Oncogene (1998) 17, 247 ± 254  1998 Stockton Press All rights reserved 0950 ± 9232/98 $12.00 http://www.stockton-press.co.uk/onc SHORT REPORT mafA, a novel member of the maf proto-oncogene family, displays developmental regulation and mitogenic capacity in avian neuroretina cells

So®a Benkhelifa1, Sylvain Provot1, Odile Lecoq1, Celio Pouponnot1,2, Georges Calothy1 and Marie-Paule Felder-Schmittbuhl1

1Unite mixte de recherche 146 du Centre National de la Recherche Scienti®que, Institut Curie, 91405 ORSAY Cedex, France

Transcription factors of the Maf proto-oncogene family (Casado et al., 1996; Guermah et al., 1991). We have been shown to participate in the regulation of showed that transcription of QR1 also correlates with several di€erentiation speci®c genes. We previously cell quiescence in vitro, in NR cells expressing a v-src reported that a member(s) of this family is involved in mutant, conditionally defective for its mitogenic the regulation of the neuroretina speci®c gene, QR1, capacity (Guermah et al., 1991). We identi®ed in the through a promoter region, designated the A box, that is QR1 promoter a quiescence-responsive element, termed closely related to the Maf recognition element (MARE). A box (Pierani et al., 1993, 1995), that displayed We undertook an identi®cation of Maf family genes important similarities with the Maf recognition element expressed in the quail neuroretina (QNR) and we report (MARE) (Kataoka et al., 1994b). We also showed that the isolation of mafA, a gene encoding a novel member transcription factors of the Maf family indeed regulate of the large Maf proteins subgroup. Expression of this transcription of the QR1 gene through the A box gene is developmentally regulated in the neuroretina. (Pouponnot et al., 1995). MafA is able to bind to MARE sequence and to The ®rst member of the Maf gene family, v-maf, was heterodimerize with v-Maf, MafB, Jun and Fos, but not identi®ed in an acutely oncogenic avian retrovirus with the small MafF and MafK proteins. Accordingly, it which induces musculoaponeurotic ®brosarcomas in is able to transactivate the QR1 promoter A box. We chicken (Nishizawa et al., 1989). Maf proteins are bZIP also show that increased expression of mafA induces transcription factors of the AP1 superfamily, exhibiting sustained proliferation of postmitotic QNR cells. high similarity in their DNA binding domain. They recognize the TGCTGACTCAGCA and TGCTG- Keywords: Maf; neuroretina; transcription factor; ACGTCAGCA DNA sequences, respectively called oncogene TRE- and CRE-type MARE on the basis of their resemblance with AP1 TRE and CRE targets (Kataoka et al., 1994b; Kerppola and Curran, 1994b). Maf family proteins can be subdivided into the large and Introduction small classes. c-Maf (the v-Maf cellular counterpart) (Kurschner and Morgan, 1995; Sakai et al., 1997), During development, cellular growth arrest and MafB/Kr (Cordes and Barsh, 1994; Kataoka et al., di€erentiation are determined by accurate gene 1994a; Sakai et al., 1997) and NRL (Farjo et al., 1993; expression programs which involve complex networks Swaroop et al., 1992) constitute presently the large Maf of transcription factors. The avian neural retina (NR) subgroup. These proteins contain at their amino- is a well characterized system to study the general termini a transactivating domain which is absent in mechanisms of growth control and di€erentiation. NR MafF, MafK (Fujiwara et al., 1993; Igarashi et al., development proceeds through well de®ned phases of 1995) and MafG (Blank et al., 1997; Kataoka et al., proliferation, followed by growth arrest and differ- 1995), members of the small class. entiation (Barnstable, 1987). Embryonic NR cells, Although most maf genes display oncogenic dissected at the proliferation stage, also become potential when overexpressed (Fujiwara et al., 1993; quiescent and undergo di€erentiation when they are Kataoka et al., 1993, 1994a), recent reports support the seeded in culture (Crisanti-Combes et al., 1977). These notion that these proteins are involved in