Ubiquitination and Degradation of ATF2 Are Dimerization Dependent

Abstract

Ubiquitination and proteasome-dependent degradation are key determinants of the half-lives of many transcription factors. Homo- or heterodimerization of basic region-leucine zipper (bZIP) transcription factors is required for their transcriptional activities. Here we show that activating transcription factor 2 (ATF2) heterodimerization with specific bZIP proteins is an important determinant of the ubiquitination and proteasome-dependent degradation of ATF2. Depletion of c-Jun as one of the ATF2 heterodimer partners from the targeting proteins decreased the efficiency of ATF2 ubiquitination in vitro, whereas the addition of exogenously purified c-Jun restored it. Similarly, overexpression of c-Jun in 293T human embryo kidney cells increased ATF2 ubiquitination in vivo and reduced its half-life in a dose-dependent manner. Mutations of ATF2 that disrupt its dimerization inhibited ATF2 ubiquitination in vitro and in vivo. Conversely, removal of residues 150 to 248, as in a constitutively active ATF2 spliced form, enhanced ATF2 dimerization and transactivation, which coincided with increased ubiquitination and decreased stability. Our findings indicate the increased sensitivity of transcriptionally active dimers of ATF2 to ubiquitination and proteasome-dependent degradation. Based on these observations, we conclude that increased targeting of a transcriptionally active ATF2 form indicates the mechanism by which the magnitude and the duration of the cellular stress response are regulated.

Posttranslational regulation of transcription factors has been recognized as the central mechanism in the mammalian response to stress and damage (16). Activating transcription factor 2 (ATF2) is a member of the ATF/CREB protein family of basic region-leucine zipper (bZIP) proteins (13, 14, 23), which are involved in the response to stress (22, 37). Amino-terminal phosphorylation of ATF2 mediated by Jun N-terminal kinase (JNK) (11) and p38 MAP kinase (29, 30) in response to stress and inflammatory cytokines results in the transactivation of ATF2, leading to increased expression of target genes. Among the target genes thought to mediate ATF2 functions in cell growth, differentiation, immune response, and response to stress are the c-jun (36), tumor necrosis factor alpha (35), transforming growth factor β (17), cyclin A (32), E-selectin (30), and DNA polymerase β (26) genes. It is known that the products of ATF2 target genes are likely to contribute to the neoplastic process and inflammation, but the physiological role of ATF2 remains largely uncharacterized.

In the absence of extracellular stimulation, ATF2 exhibits very low levels of transactivation because of an intramolecular inhibitory interaction in which the DNA binding domain binds to the amino-terminal transactivation domain (18). Several viral proteins, including adenovirus E1A (20), protein X of hepatitis B virus (24), and human T-cell leukemia virus type 1 Tax (39), interact with ATF2 and stimulate its transcriptional activity.

Transcriptionally active ATF2 recognizes and binds specific ATF/CRE motifs as a homo- or heterodimer. ATF2 interacts with its heterodimerization partners (i.e., other bZIP proteins) via the leucine zipper (40). The nature of the dimerization partners depends on the cell type and determines the specificity and the extent of transactivation of target genes. For instance, in F9 teratocarcinoma cells, which express very low levels of c-Jun in the absence of differentiation stimuli (41), ATF2 upregulates the expression of c-jun. De novo-synthesized c-Jun is capable of heterodimerization with ATF2, resulting in the activation of its own promoter and the formation of a positive-feedback regulatory loop (36, 37). The mechanisms by which this response may be downregulated remain unclear.

ATF2 is degraded in vivo via the ubiquitin-proteasome pathway (5, 7). We found that the ubiquitination of ATF2 as well as of c-Jun, JunB, and p53 is targeted by association with JNK (6–8). Our previous data demonstrated that such targeting of c-Jun and ATF2 ubiquitination occurs in a phosphorylation-dependent manner. Interestingly, the association of ATF2 with JNK is necessary but not sufficient for the targeting of ATF2 ubiquitination in vitro (7). In the present study, we sought to further elucidate the regulation of ATF2 ubiquitination. We show that leucine zipper-based heterodimerization with c-Jun as one of the key ATF2 heterodimers is required for ATF2 ubiquitination and degradation. The possible implications of ubiquitin-dependent regulation of ATF2 stability are discussed.

MATERIALS AND METHODS

Reagents.

