Oncogene (1997) 14, 873 ± 877  1997 Stockton Press All rights reserved 0950 ± 9232/97 $12.00

SHORT REPORT c-Jun inhibits NF-E2 transcriptional activity in association with p18/ in Friend erythroleukemia cells

Claire Francastel1, Yvette Augery-Bourget1, Michel Prenant1, Mark Walters2, David IK Martin2 and Jacqueline Robert-Le ze neÁ s1

1INSERM U268, HoÃpital Paul Brousse, 94800 Villejuif, France; 2Fred Hutchinson Cancer Center, Seattle, Washington, USA.

We have reported previously that antisense c-jun over- factors, the ubiquitous AP-1 complex and the erythroid comes a block of Friend erythroleukemia cells to NF-E2 complex (Mignotte et al., 1989a, b; Ney et al., di€erentiation suggesting that the factor c-Jun may be 1990a and b; Talbot et al., 1991). The importance of an important negative regulator of erythroid di€erentia- the NF-E2/AP-1 sites in globin expression has tion. The recently described erythroid transcription been inferred from mutagenesis experiments. Mutation factor NF-E2 plays an important role in the regulation of these sites within HS2 leads to a marked decrease in of the transcription of globin and recognizes a its enhancer activity on b-globin in sequence containing an AP-1 site. NF-E2 is a complex of tranfected erythroid cells and in transgenic mice two bZip , p45 and p18/Maf. In order to (Caterina et al., 1991; Liu et al., 1992; Ney et al., determine whether c-Jun can interact with NF-E2/AP-1 1990a, b; Talbot et al., 1991). A direct evidence that sites to regulate transcriptional activation from them, we the NF-E2 complex is critical for globin expression is have compared the activity of AP-1 and NF-E2 in provided by the existence of a F-MEL cell line in transient transcriptional assays, in erythroid and non- which retroviral integration results in inactivation of erythroid cells in the presence of c-jun sense and the NF-E2 transcriptional factor (Lu et al., 1994). This antisense expression vectors. In non-erythroid cells, c- line is associated with drastic reduction of globin RNA Jun activates and NF-E2p18 inhibits both AP-1 and NF- expression which can be restored by re-introducing E2 activities, suggesting that NF-E2/AP-1 sites function NF-E2. as AP-1 binding sites in these cells. In contrast, NF- The NF-E2 complex, which was ®rst isolated from E2p18 is a positive regulator of NF-E2 activity in F-MEL cells, is a heterodimer of two bZip proteins of erythroid cells. c-Jun alone is also a positive regulator of 45 and 18 kDa (Andrews et al., 1993a, b). While NF- NF-E2 activity in erythroid cells but in association with E2p18 is widely expressed in many di€erent cell types, NF-E2p18 inhibits this activity. Moreover antisense c-jun NF-E2p45 expression is restricted to cells of the increases endogenous NF-E2 activity in erythroid cells. erythroid, megakaryocyte and mast cell lineage. Since These results suggest that c-Jun could act as a repressor NF-E2p45 homodimers cannot bind to NF-E2 motifs, of NF-E2 transcriptional activity by forming inactive c- NF-E2p45 binds as an obligate heterodimer with NF- Jun/NF-E2p18 heterocomplexes which interfer with the E2p18 (Andrews et al., 1993b), to form the NF-E2 transcription of globin genes in Friend erythroleukemia complex which has been described as the transcrip- cells. tional activator of NF-E2 enhancer elements (Igarashi et al., 1994). NF-E2p18 is encoded by a gene of Keywords: c-Jun; NF-E2; transactivation; globin the gene family lacking a putative trans- regulation; erythroleukemia activator domain (Andrews et al., 1993b; Fujiwara et al., 1993; Igarashi et al., 1995). Homodimers of small Maf proteins suppress transcription from NF-E2/AP-1 Friend murine erythroleukemia (F-MEL) cells provide sites in non-erythroid cells (Igarashi et al., 1994). Maf a useful system for studying the mechanisms which proteins could also heterodimerize with bZip proteins control erythroid di€erentiation and globin gene from the Fos/Jun family (Kataoka et al., 1994, 1995; expression. When exposed to DMSO or other Kerpola and Curran, 1994). The Jun/Maf heterodimers inducers, F-MEL cells undergo phenotypic changes have binding speci®cities distinct from those of the that resemble the ®nal stages of normal erythropoiesis Jun/Jun and Maf/Maf homodimers but their transcrip- (Friend et al., 1971; Chen et al., 1982). The most tional activity has not yet been reported. It is apparent striking feature is the increased synthesis of globin that multiple forms of homo- and heterodimeric chains which results primarily from the activation of complexes of these bZip proteins can mediate a wide globin gene transcription (Charnay et al., 1984; variety of transcriptional regulation. Profous-Juchelka et al., 1983). Expression of globin We have recently reported that the proto-oncogene genes is dependent on far-upstream locus control c-Jun may be a negative regulator of F-MEL erythroid regions (LCR) (Orkin, 1995). The overall enhancer di€erentiation (Francastel et al., 1994). Overexpression activity of the b-globin complex LCR requires the of c-jun inhibits F-MEL erythroid di€erentiation integrity of a short sequence in the HS-2 site, whereas inhibition of c-jun by antisense transcripts in containing tandem binding sites for two transcription a di€erentiation-resistant F-MEL cell line restores the sensitivity of these cells to DMSO-induced differentia- tion. Because c-Jun, as a homodimer or as a Correspondence: J Robert-Le ze neÁ s Received 10 May 1996; revised 21 October 1996; accepted 21 October heterodimer together with the NF-E2p18 protein, 1996 could bind to NF-E2/AP-1 binding sites, and since c-Jun and NF-E2 activity in erythroleukemia cells C Francastel et al 874 NF-E2 binding sites seem to be required for globin gene expression during erythroid di€erentiation, we hypothesised an inhibitory e€ect of c-jun expression on NF-E2 transcriptional activity in F-MEL cells. In this study, we have investigated and compared the e€ect of c-Jun and NF-E2p18 on AP-1 and NF-E2/AP-1 activities by performing transient transcription assays in Friend erythroid cells and in non-erythroid cells.

c-jun enhances and NF-E2p18 inhibits AP-1 activity in both erythroid and non-erythroid cells Figure 1 Description of the reporter expression used in this study. 5xTRE-tk-CAT contains ®ve AP-1 DNA-binding sites Since c-Jun and NF-E2p18, as homo- or heterodimers, upstream of the thymidine kinase (tk) promoter driving are known to bind the AP-1 related sites (Kerppola and expression of the chloramphenicol acetyl transferase (CAT) reporter gene. The construct 5xTRE-tk-CAT kindly provided by Curran, 1994), we ®rst evaluated their e€ect on AP-1 Dr E Ser¯ing (WuÈ rzburg, Germany) has been described elsewhere site transcriptional activity in F-MEL cells by using the (Angel et al., 1987). NFE2-SV-CAT contains the b-globin classical 5xTRE-tk-CAT reporter vector (Figure 1). tandem NF-E2 sites, the SV40 promoter and the CAT reporter Transient transfection with this plasmid of F-MEL gene. To construct NFE2-SV-CAT, we synthesised single-strand 745A cell line, in which c-jun transcripts are oligonucleotides containing a 46-base-pair (bp) enhancer element including a tandem NF-E2/AP-1 motif according to Walters and undetectable, resulted in a 4-fold induction of AP-1 Martin (1992): (5'-ctagaTCAAGCACAGCAATGCTGA- activity, while transfection of F-MEL TFP10 cell line, GTCATGATGAGTCATGCTGAGGCTTActgca-3') and (3'-tA- expressing endogenous c-jun (Francastel et al., 1992, GTTCGTGTCGTTACGACTCAGTACTAC TCAGTACGACT- 1994), induced a 30-fold increase in AP-1 site- CCGAATg-5'). They were annealed and cloned at the XbaI/PstI dependent transcription compared to the control sites of the pSV-CAT reporter plasmid (Promega). The minimal erythroid promoter construct (GATA-NFE2-hGH) has been plasmid tk-CAT (Figure 2a). The basal AP-1 activity previously described elsewhere (Walters and Martin, 1992) was increased in both 745A and TFP10 cells by

Figure 2 E€ects of c-Jun and NF-E2p18 on AP-1 transcriptional activity in both erythroid and non-erythroid cells (a) F-MEL cells and (b) NIH3T3 cells were transfected with 10 mg of the tk-CAT control plasmid lacking AP-1 sites, or 10 mg of the 5xTRE-tk-CAT plasmid. When indicated, 5xTRE-tk-CAT was cotransfected with 40 mg of plasmids expressing full-length c-jun (pSVc-junS) or NF- E2p18 (pSVp18). CAT enzymatic activity (% of conversion) is indicated. Errors bars represent standard error of the mean. Methods: NIH3T3 and F-MEL cells, strain 745A and TFP10 (Robert-Le ze neÁ s et al., 1988) were grown in RPMI-1640 medium (Gibco) containing 10% fetal calf serum (FCS, Eurobio). For each transfection, log-phase cells were washed in phosphate-bu€ered saline (PBS), pelleted and resuspended in ice cold PBS/HEPES 50 mM. Electroporations were performed using the Bio-Rad Gene Pulser with conditions of 400V and 500 mF for NIH3T3 cells, 370 V and 500 mF for F-MEL cells, 10 mg of supercoiled reporter construct 7 (tk-CATor 5xTRE-tk-CAT) and 10 cells in 0.4 ml of PBS/50 mM HEPES/10 mg salmon sperm DNA. The tk-CAT construct, which lacks the AP-1 element, served as the control for AP-1 speci®c transcriptional activity. Transfection mixture also contained 40 mgof supercoiled expression vectors. The c-jun sense expression vector (pSVc-junS) has been described previously (Francastel et al., 1994). NF-E2p18 (pSVp18) and NF-E2p45 (pSVp45) sense expression vectors were obtained by ligation of murine NF-E2p18 full length cDNA or NF-E2p45 full length cDNA (generous gifts from N Andrews, Boston) into EcoRI-digested pSVsport1 (BRL), under the control of the SP6 or the SV40 promoters. In vitro translated proteins were prepared using 1 mg of these expression vectors, SP6 polymerase and TNT-lysate-reticulocyte kit (Promega) to check the size of the proteins produced. 48 h after electroporation, cells were washed twice in PBS and lysed in Tris-HCl 250 mM/pH 8 by three cycles of freeze/thawing. After centrifugation to eliminate cell debris, extracts were heated at 608C to eliminate endogenous acetylases. The protein concentration of the supernatant was determined by the Bradford assay. CAT activity was assayed in 200 ml reactions containing 20 mg of total cell extracts, 0.2 mCi [(14C] chloramphenicol (NEN), 2 mM acetylcoenzyme A. After a 16 h incubation at 378C chloramphenicol and acetyled products were extracted by ethyl acetate. The acetylated products were separated from unacetylated chloramphenicol by thin-layer chromatography in chloroform/5% methanol. Quanti®cation of both unacetylated (U) and acetylated (A) forms of [(14C)] was performed using a Phosphorimager (Molecular Dynamics) and % of conversion calculated as follows: A*100/(U+A). For each experiment, three plasmid preparations were used. Data represent mean+standard error (SE) for three experiments c-Jun and NF-E2 activity in erythroleukemia cells C Francastel et al 875 cotransfection with c-jun sense expression vector (pSVc- experiments in the presence of c-jun or NF-E2p18 junS). In contrast, cotransfection of both F-MEL cell expression vectors. For this analysis, an oligonucleotide lines with the NF-E2p18 expression vector (pSVp18) carrying the human b-globin enhancer tandem NF-E2 alone, or pSVp18 together with pSVc-junS, resulted in a sites was inserted into a SV-CAT expression vector to marked decrease in AP-1 activity (Figure 2a). generate SV-NFE2-CAT (Figure 1). Transfection of To compare the e€ect of c-Jun and NF-E2p18 non-erythroid NIH3T3 cells with SV-NFE2-CAT proteins on AP-1 activity in non-erythroid cells, resulted in an increase in CAT activity compared to NIH3T3 cells, used as control cells, were also transfection of cells with a control of SV-CAT plasmid transiently transfected with the 5xTRE-tk-CAT vector. containing only the SV40 promoter (Figure 3a). These As expected, we detected signi®cant endogenous AP-1 results are not surprising since the NF-E2 site activity which was slightly increased by c-jun over- encompasses a perfect AP-1 site. Moreover, NF-E2 expression (Figure 2b). As observed in F-MEL cells, a activity in NIH3T3 was increased by c-jun expression marked decrease in AP-1 activity was obtained in and dramatically reduced by NF-E2p18 expression NIH3T3 cells by cotransfection with NF-E2p18 alone (Figure 3a), the latter observation being in agreement or NF-E2p18 together with pSVc-junS (Figure 2b). with the work of Igarashi et al. (1994). This inhibitory e€ect of NF-E2p18 on NF-E2 activity was also obtained in NIH3T3 cells by cotransfection with c-jun activates and NF-E2p18 inhibits NF-E2/AP-1 pSVp18 together with pSVc-junS (Figure 3a). These activity in non-erythroid cells results suggest that in non-erythroid cells, c-Jun and We next analysed NF-E2/AP-1 dependent reporter NF-E2p18 regulate the NF-E2/AP-1 site in the same gene expression by performing transient transfection way as an AP-1 site (compare Figures 2b and 3a).

Figure 3 c-Jun increases NF-E2/AP-1 transcriptional activity, but its association with NF-E2p18 is inactive. (a) F-MEL cells and (b) NIH3T3 cells were transfected with 10 mg of the SV-CAT control plasmid lacking NF-E2/AP-1 sites, or 10 mg of the NFE2-SV- CAT plasmid. When indicated, NFE2-SV-CAT was cotransfected with 40 mg of plasmids expressing full-length c-jun (pSVc-junS), NF-E2p45 (pSVp45) or NF-E2p18 (pSVp18). CAT enzymatic activity (% of conversion) is indicated. Error bars represent standard error of the mean. (c) F-MEL cells were transfected with a minimal erythroid construct (GATA-NFE2-hGH) containing GATA and NF-E2/AP-1 DNA-binding sites upstream of a rabbit b-globin TATA box driving expression of the hGH gene. When indicated, GATA-NFE2-hGH was cotransfected with 40 mg of plasmids expressing full-length c-jun (pSVc-junS), NF-E2p18 (pSVp18) or c-jun antisense transcripts (pSVc-junAS). Forty-eight h after electroporation, medium was harvested and 10 ml was assayed for hGH using an ELISA kit (Eurogenetics) according to the manufacturer recommendations. hGH levels in culture medium are indicated. The results consist of average of three separate experiments and error bars represent standard error (SE) of the mean c-Jun and NF-E2 activity in erythroleukemia cells C Francastel et al 876 (GATA-NF-E2-hGH) contains a globin TATA box c-jun or NF-E2p18 activates NF-E2/AP-1 activity in under the control of the b-globin tandem NF-E2 sites erythroid cells but their association is inactive and a binding site for one other erythroid transcrip- As seen on Figure 3b, endogenous NF-E2 site tional factor GATA-1 (see Figure 1). As in the case of dependent activity is very weak in F-MEL cells, while experiments using the SV-NFE2-CAT plasmid, it is a little stronger in c-jun expressing TFP-10 cells. although c-jun or NF-E2p18 overexpression resulted These results are in agreement with the work of in two- or threefold increase in hGH reporter gene Walters and Martin (1992) in which they have shown expression, cotransfection with both pSVc-junS and that NF-E2 sites are strong positive regulators of pSVp18 together did not increase transcription from erythroid gene expression, but only when linked to this minimal erythroid promoter in F-MEL cells other cis-acting elements from the human b-globin (Figure 3c). locus. However, the NF-E2 element responded to c-jun overexpression, and its activity was increased by about Antisense c-jun decreases AP-1 activity but increases sixfold in 745A cells and in TFP10 cells (Figure 3b). NF-E2/AP-1 activity in erythroid cells Cotransfection with pSVp18 also led to an increase in NF-E2 activity in 745A cells (threefold) and in TFP10 Since c-Jun could activate NF-E2 activity when present cells (®vefold), whereas cotransfection with pSVp45 alone but not when associated with NF-E2p18 (Figure had no e€ect on NF-E2 activity in F-MEL cells. While 3), is was interesting to analyse the e€ect of the NF-E2p18 transfected alone or together with pSVp45 inhibition of endogenous c-jun expression by antisense activate NF-E2 driven gene expression in F-MEL cells, c-jun transcripts on the activity of NF-E2 sites in both cotransfection with c-jun together with NF-E2p18 had erythroid and non-erythroid cell lines. Inhibition of c- no e€ect on NF-E2/AP-1 activity in 745A cells and jun expression by transfection with c-jun antisense decreased this activity in TFP10 cells (Figure 3b). expression vector (pSVc-junAS) led to a dramatic In order to con®rm the results obtained on NF-E2 inhibition of both AP-1 and NF-E2/AP-1 activity in sites linked to the SV40 heterologous promoter, we non-erythroid NIH3T3 cells (Figure 4a and b). In have studied the e€ect of c-jun and p18 expression contrast, transfection with pSVc-junAS decreased vectors on the minimal erythroid promoter described endogenous AP-1 activity but led to an increase in by Walters and Martin (1992). This minimal promoter endogenous NF-E2 activity in F-MEL cells (Figure 4c

Figure 4 Inhibition of c-jun by antisense transcripts increases NF-E2/AP-1 transcriptional activity in F-MEL erythroid cells. (a) NIH3T3 cells and (c) F-MEL cells were transfected with 10 mg of the tk-CAT plasmid or 10 mg of the 5xTRE-tk-CAT plasmid. When indicated, 5XTRE-tk-CAT was cotransfected with 40 mg of plasmid expressing c-jun antisense transcripts (pSVc-junAS). (b) NIH3T3 cells and (d) F-MEL cells were transfected with 10 mg of the SV-CAT plasmid or 10 mg of the NFE2-SV-CAT plasmid. When indicated, NFE2-SV-CAT was cotransfected with 40 mg of plasmids expressing c-jun (pSVc-junAS), junB (pSVjunBAS) or junD (pSVjunDAS) antisense transcripts. The antisense expression vectors pSVc-junAS, pSVjunBAS and pSVjunDAS have been described previously (Francastel et al., 1994). CAT enzymatic activity (% of conversion) is indicated. Results represent the average (+SE) of three independent experiments c-Jun and NF-E2 activity in erythroleukemia cells C Francastel et al 877 and d). These results clearly show that in erythroid formation of c-Jun/NF-E2p18 heterodimers that are cells the activity of AP-1 and NF-E2/AP-1 sites can be unable to activate transcription from these sites. distinguished by the inhibition of endogenous c-jun Similarly, recent studies have shown that c-Fos can using antisense expression vectors, whereas they are act as a negative regulator of transcription in indistinguishable in non-erythroid cells. The e€ect of combination with the small Maf proteins (Kataoka et antisense transcripts is more dramatic in TFP10 c-jun al., 1995). Although we favor the model of inactive c- expressing cells in which we obtained a 30-fold increase Jun/p18 complexes, two alternative explanations of our in NF-E2 activity, whereas only a fourfold increase in results may be considered. The ®rst is that transcrip- NF-E2 activity was observed in 745A cells. Antisense tional activation from NF-E2 sites in F-MEL cells can junB or antisense junD expression vectors had little or be dysregulated by sequestration of NF-E2p18 by c- no e€ect on NF-E2 driven reporter gene expression in Jun, resulting in a decrease in the amount of the NF- F-MEL cells (Figure 4d). These results suggest that E2p45/p18 complex which is the functional form of the endogenous c-Jun, but not JunB nor JunD, may act as NF-E2 . A second explanation a negative regulator of NF-E2/AP-1 transcriptional could be that c-Jun directly occludes the transactiva- activity in erythroid F-MEL cells. Moreover, we found tion domain of the NF-E2 p45/p18 complex. Finally, that the inhibition of endogenous c-jun expression in F- since NF-E2 sites seem to be required for globin gene MEL cells by cotansfection with pSVc-junAS also expression, this interaction between c-Jun and NF-E2 increased the activity of the miminal erythroid may provide a mechanism by which the proto- promoter containing NF-E2/AP-1 and GATA sites oncogene c-Jun, one of the major components of the (see Figure 3c). ubiquitous transcription factor AP-1, can block the In summary, we report that although c-Jun acts as a di€erentiation of Friend erythroleukemia cells. positive regulator of NF-E2/AP-1 activity, cotransfec- tion of c-Jun together with NF-E2p18 is not able to increase this activity in erythroid F-MEL cells. These results suggest that c-Jun can indeed act as a negative Acknowledgements We thank N Andrews for providing NF-E2p45 and NF- regulator of NF-E2 transcriptional activity, when E2p18 cDNAs, and E Ser¯ing for 5xTRE-tk-CAT plasmid. associated with NF-E2p18 protein. Since it was This work was supported in part by grants from the reported that Maf/Jun heterodimers can bind NF-E2/ Association pour la Recherche sur le Cancer (ARC 3035) AP-1 like sites more eciently than c-Jun homodimers and from the Association NRB-Vaincre le Cancer. C or Maf homodimers (Kataoka et al., 1994; Kerppola Francastel was supported by grants from the Association and Curran, 1994), NF-E2 activity may be repressed by pour la Recherche sur le Cancer.

References

Andrews NC, Erdjument-Bromage H, Davidson MB, Kataoka K, Noda M and Nishizawa M. (1994). Mol. Cell. Tempst P and Orkin SH. (1993a). Nature, 362, 722 ± Biol., 14, 700 ± 712. 728. KataokaK,IgarashiK,ItohK,Fujiwara,K,NodaM, Andrews NC, Kotkow KJ, Ney PA, Erdjument-Bromage H, Yamamoto M and Nishizawa M. (1995). Mol. Cell. Biol., Tempst P and Orkin SH. (1993b). Proc. Natl. Acad. Sci. 15, 2180 ± 2190. USA, 90, 11488 ± 11492. Kerppola TK and Curran T. (1994). Oncogene, 9, 675 ± 684. Angel P, Imagawa M, Chiu R, Stein B, Imbra RJ, Liu D, Chang JC, Moi P, Liu W, Kan YW and Curtin PT. Rahmsdorf HJ, Jonat C, Herrlich P and Karin M. (1992). Proc. Natl. Acad. Sci. USA, 89, 3899 ± 3903. (1987). Cell, 49, 729 ± 739. Lu SJ, Rowan S, Bani MR and Ben-David Y. (1994). Proc. Caterina JJ, Ryan TM, Pawlik KM, Palmiter RD, Brinster Natl. Acad. Sci. USA, 91, 8398 ± 8402. RL, Behringer RR and Townes TM. (1991). Proc. Natl. Mignotte V, Eleouet JF, Raich N and Romeo PH. (1989a). Acad. Sci. USA, 88, 1626 ± 1630. Proc.Natl.Acad.Sci.USA,86, 6548 ± 6552. CharnayP,TreismanP,MellonP,ChaoM,AxelRand Mignotte V, Wall L, DeBoer E, Grosveld F and Romeo PH. Maniatis T. (1984). Cell, 55, 17 ± 26. (1989b). Nucleic Acids Res., 17, 37 ± 54. Chen ZX, Banks J, Rifkind RA and Marks PA. (1982). Proc. Ney PA, Sorrentino BP, Lowrey CH and Nienhuis AW. Natl. Acad. Sci. USA, 79, 471 ± 475. (1990a). Nucleic Acids Res., 18, 6011-6017. Francastel C, Mazouzi M and Robert-Le ze neÁ s J. (1992). Ney PA, Sorrentino BP, McDonagh KT and Nienhuis AW. Leukemia, 6, 935 ± 939. (1990b). Genes Dev., 4, 993 ± 1006. Francastel C, Groisman R, Pfarr CM and Robert-Le ze neÁ sJ. Orkin SH. (1995). Eur. J. Biochem., 231, 271 ± 281. (1994). Oncogene, 9, 1957 ± 1964. Profous-Juchelka HR, Reuben RC, Marks PA and Rifkind Friend C, Scher W, Holland JG and Sato T. (1971). Proc. RA. (1983). Mol. Cell. Biol., 3, 229 ± 232. Natl. Acad. Sci. USA, 68, 378 ± 382. Robert-Le ze neÁ s J, Meneceur P, Ray D and Moreau- Fujiwara KT, Kataoka K and Nishizawa M. (1993). Gachelin F. (1988). Cancer Res., 48, 3972 ± 3976. Oncogene 8, 2371 ± 2380. Talbot D and Grosveld F. (1991). EMBO J., 10, 1391 ± 1398. Igarashi K, Kataoka K, Itoh, K,Hayashi N, Nishizawa M Walters M and Martin DIK. (1992). Proc. Natl. Acad. Sci. and Yamamoto M. (1994). Nature, 367, 568 ± 572. USA, 89, 10444 ± 10448. Igarashi K, Itoh K, Motohashi H, Hayashi N, Matuzaki Y, Nakauchi H, Nishizawa M and Yamamoto M. (1995). J. Biol. Chem., 270, 7615 ± 7624.