Frequent Epigenetic Inactivation of RASSF1A and BLU Genes Located Within the Critical 3P21.3 Region in Gliomas

Frequent epigenetic inactivation of RASSF1A and BLU genes located within the critical 3p21.3 region in gliomas

Luke Hesson1, Ivan Bie` che2, Dietmar Krex3, Emmanuelle Criniere4,Kheˆ Hoang-Xuan4, Eamonn R Maher1,5 and Farida Latif*,1,5

1Section of Medical and Molecular Genetics, Division of Reproductive and Child Health, University of Birmingham, Birmingham B15 2TT, UK; 2Laboratoire d’Oncoge´ne´tique – INSERM E0017, Centre Rene´ Huguenin, 35, rue Dailly, F-92210 St-Cloud, France; 3Department of Neurosurgery, Universita¨tsklinikum Carl Gustav Carus, Technische Universita¨t Dresden, Fetscherstrae 74, 01307 Dresden, Germany; 4Fe´de´ration neurologique Mazarin and Unite´ INSERM U495, Hoˆpital de la Salpeˆtrie`re, Paris, France; 5Cancer Research UK Renal Molecular Oncology Research Group, University of Birmingham, Birmingham B15 2TG, UK

RASSF1A is a major tumor suppressor gene located specific events and not due to region-wide changes in DNA at 3p21.3. We investigated the role of aberrant methylation. promoter region hypermethylation of RASSF1A in a Oncogene (2004) 23, 2408–2419. doi:10.1038/sj.onc.1207407 large series of adult gliomas. RASSF1A was frequently Published online 26 January 2004 methylated in both primary tumors (36/63; 57%) and tumor cell lines (7/7; 100%). Hypermethylation of Keywords: methylation; gliomas; RASSF1A; BLU RASSF1A in glioma cell lines correlated with loss of expression and treatment with a demethylating agent- reactivated RASSF1A gene expression. Furthermore, re- expression of RASSF1A suppressed the growth of glioma cell line H4 in vitro. Next, we investigated whether other Introduction members of the RASSF gene family were also inactivated by methylation. NORE1B and RASSF3 were not Gliomas are a heterogenous group of primary brain methylated in gliomas, while NORE1A and RASSF5/ tumors that may be divided into four grades on the basis AD037 demonstrated methylation in glioma cell lines but of their histology and prognosis. Patients with the most not in primary tumors. We then investigated the methyla- malignant gliomas (grade 4), also known as glioblasto- tion status of three other candidate 3p21.3 tumor ma multiforme (GBM), have a mean survival of about a suppressor genes. CACNA2D2 and SEMA3B were not year. Patients with anaplastic gliomas (grade 3) survive frequently methylated, but the BLU gene located just for 2–3 years and those with grade 2 gliomas survive for centromeric to RASSF1 was frequently methylated in 10–15 years (Holland 2001; Louis et al., 2002). glioma cell lines (7/7) and in 80% (35/44) of glioma Genetic analysis of gliomas has identified various tumors. In these tumor cell lines, BLU expression was cytogenetic and molecular genetic changes, and in some restored after treatment with a demethylating agent. Loss cases these findings correlate with clinical grade of BLU gene expression in glioma tumors correlated with (reviewed in Holland 2001; Louis et al., 2002). BLU methylation. There was no association between Methylation of CG dinucleotides in the promoter region RASSF1A and BLU methylation. RASSF1A methylation of tumor suppressor genes (TSGs) producing transcrip- increased with tumor grade, while BLU methylation was tional silencingis a major mechanism of TSG inactiva- seen at similar frequencies in all grades. Our data tion in many human cancers. In this report, we implicate RASSF1A and BLU promoter methylation in undertook methylation analysis of 12 genes in a large ONCOGENOMICS the pathogenesis of adult gliomas, while other RASSF series of adult gliomas and glioma cell lines. family members and CACNA2D2 and SEMA3B The genes analysed include RASSF1 family members appear to have only minor roles. In addition, RASSF1A (RASSF1A, NORE1, RASSF3 and RASSF/AD037). and BLU methylation appear to be independent and RASSF1A is a major TSG located within the critical 120 kb interval at 3p21.3, defined by overlapping homozygous deletions in lung and breast tumor cell lines (Lerman and Minna, 2000). It is a novel TSG in that it is frequently inactivated in many types of human cancers by aberrant promoter methylation (Dammann *Correspondence: F Latif, Section of Medical and Molecular Genetics, et al., 2000; Agathanggelou et al., 2001; Burbee et al., Division of Reproductive and Child Health, University of Birming- ham, Birmingham B15 2TT, UK; E-mail: fl[email protected] 2001), while somatic inactivatingmutations are a rare Received 24 September 2003; revised 26 November 2003; accepted 2 event (reviewed in Pfeifer et al., 2002). The gene encodes December 2003 two major isoforms that are produced by alternative Epigenetic inactivation of RASSF1A and BLU L Hesson et al 2409 promoter selection and alternative messenger RNA same series of gliomas were analysed for five (CASP8, splicing. The longer isoform known as RASSF1A (340 DAPK1, MGMT, GSTP1, CLDN7) further genes that amino acids, 39 kD a peptide) contains a predicted N- are known to play a part in tumorigenesis. terminal diacylglycerol-binding (DAG) domain and a C- terminal predicted RAS-association domain. The short- er isoform known as RASSF1C (270 amino acids, Results 32 kDa peptide) has a different N-terminus that lacks the DAG domain, but has a similar C-terminus contain- RASSF1A, NORE1, RASSF3 and RASSF5 inga predicted RAS association domain. Exogenous methylation analysis expression of RASSF1A in cell lines lackingexpression caused suppression of the malignant phenotype both in We utilized a methylation-sensitive PCR (MSP) assay in vitro and in vivo assays. RASSF1A protein function is described previously (Honorio et al., 2003a, b) to beingactively investigated, and recently it was demon- determine the hypermethylation status of RASSF1A strated that RASSF1A blocks cell cycle progression and promoter-associated CpG island. RASSF1A was com- suppresses cyclin D1 accumulation (Shivakumar et al., pletely methylated in five glioma cell lines and partially 2002; Agathanggelou et al., 2003a,b). methylated in two (7/7; 100%) (Figure 1a). In primary Using in silico analysis followed by cloningand glioma tumors, RASSF1A was methylated in 36/63 sequencing, we and others have identified three further (57%) (Table 1, Figure 1b). RASSF1A methylation members of the RASSF gene family containing a ras increased with tumor grade (4/10, 40% grade II; 8/15, association domain (RA). NORE1 is located at 1q32.1, 53% grade III; 24/38, 63% IV), although it was not it has two major isoforms, NORE1A and NORE1B. statistically significant. None of the DNA isolated from Both isoforms have separate CpG islands spanningtheir normal adult brains demonstrated any methylation of first exons. NORE1A contains a predicted DAG domain the RASSF1A 50 CpG island. RASSF1A expression was and a predicted RAS association domain, while restored in six glioma cell lines (five of which were NORE1B contains only the predicted RAS association completely methylated) after treatment with 5-aza-2- domain. We recently demonstrated that NORE1A but deoxycytidine (Figure 2). not NORE1B promoter region CpG island was methy- For 20 GBM, correspondingblood DNA was lated in breast, lung, colorectal and kidney tumor cell available for loss of heterozygosity (LOH) analysis. lines; furthermore, NORE1A expression in tumor cell Usingmicrosatellite markers D3S1568 and D3S3667 lines was reactivated after treatment with a demethylat- (both located in the 3p21.3 region), we did not find any ingagent.In primary tumors, NORE1A was found to be evidence of LOH at D3S1568 in 15 informative cases, methylated in 24% of NSCLC but in 0% of SCLCs while D3S3667 demonstrated allele loss in one of 16 (Hesson et al., 2003). No somatic inactivatingmutations informative pairs. were found in lungcancers. RASSF1A heterodimerizes NORE1A was methylated in 3/7 (partially in two and with the mouse Nore1 gene, a known ras effector (Ortiz- completely methylated in one) of glioma cell lines, but in Vega et al., 2002). Recently, both NORE1 and none (0/20) of primary glioma tumors. NORE1B 50 CpG RASSF1A have been shown to associate with the island was found to be unmethylated in cell lines (0/7); proapoptotic kinase MST1 (Khokhlatchev et al., 2002). therefore, primary glioma tumors were not analysed. RASSF3 was recently described by Tommasi et al. Similarly, 50 CpG island associated with RASSF3 was (2002); it is located at 12q14.1 and also contains a RAS found to be unmethylated in all glioma cell lines association domain at the C-terminus. It is expressed in analysed (0/6). While RASSF5 50 CpG island was many normal tissues. Tommasi et al found no inactivating partially methylated in 3/7 of glioma cell lines, none of mutations in the codingregionin lungand breast tumors, the tumors were methylated (0/20). These data are but methylation status of RASSF3 was not analysed. summarized in Table 1 and Figure 1. RASSF5/AD037 was recently identified and characterized by us. It resides at 10p11.21 and produces a Methylation analysis of 50 CpG island associated with protein of 321 amino acids containinga RAS association three further genes located at 3p21.3 domain in its carboxy terminus. RASSF5 has a CpG island spanningthe first exon. No inactivatingmutations Recently, we have demonstrated that a 50 CpG island were found in lungcancers; however, the CpG island is associated with a candidate suppressor gene (BLU) hypermethylated in breast, lung, colorectal and kidney located immediately centromeric to RASSF1 locus tumor cell lines and in primary lungand breast tumors. (Figure 3) is frequently hypermethylated in neuroblas- Furthermore in primary tumors and tumor cell lines, tomas and less frequently in lungcancer. This methyla- loss of expression correlated with RASSF5 methylation tion correlates with loss of gene expression, and the (Hesson and Latif, personal communication). expression is restored after treatment with demethylat- In this report, we undertook systematic analysis of ing agents (Agathanggelou et al., 2003b). We have now RASSF family members in a large series of adult glioma analysed methylation status of BLU in the above series tumors and cell lines. Furthermore, we determined of glioma cell lines and tumors using an MSP assay methylation status of three other genes residing in the described earlier. BLU was found to be completely critical 3p21.3 region, namely, CACNA2D2, BLU and methylated in two glioma cell lines and partially SEMA3B in the above series of gliomas. In addition, the methylated in five (Figure 4a). Expression of BLU was