Okadaic acid and retinoic acid were purchased from Sigma Chemical Co. Anti-ATF2 monoclonal antibody, anti-Jun polyclonal antibody, anti-CREB polyclonal antibody, and anti–c-Fos antibody were purchased from Santa Cruz Biotechnology, and anti-hemagglutinin (HA) monoclonal antibody was purchased from BAbCo. Anti-ATF2 polyclonal antibody was a generous gift from N. Jones (Imperial Cancer Research Fund, London, England). Anti-JNK antibody was kindly provided by C. Monell (PharMingen). Proteasome inhibitor MG132 was purchased from Peptide International Co.

Expression plasmids.

Bacterial expression constructs pET-15b-ATF2 and pET-15b-Ub-HA were previously described (7). Mammalian expression constructs pRSV-c-Jun, pRSV-JunB, and pRSV-JunD (4) were kindly provided by M. Karin. pCMV-c-Jun, pCMV-c-Jun-HA, pCMV-Ub-HA, and pCMV-c-JunΔ31–57 (34) were generous gifts from D. Bohmann. pCMV-c-Jun LZM was obtained from M. Birer, and 5xjun2-luc (38) was a gift from H. van Dam. The bacterial expression construct encoding bZIP of ATF2 was kindly provided by M. Green.

N-terminal fusion of six histidines with ATF2 (pCMV-hisATF2) and C-terminal fusion of the HA epitope with ATF2 (pCMV-ATF2-HA) were accomplished by amplifying the entire human ATF2 sequence with plasmid pECE-ATF2 (a gift from M. Green) as a template, with primers having HHHHHH and ASYPYDVPDYASLS sequences, respectively, and with designed restriction sites. PCR products were digested with BamHI and EcoRV and cloned in pcDNA3 (Invitrogen). Point mutations and deletions were generated by oligonucleotide-directed mutagenesis with the aid of a QuickChange kit (Stratagene). The integrity of each construct was confirmed by partial DNA sequencing, in vitro translation, and immunoblotting.

Cell cultures and transfections.

NIH 3T3 mouse fibroblasts and 293T human embryo kidney cells were grown in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% calf serum and antibiotics at 37°C in 5% CO₂. F9 teratocarcinoma cells were maintained in DMEM-F12 (1:1) medium supplemented with fetal calf serum, 10 μM β-mercaptoethanol, and antibiotics. 293T cells were transfected by the CaPO₄ method. NIH 3T3 cells and F9 cells were transfected with DOTAP (Boehringer Mannheim Biochemicals) in accordance with the manufacturer’s recommendations. The total amount of DNA within the experiments was kept constant by adding the respective empty vector plasmid DNA to the transfection mixtures.

Expression, purification, and identification of proteins.

Histidine fusion proteins were expressed and purified as previously described (7) with the aid of nitrilotriacetic acid (NTA) resins (Qiagen). c-Jun was eluted under denaturing conditions and refolded by sequential dialysis. JNK2 purification, immunodepletion, and immunoblotting were performed as described elsewhere (7).

Ubiquitination assays.

The in vitro ubiquitination assay is described in detail elsewhere (7). Briefly, 50 μg of whole-cell lysates or 0.5 μg of purified JNK was incubated on ice with bacterially expressed ATF2 proteins (2 μg) bound to nickel beads for 45 min. After extensive washes, the substrate-bound beads were ubiquitinated with rabbit reticulocyte lysate depleted of JNK for 5 min at 30°C. The reaction was stopped by the addition of 0.5 ml of 8 M urea in sodium phosphate buffer (pH 6.3) with 0.1% Nonidet P-40. The beads were washed three times with stop buffer (7) and once with phosphate-buffered saline supplemented with 0.5% Triton X-100, and the protein moiety was eluted with Laemmli sample buffer at 100°C. Samples were resolved by sodium dodecyl sulfate (SDS)–8% polyacrylamide gel electrophoresis (PAGE) and analyzed by immunoblotting with anti-HA antibody and chemiluminescence detection. The blots were stripped and reprobed with antibody against ATF2, followed by alkaline phosphatase detection to ensure equal loading of the substrate.

In vivo ubiquitination was assayed as described by Treier et al. (34). His-ATF2 (4 μg) was cotransfected into mouse fibroblasts with a ubiquitin-HA vector (3 μg). Twenty-four hours later, cells were lysed with 6 M guanidinium HCl, and the His-tagged protein was purified with nickel resins as described by Treier et al. (34), separated by SDS–8% PAGE, and transferred to a Hybond C nitrocellulose filter (Amersham). The filter was cut just above the 71-kDa protein molecular mass marker, and the lower part was analyzed by means of immunoblotting with anti-ATF2 antibody to identify nickel-purified His-ATF2. The upper part of the filter was probed with anti-HA antibody, allowing detection of the smear which represented slower-migrating ubiquitin-HA conjugates.

