C/EBP family transcription factors are degraded by the ... - Nature

Oncogene (2003) 22, 1273–1280

& 2003 Nature Publishing Group All rights reserved 0950-9232/03 $25.00 www.nature.com/onc

ORIGINAL PAPER

C/EBP family transcription factors are degraded by the proteasome but stabilized by forming dimer Takayuki Hattori1, Nobumichi Ohoka, Yasumichi Inoue, Hidetoshi Hayashi and Kikuo Onozaki* Department of Molecular Health Sciences, Graduate School of Pharmaceutical Sciences, Nagoya City University, Mizuho, Nagoya 467-8603, Japan

CCAAT/enhancer-binding protein (C/EBP) family transcription factors are critical for transcription of several genes involved in tissue development and cellular function, proliferation, and differentiation. Here we show that inhibitory/regulatory C/EBP family proteins, Ig/EBP (C/ EBPc) and CHOP (C/EBPf), but not positively functioning NF-IL6 (C/EBPb), are constitutively multiubiquitinated and subsequently degraded by the proteasome. In addition, ubiquitination and degradation of these proteins are suppressed by forming dimer through their leucine zipper domains. Deletion of leucine zipper domain in NFIL6 caused the loss of its homodimerization activity and the degradation of protein by the ubiquitin–proteasome system. In addition, Ig/EBP with its leucine zipper domain substituted for that of NF-IL6 formed homodimer and was stabilized. These observations suggest that mammalian cells equip a novel regulatory system abrogating the excess C/EBP family transcription factors bereft of dimerizing partner. Oncogene (2003) 22, 1273–1280. doi:10.1038/sj.onc.1206204 Keywords: C/EBP; degradation; proteasome; ubiquitination; dimer; leucine zipper

Introduction CCAAT/enhancer-binding proteins (C/EBPs) are a family of leucine zipper transcription factors that are critical for the regulation of various aspects of cellular differentiation and function in multiple tissues. This family consists of six members: C/EBPa, b (NF-IL6), g (Ig/EBP), d, e, and C/EBP homologous protein (CHOP, C/EBPz) (Lekstrom-Himes and Xanthopoulos, 1998). The prototypic C/EBP consists of a transactivation domain, a dimerization bZIP region, and a DNAbinding domain. All family members share a strong *Correspondence: Kikuo Onozaki; E-mail: [email protected] 1 Current address: Department of Biochemistry 1, Hamamatsu University School of Medicine, 1-20-1 Handayama, Hamamatsu 4313192, Japan, and CREST, Japan Science and Technology Corporation, 4-1-8 Honcho, Kawaguchi, Saitama 332-0012, Japan Received 29 July 2002; revised 25 October 2002; accepted 30 October 2002

homology in carboxyl-terminal domain, which carries a basic DNA-binding domain and a leucine zipper motif. Recent studies on eliminating of C/EBP genes in targeted mouse model underscore the role these factors play in normal tissue development and cellular function, cellular proliferation, and functional differentiation. The pleiotropic effects of C/EBPs are in part because of tissue- and stage- specific expression, leaky ribosomal reading, post-transcription modifications, and variable DNA-binding specificities. These mechanisms result in variable amounts of the C/EBP isoforms available to dimerize and bind to cognate site in different tissues (Lekstrom-Himes and Xanthopoulos, 1998). Probably because of their key function in processes that need to be tightly controlled, C/EBP family proteins are under stringent regulation at every level. For example, two or three isoforms of C/EBPa and b are generated from single mRNA by a leaky ribosomal scanning mechanism (Poli et al., 1990; Descombes and Schibler, 1991; Lin et al., 1993; Ossipow et al., 1993). The truncated protein possesses only the DNA-binding and leucine zipper domain. Heterodimerization of it with a full-length one attenuates transcriptional activity, suggesting a dominant negative mechanism of transcriptional regulation. However, the eventual fate of unnecessary C/EBP family members remains unknown. A large number of transcription factors undergo degradation by the ubiquitin–proteasome-dependent pathway (Hochstrasser, 1995; Pahl and Baeuerle, 1996). In this study, we provide evidence that C/EBP family proteins bereft of dimerizing partner undergo ubiquitination and degradation by the proteasome. This suggests a novel mechanism for the targeted clearance of useless C/EBP family proteins, ensuring the irreversible termination of excess C/EBP family proteins signaling.

Results Ig/EBP and CHOP proteins, C/EBP family transcription factors are degraded by the proteasome system The effect of a proteasome inhibitor, MG132, on the ectopic expression of Ig/EBP (C/EBPg) in human melanoma cell line A375 and human embryonic kidney cell line293 is shown in Figure 1a. Treatment with

Degradation of C/EBP family transcription factors T Hattori et al

1274

MG132 caused accumulation of Ig/EBP protein in both A375 cells (Figure 1a; A375/MG132) and 293 cells (Figure 1a; 293). Inhibition of de novo protein synthesis with cycloheximide caused the diminution of intracellular Ig/EBP protein level (Figure 1a; A375/CHX). Pulse-chase analysis also demonstrated that Ig/EBP protein was stabilized with MG132 (Figure 1b). Exogenous CHOP (C/EBPz) protein was also accumulated by proteasome inhibition, but MG132 did not affect exogenous NF-IL6 (C/EBPb) protein level (Figure 1c). Accumulation of CHOP or Ig/EBP proteins but not NF-IL6 protein was also detected in the A375 cells treated with 5 mm of lactacystin, another proteasome selective inhibitor (data not shown). Moreover in A375 cells, the expression level of endogenous CHOP protein was also markedly augmented in the presence of MG132 similar to in the presence of methyl methanesulfonate (MMS), a potent inducer of CHOP (Figure 1d); however, that of endogenous NF-IL6 protein was not (data not shown). These observations suggest that Ig/ EBP and CHOP are constantly degraded by the proteasome system. Ig/EBP and CHOP but not NF-IL6 are constitutively multiubiquitinated

Figure 1 Inhibition of proteasome activity stabilizes C/EBP family proteins. (a) A375 cells or 293 cells were transiently transfected with pcDNA3.1-Flag-Ig/EBP. After 24 h, A375 cells were either left untreated or treated with indicated doses of MG132 or cycloheximide (CHX) for 15 h (A375). 293 cells were either left untreated or treated with 30 mm MG132 for 6 h (293). The cell lysates were analysed by immunoblotting using anti-Flag and anti-b-actin antibodies. (b) 293 cells were transiently transfected with pcDNA3.1-Flag-Ig/EBP. After 2 days, 293 cells were pulsed with [35S]methionine/cysteine, and then chased for the indicated periods with or without 10 mm MG132. The cell lysates were immunoprecipitated with anti-Flag antibody, and analysed by SDS–PAGE and autoradiography. The Flag-Ig/EBP signals were quantified and the values relative to the value at time 0 were plotted as a function of chase time. (c) A375 cells were transiently transfected with pcDNA3.1-Flag-Ig/EBP, pcDNA3.1-Flag-CHOP, or pCMV-Flag-NFIL6. After 24 h, cells were either left untreated or treated with 10 mm MG132 for 15 h. The cell lysates were analysed by immunoblotting using anti-Flag and anti-b-actin antibodies. (d) A375 cells were either left untreated or treated with 2 mm of MG132, 60 mg/ml of MMS or with both for 6 h. Then the lysates were analysed by immunoblotting using anti-CHOP and anti-b-actin antibodies Oncogene

To determine whether ubiquitins are conjugated to these C/EBP family proteins in vivo, we subjected lysates from 293 cells, transiently expressed with Flag-tagged C/EBPs and HA-tagged ubiquitin to immunoprecipitate with an anti-Flag antibody and then analysed the immunoprecipitates by immunoblotting using an anti-HA antibody. To prevent degradation of ubiquitin-conjugated proteins through proteasome-mediated proteolysis, the experiments were performed in the presence of MG132. Accumulation of high-molecular-mass species recognized by the anti-HA antibody was observed in the ectopic CHOP- or Ig/EBP-expressed 293 cells (Figure 2a; lanes 3 and 5); however, significant multiubiquitination was not observed in NF-IL6 (Figure 2a; lane 7). Immunoblotting with antibody used in immunoprecipitation (anti-Flag) showed a ladder pattern similar to anti-HA blotting (Figure 2b; lanes 2 and 4), indicating that ubiquitins were directly conjugated to CHOP and Ig/EBP. Multiubiquitination and subsequent degradation of Ig/EBP and CHOP proteins are suppressed by heterodimerization Multiubiquitination of Ig/EBP was markedly suppressed by coexpressing with other C/EBP family proteins, CHOP (Figure 2a; lanes 5 and 6) or NF-IL6 (Figure 2c). In a similar manner, multiubiquitination of CHOP was suppressed by coexpressing with NF-IL6 (Figure 2d). On the other hand, multiubiquitination of Flag-CHOP was not affected by expression of MycCHOP (Figure 2a; lanes 3 and 4). Similar results were obtained by anti-Flag immunoprecipitation and antiFlag immunoblotting (Figure 2b; right panel).


1275

Oncogene

Figure 2 Ig/EBP and CHOP proteins but not NF-IL6 protein are multiubiquitinated. (a and b) 293 cells were transiently transfected with indicated constructs. After 24 h, cells were treated with 10 mm MG132 for 16 h. The cell lysates were immunoprecipitated (IP) with anti-Flag antibody, and multiubiquitinated C/EBPs [(Ub)nC/EBPs] were detected by immunoblotting with anti-HA (a, top and b, top left) or anti-Flag (b, top right) antibodies. Expression level of each protein was assessed by anti-Flag (middle) and anti-Myc (bottom) immunoblotting of cell lysates. (c and d) NF-IL6 suppresses the multiubiquitination of Ig/EBP or CHOP. 293 cells were transiently transfected with indicated constructs. After 24 h, cells were treated with 10 mm MG132 for 16 h. The cell lysates were immunoprecipitated with anti-Flag antibody, and multiubiquitinated C/EBPs [(Ub)nIg/EBP (c) and (Ub)nCHOP (d)] were detected by immunoblotting with anti-HA antibody (top). Expression level of each protein was assessed by anti-Flag and anti-NF-IL6 immunoblotting of cell lysates (bottom)


1276

As NF-IL6 forms both homo- and heterodimer, it may prevent itself from ubiquitination by forming homodimer. Ig/EBP appeared to form heterodimer with wild-type CHOP, CHOPSA, mutated into p38 MAPKdependent phosphorylation sites essential for its trans-

activational activity (Wang and Ron, 1996) and CHOPDBR, lacking whole DNA-binding domain (Ubeda et al., 1996), but not with CHOPDLZ, lacking leucine zipper domain required for dimer formation (Ron and Habener, 1992) (Figure 3a and b; lanes 3–6), suggesting

Figure 3 Ig/EBP protein heterodimerizes with CHOP (b), but does not form a homodimer (c). (a) Schematic representation of mutated CHOP constructs. (b and c) 293 cells were transiently transfected with indicated constructs. After 24 h, cells were treated with 10 mm MG132 for 16 h and cell lysates were immunoprecipitated with an anti-Flag antibody, and precipitated Ig/EBP (b, middle and c, top) or CHOPs (b, top) were detected by immunoblotting with anti-Myc or anti-Flag antibodies. Expression level of each protein was assessed by anti-Flag or anti-Myc immunoblotting of cell lysates (b, bottom and c, middle and bottom) Oncogene


1277

the critical role of leucine zipper domain for heterodimer formation. CHOP proteins also form heterodimer with other C/EBP family proteins (Roman et al., 1990, Ron and Habener, 1992); however, there is no evidence that CHOP forms homodimer (our own data not shown, Ubeda et al., 1996). Ig/EBP also did not form homodimer (Figure 3c; lane 4). Therefore, it is likely that CHOP and Ig/EBP each alone cannot prevent itself from ubiquitination. As shown in Figure 4a, Ig/EBP protein was markedly stabilized by coexpressing with CHOP (compare lanes 2 with 3) and its muteins except for CHOPDLZ (lanes 4– 6), consistent with their formation of heterodimer (Figure 3b). A pulse-chase experiment confirmed that Ig/EBP protein was metabolically stabilized by wild-type CHOP, but not by CHOPDLZ (Figure 4b). Leucine zipper domain in NF-IL6 protein is critical for the homodimerization and its stabilization To assess the significance of the leucine zipper domain in NF-IL6 protein for its stabilization, we deleted this region and determined the stability of this protein in the cells (Figure 5a). As was expected the expression level of truncated NF-IL6 protein (NF-IL6DLZ) was much lower than that of wild-type NF-IL6 (Figure 5b; left panel), and MG132 treatment recovered the stability of NF-IL6DLZ (Figure 5c; lanes 1–4). Next, we constructed the Ig/EBP mutein with its leucine zipper domain substituted for that of NF-IL6 (Ig/EBPbLZ, Figure 5a). Ig/EBPbLZ protein was remarkably stabilized in the cells (Figure 5b; right panel) and its

expression level was not affected by MG132 (Figure 5c; lanes 5–8). Deletion of leucine zipper domain in NF-IL6 lost its activity to form homodimer, while Ig/EBPbLZ exhibited the potent homodimerizing activity in 293 cells (Figure 5d, e) and A375 cells (data not shown). These observations suggest that C/EBP family proteins stabilize themselves by evading ubiquitin–proteasomemediated degradation by forming dimer via leucine zipper domain.

Discussion C/EBP family transcription factors are involved in a multiple tissue or cellular events; therefore, the expression and disappearance of C/EBPs must be strictly and timely regulated. For example, in the preadipocyte differentiation process, expression of individual C/EBP is, in turn, induced or suppressed in a stage-dependent manner (Darlington et al., 1998; Lane et al., 1999). Indeed, expression of C/EBPs is regulated in several steps and fashions including transcription, translation, multiple isoforms, and post-transcription and translation modification (Lekstrom-Himes and Xanthopoulos, 1998). In this study, we demonstrated ubiquitin–proteasome dependent degradation of C/EBP family transcription factors, Ig/EBP and CHOP. This proteolytic mechanism may contribute rapid and target-specific degradation in the preadipocyte differentiation process. C/EBP family transcription factor is required to form homo- or heterodimer to bind DNA and exert transcriptional activity (Landschulz et al., 1989). Our

Figure 4 Ig/EBP protein is stabilized with CHOP protein dependent on its leucine zipper domain. (a) A375 cells were transiently transfected with indicated constructs. After 48 h, the cell lysates were analysed by immunoblotting using anti-Flag (top) and anti-Myc antibodies (middle). An arrowhead in the middle panel represents nonspecific bands. The pEGFP-C1 expression vector was included in each transfection as a transfection efficiency control, and its level was detected with anti-GFP monoclonal antibody (bottom). (b) 293 cells were transiently transfected with indicated constructs. After 48 h, the cells were pulsed with [35S]methionine/cysteine, and then chased for the indicated periods. The cell lysates were immunoprecipitated with anti-Myc antibody, and analysed as described in Figure 1b Oncogene

1278

Oncogene


Figure 5 Leucine zipper domain of NF-IL6 protein is crucial for its homodimerization and stabilization. (a) Schematic representation of mutated NF-IL6 and Ig/EBP constructs. (b) A375 cells were transiently transfected with indicated constructs. After 48 h, the lysates were analysed by immunoblotting using an anti-Flag antibody (top). The pEGFP-C1 expression vector was included in each transfection as a transfection efficiency control, and its level was detected with anti-GFP antibody (bottom). (c) A375 cells were transiently transfected with indicated constructs. After 24 h, cells were incubated for 15 h with or without 2 mm of MG132. Then, the cell lysates were analysed by immunoblotting using anti-Flag antibody (top) and anti-b-actin (bottom). (d and e) 293 cells were transiently transfected with indicated constructs. After 48 h, the cell lysates were immunoprecipitated with anti-Myc or anti-Flag antibodies, and precipitated NF-IL6DLZ (d) or Ig/EBPbLZ (e) were detected by immunoblotting with anti-Myc (tops) or anti-Flag (second panels) antibodies. Expression level of each protein was assessed by anti-Flag or anti-Myc immunoblotting of cell lysates (third panels and bottoms)


1279

results revealed that C/EBP proteins were stabilized by forming either homo- or heterodimer. The ubiquitin ligase or modification enzyme involved in ubiquitin conjugation to C/EBPs may specifically recognize monomer forms of C/EBPs. These findings suggest that mammalian cells possess a system for eliminating transcriptionally nonfunctioning C/EBPs existing as monomers or for maintaining the basal level of intracellular C/EBP proteins. Although Ig/EBP and CHOP cannot form homodimer (Figure 3c, Ubeda et al., 1996), they readily form heterodimer with other C/ EBP members (Roman et al., 1990; Ron and Habener, 1992). In contrast, other C/EBP members can form homodimer. Consistent with these reports and our results, the expression level of NF-IL6 was not affected by the proteasome inhibitors and significant multiubiquitination was not observed in NF-IL6. We further demonstrated that the leucine zipper domain of NF-IL6 is crucial for forming homodimer and its stabilization. These findings strongly suggest that NF-IL6 stabilizes itself by homodimerization. Since Ig/EBP lacks transactivational element and CHOP heterodimer cannot bind traditional C/EBP-binding site, they act as a dominant-negative C/EBP (Ron and Habener, 1992; Cooper et al., 1995). These inhibitory/regulatory C/ EBPs bereft of dimerizing partner may be preferentially eliminated by ubiquitin–proteasome system to avoid intracellular excess inhibitory regulation. We compared the sequences of leucine zipper domain of three C/EBP family proteins. In five leucines in the leucine zipper domain of NF-IL6, three of them were conserved in CHOP and four in Ig/EBP. It is possible that the difference in the number of leucines caused the differences of their activities to form dimer and affected their stability. As other views, other amino acids in the leucine zipper domains or the C terminal domain of each protein are critical for the dimerizing activity and stability. We need further study to analyse the detail mechanism. This is the first report that demonstrates ubiquitin– proteasome system-dependent regulation of intracellular mammalian C/EBP family members’ level. Further study is required to analyse the ubiquitin conjugation mechanism for C/EBPs, including ubiquitin ligase.

Cell culture A human melanoma cell line A375 given by Dr R Ruddon (NCI, Bethesda, MD, USA) was cultured in RPMI 1640 supplemented with 15 mm HEPES and 5% heat-inactivated FBS. This cell line was used for the study of interleukin (IL)-1 antiproliferative effect, and we observed a positive regulatory role of CHOP in IL-1-induced IL-6 transcription (Hattori et al., 2001). A human embryonic kidney cell line 293 obtained from Japanese Collection of Research Bioresources (Tokyo, Japan) was cultured in DMEM supplemented with 10% heatinactivated FBS. These cells were carried at 371C in air containing 5% CO2 supplemented with antibiotics.

Construction of expression plasmids The plasmid pcDNA3.1-Myc-CHOP, fused c-myc epitope tag into amino-terminal of human CHOP, pcDNA3.1-MycCHOPSA, replaced Ser79 and Ser82 with Ala, pcDNA3.1Myc-CHOPDLZ, lacking the leucine zipper domain (D134– 169), and pcDNA3.1-Myc-CHOPDBR, removed the whole basic region (D101–123), were constructed as described previously (Hattori et al., 2001). To construct pcDNA3.1Flag-Ig/EBP, human Ig/EBP cDNA obtained by PCR from pBluescript-hC/EBPg (gift from Dr M Li-Weber, German Cancer Research Center, Heidelberg) was ligated with pCMV5-Flag, and a fragment of Flag-Ig/EBP from pCMV5Flag-Ig/EBP was cloned into pcDNA3.1/Hygro (Invitrogen). pCMV-Flag-NFIL6 was constructed by ligating human NFIL6 cDNA from pEF-NFIL6 (Nishio et al., 1993), kindly provided by Dr S Akira (Osaka University, Japan), with pCMV-Tag2 (Stratagene). pCMV-Flag-NFIL6DLZ (1-314), lacking the part of leucine zipper domain, was generated by overlap extension using PCR with synthetic oligonucleotides encoding the terminal codon. To prepare pcDNA3.1-Flag-Ig/ EBPbLZ, leucine zipper domain of NF-IL6 was generated by PCR and subcloned into the pcDNA3.1-Flag-Ig/EBP plasmid. All constructs were verified by sequencing. pMT123, which expresses 8 HA-tagged ubiquitin, was kindly provided by Dr D Bohmann (European Molecular Biology Laboratory, Heidelberg). pEGFP-C1 was purchased from Clontech.

Transfection A375 cells were transfected using Effectene (Qiagen) according to the manufacture’s instructions. 293 cells were transfected by the Chen–Okayama (1987) method.

Materials and methods Immunoprecipitation and Western blot analysis Reagents RPMI 1640 and Dulbecco’s modified Eagle’s medium (DMEM) were purchased from Sigma; methionine/cysteinefree DMEM was from Life Technologies; fetal bovine serum (FBS) was from JRH Bioscience; anti-b-actin monoclonal antibody (AC-15) and anti-Flag monoclonal antibody (M2) from Sigma, anti-Myc monoclonal antibody (9E10) and antiHA monoclonal antibody (12CA5) from Roche, anti-NF-IL6 polyclonal antibody (C-19) and anti-GADD153 monoclonal antibody (B-3) from Santa Cruz and anti-GFP monoclonal antibody (JL-8) from Clontech; ubiquitin aldehyde was obtained from Calbiochem; and MG132 was obtained from Peptide Institute.

Cells were transiently transfected and treated as described in the figure legends. The cells were lysed in RIPA buffer (50 mm Tris-HCl, pH 8.0, 150 mm NaCl, 0.1% SDS, 0.5% deoxycholate, 1% Triton X-100), supplemented with protease inhibitors and ubiquitin aldehyde (0.25 mg/ml). The lysates were subjected to immunoprecipitation with the antibody described in figure legends. 1–5% of lysates and coimmunoprecipitates were subjected to SDS – 12.5% PAGE, transferred onto PVDF membranes and probed with antibody(s) described in the figure legends. The immunoreactive proteins were visualized by ECL Western blotting detection reagents (Amersham Pharmacia Biotech), and the light emission was quantified by lumino image analyzer LAS-1000 (FUJI). Oncogene


1280 Pulse-chase experiments 293 cells were transiently transfected as described in figure legends. After 48 hours, the cells were incubated with starvation DMEM (without methionine and cysteine, supplemented with 0.2% dialyzed serum) for 15 min, and then pulsed the cells for 15 min with 200 mCi/ml [35S]methionine/cysteine (Amersham Bioscience). The cells were then chased for various periods with normal DMEM in the absence or presence of 10 mm MG132. The cells were lysed in RIPA buffer supplemented with protease inhibitors, and the lysates were subjected to immunoprecipitation with antibody as described in the figure legends. The resulting precipitates were subjected to

SDS–12.5% PAGE, autoradiography, and image analysis (BAS-2500; FUJI).

Acknowledgements We thank Dr M Li-Weber, Dr S Akira, and Dr D Bohmann for providing expression plasmids. This work was supported in part by Grants-in-Aid for Scientific Research (B) from Japan Society for the Promotion of Science, and Grants-in-Aid for Scientific Research on Priority Areas (C) from The Ministry of Education, Science, Sports and Culture.

References Chen C and Okayama H. (1987). Mol. Cell. Biol., 7, 2745–2749. Cooper C, Henderson A, Artandi S, Avitahl N and Calame K. (1995). Nucleic Acids Res., 23, 4371–4377. Darlington GJ, Ross SE and MacDougald OA. (1998). J. Biol. Chem., 273, 30057–30060. Descombes P and Schibler U. (1991). Cell, 67, 569–579. Hattori T, Itoh S, Hayashi H, Chiba T, Takii T, Yoshizaki K and Onozaki K. (2001). J. Interferon Cytokine Res., 21, 323–332. Hochstrasser M. (1995). Curr. Opin. Cell. Biol., 7, 215–223. Landschulz WH, Johnson PS and McKnight SL. (1989). Science, 243, 1681–1688. Lane MD, Tang Q-Q and Jiang M-S. (1999). Biochem. Biophys. Res. Commun., 266, 677–683. Lekstrom-Himes J and Xanthopoulos KG. (1998). J. Biol. Chem., 273, 28545–28548.

Oncogene

Lin FT, MacDougald OA, Diehl AM and Lane MD. (1993). Proc. Natl. Acad. Sci. USA, 90, 9606–9610. Nishio Y, Isshiki H, Kishimoto T and Akira S. (1993). Mol. Cell. Biol., 13, 1854–1862. Ossipow V, Descombes P and Schibler U. (1993). Proc. Natl. Acad. Sci. USA, 90, 8219–8223. Pahl HL and Baeuerle PA. (1996). Curr. Opin. Cell. Biol., 8, 340–347. Poli V, Mancini FP and Cortese R. (1990). Cell, 63, 643–653. Roman C, Platero JS, Shuman J and Calame K. (1990). Genes Dev., 4, 1404–1415. Ron D and Habener JF. (1992). Genes Dev., 6, 439–453. Ubeda M, Wang X-Z, Zinszner H, Wo I, Habener JF and Ron D. (1996). Mol. Cell. Biol., 16, 1479–1489. Wang X-Z and Ron D. (1996). Science, 272, 1347–1349.