Frequent epigenetic inactivation of the RASSF1A tumor suppressor ...

Oncogene (2004) 23, 1326–1331

& 2004 Nature Publishing Group All rights reserved 0950-9232/04 $25.00 www.nature.com/onc

Frequent epigenetic inactivation of the RASSF1A tumor suppressor gene in Hodgkin’s lymphoma Paul G Murray*,1, Guo-Hua Qiu2, Li Fu2, Elyse R Waites1, Gopesh Srivastava3, Duncan Heys1, Angelo Agathanggelou4, Farida Latif4, Richard G Grundy5, Jillian R Mann5, Jane Starczynski6, John Crocker6, Sheila E Parkes5, Richard F Ambinder7, Lawrence S Young8 and Qian Tao*,2,7

ONCOGENOMICS

1

Department of Pathology, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK; 2Cancer Epigenetics/Tumor Virology Laboratory, Johns Hopkins Singapore, Singapore; 3Department of Pathology, University of Hong Kong, China; 4Section of Medical and Molecular Genetics, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK; 5Department of Oncology, The Children’s Hospital, Steelhouse Lane, Birmingham B4 6NH, UK; 6Department of Cellular Pathology, Birmingham Heartland’s Hospital, Bordesley Green East, Birmingham B15 2TT, UK; 7Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins School of Medicine, Baltimore, MD, USA; 8Cancer Research UK Institute for Cancer Studies, Division of Cancer Studies, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK

Epigenetic inactivation of RASSF1A, a putative tumor suppressor with proapoptotic activity, is frequently observed in a number of solid tumors, including a variety of epithelial cancers, but has not been described in hematopoietic tumors. We have analysed the expression and methylation status of RASSF1A in Hodgkin’s lymphoma (HL)-derived cell lines, primary HL tumors and serum samples from HL patients. RASSF1A transcription was detectable in only 2/6 HL cell lines. Methylation-specific PCR and bisulfite genomic sequencing revealed that the RASSF1A promoter was hypermethylated in all four RASSF1A-nonexpressing cell lines. 5-aza-20 -deoxycytidine treatment resulted in demethylation of the promoter and RASSF1A expression in these lines. Hypermethylation of RASSF1A was also detected in 34/52 (65%) primary HL tumors and in 2/22 serum samples from these patients. Microdissection of Hodgkin/ Reed–Sternberg (HRS) cells from several of these cases confirmed that the RASSF1A hypermethylation we detected in the analysis of whole tumor originated from the tumor cell population. Although hypermethylation of RASSF1A was detected in 5/6 non-Hodgkin’s lymphoma (NHL)-derived cell lines, only rare primary NHL (1/10 of Burkitt’s lymphoma, 1/12 of post-transplant lymphoma, 1/12 diffuse large B-cell lymphoma, 0/27 of nasal lymphoma, 0/8 follicular center cell lymphoma, 0/4 mantle cell lymphoma, 0/4 anaplastic large cell (Ki-1 þ ) lymphoma, 0/2 MALT lymphoma) showed hypermethylation of the promoter. No methylation was detected in any of the 14 normal PBMC. These results point to an important role for epigenetic silencing of RASSF1A in the pathogenesis of HL. Inactivation of RASSF1A could be one mechanism by which HRS cells escape the apoptosis

*Correspondence: PG Murray; E-mail: [email protected]. Q Tao, Cancer Epigenetics/Tumor Virology Laboratory, Johns Hopkins Singapore, CRC (MD11), Level 5, 10 Medical Drive, Singapore 117597, Singapore; E-mail: [email protected] Received 10 March 2003; revised 15 October 2003; accepted 25 October 2003

that should occur following nonproductive immunoglobulin gene rearrangements. Oncogene (2004) 23, 1326–1331. doi:10.1038/sj.onc.1207313 Keywords: Hodgkin’s lymphoma; RASSF1A; hypermethylation; tumor suppressor gene; Hodgkin/Reed– Sternberg cells