Oncogene Epigenetic inactivation of RASSF1A and BLU L Hesson et al 2410

Oncogene Epigenetic inactivation of RASSF1A and BLU L Hesson et al 2411 Table 1 Summary of methylation data for all genes analysed in absent or downregulated in glioma cell lines showing glioma tumor cell lines and primary tumors methylation and expression was restored after treatment Gene Cell lines (%) Tumors (%) of two glioma tumor lines with 5-aza-2-deoxycytidine N methylated/N total N methylated/N total (Figure 4b and c). In primary tumors, BLU was found to be frequently and heavily methylated (35/44; 80%) RASSF1A 7/7 (100%) 36/63 (57%) NORE1A 3/7 (43%) 0/20 since many tumors showed only the methylated alleles NORE1B 0/7 ND (Figure 4d). Using quantitative real-time reverse tran- RASSF3 0/6 ND scriptase–polymerase chain reaction (RT–PCR), we RASSF5 3/7 (43%) 0/20 demonstrated that loss of BLU gene expression corre- BLU 7/7 (100%) 35/44 (80%) lated with BLU methylation in glioma tumors CACNA2D2 4/6 (67%) 1/24 (4%) SEMA3B 1/7 (14%) ND (Figure 4e). None of the DNA from 20 corresponding CASP8 0/6 ND lymphocytes or from normal adult brain was found to DAPK1 0/7 ND be methylated for this CpG island, indicatingtumor- MGMT 6/7 (86%) 10/18 (56%) speciﬁc methylation. There was no association between GST1 0/7 ND CLDN7 4/6 (67%) 2/45 (4.4%) BLU and RASSF1A methylation in this series of P16 ND 0/18 (Dallol et al., 2003b) gliomas (P ¼ 0.45). SLIT2 5/7 (71%) 37/63 (59%) (Dallol et al., 2003b) The CACNA2D2 gene located 162.7 kb centromeric of RASSF1 locus (Figure 3) is a new member of the ND, not determined. In this study, glioma tumors were analysed only Ca(2 þ ) channel complex (Gao and Minna, personal if frequent methylation was found in glioma tumor cell lines communication; Lerman and Minna, 2000). Majority of

Figure 2 Expression of RASSF1A in methylated glioma tumor cell lines before and after treatment with the demethylating agent 50aza-2-deoxycytidine. RASSF1A expression is vastly upregulated following 50-aza-2-deoxycytidine treatment ( þ ) in methylated glioma cell lines. RASSF1F (another isoform of the RASSF1 gene) is also upregulated showing that transcription of this isoform is controlled via the same promoter region. GAPDH was used as a positive control to ensure RNA integrity and equal loading