differentia- postmitotic cultures acquire sustained proliferative tion processes. Thus, the ubiquitous small Maf capacity upon infection with acute oncogenic retro- proteins, MafG and MafK, were shown to play a viruses such as Rous Sarcoma Virus (RSV) (Pessac and role in erythroid di€erentiation in combination with Calothy, 1974). Induction of NR cell proliferation by the p45 NF-E2 tissue speci®c transcription factor pp60v-src correlates with the down-regulation of differ- (Andrews et al., 1993; Blank et al., 1997; Igarashi et entiation speci®c genes (Guermah et al., 1990). We al., 1994). Similarly, c-maf and mafB expression previously reported the isolation of a NR speci®c gene, patterns during cartilage formation, and eye lens and QR1, expressed in postmitotic di€erentiating cells spinal cord development, suggest that they are implicated in morphogenetic and di€erentiation pro- cesses (Sakai et al., 1997). Accordingly, analysis of Correspondence: MP Felder-Schmittbuhl kreisler (kr) mutant mice demonstrated that mafBis 2Current address: Cell Biology Program and Howard Hughes involved in hindbrain segmentation during early Medical Institute, Memorial Sloan-Kettering Cancer Center, New York, NY 10021, USA development (Cordes and Barsh, 1994). Likewise, c- Received 18 November 1997; revised 4 February 1998; accepted 9 Maf was shown to control T cell speci®c expression of February 1998 the interleukin 4 gene (Ho et al., 1996) and to stimulate Characterization of a novel Maf family gene S Benkhelifa et al 248 transcription of L7, a cerebellum and retina speci®c regulated during the NR development and that its gene (Kurschner and Morgan, 1995). The expression overexpression induces sustained division of postmito- pattern of nrl appears much more restricted, since it is tic QNR cells. essentially detected in postmitotic and di€erentiating neurons of the peripheric and central nervous system, including the retina (Liu et al., 1996). Functional Results and discussion importance of nrl in the retina, where it was ®rst isolated, was demonstrated by its ability to transacti- MafA is a novel member of the large Maf family vate the rhodopsin promoter (Kumar et al., 1996; proteins Rehemtulla et al., 1996). Regulation of target genes by Maf proteins probably To identify maf genes that are expressed in the NR, involves a complex network of heterodimers. Thus, the we used an RT ± PCR strategy. Four degenerated large Maf proteins are able to heterodimerize not only oligonucleotides were designed on the basis of amino among themselves but also with AP1 proteins, resulting acid sequence homology between chicken c-Maf, in the formation of complexes with altered DNA binding MafB and human NRL DNA binding and transacti- speci®cities (Kataoka et al., 1994a,b; Kerppola and vating domains (see arrows in Figure 1a for Curran, 1994a,b). Finally, MafB was recently shown to localization of primers). Maf3'1 was used as a primer inhibit erythroid cell di€erentiation through its interac- to synthesize DNA complementary to mRNA tion with Ets-1 protein (Sieweke et al., 1996) and the prepared at various stages of QNR development. bHLH-zip protein USF2 was shown to associate with c- Two successive rounds of ampli®cation were per- Maf and decrease its DNA binding capacity (Kurschner formed with the Maf5'1-Maf3'2 and Maf5'2-Maf3'2 and Morgan, 1997). Therefore, it becomes apparent from primer pairs. Southern blot analysis of PCR products the variability in tissue expression, dimer formation showed that this method generated cDNA fragments speci®city and transactivation potential that Maf corresponding to mafB and c-maf genes (data not proteins contribute to the regulation of a wide array of shown). We cloned and characterized one PCR target genes and consequently of multiple di€erentiation product generated at ED 10 (embryonic day 10: a processes, notably in the neuroretina. stage at which all NR cells are postmitotic) and after We report here the isolation of a quail NR (QNR) hatching but not at ED6 (at which the lamination gene encoding a novel member of the large Maf family process has begun but neuron precursors are still proteins, designated MafA. We characterize MafA dividing). Analysis of nucleotide and deduced amino speci®city of dimer formation and we show that it acid sequence indicated that it belonged to a thus far binds to and transactivates MARE related sequences. unidenti®ed Maf family gene, which we named mafA. We also report that transcription of this gene is The mafA PCR product was further used to screen a

a Characterization of a novel Maf family gene S Benkhelifa et al 249 randomly primed ED13 QNR cDNA library and a molecular mass of 32467. In vitro transcription and quail genomic library. The complete open reading translation of mafA orf resulted in the synthesis of a frame (orf) is presented in Figure 1a. mafA potentially protein with an apparent molecular weight of 33 kDa encodes a 286 amino acid protein with a calculated (data not shown). The sequence surrounding the

b

Figure 1 (a) Nucleotide sequence and deduced amino acid sequence of the quail mafA gene. Complete open reading frame with ¯anking sequences on the genomic clone are presented. Translation initiation ATG is shown in bold and stop codon is indicated by an asterisk. The upstream in frame stop codon is underlined and histidine stretches within the protein are indicated by double spots. The leucine zipper motif is indicated by stars and dotted lines. 5' ends of primers used for RT ± PCR are shown by arrows above the nucelotide sequence. (b) Comparison of Maf family protein sequences. Deduced amino acid sequence of quail (q) mafA gene was aligned with sequences of chicken (c) c-Maf (c-Maf long form; GenBank d28596), MafB (Kataoka et al., 1994a), MafF, MafK (Fujiwara et al., 1993), and MafG (Kataoka et al., 1995) and human (h) NRL (Swaroop et al., 1992) proteins. Amino acid residues conserved with respect to the MafA protein or among other large Maf family members are boxed. Leucine residues within the zipper are in bold and indicated by stars. The DNA binding domain is indicated by a dotted line (Kataoka et al., 1996; Kerppola and Curran, 1994b). Gaps introduced for maximal sequence alignment are indicated by dashes. mafA nucleotide and amino acid sequences have been deposited into the GenBank (accession number AF034693). Four degenerated primers were designed to amplify Maf family genes: Maf3'1 (5'-CKRCANGCYTGNGCRTANCC-3'); Maf3'2 (5'-TAGCCSCKGTTYTTSARSGTSC-3'); Maf5'1 (5'- ATGGARTAYGTNAAYGAYTTYGA-3'); Maf5'2 (5'-GAYYTGATGAARTTYGARGTNAA-3'). PolyA-selected RNA was prepared from quail NR at various stages (Fast Track Kit, Invitrogen). 1 mg of RNA was reverse transcribed with Maf3'1 primer (Reverse Transcription System, Promega) and successively ampli®ed with Maf5'2-Maf3'2 (40 cycles at 948C 1'30 ± 528C 2' ± 708C 3'; and 10' at 708C) primer pairs and Taq DNA polymerase (Appligene). mafA generated a fragment of about 500 bp which was used to screen a randomly primed ED13 QNR cDNA library (EycheÁ ne et al., 1992) and a quail genomic library (kindly provided by Dr S Saule), under stringent conditions. Positive clones were subcloned into the pBluescript vector (Stratagene) and sequenced on both strands Characterization of a novel Maf family gene S Benkhelifa et al 250 putative initiation codon is in agreement with the clusters respectively, but is absent in NRL. The Kozak consensus sequence (Kozak, 1987) and is presence of these repeats in three Maf proteins preceded by an in frame stop codon. By comparing suggests their functional importance, although they mafA cDNA and genomic clones we found that this have not been linked to transactivating or transforming gene is devoid of introns within the coding sequence, ability (Kataoka et al., 1996). Maf family proteins are a characteristic property shared with mafB and c-maf also highly similar in their leucine zipper motif, which (Kataoka et al., 1994a) but not mouse and human is constituted of six heptad repeats with the ®fth NRL (Farjo et al., 1993, 1997). leucine being replaced by a tyrosine in all large We identi®ed the functionally important regions of members. At the C-terminal end, however, MafA MafA protein by comparing its primary sequence with again diverges, whereas MafB and c-Maf remain those of other family members (Figure 1b). We found similar. In summary, the newly identi®ed mafA gene the highest sequence identity in the DNA binding encodes a bona ®de large Maf family protein. It domain. This similarity extends over a region of about appears to be more closely related to mafB with an thirty conserved amino acids located upstream of the overall identity of 51% and of 57% at the nucleic acid basic domain and shown to be required for speci®c and amino acid levels respectively, than to c-maf and binding of Maf proteins to the MARE (Kataoka et al., NRL. 1996; Kerppola and Curran, 1994b), thus con®rming the importance of this region. Striking homology is also displayed in the amino-terminal part, the reported transactivating region (Kataoka et al., 1996). It is rich in acidic and serine/threonine/proline residues and, accordingly, it contains several putative sites of phosphorylation by proline-directed kinases (Kemp and Pearson, 1990). These sites are present in the four large Maf proteins and may be signi®cant for their regulation. It should be noted that MafA has a unique sequence of eight amino acids (residues 77 ± 84) within the transactivating domain, but without relevance to motif data bases. The conserved transactivating and DNA binding domains are separated by a much more divergent linker sequence which, in MafA, contains two histidine stretches. This feature is also found in MafB and c-Maf which carry two and one histidine

Figure 3 Sequence-speci®c binding of MafA to the MARE consensus sequence. (a) Schematic representation of MafA variant proteins. Full length MafA is encoded by a 3.5 kb HindIII genomic fragment subcloned into the pBluescript vector. It served as a template to generate MafAL2PL4P, a MafA mutated in its leucine zipper (the second and fourth leucine residues, indicated by stars, are mutated to proline residues) by oligonucleotide-mediated site-directed mutagenesis. DMafA, a truncated MafA devoid of transactivating domain (residues 173 ± 286 in Figure 1a), was also obtained by PCR, with the 5' primer providing the HA1 epitope translation initiation sequences. PCR products were subcloned into the pMOSblue vector (Amersham Corp.). Conserved domains are indicated above (HIS: histidine stretches). (b) Maf proteins depicted in Figure 3a were obtained by in vitro transcription and translation in reticulocyte lysates (TNT Coupled Reticulocyte Lysate System; Promega), and used in electrophoretic mobility shift assays (EMSA) as described (Kataoka et al., 1994b). Oligonucleotides representing the CRE-type MARE (#2: 5'- Figure 2 mafA expression during neuroretina development. TCGAGCTCGGAATTGCTGACGTCAGCATTACTC-3') or a Total RNA was extracted from quail neuroretina at days 5, 7, mutated one (#31: 5'-TCGAGCTCGGAATTGCCGGCT- 9, 11 and 15 of embryonic development (ED) (RNA NOW, CATTGTTACTC-3') were generated by annealing complemen- Biogentex, Inc). 25 mg were used for Northern blot analysis. The tary strands and labeling 5' ends with polynucleotide kinase and ®lter was hybridized with a32P-labeled mafA probe encompassing g-32P-ATP [oligonucleotides were designated with respect to the the leucine zipper as well as the DNA binding domain EMSA probes used in previous studies by Kataoka et al. (1994)]. (nucleotides 710 ± 1023 in Figure 1a). The same membrane was For competition, a 50- and 500-fold excess of unlabeled also hybridized with an avian glyceraldehyde-3-phosphate oligonucleotides were used. Lysate indicates the control with dehydrogenase (GAPDH) probe (lower panel) MARE probe and reticulocyte lysate only Characterization of a novel Maf family gene S Benkhelifa et al 251 (data not shown). Its level steadily increased between mafA expression is regulated during NR development ED7 and ED15, indicating that mafA expression is We investigated mafA expression during QNR developmentally regulated in the neuroretina. More- development by Northern blot analysis (Figure 2). A over, increase in the levels of mafA transcript 3.8 kb transcript was detected in the QNR between coincides with terminal mitosis and di€erentiation in ED7 and ED15 and was still present after hatching the NR. Interestingly, nrl expression takes place

Figure 4 MafA dimer forming speci®city. Protein ± protein interactions were analysed by EMSA, based on the di€erence in mobility shift between proteins of di€erent size. MafA and DMafA were cotranslated either together, to assess the capacity of MafA to homodimerize (a), or with other Maf family members and their counterparts unable to form a leucine zipper structure [namely (b); MafBPT, v-MafPT, MafBL2PL4P, v-MafL2PL4P, (c); MafF and MafK (Kataoka et al., 1994a)] and with c-Jun and c-Fos (d) (avian c-jun and c-fos coding sequences are expressed by the pSP64Poly(A) vector (Promega)). Translation products were subjected to EMSA with the CRE-type MARE as a probe (oligonucleotide #2) as described for Figure 3b. Plasmid amounts in the transcription/translation mixture were appropriately adjusted to produce equivalent quantities of proteins. Lysate indicates the control with MARE probe and reticulocyte lysate only. Homodimers and functional heterodimers are indicated by open and black arrows respectively Characterization of a novel Maf family gene S Benkhelifa et al 252 during equivalent stages of rat neuroretina develop- tion, even by overexposing the gel (Figure 4c). ment (Liu et al., 1996). Taken together, these data Hence, the failure to interact with the small Maf suggest that, in addition to nrl, mafA plays a role in proteins appears to be a characteristic property of the NR di€erentiation. We also analysed mafA expression large Maf proteins, thus far tested. It is, moreover, in several ED7 tissues (data not shown). We could not consistent with the marked sequence divergence detect mafA expression except in the brain and the observed in the leucine zipper, between large and NR. The mafA 3.8 kb mRNA was also detectable in small Maf proteins (Figure 1b). cultured chicken embryonic ®broblasts (data not Within the bZIP family, Maf proteins are most shown). closely related to the Fos and Jun subfamilies (Kerppola and Curran, 1994a) and, accordingly, are able to heterodimerize with them (Kataoka et al., MafA binds to MARE sequence 1994a, 1995; Kerppola and Curran, 1994a). We Given the extensive similarity in the DNA binding investigated whether MafA also interacts with c-Fos domain of Maf family members, we expected that and c-Jun proteins (Figure 4d). We found that MafA MafA should also bind a consensus MARE sequence. is able to heterodimerize with c-Fos, a property shared Therefore, we analysed the binding of in vitro by v-Maf, MafB and NRL (Kataoka et al., 1994a,b; synthesized MafA protein to a CRE-type MARE Kerppola and Curran, 1994a). It was reported that (oligonucleotide #2) by using an electrophoretic MafB, unlike v-Maf and NRL, was unable to interact mobility shift assay (EMSA) (Figure 3b). MafA with v-Jun (Kataoka et al., 1994a). We found that generates a retarded band that is clearly distinct from MafA, in spite of its extensive similarity with MafB, the background bandshift seen with the lysate alone. does heterodimerize with c-Jun. As reported for v- MafA binding is speci®c, since it was eciently Maf, interaction with c-Jun and c-Fos is likely to competed for by unlabeled oligonucleotide #2 but not involve the leucine zipper motif (Kataoka et al., by its mutated version, #31. It was further con®rmed 1994b). by the presence of a faster migrating band generated with DMafA, a truncated MafA restricted to the bZIP domain (Figure 3a). We veri®ed that MafA also binds to a TRE-type MARE (data not shown). We also tested the DNA binding ability of MafAL2PL4P, a MafA derivative designed to impair formation of the coiled coil structure (Figure 3a). As expected, the probe was not shifted by MafAL2PL4P, indicating that binding to MARE DNA requires homodimerization (Figure 3b).

MafA forms heterodimers with v-Maf, MafB, Jun and Fos, but not with the small MafF and MafK proteins We next examined the ability of MafA to form homo- and heterodimers. Therefore, we cotranscribed and cotranslated in vitro MafA or its truncated version DMafA (see Figure 3a), with the putative interaction partners and we analysed the binding of translation products to oligonucleotide #2 by EMSA (Figure 4). While each of MafA and DMafA, translated alone, generated a distinguishable re- tarded band, cotranslation products generated an additional complex with an intermediate electro- phoretic mobility, typically a MafA-DMafA homo- dimer (Figure 4a). Likewise, cotranslation of DMafA with MafB or v-Maf generated an additional band Figure 5 MafA is able to activate transcription via the QR1 promoter Abox. mafA orf was obtained by PCR and subcloned (Figure 4b). Moreover, this intermediate band was into the pEF/BssHII vector (Kataoka et al., 1994a). pEF and not observed when we cotranslated DMafA with pEF-mafA indicate respectively the empty and mafA containing MafBL2PL4P or v-MafL2PL4P, the respective expression vectors. Reporter constructs carrying a multimerized mutants of MafB and v-Maf in the leucine zipper, wild type (4XWT) or mutated (4XA-) QR1 promoter A box upstream of the thymidine kinase promoter were previously that are unable to form a coiled coil structure described (Pouponnot et al., 1995). QNR cells infected with (Kataoka et al., 1994a) (Figure 4b). Thus, MafA tsNY68 RSV and actively proliferating, were seeded in Basal interacts with MafB and v-Maf through the leucine Medium of Eagle (BME) containing 10% FCS at 96105 cells per zipper domain. 100 mm diameter dish and transfected 48 h later by the calcium We also investigated whether MafA could form phosphate technique as previously described (Pouponnot et al., 1995). Each transfection mixture contained 1 mg of the respective heterodimers with MafF and MafK, members of the reporter plasmids, 3 mg of expression vectors and 2 mg of b-actin- small Maf subgroup previously shown to be unable lacZ (Beddington et al., 1989). Cells were harvested 48 h after to interact with MafB or c-Maf (Kataoka et al., transfection and CAT activity in cell extracts normalized to b- 1994a). Although we could clearly distinguish the galactosidase activity, was determined (Gorman et al., 1982). Results represent mean values from four experiments and are individually shifted bands, we could not detect the expressed as relative activities with respect to the 4XWT and pEF presence of intermediate bands following cotransla- cotransfection Characterization of a novel Maf family gene S Benkhelifa et al 253 used (Figure 5). We also found that MafA stimulated MafA transactivates the QR1 promoter A box transcription of the full-length QR1 promoter (data not We previously reported that the A box, the quiescence- shown). Taken together, our results demonstrate that responsive unit of the QR1 promoter, contains two MafA exhibits transactivating capacity and, like c-Maf repeats of a half MARE sequence (Pouponnot et al., and MafB, is able to transactivate the QR1 promoter A 1995). Both the QR1 promoter and the A box were box. shown to be transactivated by c-Maf and MafB proteins in cotransfection assays. Conversely, muta- mafA induces sustained proliferation of quiescent NR tions in the A box that impair binding by Maf proteins cells also abolish transactivation (Pouponnot et al., 1995). We investigated the possibility that MafA could also While previous reports indicated that overexpressed v- transactivate the QR1 A box. Therefore, we transfected maf, c-maf and mafB and, to a lesser extent, small maf the MafA expression vector, pEF-mafA together with a genes can transform chicken embryonic ®broblasts reporter gene construct containing four tandem repeats (Fujiwara et al., 1993; Kataoka et al., 1993, 1994a), the of the A box (4XWT) upstream of the heterologous capacity of these genes to promote division of quiescent TK promoter fused to the chloramphenicol acetyl cells has not been yet assessed. We investigated whether transferase (CAT) gene, into actively dividing QNR overexpression of mafA could induce proliferation of cells infected with tsNY68 RSV (Kawai and Hanafusa, postmitotic QNR cells. Therefore, we cloned the mafA 1971). We observed a marked increase in CAT activity coding sequence into the pRc/RSV expression vector, in cells containing the 4XWT reporter and pEF-mafA, under the control of RSV promoter and used this in comparison to those cotransfected with 4XWT and recombinant DNA to transfect NR cultures dissected the empty pEF plasmid (Figure 5). This e€ect was from ED7 quail embryos. Control cells were transfected similar to that observed with MafB and c-Maf under with the pRc/RSV empty vector. Occurrence of comparable experimental conditions (data not shown). proliferating NR cells was examined 1 week after Transactivation by MafA was dependent on the transfection and G418 selection. In NR cultures presence of the A box, since it was not observed transfected with pRc/RSV control vector, resistant cells when the empty reporter vector (TK10) or that remained isolated and did not give rise to colonies. In carrying a mutated multimerized A box (4XA-) were contrast, we observed the presence of foci of dividing

Figure 6 mafA induces sustained proliferation of postmitotic QNR cells. (a) Growth curve of mafA transfected QNR cells. (b) Cell morphology of QNR cells transfected with pRc/RSV (A), QNR cells transfected with pRc/RSV-mafA (B) and QNR cells transformed by tsNY68 RSV (C). The 3.5 kb HindIII fragment containing the mafA orf was inserted into the pRc/RSV expression vector (Invitrogen). NR from 7 day old quail embryos were transfected (20 mg of DNA per 100 mm dish) one day after dissection, by the calcium phosphate technique as previously described (Pouponnot et al., 1995) and maintained in BME containing 10% FCS. G418 (600 mg/ml) was added to transfected cells 7 days later. Proliferating G418 resistant cells were pooled 2 weeks later and seeded at a density of 2.56105 cells in 60 mm dishes. Cells were counted at indicated intervals following extensive dissociation with trypsin/ EDTA. Control untransfected QNR cells, dissected at ED7, were treated similarly Characterization of a novel Maf family gene S Benkhelifa et al 254 cells in mafA transfected cultures (data not shown). We displays extensive similarities with mafB and c-maf. then seeded these proliferating cells at low density and Interestingly, increased transcription of this gene appears analysed their growth capacity, in comparison to that of to correlate with late stages of development, suggesting untransfected normal QNR cells (Figure 6a). We found that mafA plays a role during NR di€erentiation. that the number of mafA transfected cells steadily However, we showed that overexpression of this gene increased with time, whereas that of normal cells induces long term proliferation of QNR cells. Whether remained constant. Moreover, proliferation of mafA this mitogenic e€ect results from a direct or indirect e€ect transfected cells was sustained during several passages, of MafA on target genes remains to be determined. as observed with other oncogenes. Multiplying cells Taken together, our data raise the interesting question of expressed high levels of exogenous mafA transcripts the biological role of mafA in vivo. Answers to this (data not shown). Cultures transfected with pRc/RSV- question should be provided by detailed analysis of mafA mafA contained cells with di€erent shapes, some of expression in the di€erentiating NR and identi®cation of which extending processes. These cells were essentially its putative partners and target genes. ¯at like normal cells and, clearly, distinct from cells infected with transforming viruses such as tsNY68 RSV (Figure 6b). Moreover, mafA containing cells could not Acknowledgements form colonies when seeded in soft agar (data not shown). We thank M Nishizawa for the generous gift of v-Maf, MafB, MafF and MafK encoding vectors and pEF/BssHII This result provides the ®rst evidence that increased plasmid, and M Castellazzi for kindly providing constructs expression of a Maf family gene and, and more generally, encoding c-Jun and c-Fos. We also thank M Marx and A that of a nuclear proto-oncogene, is sucient to induce EycheÁ ne for critical reading of the manuscript. This work division of quiescent and di€erentiated NR cells. was supported by the CNRS, the Institut Curie and by In conclusion, we have identi®ed in the avian NR grants from the Association pour la Recherche sur le mafA, a novel proto-oncogene of the Maf family, that Cancer and Association FrancË aise Retinitis Pigmentosa.

References

Andrews NC, Kotkow KJ, Ney PA, Erdjument-Bromage H, Kataoka K, Noda M and Nishizawa M. (1994b). Mol. Cell. Tempst P and Orkin SH. (1993). Proc. Natl. Acad. Sci. Biol., 14, 700 ± 712. USA, 90, 11488 ± 11492. Kataoka K, Igarashi K, Itoh K, Fujiwara KT, Noda M, Barnstable CJ. (1987). Mol. Neurobiol., 1, 9 ± 46. Yamamoto M and Nishizawa M. (1995). Mol. Cell. Biol., Beddington RSP, Morgenstern J, Land H and Hogan A. 15, 2180 ± 2190. (1989). Development, 106, 37 ± 46. Kataoka K, Noda M and Nishizawa M. (1996). Oncogene, Blank V, Kim MJ and Andrews NC. (1997). Blood, 89, 12, 53 ± 62. 3925 ± 3935. Kawai S and Hanafusa H. (1971). Virology, 46, 470 ± 479. Casado FJ, Pouponnot C, Jeanny JC, Lecoq O, Calothy G Kerppola TK and Curran T. (1994a). Oncogene, 9, 675 ± 684. and Pierani A. (1996). Mech. Dev., 54, 237 ± 250. Kerppola TK and Curran T. (1994b). Oncogene, 9, 3149 ± Cordes SP and Barsh GS. (1994). Cell, 79, 1025 ± 1034. 3158. Crisanti-Combes P, Privat A, Pessac B and Calothy G. Kemp BE and Pearson RB. (1990). Trends Biochem. Sci., 15, (1977). Cell Tissue Res., 185, 159 ± 173. 342 ± 346. EycheÁ ne A, Barnier JV, Deze le e P, Marx M, Laugier D, Kozak M. (1987). Nucleic Acids Res., 15, 8125 ± 8148. Calogeraki I and Calothy G. (1992). Oncogene, 7, 1315 ± Kumar R, Chen S, Scheurer D, Wang QL, Duh E, Sung CH, 1323. Rehemtulla A, Swaroop A, Adler R and Zack DJ. (1996). Farjo Q, Jackson AU, Xu J, Gryzenia M, Skolnick C, J. Biol. Chem., 271, 29612 ± 29618. Agarwal N and Swaroop A. (1993). Genomics, 18, 216 ± Kurschner C and Morgan JI. (1995). Mol. Cell. Biol., 15, 222. 246 ± 254. Farjo Q, Jackson A, Pieke-Dahl S, Scott K, Kimberling WJ, Kurschner C and Morgan JI. (1997). Biochem. Biophys. Res. Sieving PA, Richards JE and Swaroop A. (1997). Commun., 231, 333 ± 339. Genomics, 45, 395 ± 401. Liu Q, Ji X, Breitman ML, Hitchcock PF and Swaroop A. Fujiwara KT, Kataoka K and Nishizawa M. (1993). (1996). Oncogene, 12, 207 ± 211. Oncogene, 8, 2371 ± 2380. Nishizawa M, Kataoka K, Goto N, Fujiwara KT and Kawai Gorman CM, Mo€at LF and Howard BH. (1982). Mol. Cell. S. (1989). Proc. Natl. Sci. USA, 86, 7711 ± 7715. Biol., 2, 1044 ± 1051. Pessac B and Calothy G. (1974). Science, 185, 709 ± 710. Guermah M, Gillet G, Michel D, Laugier D, Brun G Pierani A, Pouponnot C and Calothy G. (1993). Mol. Cell. and Calothy G. (1990). Mol. Cell. Biol., 10, 3584 ± Biol., 13, 3401 ± 3414. 3590. Pierani A, Pouponnot C and Calothy G. (1995). Mol. Cell. Guermah M, Crisanti P, Laugier D, Deze le e P, Bidou L, Biol., 15, 642 ± 652. Pessac B and Calothy G. (1991). Proc. Natl. Acad. Sci. Pouponnot C, Nishizawa M, Calothy G and Pierani A. USA, 88, 4503 ± 4507. (1995). Mol. Cell. Biol., 15, 5563 ± 5575. Ho IC, Hodge MR, Rooney JW and Glimcher LH. (1996). Rehemtulla A, Warwar R, Kumar R, Xiaodong J, Zack DJ Cell, 85, 973 ± 983. and Swaroop A. (1996). Proc. Natl. Acad. Sci. USA, 93, Igarashi K, Kataoka K, Itoh K, Hayashi N, Nishizawa M 191 ± 195. and Yamamoto M. (1994). Nature, 367, 568 ± 572. Sakai M, Imaki J, Yoshida K, Ogata A, Matsushima-Hibiya Igarashi K, Itoh K, Motohashi H, Hayashi N, Matuzaki Y, Y, Kuboki Y, Nishizawa M and Nishi S. (1997). Nakauchi H, Nishizawa M and Yamamoto M. (1995). J. Oncogene, 14, 745 ± 750. Biol. Chem., 270, 7615 ± 7624. Sieweke MH, Tekotte H, Frampton J and Graf T. (1996). Kataoka K, Nishizawa M and Kawai S. (1993). J. Virol., 67, Cell, 85, 49 ± 60. 2133 ± 2141. Swaroop A, Xu JZ, Pawar H, Jackson A, Skolnick C and Kataoka K, Fujiwara KT, Noda M and Nishizawa M. Agarwal N. (1992). Proc. Natl. Acad. Sci. USA, 89, 266 ± (1994a). Mol. Cell. Biol., 14, 7581 ± 7591. 270.