RASSF1 had been suggested as an important candidate tumor suppressor gene on the basis of its frequent loss of heterozygosity (LOH) and epigenetic silencing in lung cancers (Dammann et al., 2000). RASSF1 encodes several isoforms, including RASSF1A, RASSF1B and RASSF1C, which are derived from alternative mRNA splicing and promoter usage (Dammann et al., 2000). The three isoforms share four common exons (exons 3–6), which encode a Ras association domain (Ponting and Benjamin, 1996). RASSF1A has two 50 exons (1a and 2ab) and encodes a 39 kDa peptide (Dammann et al., 2000). The tumor suppressor function of RASSF1A is suggested by studies showing that exogenous expression of RASSF1A decreases in vitro colony formation, suppresses anchorage-independent growth and dramatically reduces tumorigenicity in vivo (Burbee et al., 2001; Dreijerink et al., 2001). RASSF1A can induce cell cycle arrest by engaging the Rb family cell cycle checkpoint and inhibiting the accumulation of cyclin D1 (Shivakumar et al., 2002). The ability of RASSF1A to heterodimerize with Nore and thereby associate with Ras-like GTPases is likely to be important for its tumor suppressor and proapoptotic activities (Vos et al., 2000; Khokhlatchev et al., 2002; Ortiz-Vega et al., 2002). Inactivation of RASSF1A has so far only been reported in epithelial tumors, including breast, ovarian kidney, bladder and nasopharyngeal carcinomas (Agathanggelou et al., 2001; Burbee et al., 2001; Dammann et al., 2001; Lo et al., 2001; Lee et al, 2001;

RASSF1A hypermethylation in Hodgkin’s lymphoma PG Murray et al

1327

Figure 1 (a) Downregulation and hypermethylation of the RASSF1A isoform and its promoter in HL and NHL cell lines (NPC cell lines were used as controls). RT–PCR analysis (upper panel) demonstrated expression of RASSF1C in all cell lines, but 4/6 HL- and 5/ 6 NHL-derived cell lines lacked RASSF1A transcription. MSP analysis showed methylation of the RASSF1A promoter in nonexpressing cell lines (m: methylated; u: unmethylated). Total RNA was extracted by TRI REAGENT (Molecular Research Center, Inc., Cincinnati, OH, USA). RT–PCR was performed using AmpliTaq Gold (PE Biosystem, Foster City, CA, USA) (Tao et al., 2002). The primers for RASSF1A and C isoforms were: RASSF1AC (for RASSF1A, in exon 1a), 50 -gacctctgtggcgacttc; RDAA (for RASSF1C, in exon 2g), 50 -gaggcgccttctttcgaa; RDAB (reverse primer, for both isoforms in exon 4), 50 -caaggagggtggcttctt. The PCR program utilized initial denaturation at 951C for 10 min, followed by 35 cycles (941C for 30 s, 531C for 30 s and 721C for 30 s), with a final extension at 721C for 10 min. Genomic DNA for MSP was extracted from tumors, PBMCs and cell pellets with TRI REAGENT and from serum using the QIAamp UltraSens kit (Qiagen Ltd, catalog no. 53704). Genomic DNA was treated by bisulfite as previously described (Tao et al., 2002). MSP was conducted according to our previous report (Tao et al., 1999). Primers used were: for the methylated promoter, RASSF1Am1, 50 -ggttttttttagttttttttcgtc; RASSF1Am6, 50 -ctaccgtataaaattacacgcg; for the unmethylated promoter, RASSF1Au1, 50 -tggttttttttagtttttttttgtt; RASSF1Au6, 50 -actaccatataaaattacacaca. PCR was conducted at 951C for 10 min, then 40 cycles (941C, 30 s; 57 or 601C, 30 s; 721C, 30 s), followed by 721C for 5 min. For serum samples, 45 cycles were used. (b) In our hands, the sensitivity of the MSP analysis for the detection of RASSF1A methylation was determined to be 1% (m/u þ m). (c) Induction of RASSF1A expression and demethylation of its promoter in HL cell lines by 5-aza-dC treatment. Cells were suspended at 1 105 cells/ml in cRPMI-1640 and allowed to grow overnight. 5-aza-dC was added to the desired concentrations (5 mM, dissolved in H2O). Controls were carried out by adding distilled H2O instead of drug. Cells were treated for up to 6 days. Fresh medium containing 5-aza-dC was replaced every 24 h. After the treatment, cells were pelleted and washed with PBS, and DNA and RNA extracted. (d) Detection of hypermethylated RASSF1A promoter by MSP in HL tumors, serum samples from HL patients, and tumor tissues of diffuse large B-cell lymphomas (DLCL)

Oncogene


1328

Maruyama et al., 2001; Morrissey et al., 2001), and in neuroblastomas and pheochromocytomas (Astuti et al., 2001). A recent survey of pediatric tumors also showed frequent hypermethylation of RASSF1A in rhabdomyosarcoma, retinoblastoma and medulloblastoma (Harada et al., 2002). However, to date there are no studies that demonstrate a role for RASSF1A inactivation in the pathogenesis of lymphoid malignancies. Hodgkin’s lymphoma (HL) is characterized by malignant Hodgkin and Reed–Sternberg (HRS) cells in a background of reactive cells. In most cases, HRS

cells derive from germinal center or postgerminal center B cells with nonfunctional immunoglobulin gene rearrangements. Therefore, HRS cells must have acquired cellular changes that enable them to avoid the apoptosis that is the normal fate of such B cells (Kanzler et al., 1996). We studied methylation of the RASSF1A promoter in HL- and non-Hodgkin’s lymphoma (NHL)-derived cell lines and in the tumor tissues of 52 HL patients, including eight pediatric cases. In 22 cases, serum was also available. Tissues from post-transplant lymphopro-

Figure 2 Detailed methylation analysis of the RASSF1A promoter in (a) HL cell lines, (b) normal PBMCs, and (c) primary tumors of HL and NHL. No methylated allele was present in any PBMC sample. All the RASSF1A alleles in the four nonexpressing HL cells lines were methylated, but they became partially unmethylated after the 5-aza-dC treatment (as illustrated here for L428 and KM-H2). The presence of methylated alleles in 6/8 primary HL tumors (HDB10, HDB11, HD3ad, HD9, HDB19, HDB20) is shown. Also a single BL showing methylated RASSF1A and PTLD8 with unmethylated RASSF1A gene is shown. Note that for samples HDB2 and HDB3, the initial BGS analysis could not detect the presence of methylated alleles despite the detection of methylated RASSF1A sequences by MSP analysis (shown in Figure 1d). However, by screening 96 colonies, we identified two methylated alleles in HDB2 and onw methylated allele in HDB3 (d). The failure to detect methylated alleles in the initial BGS screen was almost certain due to the low number of tumor cells in these samples. CpG sites are shown on the top row as numbers. The RASSF1A transcription start sites are shown as bent arrows. Each row of circles represents a single allele of the RASSF1A promoter analysed. Open circles: unmethylated CpG sites; filled circles: methylated CpG sites. BGS was performed as described previously (Tao et al., 2002). The strand-specific primers for the bisulfite-converted single-stranded DNA of the RASSF1A promoter were: BGS5, 50 -GTTAAGTGTGTTGTTTTAGTAAAT; BGS4, 50 -CCCRCAACTCAATAAACTCAAA. These two primers amplify a 388-bp region with 34 CpG sites in the RASSF1A promoter. For BGS, at least 10 bacterial colonies were analysed for each DNA sample Oncogene


1329

liferative disease (PTLD), endemic Burkitt’s lymphoma (BL) and a range of other NHL were also studied. We initially evaluated the expression levels of RASSF1A and RASSF1C mRNA in HL and NHL cell lines by RT–PCR (Figure 1a, upper panel). RASSF1C was expressed in all cell lines. However, the expression of RASSF1A was absent in 4/6 HL and 5/6 NHL cell lines. This lack of RASSF1A expression was not the consequence of gene deletion since further PCR analysis detected the presence of the gene in all of these cell lines (data not shown). To investigate whether aberrant DNA methylation was involved in the transcriptional silencing of RASSF1A, the methylation status of RASSF1A was assessed. MSP detected hypermethylation of the RASSF1A promoter in all the cell lines lacking RASSF1A expression, while the RASSF1A-expressing cell lines all had unmethylated promoter (Figure 1a, lower panel). To further test whether hypermethylation of the RASSF1A promoter mediates its gene repression, HL cell lines were treated with the demethylating agent 5aza-20 -deoxycytidine (5-aza-dC) (Figure 1c). Following this treatment, transcription of the RASSF1A gene was restored in all three cell lines studied that lacked RASSF1A expression. Accordingly, the RASSF1A promoter became demethylated. These results indicate that hypermethylation of the RASSF1A promoter directly mediates its transcriptional repression in HL cell lines. To additionally define the methylation status of the gene, we performed bisulfite genomic sequencing (BGS) to analyse the methylation status of every individual CpG site of the CpG island in the RASSF1A promoter. The region (338 to þ 50) spanning 34 CpG sites was amplified by PCR using sodium bisulfite-modified genomic DNA as templates. The 34 CpG sites were virtually completely methylated in all RASSF1A-nonexpressing cell lines (Figure 2a). Demethylated alleles were seen in the 5-aza-dC-treated cell lines. We then studied whether this hypermethylation of RASSF1A promoter exists in primary HL and NHL tumors. MSP analysis showed evidence of promoter hypermethylation in 29/44 (66%) cases of adult HL and 5/8 cases of pediatric HL (Figure 1d; Tables 1 and 2), but only rarely in PTLD (1/12 weakly methylated), BL (1/10), diffuse large B-cell lymphoma (1/12) and not in any of 27 nasal lymphomas, eight follicular center cell lymphomas, four mantle cell lymphomas, nor in two MALT lymphomas (Table 2), despite the detection of hypermethylation in the majority of NHL cell lines. Interestingly, we could not detect hypermethylated RASSF1A in four cases of anaplastic large cell (Ki1 þ ) lymphoma (ALCL), which closely resembles HL in terms of the tumor cell phenotype. Hypermethylation of RASSF1A could not be detected in 14 peripheral blood mononuclear cell (PBMC) samples from normal donors either. Detailed methylation profiling by BGS confirmed the MSP results in some cases (Figure 2b–d). Furthermore, we performed microdissection of HRS cells from several biopsy specimens of HL to confirm that the RASSF1A hypermethylation we had detected by analysis of whole tumor material originated from the tumor

Table 1 Histology, EBV positivity and RASSF1A methylation status of primary HL tumors HL cases

1 2 3 4 5 6 8 9 10 11 12 B1 B2 B3 B4 B7 B8 B9 B10 B11 B12 B13 B15 B16 (pediatric) B17 B18 B19 B20 (pediatric) B22 B23 B25 B26 B27 B28 B29 B30 HDad1 HDad2 HDad3 HDad4 HDad5 HDad6 HDad7 HDad8 HDad9 HDad10 P1 (pediatric) P2 (pediatric) P3 (pediatric) P4 (pediatric) P5 (pediatric) P8 (pediatric)

Histology

EBV positivity (subtype)

RASSF1A methylation status

NS NS NS MC MC NS UC UC UC UC UC NS NS NS NS LP NS NS LD NS LP Interfollicular NS NS NS NS NS NS NS NS NS NS LP LP NS NS NS MC NS NS MC MC LP MC NS NS MC NS MC UC UC UC

A A A A A A A — — — — A A A A — A A A A — A — A — A — — — A A A — — — — — A — — A A — A — — A — A NA NA NA

M M M u M M u m (m) U M U m M M U (m) u M M (M) u u U m m M M M M M U U u m m M U M U M M M M U U U M U M M M

—: no EBV detected; A: EBV type A present; NA: not available; NS: nodular sclerosis type; MC: mixed cellularity type; LP: lymphocyte predominant type; LD: lymphocyte depletion type; UC: not classified; M: methylated; (M): weakly methylated; U: unmethylated

cells (Figure 3). There was no correlation between the RASSF1A promoter status and either histological subtype or Epstein–Barr virus (EBV) status in the HL tumors (Table 1). Serum DNA samples from a total of 22 patients with HL were available. MSP detected RASSF1A methylation in two of these samples Oncogene


1330 Table 2 Summary of RASSF1A promoter methylation in primary HL and NHL tumors, sera from HL patients, and normal PBMC Samples

RASSF1A methylated (%)

HL

Adult HL Pediatric HL HL sera

NHL

PTLD BL Nasal lymphoma DLCL FCCL MALT lymphoma ALCL Mantle cell lymphoma

Normal PBMC

29/44 (66%) 5/8 (63%) 2/22 (9%) 1/12 (8%) 1/10 (10%) 0/27 1/12 0/8 0/2 0/4 0/4 0/14

PTLD: post-transplant lymphoproliferative disease; BL: Burkitt’s lymphoma; DLCL: diffuse large B-cell lymphoma; FCCL: follicular center cell lymphoma; ALCL: anaplastic large cell (Ki-1+) lymphoma

(Figure 1d). In both cases, their corresponding tumors also showed RASSF1A methylation. An important pathogenic event in HL is bypass from the apoptosis that would otherwise eliminate B cells with defectively rearranged Ig genes. Therefore, genes whose inactivation provides an antiapoptotic mechanism are likely candidates for a pathogenic role in HL. RASSF1 overexpression induces apoptosis through its interaction with Ras (Vos et al., 2000), an effect that may be mediated by heterodimerization with NORE1 (OrtizVega et al., 2002). There is also evidence for an association of both NORE1 and RASSF1 with the proapoptotic kinase MST1, and that this interaction is involved in apoptosis induced by activated Ras (Khokhlatchev et al., 2002). Therefore, inactivation of RASSF1A could be one mechanism by which HRS progenitors escape the apoptosis that should follow nonproductive Ig rearrangements. The latent membrane protein-1 (LMP1) of EBV induces expression and activity of the DNA methyltransferases 1, 3A and 3B, causing hypermethylation of the E-cadherin promoter (Tsai et al., 2002). LMP1 is highly expressed in EBV-associated HL (Murray et al., 1992). Although our analysis revealed no correlation between EBV status and RASSF1A methylation, this does not rule out the possibility that LMP1 contributes to RASSF1A hypermethylation in EBV-associated HL, but other mechanisms are operational in the EBVnegative tumors. We also detected hypermethylation of RASSF1A in sera of two HL patients. Aberrant methylation of TSGs in the serum of cancer patients has been reported previously in a number of studies (Lee et al., 2002; Zou et al., 2002). Despite the fact we had not optimized serum collection, storage or extraction of DNA specifically for this analysis, our results suggest that the detection of RASSF1A hypermethylation in the serum of HL patients should be evaluated as a potential noninvasive molecular marker of treatment response or Oncogene

Figure 3 Detailed methylation analysis of the RASSF1A promoter in microdissected HL tumors. (a) Microdissection of HRS cells was performed on several biopsy specimens using the Leica LMD System following CD30 staining of HRS cells. DNA was extracted from about 50 pooled microdissected HRS cells in each case using the QIAamp MinElute virus kit (Qiagen), and bisulfite-treated. Bisulfited DNA was PCR amplified using strand-specific primers for the converted single-strand DNA of the RASSF1A promoter: BGS6, 50 -CCAAABGS3: 50 -tatagtaaagttggtttttagaaat; TAAAATCRCCACAAAAAT. (b) The PCR product was then amplified by nested MSP using primers RASSF1Am1 and RASSF1Am4 (100-bp product): 50 -CGCTAACAAACGCGAACCG; RASSF1Au1 and RASSF1Au4 (104-bp product): 50 AAACACTAACAAACACAAACCA. PCR cycles were 45 with an annealing temperature of 571C. Two HL tumors (P5, P8) showed methylated RASSF1A and one tumor (P1) lacked methylation when these tumors were studied by MSP analysis of homogenized tumor. The nested MSP analysis of HRS cells microdissected from these tumors is shown. (c) Nested BGS analysis of microdissected HRS cells. The BGS3/BGS6 PCR product was amplified again using primers BGS3 and SCR2: 50 cccccaaaatccaaactaaa. The final PCR product containing CpG sites #3–22 of the RASSF1A promoter was cloned and sequenced. CpG sites are shown on the top row as numbers. Each row of circles represents a single clone analysed. Open circles: unmethylated CpG sites; filled circles: methylated CpG sites; circles with cross inside: CpG site abolished by mutations. All the colonies analysed in P5 showed methylation, implying that both alleles of the RASSF1A promoter were methylated in this case

disease relapse in patients whose tumors harbor such changes. In summary, our results demonstrate the frequent inactivation of the RASSF1A gene by promoter hypermethylation in the majority of cases of HL, but rarely in NHL. Clearly, LOH or mutational analysis will be required to evaluate the additional contribution of other genetic events to RASSF1A silencing in this disease. Acknowledgements We are grateful to Pamela Edwards, Marie Smith and Risa Mann for help in the collection of tissue specimens, and Jianming Ying, Jing Tan and Cynthia Lee for their technical support.


1331 References Agathanggelou A, Honorio S, Macartney DP, Martinez A, Dallol A, Rader J, Fullwood P, Chauhan A, Walker R, Shaw JA, Hosoe S, Lerman MI, Minna JD, Maher ER and Latif F. (2001). Oncogene, 20, 1509–1518. Astuti D, Agathanggelou A, Honorio S, Dallol A, Martinsson T, Kogner P, Cummins C, Neumann HP, Voutilainen R, Dahia P, Eng C, Maher ER and Latif F. (2001). Oncogene, 20, 7573–7577. Burbee DG, Forgacs E, Zochbauer-Muller S, Shivakumar L, Fong K, Gao B, Randle D, Kondo M, Virmani A, Bader S, Sekido Y, Latif F, Milchgrub S, Toyooka S, Gazdar AF, Lerman MI, Zabarovsky E, White M and Minna JD. (2001). J. Natl. Cancer Inst., 93, 691–699. Dammann R, Li C, Yoon JH, Chin PL, Bates S and Pfeifer GP. (2000). Nat. Genet., 25, 315–319. Dammann R, Yang G and Pfeifer GP. (2001). Cancer Res., 61, 3105–3109. Dreijerink K, Braga E, Kuzmin I, Geil L, Duh FM, Angeloni D, Zbar B, Lerman MI, Stanbridge EJ, Minna JD, Protopopov A, Li J, Kashuba V, Klein G and Zabarovsky ER. (2001). Proc. Natl. Acad. Sci. USA, 98, 7504–7509. Harada K, Toyooka S, Maitra A, Maruyama R, Toyooka KO, Timmons CF, Tomlinson GE, Mastrangelo D, Hay RJ, Minna JD and Gazdar AF. (2002). Oncogene, 21, 4345–4349. Kanzler H, Kuppers R, Hansmann ML and Rajewsky K. (1996). J. Exp. Med., 184, 1495–1505. Khokhlatchev A, Rabizadeh S, Xavier R, Nedwidek M, Chen T, Zhang XF, Seed B and Avruch J. (2002). Curr. Biol., 12, 253–265. Lee MG, Kim HY, Byun DS, Lee SJ, Lee CH and Kim JI et al. (2001). Cancer Res., 61, 6688–6692.

Lee TL, Leung WK, Chan MWY, Ng EKW, Tong JHM, Lo KW, Chung SCS, Sung JJY and To KF. (2002). Clin. Cancer Res., 8, 1761–1766. Lo KW, Kwong J, Hui AB, Chan SY, To KF, Chan AS, Chow LS, Teo PM, Johnson PJ and Huang DP. (2001). Cancer Res., 61, 3877–3881. Maruyama R, Toyooka S, Toyooka KO, Harada K, Virmani AK, Zochbauer-Muller S, Farinas AJ, Vakar-Lopez F, Minna JD, Sagalowsky A, Czerniak B and Gazdar AF. (2001). Cancer Res., 61, 8659–8663. Morrissey C, Martinez A, Zatyka M, Agathanggelou A, Honorio S, Astuti D, Morgan NV, Moch H, Richards FM, Kishida T, Yao M, Schraml P, Latif F and Maher ER. (2001). Cancer Res., 61, 7277–7281. Murray PG, Young LS, Rowe M and Crocker J. (1992). J. Pathol., 166, 1–5. Ortiz-Vega S, Khokhlatchev A, Nedwidek M, Zhang XF, Dammann R, Pfeifer GP and Avruch J. (2002). Oncogene, 21, 1381–1390. Ponting CP and Benjamin DR. (1996). Trends Biochem. Sci., 21, 422–425. Shivakumar L, Minna J, Sakamaki T, Pestell R and White MA. (2002). Mol. Cell. Biol., 22, 4309–4318. Tao Q, Huang H, Geiman TM, Lim CY, Fu L, Qiu GH and Robertson KD. (2002). Hum. Mol. Genet., 11, 2091–2102. Tao Q, Swinnen LJ, Yang J, Srivastava G, Robertson KD and Ambinder RF. (1999). Am. J. Pathol., 155, 619–625. Tsai CN, Tsai CL, Tse KP, Chang HY and Chang YS. (2002). Proc. Natl. Acad. Sci. USA, 99, 10084–10089. Vos MD, Ellis CA, Bell A, Birrer MJ and Clark GJ. (2000). J. Biol. Chem., 275, 35669–35672. Zou HZ, Yu BM, Wang ZW, Sun JY, Cang H, Gao F, Li DH, Zhao R, Feng GG and Yi J. (2002). Clin. Cancer Res., 8, 188–191.

Oncogene