Figure 1 (a) The methylation status of the RASSF family members RASSF1A, NORE1A, NORE1B, RASSF3 and RASSF5 NSCLC cell lines show loss of CACNA2D2 gene in glioma tumor cell lines. RASSF1A and NORE1A methylation expression. We have recently demonstrated that this status was determined usingMSP. NORE1B, RASSF3 and RASSF5 methylation was determined usingthe COBRA method loss of gene expression (both at RNA and protein levels) (TaqI/BstUI digestion and direct sequencing). The undigested PCR correlates with hypermethylation of a CpG island in the product and TaqI-digested PCR product are shown in the promoter of CACNA2D2 in NSCLC (Gao and Minna, NORE1B and RASSF3 diagrams. RASSF5Taq I- and BstUI- personal communication). It has recently been shown digested PCR products are also shown. Frequent methylation was observed in RASSF1A (7/7), NORE1A (3/7) and partial methyla- that overexpression of CACNA2D2 induced apoptosis tion of three lines was observed for RASSF5 (3/7). N1, N2 and N3 in NSCLC lines (Carboni et al., 2003). Usingdirect represent the DNA from three individual normal brain tissues. No sequencingof bisulfite-modified DNA (see materials and methylation of the RASSF3 and NORE1B CpG islands was methods for primer sequence), we analysed part of the 50 observed in any of the glioma tumor cell lines analysed. (b) The CpG island of CACNA2D2 in glioma cell lines. Four of methylation status of RASSF1A, NORE1A and RASSF5 in glioma primary tumors usingMSP ( RASSF1A and NORE1A) and the six lines were partially methylated, but only one of COBRA (RASSF5). Frequent RASSF1A methylation was ob- 24 primary glioma tumors was methylated. served in glioma tumors, whereas NORE1A and RASSF5 SEMA3B located 72 kb telomeric of the RASSF1 methylation was not observed in any tumors analysed. U343 and locus (Figure 3) is a member of the semaphorin gene U373 were included as RASSF5 methylation positive controls. PCR ¼ undigested PCR product; T ¼ TaqI digest; B ¼ BstUI digest. family comprised of secreted and membrane-associated Sss1 in vitro methylated normal blood DNA was used as a positive proteins that contribute to axonal path findingduring control neural development. SEMA3B has recently been shown

Oncogene Epigenetic inactivation of RASSF1A and BLU L Hesson et al 2412

Figure 3 Schematic representation of the 3p21.31 homozygously deleted region and the location and distances between four candidate genes analysed. Previously, the 3p21.3 region was implicated in lung cancer by the identification of three independent lung cancer homozygous deletions (Lerman and Minna, 2000) represented by the 370 kb deletion region (wavy lines; NCI-H1450, NCI-H740 and GLC20). Identification of a breast cancer homozygous deletion (HCC1500) refined the candidate region to 120 kb. The causative TSG/ TSGs could, however, reside in the minimal 120 kb indicated by the breast cancer homozygous deletion or the additional 250 kb indicated by the lungcancer homozygousdeletions. Several promisingcandidates have been identified and their involvement in several other cancers has been recently established. We undertook epigenetic analysis of four such candidates to assess their involvement in gliomagenesis; SEMA3B (NM_004636), RASSF1A (NM_007182), BLU (NM_015896) and CACNA2D2 (NM_006030). Distances between genes were calculated from their transcriptional start sites as indicated by the respective Genbank accession numbers. LOH status was determined usingthe nearest flankingpolymorphic loci indicated by two asterisks. D3S1568 is located within intron 2 of the gene CACNA2D2. D3S3667 is located within intron 5 of the gene RBM6 (B275 kb upstream of the gene SEMA3B)

to inhibit lungcancer cell growth and re-expression- tumor cell lines and primary tumors (Tomizawa et al., induced apoptosis in NSCLC cell lines (Tomizawa et al., 2001; Kuroki et al., 2003). By direct sequencingof 2001). SEMA3B was found to be methylated in NSCLC bisulﬁte-modiﬁed DNA with appropriate primers

Oncogene Epigenetic inactivation of RASSF1A and BLU L Hesson et al 2413

Figure 4 BLU is frequently methylated in glioma tumor cell lines and primary tumors with associated downregulation or loss of BLU expression. (a) Methylation status of the BLU promoter region in seven glioma tumor cell lines. All seven glioma tumour cell lines were methylated. (b) Methylation correlates with downregulation or loss of BLU expression. (c) Treatment of methylated cell lines with 50- aza-2-deoxycytidine ( þ ) restores expression of BLU.(d) BLU methylation status in grade II, III and IV glioma primary tumors using MSP. Heavy methylation of the BLU promoter was also observed in many primary tumors analysed includinglow-gradetumors. No methylation was observed in the DNA from 20 correspondinglymphocytes and six normal brain tissues. Quantitative real-time RT– PCR shows that heavy methylation of BLU ( þþ) correlates with very low levels of BLU expression in primary tumors relative to unmethylated tumors (À). In tumors where both methylated and unmethylated alleles were seen ( þ ), BLU expression was also downregulated compared to unmethylated tumors

(see Materials and methods), we found the SEMA3B 50 in one) and tumors (10/18; 56%). None of the DNA CpG island to be partially methylated in 1/7 glioma from normal adult brain was methylated for MGMT cell lines. Since the methylation frequency was low (Figure 5a, b). in glioma tumor lines, primary tumors were not Claudin-7 (CLDN-7) is a member of a large gene analysed. family (20 members so far) (Mitic et al., 2000) located at 17p13.1. Claudins are transmembrane proteins that seal tight junctions and are critical for maintaining cell-to- Methylation analysis of five other genes known to play cell adhesion. Recently, Claudin-7 expression was found roles in tumorigenesis to be lower in breast carcinomas than in normal breast epithelium (Kominsky et al., 2003). An MSP assay was We went onto analyse the same large series of glioma developed to determine the methylation status of the 50 tumors and cell lines for methylation in CASP8, CpG island associated with CLDN-7. Three out of three DAPK1, MGMT, GSTP1 and CLDN7 genes. breast tumor lines analysed were found to be methylated Previously, we and others had described frequent and gene expression was restored in these tumor lines methylation of CASP8 in neuroblastomas (Teitz et al., after treatment with 5-aza-dC. However, no methylation 2000; Astuti et al., 2001). Recently, CASP8 methylation was detected in five invasive ductal carcinomas (Ko- has been described in other pediatric tumors and in lung minsky et al., 2003). Claudins have tissue-specific cancer (Harada et al., 2002; Shivapurkar et al., 2002). expression patterns. CLDN-7 is well expressed in brain, We used an MSP assay to determine methylation status heart, liver, pancreas, prostate, kidney and lung. Hence, of CASP8 in glioma cell lines. None of the glioma cell we utilized the MSP primers described in the above lines (0/7) were found to be methylated, similarly no reference to determine the status of the 50 CpG island glioma cell line was methylated for DAPK1 (0/7) and associated with CLDN-7 in gliomas. CLDN-7 was GSTP1 (0/7) (Figure 5a). MGMT was frequently methylated in four of six glioma cell lines (two partial methylated in glioma cell lines (completely methylated and two completely methylated), but in only 2 of 45 in three, partially methylated in three and unmethylated glioma tumors (Figure 5a, b).

Oncogene Epigenetic inactivation of RASSF1A and BLU L Hesson et al 2414

Figure 5 (a) Methylation status of the promoter regions of MGMT, CLDN7, CASP8, GSTP1 and DAPK1 in glioblastoma tumor cell lines usingMSP. Frequent methylation of the MGMT promoter (6/7) and CLDN7 promoter (4/6) was observed. N1–N6 represent the DNA from six individual normal brain tissues. (b) Methylation status of MGMT and CLDN7 in glioma primary tumors using MSP. Frequent methylation of MGMT was observed (56%) includinglow-gradetumors, whereas and CLDN7 was methylated in only 4% of cases

Tumor suppressor activity of RASSF1A expression. Furthermore, exogenous expression of RASSF1A decreased in vitro colony formation. Exogenous expression of RASSF1A has been shown to RASSF3 is not methylated in glioma tumor cell decrease in vitro colony-formation ability in lungtumor lines, while NORE1A and RASSF5 are frequently RASSF1A cell lines exhibiting hypermethylation and methylated in glioma tumor cell lines but not in loss of expression (Dammann et al., 2000; Burbee et al., primary tumors. Hence, NORE1A and RASSF5 may 2001). Wild-type RASSF1A in the pCMVHA-tagged play a role in very late gliomagenesis. We also construct or empty pCMVHA construct were trans- demonstrated that allele loss at 3p21.3 in gliomas was fected into glioma cell line H4 (completely methylated a rare event. Although we did not analyse for RASSF1A for and lackingexpression of RASSF1A). Usinga somatic mutations in our series of adult gliomas, colony-formation assay, we show that re-expression of previous studies in various tumor types have demon- RASSF1A significantly (P 0.028) reduces colony-for- ¼ strated that inactivation of RASSF1A by somatic mation ability by approximately 64% (Figure 6). This mutations is a very rare event. It is clear that majority effect was consistent through three independent experi- of glioma tumor cell lines we analysed exhibit only ments. methylated alleles. Even though the glioma tumors used in this study were not microdissected, several tumors exhibited only methylated alleles. Hence, its seems likely Discussion that major mechanism of RASSF1A inactivation in adult gliomas is via biallelic promoter region hyper- To our knowledge, this is the first report describing methylation as has been recently demonstrated for systematic epigenetic analysis of members of the RASSF RASSF1A methylation in medulloblastomas (Lusher gene family in any tumor type. Our data indicate that et al., 2002). RASSF1A is frequently inactivated by promoter region The critical region at 3p21.3 where the RASSF1 gene hypermethylation in adult gliomas and promoter is located was delineated by findings of overlapping hypermethylation is associated with loss of gene homozygous deletions in lung and breast cancer cell

Oncogene Epigenetic inactivation of RASSF1A and BLU L Hesson et al 2415 phila developmental proteins DEAF-1 (Michelson et al., 1999) and nervy (Feinstein et al., 1995). The MYND motif is essential to the function of many of these proteins. BLU as an MYND-domain containingprotein is likely to be involved in important transcriptional regulation pathways. Previously we have demonstrated that BLU is frequently methylated in neuroblastomas and less frequently in lungcancer, and methylation corresponded with gene silencing in tumor lines. Furthermore, in tumor cell lines exogenous expression of BLU reduced colony-formation efficiency by 49–75% (Agathanggelou et al., 2003b). We now demonstrate that BLU is frequently methylated in adult gliomas, and BLU expression can be restored after treatingmethy- lated glioma cell lines with a demethylating agent. BLU methylation in primary tumors also correlated with a reduction in BLU expression. In two tumors where only the methylated allele was detected, BLU expression was markedly downregulated relative to unmethylated tumors. The residual expression in these two tumors may be due to normal infiltratinglymphocytes since the tumors were not microdissected. As was the case in neuroblastomas, there was no association between BLU and RASSF1A methylation in adult gliomas, suggesting that inactivation of BLU is an independent event and not the result of methylation of the 3p21.3 region in Figure 6 RASSF1A inhibits the growth of gliomablastoma cell these tumor types. In fact, in gliomas methylation of the line H4 in vitro.(a) Cells were transfected with the empty vector or BLU gene was more frequent than RASSF1A methyla- vector containingRASSF1A. In total, 64% reduction in colony- tion (P ¼ 0.022). Furthermore, many glioma tumors and formation ability was observed in cells transfected with the wild- tumor lines exhibited only methylated alleles, indicating type RASSF1A construct compared to vector alone. (b) Western blot showingre-expression of RASSF1A (39 kDa) in stably that epigenetic inactivation of BLU plays an important transfected H4 colonies (b). Expression of RASSF1A was not seen role in gliomas. We demonstrated high frequency of in H4 transfected with vector alone (a) BLU methylation in all grades of gliomas, while RASSF1A methylation frequency increased with tumor grade. In a recent study, Liu et al. (2003) also reported that 74% of nasopharyngeal carcinoma underwent BLU lines. The 120 kb minimal critical region contains eight methylation and that 78% of the primary tumors genes, namely, CACNA2D2, PL6, 101F6, NPRL2/G21, demonstrated absent or reduced BLU expression. BLU, RASSF1, FUS1 and HYAL2 (Lerman and CACNA2D2 is a large gene encompassing 141 kb of Minna, 2000) (Figure 3). Further, 11 well-characterized genomic DNA, it lies centromeric of BLU and RASSF1 genes reside in the larger homozygous deleted region, loci and contains a large 50 CpG island. Usingdirect namely, HYAL1, FUS2, HYAL3, IFRD2, SEMA3B, sequencingof bisulfite-modified DNA, we demonstrated GNA12, SLC38A3, GNAT1, SEMA3F, RBM5 and that four of six glioma cell lines were partially RBM6. None of these genes demonstrated frequent methylated, but in primary glioma tumors methylation somatic mutations in lungcancer, but several of these of CACNA2D2 was a rare event. genes demonstrated loss or reduced expression in lung We have recently demonstrated that the axon cancer cell lines (Lerman and Minna, 2000). Epigenetic guidance molecule SLIT2 is frequently methylated in inactivation of RASSF1A has been reported in many lung, breast, colorectal and glioma tumors (Dallol et al., studies in various tumor types, but there has been no 2002, 2003a, b). Semaphorins comprise another large study of systematic epigenetic analysis of other genes family of proteins that play important roles in axonal showingloss of expression from within the 3p21.3 pathfindingduringneural development. SEMA3B is region. We now report methylation status of three located 72 kb telomeric of RASSF1 locus at 3p21.3. (CACNA2D2, BLU, SEMA3B) further genes from this SEMA3B shows loss of expression in many NSCLC cell region in the same series of glioma cell lines and tumors. lines. There have been two previous studies demonstrat- BLU is a candidate 3p21.3 TSG immediately centro- inghypermethylation of the 5 0 CpG island associated meric of RASSF1. BLU contains a predicted MYND with SEM3B gene in NSCLC cancer cell lines and this domain at its C-terminus that shares 45% identity with hypermetylation correlated with loss of gene expression the MYND domains found in important transcription (Tomizawa et al., 2001). Recently, SEMA3B was regulatory proteins such as ETO(MTG8) (Wolford and demonstrated to be methylated in primary non-small- Prochazka, 1998), suppressins (LeBoeuf et al., 1998), cell lungtumors (Kuroki et al., 2003). By direct BS69 (Masselink and Bernards, 2000) and the Droso- sequencingof bisulfite-modified DNA, only one of

Oncogene Epigenetic inactivation of RASSF1A and BLU L Hesson et al 2416 seven glioma cell lines was found to be partially DNA was then ethanol precipitated and resuspended in 50 ml methylated for SEMA3B. Our CACNA2D2 and SE- water. MA3B data further support the importance of RASS- F1A and BLU methylation in gliomas in that it is not a MSP regional methylation effect. The promoter methylation status of RASSF1A, NORE1A, We also demonstrated frequent epigenetic inactiva- BLU, MGMT, CASP8, CLDN7, DAPK1 and GSTP1 were tion of MGMT gene in glioma tumors, while CASP8, determined usingMSP (Herman et al., 1996). The conditions DAPK1, GSTP1 and CLDN7 methylation was a rare and primers used for the MSP reactions are described in event. Previously we had analysed the same series of Table 2. For each gene, two sets of primers were used that gliomas for p16 (0/18) and SLIT2 (37/63; 59%) distinguish between methylated and unmethylated DNA methylation (Dallol et al., 2003b). There was no sequences. All reactions (includingMSP and COBRA) were correlation between RASSF1A, BLU or SLIT2 methy- performed on a GeneAmp 9700 thermocycler (Perkin Elmer) and with HotStar Taq DNA polymerase (Qiagen). SssI in vitro lation status in this series of tumors. Our data also methylated (IVM) normal blood DNA was used as a positive highlight the importance of analysing primary tumors control. since frequent methylation in tumor cell lines does not necessary mean frequent in vivo methylation (in primary Combined bisulfite restriction analysis (CoBRA) and direct tumors). For example, NORE1A, RASSF5, CAC- sequencing NA2D2 and CLDN7 were frequently methylated in glioma cell lines but rarely in tumors. The promoter methylation status of RASSF3, RASSF5, NORE1B and CACNA2D2 were determined usingthe This is a comprehensive methylation profile of a large COBRA method followed by direct sequencingto confirm series of well-characterized gliomas, suggesting that methylation status and ascertain the extent of methylation, if epigenetic inactivation of specific genes may be im- present. SEMA3B was directly sequenced without prior assay portant in glioma development. Our data indicate that usingrestriction enzymes. All primers and a summary of PCR not only RASSF1A but also the BLU gene residing just conditions used for methylation analysis are listed in Table 2. centromeric of RASSF1 is frequently inactivated by The assays used to determine methylation status at the promoter region hypermethylation in gliomas and that CACNA2D2 and RASSF3 CpG islands were novel assays. in many cases it may be biallelic hypermethylation. Our For RASSF3, primers were designed to amplify the putative data warrant further genetic/epigenetic and functional promoter region at À3514À44 bp relative to the transcription analysis of this novel gene (BLU) in order to understand start point of EST BU182700 (PromoterInspector at www.ge- nomatix.de). The region examined for methylation was also its role in tumorigenesis. The elucidation of the within a CpG island located À6334 þ 482 bp relative to the functions of the RASSF1A and BLU proteins may lead transcription start point of EST BU182700 (CpGplot at to new strategies for molecular targeted therapy for www.ebi.ac.uk). This CpG island was 1116 bp in length with certain cancers. an ObsExp of 1.05 and a percentage CG of 74.1% (CpGreport at www.ebi.ac.uk). The conditions and primers were as follows; initial denaturation for 10 min at 951C, followed by 25 cycles of 1 min at 941C, 1 min at 541C and 2 min at 741C Materials and methods with a final extension for 10 min at 721C usingthe primers RASSF3 F 50-TTGAGGAAGYGATTYGAGTATAGTTT- Tumor samples TAGT-30 and RASSF3 R 50-CAAAAACACRTAAAAA- 0 A total of seven glioblastoma tumor cell lines (U87, A172, CAAAAAACRCRAACTAA-3 (where Y ¼ C or T and U343, U373, T17, H4 and Hs683) and 65 primary glioma R ¼ A or G to allow unbiased PCR amplification with regards tumors (10 grade II, 15 grade III and 40 grade IV) were to methylation status). The reaction volume of 20 ml contained analysed for promoter region methylation. All grade II and III 40 ngbisulfite-modified DNA, 1.5 m M MgCl2, 0.25 mM gliomas are oligodendrogliomas. In addition, six normal adult dNTPs, 1 mM each primer and 0.5 U HotStar Taq (Qiagen). brain samples were also analysed for promoter hypermethyla- A measure of 1 ml of this reaction was then used in a tion. These were designated N1, N2, N3, N4, N5 and N6. The seminested PCR reaction (50 mls) usingRASSF3 F (described above) and RASSF3 RN 50-CCACTCACTTATTAAC- typical normal brains collected are those from autopsy for 0 whatever reason (e.g. from auto accidents) and are examined CRAAACRAAACTTA-3 and the same PCR program but by a neuropathologist to document that the brain is normal. with 1 U HotStar Taq, 0.4 mM each primer and an annealing temperature of 561C. This produced a 533 bp product.

Sodium bisulfite modification CACNA2D2 promoter Sodium bisulfite modification was performed as described The sequence AF042794 submitted to Genbank corresponds previously (Agathanggelou et al., 2001). Briefly, 0.5–1.0 mgof to the genomic sequence encompassing exon 1 of CACNA2D2 genomic DNA was denatured in 0.3 M NaOH for 15 min at and has previously been shown to have promoter activity (Gao 371C. Unmethylated cytosine residues were then sulphonated and Minna, personal communication). Primers were designed by incubation in 3.12 M sodium bisulfite (pH5.0) (Sigma) and to amplify the promoter region at À3774À9 bp relative to the 5mM hydroquinone (Sigma) in a thermocycler (hybaid Omn- transcription start point of NM_006030. The region examined E) for 15 s at 991C and 15 min at 501C for 20 cycles. for methylation was also within a CpG island located Sulphonated DNA was then recovered usingthe Wizard À9334 þ 960 bp relative to the transcription start point of DNA cleanup system (Promega), according to the manufac- NM_006030 (CpGplot at www.ebi.ac.uk). This CpG island turer’s instructions. The DNA was desulphonated by addition was 1893 bp in length with an ObsExp of 1.09 and a percentage of 0.3 M NaOH for 10 min at room temperature. The converted CG of 69.65% (CpGreport at www.ebi.ac.uk). The primers

Oncogene Table 2

Gene Genbank Assay Primer set forward Primer set reverse Annealing [MgCl2] Reference Accession temperature (mM) no.

RASSF1A NM_007182 MSP CGAGAGCGCGTTTAGTTTCGTT CGATTAAACCCGTACTTCGCTAA 581C 1.5 Honorio et al., 2003a, USP GGGGGTTTTGTGAGAGTGTGTTT CCCAATTAAACCCATACTTCACTAA Honorio et al., 2003b RASSF3 BU182700 COBRA TTGAGGAAGYGATTYGAGTATAGTTTTAGT CAAAAACACRTAAAAACAAAAAACRCRAACTAA 54 and 561C 1.5 This study RN-CCACTCACTTATTAACCRAAACRAAACTTA RASSF5 NM_032023 COBRA AGGATAYGATATATGTAGTGG TTTTTGGATT ATTATAACCCCTAAATTACTTAACAAA 581C 1.5 This study AATACCAAA NORE1A NM_031437 MSP CGTCGTTTGGTACGGATTTTATTTTTTTCGGTTC GACAACTTTAACAACGACGACTTTAACGACTACG 621C 3.0 Hesson et al., 2003 USP ATTTATATTTGTGTAGATGTTGTTTGGTAT ACTTTAACAACAACAACTTTAACAACTACA 541C NORE1B AF445801 COBRA GTAGTGATAGTYGAGTYGGTTTTGAGGAATT TCCAAACTACAATACCCACTACTCATACTA 581C 1.5 Hesson et al., 2003 RN-ACCRCCAACRCATCAAAAAAACCRAACTTAA BLU NM_015896 MSP TTCGTGGGTTATAGTTCGAGAAAGCG AACGAATTAACCGCGCCTACGC MSP 72–661C 1.5 Agathanggelou et al., 2003b

USP TTTGTGGGTTATAGTTTGAGAAAGTG AACAAATTAACCACACCTACAC USP 66–601C BLU Hesson and L RASSF1A of inactivation Epigenetic SEMA3B NM_004636 Direct sequencingTAACCCTAAAAATATACCCA TATTTTAGTAGTTTAGGGTG 50 1C 3.0 Tomizawa et al., 2001 CACNA2D2 NM_006030 Direct sequencingTTATTATTAAATTTGTGATTTTAGGTTTTAAGTT TCCCTACAACRCTAACTCCAAA 52 1C 1.5 This study

FN-GTTTTTAAAAGGGTTATATATGTTTTTGTT al et MGMT NM_002412 MSP TTTCGACGTTCGTAGGTTTTCGC GCACTCTTCCGAAAACGAAACG 581C 3.0 Esteller et al., 1999 USP TTTGTGTTTTGATGTTTGTAGGTTTTTGT AACTCCACATCTTTCCAAAAACAAAACA CLDN7 NM_001307 MSP GACGTTAGGTTATTTTCGGTC AAACGCGTTTCTAAACGCCG 561C MSP 5.0 Kominsky et al., 2003 USP TGGGGAAAGGGTGGTGTTG TTACCCAATTTTAACCACCAC USP 1.5 CASP8 AF210257 MSP TAGGGGATTCGGAGATTGCGA CGTATATCTACATTCGAAACGA MSP 601C MSP 5.0 Teitz et al., 1998 USP TAGGGGATTTGGAGATTGTGA CCATATATATCTACATTCAAAACAA USP 551C USP 3.0 GSTP1 NM_000852 MSP TTCGGGGTGTAGCGGTCGTC GCCCCAATACTAAATCACGACG 581C 1.5 Esteller et al., 1998 USP GATGTTTGGGGTGTAGTGGTTGTT CCACCCCAATACTAAATCACAACA DAPK1 NM_004938 MSP GGATAGTCGGATCGAGTTAACGTC CCCTCCCAAACGCCGA 591C 5.0 Katzenellenbogen et al., 1999 USP GGAGGATAGTTGGATTGAGTTAATGTT CAAATCCCTCCCAAACACCAA Oncogene 2417 Epigenetic inactivation of RASSF1A and BLU L Hesson et al 2418 used to determine the methylation status of the CACNA2D2 ciences). Primers and conditions used for RASSF1A RT– promoter were CACNA2D2 F50-TTATTATTAAATTTGT- PCR were as described in Burbee et al. (2001). The PCR GATTTTAGGTTTTAAGTT-30 and CACNA2D2 R50- thermocycle (Hybaid Onm-E) consisted of an initial denatura- TCCCTACAACRCTAACTCCAAA-30 usingthe following tion of 10 min at 951C followed by 35 cycles of 951C for 30 s, conditions; initial denaturation for 10 min at 951C followed by 63.51C for 30 s, 721C for 30 s and a final extension of 10 min at 30 cycles of 1 min at 941C, 1 min at 521C, and 2 min at 741C 721C. Primers used for BLU RT–PCR were lungisoform with a final extension for 10 min at 721C. The concentration of specific with primers located in exon 6.2 (50-GCTTAGCACA- 0 0 MgCl2, dNTPs, primers and HotStar Taq DNA polymerase CACAACCTGC-3 ) and exon 8 (5 -CCTTGGCAAAACTTG used were as for RASSF3. A measure of 1 ml of this reaction TGAGG-30) to give an expected product size of 233 bp. The was then used in a seminested PCR reaction (50 mls) using PCR thermocycle (Hybaid Onm-E) consisted of an initial CACNA2D2 FN 50-GTTTTTAAAAGGGTTATATATG- denaturation of 10 min at 951C followed by 35 cycles of 951C TTTTTGTT-30 and CACNA2D2 R (described above) and for 30 s, 581C for 30 s, 721C for 301C and a final extension of the same PCR program but with 1 U HotStar Taq and 0.8 mM 10 min at 721C. Primers used for the GAPDH control were 50- each primer. This produced a 368 bp product. All RASSF3, TGAAGGTCGGAGTCAACGGATTTGGT-30 and 50-CAT CACNA2D2, RASSF5 and NORE1B PCR products were GTGGGCCATGAGGTCCACCAC-30. PCR products were assayed for methylation by incubation with TaqI and BstUI visualized on 2% agarose gel with added ethidium bromide. for 2 h at 65 and 601C, respectively, before visualization on a 2% agarose gel with added ethidium bromide. The SEMA3B Quantitative real-time RT–PCR in primary glioma tumors CpG island methylation status was determined by direct sequencingof PCR products obtained usingthe primers and The theoretical and practical aspects of real-time quantitative conditions described in Tomizawa et al. (2001). The PCR RT–PCR usingthe ABI Prism 7700 Sequence Detection products obtained were purified usingthe QIAquick PCR System (Perkin-Elmer Applied Biosystems) have been de- Purification Columns (Qiagen – according to manufacturer’s scribed in detail elsewhere (Bieche et al., 2001). Briefly, total instructions), reamplified usingABI BigDyeCycle Sequencing RNA is reverse-transcribed before real-time PCR amplifica- Kit (Perkin Elmer) and analysed usingan ABI Prism 377 DNA tion. Quantitative values are obtained from the threshold cycle sequencer (Perkin Elmer). (Ct) number at which the increase in the signal associated with SssI in vitro methylated (IVM) normal blood DNA was used exponential growth of PCR products begins to be detected as a positive control. usingPE Biosystems analysis software, accordingto the manufacturer’s manuals. The precise amount of total RNA LOH analysis added to each reaction mix (based on optical density) and its quality (i.e. lack of extensive degradation) are both difficult to Possible allele loss of CACNA2D2, RASSF1, BLU or assess. We therefore also quantified transcripts of the gene SEMA3B was determined usingtwo polymorphic markers codingfor the TATA box-bindingprotein ( TBP) (a compo- from within the 3p21.3 region. D3S1568 and D3S3667 nent of the DNA-bindingprotein complex TFIID) as the represent the nearest flankingpolymorphic loci. Primers used endogenous RNA control, and each sample was normalized on to amplify D3S1568 were 50-AGCAAGGTTTGTCTGT- 0 0 0 the basis of its TBP content. Results, expressed as N-fold GAAA-3 and 5 -GAACCAGCAAGTTCATCAGA-3 to differences in BLU gene expression relative to the TBP gene, produce a product between 276–296 bp. Primers used to 0 termed NBLU, were determined by the formula: amplify D3S3667 were 5 -TTTAAGAAGCCTTTGTCA- sample 0 0 0 NBLU ¼ 2DCt , where DCt value of the sample was GAA-3 and 5 -CTATTGATGTTTCATGGTGC-3 to pro- determined by subtractingthe C value of the target gene duce a product between 167–183 bp. PCR conditions for both t from the Ct value of the TBP gene. The nucleotide sequences primer pairs were 1.5 mM MgCl2, 0.25 mM dNTPs, 0.8 mM each of the primers used for PCR amplification were: BLU-U (50- primer, 100 ngDNA and 0.5 U HotStar Taq. The PCR GACAGTGTTCTTCCACAAGGAGGT-30) and BLU-L (50- thermocycle for both primer pairs consisted an initial CAGTTTGCGGTGGCAATAGTCT-30) with a BLU-specific 1 denaturation of 10 min at 95 C followed by 35 cycles of 30 s product size of 82 bp, and TBP-U (50-TGCACAGGAGCCAA at 951C, 30 s at 501C, 30 s at 721C with a final extension for GAGTGAA-30) and TBP-L (50-CACATCACAGCTCCC- 10 min at 721C. PCR products were then resolved in a 7% CACCA-30) with a TBP-specific product size of 132 bp. PCR PAGE gel and visualized via silver staining. was performed usingthe SYBR s Green PCR Core Reagents kit (Perkin-Elmer Applied Biosystems). To avoid amplification Cell lines and 5-aza-20deoxycytidine treatment of contaminatinggenomicDNA, one of the two primers was Glioblastoma cell lines were routinely maintained in RPMI placed at the junction between two exons. The thermal cycling 1640 growth media (Invitrogen) supplemented with 10% FCS conditions comprised an initial denaturation step at 951C for 5 10 min and 50 cycles at 951C for 15 s and 651C for 1 min. at 371C, 5% CO2. Cells (5–10 Â 10 ) were plated and allowed 24 h growth before addition of 5 mM 5-aza-20-deoxycytidine (Sigma) freshly prepared in ddH20 and filter-sterilized. The Growth suppression analysis medium (including5 mM 5-aza-20-deoxycytidine) was changed 24 h after seedingand then every 2 days. Controls without 5- Glioma cell line H4 (not expressing RASSF1A due to aza-20-deoxycytidine were cultured concomitantly in the same methylation) was used in a colony-formation assay. Using manner. RNA was prepared 5 and 7 days after treatment using Fugene6 Transfection Reagent (Roche), 2 mgof pCMVHA- the RNeasy kit (Qiagen), according to the manufacturer’s RASSF1A mammalian expression construct or pCMVHA- 5 instructions. empty construct was used to transfect 1 Â 10 cells and allowed 48 h for expression of G418 selection marker. At 48 h after transfection, cells were maintained in DMEM–10% RT–PCR fetal bovine serum supplemented with 600 mggeneticin We analysed the expression of RASSF1A and BLU mRNA in (Sigma). Surviving colonies were counted 14 days later after glioblastoma tumor cell lines. RT–PCR was performed using stainingwith crystal violet. The experiment was repeated three Ready-to-go RT–PCR reaction beads (Amersham Bios- times.

Oncogene Epigenetic inactivation of RASSF1A and BLU L Hesson et al 2419 Statistical analysis Acknowledgements This work was in part supported by Cancer Research UK. Comparisons were made usingFisher’s exact test or w2 test as appropriate. P-values o0.05 were taken as statistically signiﬁcant.

References

Agathanggelou A, Bieche I, Ahmed-Choudhury J, Nicke B, Katzenellenbogen RA, Baylin SB and Herman JG. (1999). Dammann R, Baksh S, Gao B, Minna JD, Downward J, Blood, 93, 4347–4353. Maher ER and Latif F. (2003a). Cancer Res., 63, 5344–5351. Khokhlatchev A, Rabizadeh S, Xavier R, Nedwidek M, Chen Agathanggelou A, Dallol A, Zochbauer-Muller S, Morrissey T, ZhangXF, Seed B and Avruch J. (2002). Curr. Biol., 12, C, Honorio S, Hesson L, Martinsson T, Kuo M, Yuen PW, 253–265. Maher ER, Minna JD and Latif F. (2003b). Oncogene, 22, Kominsky SL, Argani P, Korz D, Evron E, Raman V, Garrett 1580–1588. E, Rein A, Sauter G, Kallioniemi OP and Sukumar S. Agathanggelou A, Honorio S, Macartney DP, Martinez A, (2003). Oncogene, 22, 2021–2033. Dallol A, Rader J, Fullwood P, Chauhan A, Walker R, Kuroki T, Trapasso F, Yendamuri S, Matsuyama A, Alder H, Shaw JA, Hosoe S, Lerman MI, Minna JD, Maher ER and Williams NN, Kaiser LR and Croce CM. (2003). Cancer Latif F. (2001). Oncogene, 20, 1509–1518. Res., 63, 3352–3355. Astuti D, Agathanggelou A, Honorio S, Dallol A, Martinsson LeBoeuf RD, Ban EM, Green MM, Stone AS, Propst SM, T, Kogner P, Cummins C, Neumann HP, Voutilainen R, Blalock JE and Tauber JD. (1998). J. Biol. Chem., 273, Dahia P, EngC, Maher ER and Latif F. (2001). Oncogene, 361–368. 20, 7573–7577. Lerman MI and Minna JD. (2000). Cancer Res., 60, Bieche I, Parfait B, Le Doussal V, Olivi M, Rio MC, Lidereau 6116–6133. R and Vidaud M. (2001). Cancer Res., 61, 1652–1658. Liu XQ, Chen HK, ZhangXS, Pan ZG, Li A, FengQS, Long Burbee DG, Forgacs E, Zochbauer-Muller S, Shivakumar L, QX, WangXZ and ZengYX. (2003). Int. J. Cancer, 106, FongK, Gao B, Randle D, Kondo M, Virmani A, Bader S, 60–65. Sekido Y, Latif F, Milchgrub S, Toyooka S, Gazdar AF, Louis DN, Pomeroy SL and Cairncross JG. (2002). Cancer Lerman MI, Zabarovsky E, White M and Minna JD. (2001). Cell, 1, 125–128. J. Natl. Cancer Inst., 93, 691–699. Lusher ME, Lindsey JC, Latif F, Pearson ADJ, Ellison DW Carboni GL, Gao B, Nishizaki M, Xu K, Minna JD, Roth JA and Clifford SC. (2002). Cancer Res., 62, 5906–5911. and Ji L. (2003). Oncogene, 22, 615–626. Masselink H and Bernards R. (2000). Oncogene, 19, Dallol A, Da Silva NF, Viacava P, Minna JD, Bieche I, Maher 1538–1546. ER and Latif F. (2002). Cancer Res., 62, 5874–5880. Michelson RJ, Collard MW, Ziemba AJ, Persinger J, Dallol A, Krex D, Hesson L, EngC, Maher ER and Latif F. Bartholomew B and Huggenvik JI. (1999). J. Biol. Chem., (2003b). Oncogene, 22, 4611–4616. 274, 30510–30519. Dallol A, Morton D, Maher ER and Latif F. (2003a). Cancer Mitic LL, Van Itallie CM and Anderson JM. (2000). Res., 63, 1054–1058. Am. J. Physiol. Gastrointest. Liver Physiol., 279, Dammann R, Li C, Yoon JH, Chin PL, Bates S and Pfeifer G250–G254. GP. (2000). Nat. Genet., 25, 315–319. Ortiz-Vega S, Khokhlatchev A, Nedwidek M, Zhang XF, Esteller M, Corn PG, Urena JM, Gabrielson E, Baylin SB and Dammann R, Pfeifer GP and Avruch J. (2002). Oncogene, Herman JG. (1998). Cancer Res., 58, 4515–4518. 21, 1381–1390. Esteller M, Hamilton SR, Burger PC, Baylin SB and Herman Pfeifer GP, Yoon JH, Liu L, Tommasi S, Wilczynski SP and JG. (1999). Cancer Res., 59, 793–797. Dammann R. (2002). Biol. Chem., 383, 907–914. Feinstein PG, Kornfeld K, Hogness DS and Mann RS. (1995). Shivakumar L, Minna J, Sakamaki T, Pestell R and White Genetics, 140, 573–586. MA. (2002). Mol. Cell. Biol., 22, 4309–4318. Harada K, Toyooka S, Shivapurkar N, Maitra A, Reddy JL, Shivapurkar N, Toyooka S, Eby MT, HuangCX, Sathyanar- Matta H, Miyajima K, Timmons CF, Tomlinson GE, ayana UG, Cunningham HT, Reddy JL, Brambilla E, Mastrangelo D, Hay RJ, Chaudhary PM and Gazdar AF. Takahashi T, Minna JD, Chaudhary PM and Gazdar AF. (2002). Cancer Res., 62, 5897–5901. (2002). Cancer Biol. Ther., 1, 65–69. Herman JG, Graff JR, Myohanen S, Nelkin BD and Baylin Teitz T, Wei T, Valentine MB, Vanin EF, Grenet J, Valentine SB. (1996). Proc. Natl. Acad. Sci. USA, 93, 9821–9826. VA, Behm FG, Look AT, Lahti JM and Kidd VJ. (2000). Hesson L, Dallol A, Minna JD, Maher ER and Latif F. (2003). Nat. Med., 6, 529–535. Oncogene, 22, 947–954. Tomizawa Y, Sekido Y, Kondo M, Gao B, Yokota J, Roche J, Holland EC. (2001). Nat. Rev. Genet., 2, 120–129. Drabkin H, Lerman MI, Gazdar AF and Minna JD. (2001). Honorio S, Agathanggelou A, Schuermann M, Pankow W, Proc. Natl. Acad. Sci. USA, 98, 13954–13959. Viacava P, Maher ER and Latif F. (2003a). Oncogene, 22, Tommasi S, Dammann R, Jin SG, ZhangXF, Avruch J and 147–150. Pfeifer GP. (2002). Oncogene, 21, 2713–2720. Honorio S, Agathanggelou A, Wernert N, Rothe M, Maher Wolford JK and Prochazka M. (1998). Gene, 212, ER and Latif F. (2003b). Oncogene, 22, 461–466. 103–109.

Oncogene