulates p53- Tao Q, Swinnen LJ, Yang J, Srivastava G, Robertson KD, Ambinder mediated transcriptional transactivation, cell growth, and apoptosis. RF. (1999). Methylation status of the Epstein-Barr virus major Cancer Res 59: 843–848.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene