Developmentally-Related Candidate Retinoic Acid Target Genes Regulated Early During Neuronal Di€Erentiation of Human Embryonal Carcinoma

Developmentally-related candidate retinoic acid target genes regulated early during neuronal dierentiation of human embryonal carcinoma

Sarah J Freemantle1,2,4, Joanna S Kerley1,2,4, Shannon L Olsen1,2, Robert H Gross3 and Michael J Spinella*,1,2

1The Department of Pharmacology and Toxicology, Dartmouth Medical School, Hanover, New Hampshire, NH 03755, USA; 2The Norris Cotton Cancer Center, Dartmouth Hitchcock Medical Center, Lebanon, New Hampshire, NH 03755, USA; 3The Department of Biological Sciences, Dartmouth College, Hanover, New Hampshire, NH 03755, USA

Embryonal carcinoma is a model of embryonic develop- Introduction ment as well as tumor cell dierentiation. In response to all-trans retinoic acid (RA), the human embryonal Retinoids are natural or synthetic derivatives of carcinoma (EC) cell line, NT2/D1, dierentiates toward vitamin A and exert potent growth and dierentiation a neuronal lineage with associated loss of cell growth and eects on normal, embryonic and neoplastic cells tumorigenicity. Through the use of cDNA-based micro- (Gudas et al., 1994). Vitamin A (retinol) is essential arrays we sought to identify the early downstream for vision, reproduction, normal embryonic develop- targets of RA during dierentiation commitment of ment, and the regulation of the growth and dierentia- NT2/D1 cells. A total of 57 genes were induced and 37 tion of diverse cell types in adult tissues (Gudas et al., genes repressed by RA. RA regulated genes were 1994; Zile, 2001). In addition there is strong evidence restricted at 8 h with 27 genes induced and ®ve repressed. for bene®cial retinoid responses in preventing or The total number of RA-responsive transcripts increased treating clinical tumors (as reviewed in Nason- at 24 and 48 h and their pattern of expression was more Burchenal and Dmitrovsky, 1999). symmetrical. For a given time point less than 1% of the Mechanisms active in signaling all-trans retinoic acid 9128 cDNAs on the expression array were regulated by (RA) eects involve the retinoid receptors: retinoic acid RA. Many of these gene products are associated with receptors (RARs) and retinoid-X-receptors (RXRs) developmental pathways including those of TGF-b (Lefty (Mangelsdorf et al., 1994; Chambon, 1996). RA A, NMA, follistatin), homeo domain (HoxD1, Meis2, regulates gene expression via binding of RA-bound Meis1, Gbx2), IGF (IGFBP3, IGFBP6, CTGF), Notch RAR/RXR heterodimers to consensus retinoic acid (manic fringe, ADAM11), Hedgehog (patched) and Wnt response elements (RAREs) in the proximal promoter (Frat2, secreted frizzled-related protein 1) signaling. In region of RA target genes (Mangelsdorf et al., 1994; addition a large cassette of genes induced by RA at 24 ± Chambon, 1996). Three RARs (RARa, RARb,and 48 h are associated with cell adhesion, cytoskeletal and RARg) and three RXRs (RXRa,RXRb, and RXRg) matrix remodeling, growth suppression and intracellular exist and have cell context-dependent expression signaling cascades. The majority of repressed genes are patterns re¯ecting tissue-speci®c retinoid functions associated with protein/RNA processing, turnover or and developmental programs (Mark et al., 1999). The metabolism. The early induced genes identi®ed may play identity of target genes, which mediate the dierentia- a regulatory role in RA-mediated growth suppression and tion, growth suppressive and antitumorigenic eects of terminal dierentiation and may have physiologic or retinoid-activated RARs are largely unknown. pharmacologic importance during normal human devel- Testicular germ cell tumors (TGCTs) are the most opment and retinoid-based cancer therapy or prevention. common carcinomas of males between the age of 15 Oncogene (2002) 21, 2880 ± 2889. DOI: 10.1038/sj/ and 35 years (Schottenfeld et al., 1980). Despite recent onc/1205408 progress, the molecular basis for germ cell tumorigenesis is poorly understood (Heidenreich et al., 2000). Keywords: retinoic acid; embryonal carcinoma; devel- Embryonal carcinoma (EC) cells are the stem cell of opment; microarray; dierentiation; mammalian em- nonseminomatous TGCTs (Chaganti and Houlds- bryogenesis worth, 2000). In this study, the human TGCT derived EC cell line NT2/D1 was examined. This line was developed from a metastatic lesion of a testicular *Correspondence: MJ Spinella, Department of Pharmacology and cancer patient (Andrews, 1998). Human EC cells, Toxicology, 7650 Remsen, Dartmouth Medical School, Hanover, including NT2/D1 cells, appear distinct from murine New Hampshire, NH 03755, USA; EC in terms of genetic, phenotypic and biologic E-mail: [email protected] markers (Pera et al., 1990; Matin et al., 1998). Based 4These authors contributed equally to this work Received 26 November 2001; revised 28 January 2002; accepted 19 on these dierences, it has been proposed that NT2/D1 February 2002 cells may represent an earlier stage of dierentiation Retinoic acid target genes in human embryonal carcinoma SJ Freemantle et al 2881 compared to most murine EC cells available (Pera et Kurie et al., 1993; Spinella et al., 1998, 1999). al., 1990; Matin et al., 1998). Commercial cDNA based microarrays were employed. Pluripotent human EC cells are an important model The array contains 9128 cDNAs representing 8524 of early human development (Andrews, 1998). They unique genes and ESTs. The strategy was to compare resemble early embryonic stem cells in morphology, mRNA expression from NT2/D1 cells treated with antigen expression patterns, biochemistry, developmen- 10 mM RA versus NT2/D1 cells treated with vehicle tal potential, and gene regulation (Andrews, 1998). control, dimethyl sulfoxide (DMSO). Three indepen- NT2/D1 cells are multipotent, dierentiating into well dent time points were studied; 8 h +/7 RA, 24 h +/ developed, morphologic and immunophenotypic CNS 7 RA, and 48 h +/7 RA. The dose of RA and time neurons upon culture in RA, with associated loss of points were chosen since previous studies demonstrated cell growth and tumorigenicity (Andrews, 1984; Kurie that under these conditions NT2/D1 cells will commit et al., 1993; Spinella et al., 1999). The neuron-like cells to dierentiate within 24 ± 48 h (Andrews, 1984; express the major neuro®lament proteins, a functional Spinella et al., 1999). Each RA time point was glutamate receptor and sodium ion channels. They also harvested and hybridized on independent days and form functional synapse and in general possess many each time point was matched to an individual vehicle of the electro-chemical properties of CNS neurons (DMSO) control. Dierences in gene expression (reviewed in Andrews, 1998; Przyborski et al., 2000). between DMSO and RA treated mRNA populations Comparisons of human ES, murine ES, and human EC was compared by scatter plot analysis (Figure 1), cells have demonstrated that human EC and human ES similar to the recent analysis of Kelly and Rizzino, lines share many phenotypic and morphologic features 2000. In this representation, each point represents a not possessed by murine ES cells (Thomson et al., particular gene. The location of the point is determined 1998; Pera et al., 2000). Since NT2/D1 cells have been by the Cy3 probe intensity value generated from the widely characterized, can be propagated without the vehicle treated mRNA population represented by the use of feeder lines or leukemia inhibitory factor (LIF) x-axis with its Cy5 probe intensity value of the RA and can be readily transfected, these cells are a treated mRNA population represented by the y-axis. In valuable and practical complement to human ES cells each plot, points that lie close to the linear regression as an in vitro model of human embryology. In line represent genes that are expressed at similar levels addition, RA treatment of EC is an important model in the two RNA populations. The distance that a gene of tumorigenesis, since RA causes a reversal of the lies along the regression line represents its level of malignant phenotype. However, the genes and path- expression in relation to other genes. The perpendicular ways regulated during neuronal dierentiation, lineage distance that a gene lies from the regression line commitment, and repression of tumorigenic potential denotes the degree of its dierential expression between of pluripotent EC cells are not well de®ned. the DMSO and RA treated samples. In an eort to identify critical RA-responsive genes This analysis reveals a very tight distribution pattern during induced tumor cell dierentiation of human EC along the linear regression line of the data points, cells we employed large-scale cDNA microarray indicating that expression of the majority of genes are technology focused on a series of time points prior unchanged with RA treatment. Fold-changes of 1.7-, to, or concurrent with, commitment of NT2/D1 cells to 1.8- and 1.8-fold or higher were chosen to represent enter a program of neuronal dierentiation. The results changes above background for the 8, 24 and 48 h time suggest that RA alters the expression of a small points, respectively. These cutos are represented as dynamic set of regulatory genes at this early stage of parallel lines from the central, linear regression line in lineage commitment. The transcripts identi®ed encode Figure 1. Under these criteria a total of 94 genes were regulatory proteins that are known to play a role in regulated by RA (57 induced and 37 repressed). The vertebrate neuronal development plus other genes and percentage of cDNAs regulated for each experiment novel genes not previously associated with this process. was 0.27% (22 induced/two repressed out of 9036) at While several of the genes are known direct targets of 8 h; 0.55% (31 induced/18 repressed out of 8917) at RA, the majority have not been described in this 24 h and 0.77% (40 induced/24 repressed out of 8264) context before. The genes identi®ed are likely candi- at 48 h. Figure 1 also demonstrates that the number of dates in controlling early steps in human development RA mediated changes in gene expression is severely and cellular dierentiation in particular neuronal restricted at 8 h with a bias toward genes induced by dierentiation and may also be pertinent to malignant RA. The total number of RA-responsive genes progression of TGCTs and to the clinically relevant increases at 24 and especially 48 h and their pattern antitumor eects of RA. of expression is more symmetrical. The identity and associated function of each gene is provided in Table 1. Genes were clustered according to Results their temporal expression patterns in response to RA (Figure 2 and Table 1). While several of the genes have We sought to identify RA target genes in the human been previously shown to be RA-regulated, the EC cell line NT2/D1. RA treatment of NT2/D1 cells majority have not been previously associated as RA results in G1 arrest of the cell cycle, growth inhibition targets (see Table 1). We noticed that many of the and terminal neuronal dierentiation (Andrews, 1984; same genes were highly ranked in all three experiments

Oncogene Retinoic acid target genes in human embryonal carcinoma SJ Freemantle et al 2882 a the 94 genes in Table 1. These values still represent less than 3% of the genes for each experiment and are non- bolded in Table 1. This reveals that 27 genes were upregulated by RA at 8 h and ®ve repressed. Many of the same genes remain induced at 24 ± 48 h with the total number of regulated genes increasing (53 induced and 36 repressed). Functional categorization of the genes reveals that many of the acutely induced genes are associated with development, retinoid metabolism, cell signaling and growth suppression (Table 2). These include extracellular signaling molecules and their receptors, transcription factors, and second messenger signaling modi®ers. A larger percentage of the genes induced later (at 24 and 48 h) are involved in matrix remodeling, cell adhesion and regulation of the b cytoskeleton (Table 2 and Figure 2). The majority of repressed genes are associated with protein/RNA processing, turnover or metabolism. In addition, several uncharacterized or partially characterized ESTs of unknown function were identi®ed as regulated by RA in NT2/D1 cells (Table 2). Thus far 22 of 24 genes have been con®rmed by RT ± PCR indicating a very low level of false positives (Figure 3, Table 1 and data not shown). Furthermore the kinetics of induction predicted by the array analysis was con®rmed for many of the genes. For example, SFRP1 was con®rmed to be repressed with RA treatment beginning at 24 h. Induction of IGFBP3 and MMP11 did not commence until 24 and 48 h, respectively. Chimerin 2, ESTAA251635, HAS2, CNR1, NRP1, BTG2 and SDR1 were con®rmed to c be induced within 8 h (Figure 3).

Discussion

Extensive knowledge exists in regard to the molecular interactions occurring during nuclear receptor- mediated target gene activation (Edwards, 1999; Glass and Rosenfeld, 2000). However, the identity of all but a few of the target genes of retinoic acid receptors remains undiscovered. The current study has identi®ed a variety of regulatory genes, which are candidate targets of RA during induced tumor cell dierentiation of human EC cells. Since these genes are regulated by RA during the commitment phase of NT2/D1 neuronal dierentiation, it is likely that many play a regulatory Figure 1 Scatter plot analysis comparing expression data from role in promoting neuronal dierentiation, suppressing vehicle (DMSO) and RA treated NT2/D1 cells. For each stem cell renewal and growth, and directly mediating experimental time point, individual genes were plotted using the the antitumor eects of RA in this model. It should be log of the non-normalized signal intensity for Cy3 and Cy5 probes. The y-axis represents the Cy5 probe signal generated from noted that NT2/D1 cells are particularly informative as RA-treated cells. The x-axis represents the Cy3 probe signal compared to other dierentiation models since commit- generated from vehicle (DMSO) treated control. The thick line is ment to dierentiation occurs with a single agent, RA, the linear regression line. Points above this line represent genes without additional specialized conditions such as serum induced by RA and genes below the line represent genes repressed starvation, application of mitotic inhibitors or cAMP, by RA. The cutos used are indicated by thin lines or con¯uent growth and aggregation (Andrews, 1984; Spinella et al., 1999). but sometimes just missed the 1.7, 1.8, 1.8 standard. To A total of 57 distinct transcripts were induced and 37 provide information on the trends of gene expression, repressed by RA. At the early 8 h time point the fold-changes of 1.5-, 1.6- and 1.7-fold are included for expression of many more genes were induced than

Oncogene Retinoic acid target genes in human embryonal carcinoma SJ Freemantle et al 2883

Table 1 Description and kinetics of genes regulated by RA in NT2/D1 cells Gene symbol Gene description Accession No. 48 h 24 h 8 h Confirmation Continuous and early induced genes EBAF Endometrial bleeding assoc. factor (LEFTYA); TGFb member NM_003240 47.6 10.4 3.6 RT/N/a MEIS2 Meis (mouse) homolog 2; homeo box protein NM_020149 9.4 2.1 1.8 RT/a CYP26A1 RA metabolizing cytochrome P450 NM_000783 3.6 8 4.2 RT/N/a CRABP2 Cellular retinoic acid-binding protein 2 NM_001878 6.9 2.9 1.9 RT/a CHN2 Chimerin 2; GTPase-activating protein for Rac NM_004067 4.1 2.6 1.8 RT HOXD1 Homeo box D1 NM_024501 3.5 2.3 2.3 a SLC5A3 Solute carrier 5 inositol transporter 3; osmoregulation AF027153 2.9 3 2.4 RT SDR1 Short-chain dehydrogenase/reductase 1; retinal reductase NM_004753 2.3 2.2 2.2 RT NRIP1 RIP140; nuclear receptor corepressor NM_003489 2.2 1.8 1.8 RT/N/W SERPINB9 Serine protease inhibitor B, member 9; granzyme B inhibitor BC002538 2 2.2 1.8 RT HOXA5 Homeo box A5 NM_019102 3.6 1.6 1.9 a HAS2 Hyaluronan synthase 2; ECM maintenance NM_005328 1.6 4.1 2.2 RT HOXB2 Homeo box B2 NM_002145 5 1.9 1.6 a SOX3 Sox3 NM_005634 3.5 1.6 1.6 RT SDCCAG28 Similar to phosphatidyl choline transfer protein 2; tumor antigen NM-006645 2.6 NI 1.5 RT BTG2 B-cell translocation gene 2; antiproliferative, p53 regulated NM_006763 1.9 2.1 1.5 RT CTGF Connective tissue growth factor (IGFBP8) NM_001901 1.8 1.6 1.5 b ESTAA251635 EST, weakly similar to Sprouty 2 AA251635 7 2.1 2.7 RT KIAA0752 KIAA0752 protein AB018295 7 1.8 1.7 ND NRP1 Neuropilin 1; receptor mediating neuronal guidance AL121748 7 2 1.7 RT CNR1 Cannabinoid receptor 1 (brain); CNS specific GPCR AL136096 7 1.6 1.7 RT NRGN Neurogranin (protein kinase C substrate, RC3) NM_006176 7 2.2 7 b MAPKAPK2 MAPK-activated protein kinase 2 NM_004759 NI 1.7 2.1 ND ESTAA780042 EST-AA780042 AA780042 NI 7 2.5 ND PRKCH Protein kinase C, eta NM_006255 771.7 a HGS HGF-regulated tyrosine kinase substrate NM_004712 771.8 ND MFNG Manic fringe (Drosophila) homolog NM_002405 771.7 b

Late induced genes MLN Motilin; hormone for enteric neurons of GI tract NM_002418 7 2.3 7 ND PIK3C3 Phosphoinositide-3-kinase, class 3; related to vps34 NM_002647 7 2.1 7 ND GBX2 Gastrulation brain homeo box protein 2 NM_001485 7 2 7 a ADAM11 A disintegrin and metalloproteinase domain 11, brain NM_002390 NI 2 7 ND NMA Bambi-related, putative role in TGDb signaling NM_012342 1.6 2.4 7 ND COL1A1 Collagen, type I, alpha 1 NM_000088 1.6 1.9 7 a IGFBP3 Insulin-like growth factor binding protein 3; p53 regulated NM_000598 6.1 3.1 7 RT/N/a IGFBP6 Insuling-like growth factor binding protein 6; p53 regulated NM_002178 2.2 3.2 7 RT/b CPE Carboxypeptidase E NM_001873 5 2.3 7 ND AHNAK AHNAK nucleoprotein (desmoyokin) M80899 3.4 1.8 7 ND CLDN10 Claudin 10; cell adhesion protein NM_006984 2.5 2.1 7 ND P5CR2 Pyrroline 5 carboxylate reductase 2; proline synthesis NM_013328 2.5 1.8 7 ND IL6ST Interleukin 6 signal transducer (gp130, oncostatin M receptor) NM_002184 2 1.8 7 ND ITGA4 Integrin, alpha 4 NM_000885 1.8 1.8 7 ND MDK Midkine (neurite outgrowth-promoting factor 2) NM_002391 2.2 1.6 7 a CFTR Cystic fibrosis transmembrane conductance regulator NM_000492 3 1.7 7 ND PKP2 Plakophilin 2; intracellular junction protein NM_004572 4.3 1.6 7 ND PEG10 Paternally expressed 10 NM_015068 2.3 1.5 7 ND HPTCRA Human pre TCR alpha mRNA AL035587 3.5 77 ND SPP1 Secreted phosphoprotein 1 (osteopontin); bone ECM protein NM_000582 2.5 77 b MEIS1 Meis1 (mouse) homolog; homeodomain protein NM_002398 2.4 77 b IFI16 Interferon gamma-inducible protein 16; myeloid differentiation AF208043 2.3 ±± b MMP11 Matrix metalloproteinase 11 (stromelysin 3) NM_005940 2.2 77 RT/b SERPINB8 Serine protease inhibitor B, member 8 NM_002640 2.2 77 ND GCP2 Gamma-tubulin complex protein 2; microtubule assembly NM_006659 2.2 77 ND OSF-2 Osteoblast specific factor 2; cell adhesion, skeletal dev. NM_006475 2.2 77 ND ESTBF435432 EST-BF435432 BF435432 2 77 ND ZFHX1B Zinc finger homeo box 1B (SMAD-interacting protein 1) NM_014795 1.9 77 ND PTCH Patched (Drosophila) homolog U43148 1.9 77 b SERPING1 Serine protease inhibitor G, member 1 NM_000062 1.8 7 ND

Continuous and early repressed genes HXB Hexabrachion (tenascin C, cytotactin); cell adhesion receptor NM_002160 74.4 72.7 71.5 R/b OAT Ornithine aminotransferase (gyrate atrophy) NM_000274 72.2 71.6 71.5 ND FRAT2 GSK-3 binding protein (Frat2) AB045118 772.4 71.6 RT/N/a FST Follistatin; activin binding protein NM006350 771.7 71.7 a ESTAI884571 EST, similar to 45 kDa splicing factor AI884571 7772.5 ND Continued

Oncogene Retinoic acid target genes in human embryonal carcinoma SJ Freemantle et al 2884 Table 1 (Continued ) Gene symbol Gene description Accession No. 48 h 24 h 8 h Confirmation Late repressed genes CAPN2 Calpain 2 (m/ll) large subunit; protease NM-001748 772 7 ND PLS3 Plastin 3 (T isoform); actin binding protein NM-005032 772 7 ND DPYSL3 Dihydropyrimidinase-like 3 (UNC-33 like); axonal guidance NM_001387 771.9 7 b YWHAZ 14-3-3-zeta; phosphoserine binding protein NM_003406 771.8 7 ND YWHAB 14-3-3 beta; phosphoserine binding protein NM-003404 771.9 7 ND TPT1 Histamine-releasing factor (TPT1 family member) NM_003295 771.8 7 ND DDX15 DEAD/H Box 15 (RNA Helicase 2); pre mRNA splicing NM_001358 771.8 7 ND VIM Vimentin; mesenchymal specific intermediate filament NM_003380 71.5 72.2 7 a MTHFD2 Methylene tetrahydrofolate dehydrogenase NM_006636 71.5 71.8 7 ND NCL Nucleolin; nucleolar protein, ribosomal RNA processing NM_005381 71.6 71.8 7 b SFRP1 Secreted frizzled-related protein 1; Wnt binding protein NM_003012 72.4 72.9 7 RT AP1S2 Adaptor-related protein complex1, sigma2; vesicle coat protein AF251295 72.9 72.3 7 ND SPS Selenophosphate synthetase; human selenium donor protein NM_012247 73 71.9 7 ND PMAIP1 Noxa, PMA-induced protein 1; BCL2 member, p53 regulated NM_021127 72.5 71.9 7 ND DSP Desmoplakin (DPI, DPII); cytoskeletal structural protein NM_004415 71.9 71.9 7 b ROR1 RTK-like orphan receptor (NTRK1) NM_005012 71.8 71.8 7 ND TACSTD1 Tumor-associated calcium signal transducer 1 NM_002354 73.5 71.7 7 ND RNAHP RNA helicase-related protein; DEAD box protein NM_007372 73.4 71.7 7 ND EPHA1 Epha1 (Oncogene EPH); RTK, neuronal development NM_005232 73.2 71.7 7 ND NP Nucleoside phosphorylase; adenonine nucleotide pathway NM-000270 72 71.7 7 ND ESTAI355302 EST-AI355302 AI355302 71.9 71.6 7 ND MT1L Metallothionein 1L; heavy metal binding protein NM-002450 71.9 71.6 7 ND PLP1 Proteolipid protein 1 (lipophilin); major constituent of myelin M54927 72.2 71.5 7 ND TERF1 Telomeric repeat binding factor 1; telomerase inhibitor NM_017489 72 71.5 7 ND ITPR2 Inositol 1,4,5-triphosphate receptor 2; calcium release NM_002223 71.9 71.5 7 ND HSPD1 Heat shock 60 kDa protein 1 (chaperonin) (GroEL) BC_010112 71.8 71.5 7 b PTPRZ1 Protein tyrosine phosphatase, receptor-type, zeta 1; CNS dev. NM_002851 72.8 77 ND IFITM1 Interferon induced transmembrane protein 1 NM_003641 72 77 ND ESTAW973290 EST-AW973290 AW973290 72 77 ND HK1 Hexokinase 1 NM_000188 71.9 77 ND M6PR Mannose-6-phosphate receptor; lysosomal targeting NM_002355 71.8 77 ND CTSC Cathepsin C; lysosomal dipeptidyl aminopeptidase I NM_001814 71.8 77 ND

Bolded numbers are balanced dierential expression (fold-changes) of 1.7, 1.8 and 1.8 for 8, 24 and 48 h experiments, respectively. Non-bolded numbers represent fold-changes between 1.5 and 1.6 for the 8 h experiment and between 1.5 and 1.7 for the 24 and 48 h experiments. Dash (7) represents change of 1.0 to 1.4. NI represents a non-informative gene for a given array. RT, N, W represents con®rmation by RT ± PCR, Northern, or Western analysis, respectively. (a) Indicates a gene previously shown to be regulated by RA including EC cells. (b) Indicates a gene previously shown to be induced by RA in cells other than EC cells. ND is not determined

repressed. This is consistent with RARs being direct in NT2/D1 and MCF-7 cells and that RIP140 in turn inducers and not direct repressors of transcription. represses RAR activity in these cells (Kerley et al., These distributions become more symmetrical at 24 2001). Thus, an early-programmed response to RA in and 48 h and may represent a second wave of gene NT2/D1 cells appears to be modulation of the RA regulation mediated by the early induced genes. Several signal itself. of the genes have been previously shown to be direct Many of the early induced genes are known to be targets of RARs, in EC and other cell contexts and associated with development and in many cases include LeftyA, CYP26A1, CRABP2, Meis2, HoxD1, neuronal development. Surprisingly, in this cell auton- midkine, Gbx2 and neurogranin (see Table 1) (Oulad- omous dierentiation model, many genes involved in Abdelghani et al., 1997, 1998; White et al., 1997; cell-cell communication, patterning, and fate determi- Durand et al., 1992; Simeone et al., 1990; Michikawa nation are regulated by RA. These include develop- et al., 1993; Bouillet et al., 1995a; Iniguez et al., 1994). mental pathways such as TGF-b (LeftyA, NMA, This supports the validity in using cDNA microarray follistatin), homeo domain (HoxD1, Meis2, Mesi1, analysis to identify RA targets in EC cells. The Gbx2) IGF (IGFBP3, IGFBP6, CTGF), Notch (manic majority of genes have not been previously associated fringe, ADAM11), Hedgehog (patched) and Wnt as RA inducible and thus may play previously (Frat2, secreted frizzled-related protein 1) signaling. unrecognized roles in mediating RA eects in EC cells. (For recent reviews of these pathways see Schier and Four genes that are induced within 8 h (CYP26A1, Shen, 2000; Cillo et al., 2001; Allan et al., 2001; CRAPB2, short-chain-dehydrogenase/reductase 1, and Mumm and Kopan, 2000; Taipale and Beachy, 2001; the corepressor RIP140) are associated with modula- Altmann and Brivanlou, 2001). This suggests multiple tion/attenuation of retinoid-signaling (White et al., signals from diverse signaling pathways are involved in 1997; Durand et al., 1992; Haeseleer et al., 1998; Lee the maintenance of stem cell renewal and subsequent and Wei, 1999). We have recently shown that RIP140 RA-induced lineage commitment of human EC. is a primary, cycloheximide insensitive, target of RARs Interestingly, gp130, a receptor component of c-kit

Oncogene Retinoic acid target genes in human embryonal carcinoma SJ Freemantle et al 2885 and LIF, two ligands associated with stem cell renewal (Nichols et al., 2001), is induced by RA. The data not only indicates which pathways may be involved but suggests how RA, a known morphogen critical for development, reprograms the dynamic balance of these pathways by altering the transcription of speci®c pathway components. Alternatively, although the major cell type to emerge from RA treated NT2/D1 cells are neurons, the regulation of the diverse pathway genes seen here may represent the early emergence of multiple cell types in the NT2/D1 culture. Deregulation of developmentally related genes has been associated with tumorigenesis (Calvo and Drabkin, 2000; Taipale and Beachy, 2001). A large cassette of genes upregulated at 24 ± 48 h are associated with matrix remodeling and cell adhesion. These include members of the serpin family of serine protease inhibitors, MMP11, integrin alpha-4, the cell adhesion proteins Claudin 10 and OSF-2, and secreted matrix proteins osteopontin and collagen type 1 (Janciauskiene, 2001; Basset et al., 1990; Takada et al., 1989; Morita et al., 1999; Takeshita et al., 1993; Kiefer et al., 1989). This implies that a substantial remodeling of the extracellular matrix takes place during the early phase of induced dierentiation of NT2/D1 cells. Other upregulated genes are associated with intercellular signaling (see Table 2). In general, few of the genes upregulated by RA are structural, metabolic or housekeeping genes. In contrast, the majority of downregulated genes are associated with protein/RNA processing, turnover or metabolism. This suggests a shutting down of cellular activity during the transition to a postmitotic state established to commence approximately 48 h after RA treatment of NT2/D1 cells (Spinella et al., 1999). The limited number of global expression studies reported utilizing NT2/D1 cells or murine ES cells show minimal overlap with the RA-regulated genes found here (Satoh and Kuroda, 2000; Bani-Yaghoub et al., 2001; Leypoldt et al., 2001; Kelly and Rizzino, 2000). This is perhaps due to a focus on gene expression of end-stage neuronal cells which require prolonged RA treatments (up to several weeks) and/or addition of mitotic inhibitors to purify neuronal cells from undierentiated cells. In addition, these studies utilized smaller-scale expression arrays containing few- er than 600 genetic elements. In contrast, Bouillet et al. (1995b) identi®ed 12 early RA-induced genes using a subtractive hybridization technique in murine P19 EC cells. Three of these genes LeftyA, Gbx2, and Meis2 were likewise identi®ed as early RA-responsive genes in Figure 2 Hierarchical cluster analysis of temporal expression the current study. pro®le of RA-regulated genes. The 94 genes which demonstrated RA induces G1 cell cycle arrest and growth RA regulation equal to or higher than 1.7, 1.8, and 1.8, at 8, 24 suppression of NT2/D1 cells (Andrews, 1984, Spinella or 48 h respectively, were hierarchical clustered for similarity. The cluster image shows the trend in gene expression over time. Green et al., 1999). Several genes identi®ed in the current represents genes repressed with RA at a given time point. Red study may potentially contribute to this response. represents genes induced with RA at a given time point. The Genes associated with antiproliferation that are greater the color intensity the greater the fold-change. Black induced within 8 h of RA treatment include CHN2, represents no change with RA de®ned as below the level of 1.5. BTG2 and HGS while IGFBP3 and IGFPB6 which Grey represents an uninformative data point due to failure to meet probe selection criteria as described in Materials and repress mitogenic IGF signaling are induced by RA at methods later time points. CHN2, chimerin 2 (beta-2-chimerin)

Oncogene Retinoic acid target genes in human embryonal carcinoma SJ Freemantle et al 2886 Table 2 Functional categorization of RA-regulated genes in NT2/D1 cells Development/differentiation Antiproliferation/antitumor *LeftyA (EBAF) *Chimerin 2 *Meis2 (N) *B-cell translocation gene 2 *HoxD1 (N) *HGF-regulated tyrosine kinase substrate *HoxA5 IGFBP3 *HoxB2 (N) IGFBP6 Meis1 (N) Noxa (pro-apoptotic) Gastrulation brain homeobox protein 2 (N) Zinc finger homeo box 1B (N) Protein/RNA turnover and processing *Sox3 (N) Carboxypeptidase E *Manic Fringe (N) Calpain 2 Patched (N) Adaptor-related protein complex 1 *CTGF Mannose-6-phosphate receptor Interferon gamma inducible protein 16 Cathepsin C NMA Heat shock 60 kD protein 1 *Neurophilin 1 (N) DEAD/H Box 15 *Neurogranin (N) RNA helicase-related protein Desmoyokin (AHNAK) (N) Nucleolin Midkine (N) EphA1 (N) Metabolism Dihydropyrimidinase-like 3 (N) Pyrroline 5 carboxylate reductase 2 Protein tyrosine phosphatase, receptor-type zeta 1 (N) *Ornithine aminotransferase *Follistatin (N) Methylene tetrahydrofolate dehydrogenase *Frat2 (N) Selenophosphate synthestase *Secreted frizzled-related protein 1 (N) Nucleoside phosphorylase Hexokinase Retinoid/retinoid receptor modulation *Cytochrome P450 26A1 Structural *Cellular retinoic acid-binding protein 2 Proteolipid protein 1 *Short-chain dehydrogenase/reductase 1 Vimentin *RIP140 Desmoplakin

Signaling Miscellaneous *MAPK activated protein kinase 2 *Solute carrier 5 inositol transporter 3 *Protein Kinase C, eta CFTR *Cannabinoid receptor 1 Histamine-releasing factor Phosphoinositide-3-kinase, class 3 Metallothionein IL Motilin Telomeric repeat binding factor IL-6 signal receptor transducer (gp130) Interferon induced transmembrane protein RTK-like orphan receptor 1 Inositol 1,4,5-triphospate receptor Unknown function 14-3-3 zeta *Similar to phosphatidyl choline transfer protein 2 14-3-3 beta *KIA0752 *EST AA780042 Matrix remodeling/cell adhesion/cytoskeletal regulation *EST AA251635 *Hyaluronan synthase 2 PEG10 *Serine protease inhibitor B, member 9 HPTCRA Serine protease inhibitor B, member 8 *EST AI884571 Serine protease inhibitor G, member 1 Tumor associated calcium signal transducer 1 Matrix metalloproteinase 11 EST AI355302 ADAM11 EST AW973290 Claudin 10 Integrin, alpha 4 Gamma tubulin complex protein 2 Plakophilin 2 Collagen type1 Osteopontin (SPP1) Osteoblast specific factor 2 *Hexabrachion Plastin 3

Genes in bold type are induced with RA. Genes in non-bold type represses with RA. (*) Represents genes regulated within 8 h of RA treatment. (N) Represents genes associated with neuronal development or neuronal dierentiation

is a GTPase activating protein (GAP) speci®c for the suppression of cytokine-induced cell growth (Asao et inactivation of p21-rac and is downregulated in high- al., 1997). BTG2 is a p53 target gene associated with grade gliomas compared to normal brain and low- negative cell cycle regulation (Rouault et al., 1996). In grade astrocytomas (Yuan et al., 1995). HGS, HGF- contrast, a pro-apoptotic p53 target gene, Noxa, is regulated tyrosine kinase substrate, is associated with a repressed by RA, which could contribute to the

Oncogene Retinoic acid target genes in human embryonal carcinoma SJ Freemantle et al 2887 with array expression studies (for example, as reported by Chang et al., 2000) and may be related in part to dierences in hybridization stringency between North- ern and microarray analyses. Furthermore, additional known or suspected RA target genes were consistently shown to be induced by RA across array experiments but had a fold-induction that were below the cuto thresholds used here. These genes when ranked by fold- induction, are still within the top 2 ± 3% of all the genes of a given array experiment. The entire expression data sets will be freely available upon request from the authors. An additional noted peculiarity was that several genes known to be highly induced by RA in NT2/D1 cells did not score positive in the array experiments. Thus, it is likely that the 94 genes identi®ed here represent a subset of the RA- regulated genes contained on each array. Despite recent progress, the molecular basis for germ cell tumorigenesis with regard to alterations in tumor suppressor genes or oncogenes is poorly understood. However, most TGCTs have an increased copy number of 12p, either in the form of multiple copies of i(12p) or as tandem duplications of 12p (Dmitrovsky et al., 1990; Chaganti and Houldsworth, 2000; Heidenreich et al., 2000). Further, this chromosomal abnormality is present in the early premalignant carcinoma in situ lesions of TGCT patients (Vos et al., 1990). This Figure 3 Eect of RA on transcript levels of selected genes in suggests that overexpression of a gene within 12p may NT2/D1 cells. RT ± PCR analysis depicting time course (in hours) drive tumorigenesis of TGCTs. Interestingly, two of of indicated genes in response to 10 mM RA or DMSO vehicle the genes which are repressed by RA in the present control in NT2/D1 cells. RT ± PCR analysis was performed on independent samples from those used in array experiments. study, mannose-6-phosphate receptor and inositol Regulation of indicated mRNA transcripts by RA is similar to 1,4,5-triphosphate receptor 2, reside on 12p (Pohlmann that predicted by cDNA expression arrays. In indicated samples et al., 1987; Yamamoto-Hino et al., 1994). Since RA (7RT) reverse transcriptase was omitted to control for genomic has potent eects in reversing the tumorigenic proper- DNA contamination ties of NT2/D1 cells these genes may have a role in contributing to the malignant potential of TGCTs. In summary, we have identi®ed a cassette of survival of postmitotic, NT2D1 cell-derived neurons regulatory genes whose expression is acutely altered (Oda et al., 2000). TGCTs rarely possess p53 during the initial events of RA induced terminal mutations, which has been associated with the high dierentiation of human EC cells. The majority of cure rate of TGCT patients in response to cisplatin- the genes have not been previously implicated as RA based therapy (Peng et al., 1993; Heimdal et al., 1993). targets. We propose a role for these genes in mediating We have previously shown that in NT2/D1 but not the dierentiation, growth suppressive and antitumori- cisplatin- and RA-resistant NT2/D1-R1 cells, RA genic eects of RA. In addition, the results of these in activates the transactivation function of p53 (Curtin vitro studies may provide insight into normal human et al., 2001). Based on these ®ndings, we proposed RA development and cancer. Much work needs to be done regulation of a subset of target genes is through to de®ne the physiologic or pharmacologic importance activation of p53 which contributed to G1 arrest and of these genes in normal development and retinoid- subsequent dierentiation (Curtin et al., 2001). Other based cancer therapy or prevention. p53 targets highlighted as regulated by RA in this study besides Noxa and BTG2 are IGFBP3 and IGFBP6 (Buckbinder et al., 1995; Kannan et al., 2001). It will be of interest to discover whether a larger Materials and methods subset of p53 targets are regulated by RA and to what extent this pathway contributes to RA-mediated gene Cell culture, induction protocol and RNA isolation regulation. The RA-sensitive human EC cell line, NT2/D1 is a clonal line We noticed compression of the array data. For many derived from a xenograft of Tera-2 cells (Andrews, 1998). of the genes, the array analysis underestimated the Cells were cultured in high glucose DME with 10% fetal fold-induction seen via subsequent Northern or RT ± bovine serum supplemented with penicillin, streptomycin, and PCR analysis (data not shown). The reason for this is glutamine under humidi®ed 5% CO2. The RA induction unclear, however, compression is a common ®nding protocols were as follows. On day-2 cells in log phase growth

Oncogene Retinoic acid target genes in human embryonal carcinoma SJ Freemantle et al 2888 were plated in DME media with 10% charcoal absorbed Cluster program by Michael Eisen (Eisen et al., 1998) and serum to deplete endogenous serum retinoids. 16106 cells Xcluster by Gavin Sherlock (modi®ed by RH Gross). This were plated per 15-cm dish to assure that cells would not software is available from the Stanford MicroArray Database reach con¯uence during the experiment. On day 0, medium site, http://genome-www5.standofrd.edu/MicroArray/SMD/ was replaced with fresh medium containing 10 mM RA or the restech.html. The clustered data was visualised using Tree- vehicle, DMSO. Cells were harvested for RNA isolation 8, 24 View (Eisen et al., 1998). or 48 h later. Each RA time point had a matching DMSO control. RA was stored under liquid N in the dark as a 2 RT ± PCR analysis 10 mM stock solution dissolved in DMSO. Total RNA was isolated using TriReagent (Invitrogen Life Technologies). Expression levels of several genes in response to RA were Poly A+ RNA was isolated from a minimum of 500 mgof measured by semi-quantitative RT ± PCR. RNA samples were total RNA using the Oligotex mRNA midi kit (Qiagen) independently generated from those used in array experi- according to manufacturer's recommendations. Poly A+ ments for use in RT ± PCR analysis. The cDNA was RNA was monitored for integrity on denaturing formalde- synthesized from 5 mg total RNA using Superscript II reverse hyde gels. As a further quality control, semiquantitive PCR transcriptase (GIBCO-BRL), as previously described (Spinel- using primers for a known RA target gene, retinoic acid la et al., 1998; Curtin et al., 2001). Ampli®cation of receptor b (RAR-b), was performed to con®rm RA gene highlighted array genes and b-actin was performed using induction (RAR-b sense primer-TGC CAA CCA GCT CCT Taq polymerase as described previously (Curtin et al., 2001). GCC TTT and antisense primer-GCC ATT GAT CCA GGA Experiments were performed to determine the optimal ATT TCC A). Ampli®cation was performed using recombi- number of cycles that would yield linear phase of ampli®ca- nant Taq DNA polymerase (Invitrogen Life Technologies) tion, which ranged from 17 to 30 cycles. PCR products were according to the manufacturer's instructions for 25 cycles. resolved on a 1.5% agarose gel stained with ethidium bromide. Ampli®cation of actin cDNA was used as a normalizing control and reverse transcriptase was omitted cDNA array hybridization and gene expression analysis from select samples to control for genomic DNA contamina- cDNA probe synthesis, hybridization and signal analysis were tion. For certain products detection in the linear range conducted by Incyte Genomics (St. Louis, MO, USA) as required transfer to nitrocellulose and probing with a described on the company's website, http://www.incyte.com. radiolabeled internal primer. Primers were chosen from the The cDNA-based Human Unigene 1 Life array was veri®ed sequence of each cDNA clone to yield a product of employed. This array is 100% sequence veri®ed and contains 200 ± 300 bp. Sequence of PCR primers are available upon 9128 independent clones representing 8524 unique genes and request. As reported in Table 1, in select cases Northern and ESTs. Initial data analysis was performed using GemTools Western analysis were performed by standard methods as 2.4. In all cases, cDNA generated from RA treated samples described previously (Curtin et al., 2001; Kerley et al., 2001). were labeled with Cy5 while matching DMSO controls were labeled with Cy3. All reported fold-changes are normalized balanced dierential expression. Spots on the array in which either probe failed selection criteria as de®ned by Incyte were Acknowledgments discarded prior to analysis. As described in the text, the We thank Dr Ethan Dmitrovsky (Dartmouth Medical cutos for balanced dierential expression (fold-change) School) for helpful discussion. This work was supported by between Cy3 and Cy5 probes used were 1.7, 1.8 and 1.8 for the National Cancer Institute Howard Temin Award K01- the 8, 24 and 48 h experiments, respectively. Cluster analysis CA75154 (MJ Spinella), by American Cancer Society was performed only on those genes that met the above cuto Research Scholar Grant RSG-01-144-01 (MJ Spinella) for at least one time point. Hierarchical clustering was and by a grant from the Lance Armstrong Foundation performed with an euclidean metric for similarity using the (SJ Freemantle).

References

Allan GJ, Flint DJ and Patel K. (2001). Reproduction, 122, Buckbinder L, Talbott R, Velasco-Miguel S, Takenaka I, 31 ± 39. Faha B, Seizinger BR and Kley N. (1995). Nature, 377, Altmann CR and Brivanlou AH. (2001). Int. Rev. Cytol., 646 ± 649. 203, 447 ± 482. Calvo R and Drabkin HA. (2000). Ann. Oncol., 11, 207 ± 218. Andrews PW. (1984). Dev. Biol., 103, 285 ± 293. Chaganti RS and Houldsworth J. (2000). Cancer Res., 60, Andrews PW. (1998). APMIS, 106, 158 ± 168. 1475 ± 1482. Asao H, Sasaki Y, Arita T, Tanaka N, Endo K, Kasai H, Chambon P. (1996). FASEB J., 10, 940 ± 954. Takeshita T, Endo Y, Fujita T and Sugamura K. (1997). J. Chang BD, Watanabe K, Broude EV, Fang J, Poole JC, Biol. Chem., 272, 32785 ± 32791. Kalinichenko TV and Roninson IB. (2000). Proc. Natl. Bani-Yaghoub M, Felker JM, Ozog MA, Bechberger JF and Acad.Sci.USA,97, 4291 ± 4296. Naus CC. (2001). Biochem. Cell Biol., 79, 387 ± 398. Cillo C, Cantile M, Faiella A and Boncinelli E. (2001). J. Basset P, Bellocq JP, Wolf C, Stoll I, Hutin P, Limacher JM, Cell. Physiol., 188, 161 ± 169. Podhajcer OL, Chenard MP, Rio MC and Chambon P. Curtin JC, Dragnev KH, Sekula D, Christie AJ, Dmitrovsky (1990). Nature, 348, 699 ± 704. E and Spinella MJ. (2001). Oncogene, 20, 2559 ± 2569. Bouillet P, Chazaud C, Oulad-Abdelghani M, Dolle P and Dmitrovsky E, Murty VV, Moy D, Miller Jr WH, Nanus D, Chambon P. (1995a). Dev. Dyn., 204, 372 ± 382. Albino AP, Samaniego F, Bosl G and Chaganti RS. Bouillet P, Oulad-Abdelghani M, Vicaire S, Garnier JM, (1990). Oncogene, 5, 543 ± 548. Schuhbaur B, DolleÂ P and Chambon P. (1995b). Dev. Durand B, Saunders M, Leroy P, Leid M and Chambon P. Biol., 170, 420 ± 433. (1992). Cell, 71, 73 ± 85.

Oncogene Retinoic acid target genes in human embryonal carcinoma SJ Freemantle et al 2889 Edwards DP. (1999). Vitam. Horm., 55, 165 ± 218. Oulad-Abdelghani M, Chazaud C, Bouillet P, Sapin V, Eisen MM, Spellmann PT, Brown PO and Botstein D. Chambon P and Dolle P. (1997). Dev. Dyn., 210, 173 ± 183. (1998). Proc. Natl. Acad. Sci. USA, 95, 14863 ± 14868. Oulad-Abdelghani M, Chazaud C, Bouillet P, Mattei MG, Glass CK and Rosenfeld MG. (2000). Genes Dev., 14, 121 ± Dolle P and Chambon P. (1998). Int. J. Dev. Biol., 42, 23 ± 141. 32. Gudas LJ, Sporn MB and Roberts AB. (1994). The Peng HQ, Hogg D, Malkin D, Bailey D, Gallie BL, Bulbul Retinoids: Biology, Chemistry, and Medicine. Sporn MB, M, Jewett M, Buchanan J and Gross PE. (1993). Cancer Roberts AB and Goodman DS. (eds). New York: Raven Res., 53, 3574 ± 3578. Press, Ltd. pp. 443 ± 520. Pera MF, Roach S and Elliss CJ. (1990). Cancer Surv., 9, Haeseleer F, Huang J, Lebioda L, Saari JC and Palczewski 243 ± 262. K. (1998). J. Biol. Chem., 273, 21790 ± 21799. Pera MF, Reubino B and Trounson A. (2000). J. Cell Sci., Heidenreich A, Srivastava S, Moul JW and Hofmann R. 113, 5±10. (2000). Eur. Urol., 37, 121 ± 135. Pohlmann R, Nagel G, Schmidt B, Stein M, Lorkowski G, Heimdal K, Lothe RA, Lystad S, Holm R, Fossa SD and Krentler C, Cully J, Meyer HE, Grzeschik KH, Mersmann Borresen AL. (1993). Genes Chromosomes Cancer, 6, 92 ± G, Hasilik A and Von®gura K. (1987). Proc. Natl. Acad. 97. Sci. USA, 84, 5575 ± 5579. Iniguez MA, Morte B, Rodriguez-Pena A, Munoz A, Przyborski SA, Morton IE, Wood A and Andrews PW. Gerendasy D, Sutclie JG and Bernal J. (1994). Mol. (2000). Eur. J. Neurosci., 12, 3521 ± 3528. Brain Res., 27, 205 ± 214. Rouault JP, Falette N, Guehenneux F, Guillot C, Rimokh R, Janciauskiene S. (2001). Biochim. Biophys. Acta, 1535, 221 ± Wang Q, Berthet C, Moyret-Lalle C, Savatier P, Pain B, 235. Shaw P, Berger R, Samarut J, Magaud JP, Ozturk M, Kannan K, Amariglio N, Rechavi G, Jakob-Hirsch J, Kela Samarut C and Puisieux A. (1996). Nat. Genet., 14, 482 ± A, Kaminski N, Getz G, Domany E and Givol D. (2001). 486. Oncogene, 20, 2225 ± 2234. Satoh J and Kuroda Y. (2000). J. Neurosci. Methods, 94, Kelly DL and Rizzino A. (2000). Mol. Reprod. Dev., 56, 155 ± 164. 113 ± 123. Schier AF and Shen MM. (2000). Nodal signaling in Kerley JS, Olsen SL, Freemantle SJ and Spinella MJ. (2001). vertebrate development. Nature, 403, 385 ± 389. Biochem. Biophys. Res. Commun., 285, 969 ± 975. Schottenfeld D, Warshauer ME, Sherlock S, Zauber AG, Kiefer MC, Bauer DM and Barr PJ. (1989). Nucleic Acids Leder M and Payne R. (1980). Am.J.Epidemiol.,112, Res., 17, 3306. 232 ± 246. Kurie JM, Buck J, Eppinger TM, Moy D and Dmitrovsky E. Simeone A, Acampora D, Arcioni L, Andrews PW, (1993). Dierentiation, 54, 123 ± 129. Boncinelli E and Mavilio F. (1990). Nature, 346, 763 ± 766. Lee CH and Wei LN. (1999). J. Biol. Chem., 274, 31320 ± Spinella MJ, Kitareewan S, Mellado B, Sekula D, Khoo KS 31326. and Dmitrovsky E. (1998). Oncogene, 16, 3471 ± 3480. Leypoldt F, Lewerenz J and Methner A. (2001). J. Spinella MJ, Freemantle SJ, Sekula D, Chang JH, Christie Neurochem., 76, 806 ± 814. AJ and Dmitrovsky E. (1999). J. Biol. Chem., 274, 22013 ± Mangelsdorf D, Umesono K and Evans RM. (1994). The 22018. Retinoids: Biology, Chemistry, and Medicine. Sporn MB, Taipale J and Beachy PA. (2001). Nature, 411, 349 ± 354. Roberts AB and Goodman DS. (eds). New York: Raven Takada Y, Elices MJ, Crouse C and Hemler ME. (1989). Press, Ltd. pp. 319 ± 349. EMBO J., 8, 1361 ± 1368. Mark M, Ghyselinck NB, Wendling O, Dupe V, Mascrez B, Takeshita S, Kikuno R, Tezuka K and Amann E. (1993). Kastner P and Chambon P. (1999). Proc. Nutr. Soc., 58, Biochem. J., 294, 271 ± 278. 609 ± 613. Thomson JA, Itskovitz-Eldor J, Shapiro SS, Waknitz MA, Matin A, Collin GB, Varnum DS and Nadeau JH. (1998). Swiergiel JJ, Marshall VS and Jones JM. (1998). Science, APMIS, 106, 174 ± 182. 282, 1145 ± 1147. Michikawa M, Xu RY, Muramatsu H, Muramatsu T and Vos A, Oosterhuis JW, de Jong B, Buist J and Schraordt Kim SU. (1993). Biochem. Biophys. Res. Commun., 192, Koops H. (1990). Cancer Genet. Cytogenet., 46, 75 ± 81. 1312 ± 1318. White JA, Beckett-Jones B, Guo YD, Dilworth FJ, Morita K, Furuse M, Fujimoto K and Tsukita S. (1999). Bonasoro J, Jones G and Petkovich M. (1997). J. Biol. Proc. Natl. Acad. Sci. USA, 96, 511 ± 516. Chem., 272, 18538 ± 18541. Mumm JS and Kopan R. (2000). Dev. Biol., 228, 151 ± 165. Yamamoto-Hino M, Sugiyama T, Hikichi K, Mattei MG, Nason-Burchenal K and Dmitrovsky E. (1999). Retinoids: Hasegawa K, Sekine S, Sakurada K, Miyawaki A, Handbook of Experimental Pharmacology. Nau H and Furuichi T, Hasegawa M and Mikoshiba K. (1994). Blaner WS. (eds). Berlin: Springer-Verlag. pp. 301 ± 322. Receptor Channels, 2, 9 ± 22. Nichols J, Chambers I, Taga T and Smith A. (2001). Yuan S, Miller DW, Barnet GH, Hahn JF and Williams BR. Development, 128, 2333 ± 2339. (1995). Cancer Res., 55, 3456 ± 3461. Oda E, Ohki R, Murasawa H, Nemoto J, Shibue T, Zile MH. (2001). J. Nutr., 131, 705 ± 708. Yamashita T, Tokini T, Taniguchi T and Tanaka N. (2000). Science, 288, 1053 ± 1058.

Oncogene