Electrophoretic mobility shift assay and calpain digestion.

ATF2 proteins were translated in vitro with a T7-wheat germ extract-based transcription-translation kit (Promega) in accordance with the manufacturer’s recommendations. For gel shift assays, equal amounts of ATF2 proteins (equilibrated to ∼5 to 10 ng based on immunoblotting analysis) were incubated with bacterially expressed c-Jun (45 ng) for 1 h on ice. The proteins were reacted with the ³²P-labeled heteroduplex UV response element (URE) target sequence (ACTATGACAACAGCTTGACAACAGT; the actual URE sequence is underlined) in the presence of 10 mM HEPES (pH 7.6)–50 mM KCl–0.1 mM EGTA–0.1 mM dithiothreitol–4 mM MgCl₂–10% glycerol–50 ng of dI · dC for 30 min on ice prior to separation on a 7.5% polyacrylamide gel and autoradiography.

To analyze patterns of ATF2 digestion by calpain, ATF2 proteins were labeled with ³⁵S-methionine during in vitro translation, preincubated with bacterially expressed c-Jun (45 ng), and subjected to digestion with 0.05 U of calpain (mCANP; Sigma) in the presence of 10 mM HEPES (pH 7.6)–1 mM CaCl₂–1% Triton X-100 at 37°C for various times. The reaction was stopped by boiling in Laemmli sample buffer, and the cleavage products were separated by SDS–12.5% PAGE and analyzed by autoradiography.

In vivo degradation assay.

293T cells were transfected with the pCMV-ATF2-HA (5 μg), pCMV-ATF2^Δ150–248-HA (5 μg), and pRSV-c-Jun (2 μg) constructs. Twenty-four hours later, cells were incubated with methionine- and cysteine-free DMEM supplemented with 10% dialyzed calf serum for 1 h and then were metabolically labeled with 0.5 mCi of [³⁵S]methionine–[³⁵S]cysteine mix (PRO-MIX; Amersham) per ml. The label was chased with DEM plus 10% calf serum supplemented with 2 mM cold methionine and cysteine for various times. The cells were lysed as previously described (8), and equal amounts of trichloroacetic acid-insoluble material were analyzed by immunoprecipitation with anti-HA antibody. Immunopurified proteins were resolved by SDS-PAGE. The gels were fixed, impregnated with Amplify reagent (Amersham), and subjected to autoradiography. Quantification was performed with a GS363 phosphorimager (Bio-Rad).

Transcriptional analysis.

Luciferase assays were performed with a kit from Promega and whole-cell extracts (WCE) prepared from cells transfected with the 5×jun2-driven luciferase gene.

RESULTS

JNK is necessary but not sufficient for targeting of ATF2 ubiquitination in vitro.

We have been studying the role of JNK in targeting the ubiquitination of stress-responsive transcription factors by using an in vitro ubiquitination assay (6, 7). In this assay, resin-bound substrates are first incubated with WCE. Unbound proteins are washed away, and the targeting activity of the substrate-bound proteins is monitored on the basis of the degree of substrate ubiquitination in the presence of a rabbit reticulocyte lysate depleted of JNK (7). We previously demonstrated that the inactive form of JNK targets its associated proteins c-Jun, JunB, ATF2, and p53 for ubiquitination (7, 8). In contrast to the situation for c-Jun, the binding of JNK is necessary but not sufficient for the targeting of ATF2 ubiquitination in vitro (Fig. 1A, compare lane 2 with lane 1). This observation suggested that other factors present in WCE are required for targeting. ATF2 interacts with a number of proteins via bZIP (14). JNK is known to associate with the amino-terminal region of ATF2 (11, 22). While deletion of the amino-terminal region impairs ATF2 ubiquitination in vitro (7), additional targeting factors may utilize other ATF2 domains, including bZIP.

FIG. 1.

JNK is necessary but not sufficient for the targeting of ATF2 ubiquitination in vitro. Purified JNK2 or NIH 3T3 WCE immunodepleted with NRS or anti-JNK antibody were used to target the in vitro ubiquitination of ATF2. Negative control ubiquitination of NTA resins without a substrate is shown in the rightmost lane. In vitro ubiquitination of ATF2 was analyzed by immunoblotting. The upper part of the blot was probed with anti-HA antibody to detect ubiquitin (Ub)-HA conjugates (upper panel). The lower part of the blot was probed with anti-ATF2 antibody (lower panel).

The leucine zipper is required for ubiquitination of ATF2 in vitro.

To test the possible role of bZIP in the targeting of ATF2 ubiquitination, we depleted ATF2 bZIP binding proteins by passing WCE through an NTA column carrying bacterially expressed bZIP polypeptide derived from ATF2. Flowthrough fractions were incubated with full-length ATF2, and its ubiquitination was assessed. The ability of WCE depleted of bZIP binding proteins to target ATF2 ubiquitination was impaired compared with that of mock-depleted proteins (Fig. 2A). Immunoblotting analysis with antibodies against known ATF2 heterodimerization partners revealed that c-Jun, c-Fos, ATF2, and CREB are among the proteins bound to bZIP resins (data not shown).

FIG. 2.

Additional factors targeting ATF2 ubiquitination in vitro require the bZIP region of ATF2. (A) Targeting activity of WCE is depleted after preincubation with the ATF2 bZIP polypeptide. NIH 3T3 WCE were passed through columns packed with empty NTA beads (NTA FT) or with beads bound to bacterially expressed ATF2 bZIP polypeptide (bZIP FT) and then were used for the targeting of ATF2 ubiquitination in vitro. The upper part of the blot was probed with anti-HA antibody to detect ubiquitin (Ub)-HA conjugates (upper panel). The lower part of the blot was probed with anti-ATF2 antibody (lower panel). (B) Basal and WCE-targeted ubiquitination of a dimerization-deficient mutant of ATF2 is impaired. Beads carrying wild-type ATF2 or mutant ATF2^L408P were incubated with WCE, washed, and ubiquitinated in vitro. The upper part of the blot was probed with anti-HA antibody to detect ubiquitin (Ub)-HA conjugates (upper panel). The lower part of the blot was probed with anti-ATF2 antibody (lower panel).

To confirm the role of the leucine zipper in the targeting of ATF2 ubiquitination, we introduced a point mutation encoding the L408P substitution, which was previously shown to abrogate leucine zipper-mediated dimerization of ATF2 in vitro (1). Targeting of ubiquitination by WCE was substantially attenuated by this mutation (Fig. 2B). These results indicate the role of the leucine zipper domain in the targeting of ATF2 ubiquitination in vitro. These observations also suggest that WCE contain ATF2 dimerization partners which contribute to ATF2 ubiquitination.

c-Jun targets ATF2 ubiquitination in vitro.

To identify prospective ATF2 heterodimerization partners which may participate in the targeting of ubiquitination, we immunodepleted WCE of c-Jun, c-Fos, or CREB by using respective antibodies. These extracts were analyzed by immunoblotting to confirm the depletion of the respective proteins (Fig. 3B). The targeting activities of the resulting extracts were compared with that of WCE treated with naive rabbit serum (NRS). Depletion of c-Jun reduced the degree of ATF2 ubiquitination (Fig. 3A, compare lane 2 with lane 3). The addition of recombinant c-Jun to the depleted extract restored the degree of ubiquitination. Targeting of ATF2 ubiquitination was also attenuated by depletion of c-Fos (Fig. 3A, lane 6). Although the analysis of WCE depleted with anti-Fos antibody by immunoblotting with anti-Jun antibody revealed that up to 80% of c-Jun was removed from the extract (Fig. 3B), we cannot rule out the contribution of Fos by itself in the targeting of ATF2 ubiquitination. Nevertheless, the addition of c-Jun to a Fos-free extract efficiently reconstituted the targeting of ATF2 ubiquitination (Fig. 3A, compare lane 6 with lane 7). Conversely, depletion of CREB did not affect the targeting activity of WCE. This result indicates that WCE depleted of CREB still contains factors sufficient to target ATF2 ubiquitination. It has been previously demonstrated that heterodimerization of CREB with ATF2 in vitro does not disrupt the intramolecular interaction of the ATF2 leucine zipper and its amino terminus (1). It is therefore possible that heterodimerization-dependent changes in ATF2 conformation promote the susceptibility of ATF2 to ubiquitination in vitro.

FIG. 3.

c-Jun present in WCE targets ATF2 for ubiquitination in vitro. (A) WCE were immunodepleted with NRS or with the indicated antibody and analyzed for their ability to target ATF2 ubiquitination in vitro. Bacterially expressed c-Jun (0.5 μg) was added to the WCE to reconstitute the targeting activity. The upper part of the blot was probed with anti-HA antibody to detect ubiquitin (Ub)-HA conjugates (upper panel). The lower part of the blot was probed with anti-ATF2 antibody (lower panel). (B) WCE (100 μg) immunodepleted with the indicated antibodies (ID) were analyzed for the levels of ATF2 heterodimerization partners via immunoblotting with the respective antibodies (IB). Arrowheads indicate the positions of the respective proteins. (C) NIH 3T3 and F9 cells were transfected with a pRSV-c-Jun construct or empty vector. WCE prepared from transfected cells were used to target ATF2 ubiquitination in vitro.

To confirm that c-Jun is necessary for the targeting of ATF2 ubiquitination, we compared the targeting activity in lysates from NIH 3T3 mouse fibroblasts with those prepared from F9 teratocarcinoma cells, which do not express c-Jun under nondifferentiating conditions (41). WCE prepared from F9 cells exhibited a lower ability to target ATF2 ubiquitination than WCE prepared from NIH 3T3 cells (Fig. 3C, compare lane 4 with lane 2). Both extracts exhibited significantly increased targeting of ATF2 ubiquitination when prepared from cells that were transfected with c-Jun (Fig. 3C). Together, these data suggest that heterodimerization with c-Jun promotes ATF2 ubiquitination.

Overexpression of Jun proteins targets ATF2 ubiquitination in vivo.

To evaluate the role of ATF2 dimerization in ATF2 ubiquitination in vivo, we transfected cells with constructs expressing six-histidine-tagged ATF2 proteins together with HA-tagged ubiquitin (34). Using nickel beads, we purified ATF2 under denaturing conditions and assessed the amount of HA-tagged polyubiquitin chains covalently linked to ATF2 by immunoblotting. Although the ubiquitination of endogenous ATF2 has been recently documented (5), the ubiquitination of exogenous ATF2 in MeWo and WM35 human melanoma cells, HeLa cells, and BALB/c/3T3 cells (data not shown) and in NIH 3T3 mouse fibroblasts (Fig. 4A) could not be detected even in the presence of proteasome inhibitors. Nevertheless, cotransfection of c-Jun led to noticeable ubiquitination of exogenous ATF2 in vivo (Fig. 4A).

FIG. 4.

C-Jun overexpression alleviates ubiquitination of ATF2 in vivo. (A) NIH 3T3 cells were transfected as indicated and treated with MG132 for 8 h before being harvested. ATF2 proteins were purified with NTA beads and analyzed by immunoblotting with anti-HA (upper panel) and anti-ATF2 (lower panel) antibodies. The anti-HA blot was overexposed to detect low levels of ATF2 ubiquitination. Ub, ubiquitin. (B) 293T cells were transfected as indicated; 1 μg of pRSV-JunD and 0.25 to 1.0 μg of pRSV-c-jun were used. In vivo ubiquitination of purified ATF2 proteins was assessed as described above. Immunoblots probed with anti-HA (upper panel) and anti-ATF2 (middle panel) antibodies are shown. The levels of Jun proteins expressed in 293T cells were analyzed with 100 μg of WCE by immunoblotting (lower panel) with an anti-Jun polyclonal antibody that recognizes both c-Jun and JunD (sc-44; Santa Cruz Biotechnology). (C) 293T cells were transfected as indicated; 1 μg of pCMV-c-Jun, pCMV-c-JunΔ31–57, and pCMV-c-Jun LZM was used. In vivo ubiquitination of purified ATF2 proteins was assessed as described above. Immunoblots probed with anti-HA (upper panel) and anti-ATF2 (middle panel) antibodies are shown. The levels of Jun proteins expressed in 293T cells were analyzed with 100 μg of WCE by immunoblotting (lower panel) with a mixture of anti-Jun polyclonal antibodies (sc-44 and sc-45; Santa Cruz Biotechnology).

To overcome the possible problems of low expression and strong intramolecular interactions of ATF2, we used the in vivo ubiquitination assay with E1A-expressing 293T cells. We found that exogenously expressed ATF2 can be ubiquitinated in vivo in these cells. Cotransfection of c-Jun led to a dose-dependent increase in ATF2 ubiquitination (Fig. 4B). Since ATF2 was purified under stringent denaturing conditions (6 M guanidine hydrochloride), c-Jun could not be copurified with His-ATF2 and serve as a substrate in this assay. Interestingly, the coexpression of JunD also resulted in increased ATF2 ubiquitination, although to a lesser extent. Transfection of JunB did not affect ATF2 ubiquitination (data not shown), although the level of expression of this protein was negligible compared with those of c-Jun and JunD (Fig. 4B, bottom panel). These data suggest that ATF2 heterodimerization with c-Jun and JunD results in more efficient ATF2 ubiquitination. Along these lines, the overexpression of c-Jun led to a noticeable decrease in the level of six-histidine-tagged ATF2 (Fig. 4B, middle panel), suggesting that ATF2 ubiquitination targeted by heterodimerization with c-Jun results in accelerated degradation of ATF2.

To test whether the effect of c-Jun expression on ATF2 ubiquitination requires c-Jun–ATF2 heterodimerization, we used a c-Jun mutant lacking the leucine zipper (c-Jun LZM). Expression of this construct led to a considerable decrease in ATF2 ubiquitination (Fig. 4C). A c-Jun mutant lacking the δ domain (c-Jun Δ31–57) noticeably increased ATF2 ubiquitination, although less efficiently than wild-type c-Jun. As the different effects of Jun proteins on the ubiquitination of ATF2 cannot be attributed to variations in their expression levels (Fig. 4C, bottom panel), these data suggest that heterodimerization is required for c-Jun to promote ATF2 ubiquitination in vivo.

ATF2 mutants that exhibit various levels of dimerization and transactivation differ in their degrees of ubiquitination.

To confirm that the dimerization of ATF2 is required for the ubiquitination of ATF2, we designed mutant forms of ATF2 in which dimerization is affected. To create ATF2 with impaired dimerization ability, leucine at amino acid 408 was replaced with proline (ATF2^L408P), a substitution that was shown to abrogate the dimerization of ATF2 (1) and the targeting of ATF2 ubiquitination in vitro (Fig. 2B). In vivo interactions of this mutant with c-Jun were abrogated in 293T cells (Fig. 5A).

FIG. 5.

Characterization of dimerization and transactivation of mutant ATF2 proteins. (A) In vivo interaction of c-Jun with ATF2 proteins. 293T cells were transfected as indicated. The upper panel depicts the level of c-Jun–HA expression in 100 μg of WCE analyzed by anti-HA immunoblotting. ATF2 proteins were purified from 2 mg of WCE with NTA resins under native conditions and analyzed by immunoblotting with anti-HA (middle panel) and anti-ATF2 (lower panel) antibodies. (B) Transactivation of ATF2 proteins evaluated with the 5×jun2-driven luciferase reporter assay. HA-tagged ATF2 constructs (a, pCDNA3; b, ATF2^Δ150–248; c, ATF2^L408P; d, wild-type (ATF2) were coexpressed with the 5×jun2-luc plasmid, and luciferase activity was analyzed with a Promega kit. The left panel depicts ATF2 transactivation in NIH 3T3 cells. The inset shows relative levels of ATF2 proteins analyzed by HA immunoprecipitation followed by immunoblotting with anti-ATF2 antibody. ATF2 transactivation in 293T cells is shown in the right panel. The data represent fold activation over the values for cells transfected with the reporter only. The average of three independent experiments (each in duplicate) is shown; error bars indicate standard deviations.

The DNA binding activity of ATF2^L408P (translated in vitro in a wheat germ extract, measured with an electrophoretic mobility shift assay, was lower than that of its wild-type counterpart. The addition of bacterial c-Jun to this reaction significantly increased the DNA binding activity of wild-type ATF2 but not of ATF2^L408P (data not shown).

As ATF2 dimerization is a prerequisite for the activity of ATF2 as a transcription factor, we determined the transactivation mediated by the 5× TPA-responsive element derived from the jun2 promoter (38) by using a luciferase reporter assay. In both NIH 3T3 and 293T cells, transactivation by mutant ATF2^L408P was lower than that by wild-type ATF2 (Fig. 5B). Immunoblotting analysis demonstrated that the difference in transcriptional activity cannot be attributed to variations in the levels of protein expression (Fig. 5B, inset).

To enhance the ability of ATF2 to dimerize, we relied on a splicing variant of murine ATF2 with a 294-bp internal deletion which constitutively activates the δA enhancer-driven transcription of the CD3 delta gene (10). To this end, we deleted from the human ATF2 sequence 98 amino acids that correspond to the murine deletion ATF2^Δ150–248. The deleted region is relatively hydrophobic and is capable of forming the beta-sheet structure (10). Deletion of this region was shown to keep ATF2 free from the intramolecular inhibition of transcriptional activity in CCL64 cells (18). The level of ATF2^Δ150–248 protein expressed in both 293T (Fig. 5A) and NIH 3T3 (Fig. 5B) cells was lower than the level of wild-type protein. Nevertheless, in spite of the lower expression level, ATF2^Δ150–9248 was capable of association with c-Jun in vivo (Fig. 5A). This mutant protein translated in vitro exhibited enhanced DNA binding in the electrophoretic mobility shift assay; the addition of c-Jun did not noticeably affect this binding (data not shown). This result suggests that leucine zipper domains on ATF2^Δ150–248 are highly prone to forming homodimers. Indeed, cotransfection of ATF2^Δ150–248 mediated stronger transactivation of the jun2 element than did that of wild-type ATF2 in both NIH 3T3 and 293T cells (Fig. 5B).

The ubiquitination of ATF2^Δ150–248 and ATF2^L408P mutants in vivo was distinctly different from that of wild-type ATF2. To detect the ubiquitination of ATF2^Δ150–248, we treated cells with proteasome inhibitors. While ATF2^Δ150–248 exhibited an increase in the extent of basal ubiquitination, cotransfection of c-Jun did not significantly affect this level (Fig. 6). The extent of ATF2^Δ150–248 ubiquitination shown here is likely to be underestimated as a result of the lower level of expression of this mutant protein. These data suggest that homo- and heterodimers prone to ubiquitination already prevail in the pool of ATF2^Δ150–248 molecules without c-Jun overexpression.

FIG. 6.

In vivo ubiquitination of mutant ATF2 proteins. 293T cells were transfected as indicated (1 μg of pRSV-c-Jun was used) and treated with MG132 (40 μM) for 4 h before being harvested. The in vivo ubiquitination assay was performed as described above. Immunoblots probed with anti-HA (upper panel) and anti-ATF2 (lower panel) antibodies are shown. Ub, ubiquitin.

In contrast to that of ATF2^Δ150–248, in vivo ubiquitination of ATF2^L408P was markedly impaired. Overexpression of c-Jun did not increase the level of ubiquitination of this dimerization-deficient mutant (Fig. 6). These findings further support our in vitro data indicating that the dimerization of ATF2 with Jun is indispensable for the ubiquitination of ATF2.

Dimerization modulates the conformation of ATF2.

The ATF2 mutant with an enhanced ability to dimerize (ATF2^Δ150–248) was subjected to more efficient ubiquitination than the wild-type protein (Fig. 6). This result may have been due to differences in conformation between ATF2 dimers and ATF2 monomers. To test this hypothesis, we analyzed the susceptibility of different in vitro-translated ATF2 forms to digestion by calcium-dependent calpain protease in vitro. As evident in Fig. 7, the appearance of lower-molecular-weight cleavage products was facilitated by preincubation of wild-type ATF2 with bacterially expressed c-Jun. Cleavage of ATF2^Δ150–248 was very efficient even without c-Jun. Since an ATF2 mutant with impaired dimerization ability exhibited virtually no digestion by calpain (Fig. 7), we conclude that the observed partial cleavage of wild-type ATF2 occurs with dimerized forms of the protein. These results imply that the dimeric conformation of ATF2^Δ150–248 may render this protein susceptible to ubiquitination and degradation independently of its interaction with c-Jun.

FIG. 7.

ATF2 dimerization affects ATF2 conformation. In vitro-translated ³⁵S-methionine-labeled ATF2 proteins were preincubated with bacterially expressed c-Jun as indicated and subjected to digestion with calpain protease for 15 or 30 min at 37°C. The results of digestion were analyzed by electrophoresis and autoradiography. Arrows indicate cleavage products. The control reaction performed in the presence of 2 mM EDTA is marked by an asterisk.

Dimerization-dependent ATF2 ubiquitination leads to ATF2 degradation in vivo.

Analysis of our in vivo ubiquitination assays repeatedly revealed an inverse correlation between the level of substrate expressed and the intensity of the ubiquitin-HA-reactive smear. For example, cotransfection of c-Jun coincided with a decrease in the ATF2 level and an increase in the amount of copurified ubiquitin chains (Fig. 4B). Deletion of residues 150 to 248 decreased the level of ATF2 mutant proteins and enhanced susceptibility to ubiquitination (Fig. 6). To confirm that the decrease in the ATF2 level is due to decreased stability, we used pulse-chase metabolic labeling. While the half-life of wild-type ATF2 expressed in 293T cells was estimated to be more than 2 h, elevated c-Jun expression shortened the half-life to less than 1 h (Fig. 8). Mutant ATF2^Δ150–248 exhibited a shorter half-life (∼40 min), reflecting its lower stability compared with that of its wild-type counterpart. The overexpression of c-Jun did not significantly affect the stability of mutant ATF2^Δ150–248 (Fig. 8), which exists in the form of a dimer even in the absence of c-Jun (Fig. 5B and 7). The dimerization-deficient mutant ATF2^L408P was found to be a substantially more stable protein, and the coexpression of c-Jun did not lead to a significant acceleration of ATF2^L408P degradation (Fig. 8). These results are in agreement with our in vivo ubiquitination data (Fig. 6). Together with the latter data, these findings suggest that dimerization-dependent ubiquitination marks ATF2 for efficient degradation.

FIG. 8.

c-Jun affects in vivo stability of ATF2 proteins. (A) HA-tagged ATF2 constructs were expressed with or without pRSV-c-Jun in 293T cells. Cells metabolically labeled with ³⁵S-methionine–³⁵S-cysteine were chased for the times indicated, followed by immunopurification of ATF2 proteins under stringent conditions (0.5 M LiCl), separation by SDS-PAGE, and autoradiography. wt, wild type. (B) Quantitative analysis of the representative experiment shown in panel A.

Degradation of endogenous ATF2.

To confirm that the rapid degradation of ATF2 forms which exhibit enhanced dimerization and transcriptional activity (10) also occurs for endogenous proteins in vivo, we monitored the accumulation of endogenous ATF2 in NIH 3T3 cells treated with the proteasome inhibitor MG132. The level of full-length mouse ATF2 (∼68 kDa) remained unchanged for up to 6 h after the addition of MG132 to the medium (Fig. 9A). Conversely, proteasome inhibitor treatment led to a noticeable increase in the level of a protein with an apparent molecular mass of ∼42 kDa (Fig. 9A); this protein corresponds to the in vitro-translated product of the transcriptionally active splicing form of ATF2 (10). These data suggest that the endogenous analogue of mutant ATF2^Δ150–248 exhibits less stability than the full-length splicing counterpart and further support the notion that the ability of ATF2 to form dimers is correlated with the rate of ATF2 degradation in vivo.

FIG. 9.

Stability of endogenous ATF2. (A) Accumulation of endogenous ATF2 in NIH 3T3 cells after treatment with a proteasome inhibitor. The level of ATF2 proteins in WCE (100 μg) at the indicated times after treatment with MG132 (40 μM) was assessed by immunoblotting with an antibody against ATF2. (B) Degradation of endogenous ATF2 in F9 cells treated with retinoic acid (RA, 2 × 10⁻⁷ M) for 20 to 40 h and MG132 (40 μM) for 4 h before harvest as indicated. The levels of ATF2 (upper panel) and c-Jun (lower panel) were analyzed with nuclear extracts (50 μg). The double arrow marks the position of a putative ATFa protein.

In this paper, we demonstrate that the expression of c-Jun promotes the ubiquitination and degradation of coexpressed ATF2. In order to test whether the expression of endogenous c-Jun is indeed required for the degradation of endogenous ATF2, we used an F9 teratocarcinoma cell model. These cells begin to express detectable levels of c-Jun after the induction of differentiation by retinoic acid treatment (Fig. 9B) (41). Conversely, the levels of ATF2 in nuclear extracts from F9 cells substantially decreased within 20 to 40 h after the addition of retinoic acid). Interestingly, in addition to the 68-kDa ATF2 protein, we detected an ∼59-kDa protein (Fig. 9B) whose levels underwent similar changes. The characteristics of ATF2-homologous protein ATFa are consistent with this molecular mass and the conserved N-terminal epitope recognized by the ATF2 antibody used in this analysis. Treatment of differentiating F9 cells with the proteasome inhibitor MG132 completely restored the ATF2 levels (Fig. 9B). These data suggest that the upregulation of c-Jun expression results in ubiquitin-proteasome-dependent degradation of endogenous ATF2 in nontransfected cells.

DISCUSSION

One of the key issues for understanding the cellular regulation of gene expression has to do with how cells restrict the duration and magnitude of transcription factor activities. ATF2 as well as the other members of the bZIP family require leucine zipper-dependent dimerization for transactivation. Using in vitro and in vivo ubiquitination and degradation assays, we have demonstrated that such heterodimerization with c-Jun contributes to the efficient ubiquitination of ATF2, which in turn results in the rapid degradation of ATF2. These data suggest that ubiquitination-dependent elimination of transcriptionally active ATF2 species is a putative mechanism by which ATF2 activity in cells may be regulated.

That ATF2 is transcriptionally inactive as a result of intramolecular inhibition (18) has been documented. Biochemical evidence from in vitro experiments showed that the DNA binding domain of ATF2 is capable of intramolecular interaction with its amino terminus (1). This intramolecular inhibition is assumed to be disrupted and transcriptional activities are assumed to be restored when ATF2 interacts with other proteins, such as E1A (20, 21) and c-Jun (2). Phosphorylation of ATF2 by stress-activated protein kinases was also suggested to relieve intramolecular inhibition and induce leucine zipper-dependent homodimerization (1, 18). We previously showed that the binding of inactive JNK to the amino terminus of ATF2 targets the ubiquitination of ATF2 in vitro (7). Deletion of residues 40 to 66 within the JNK binding site abrogated ATF2 ubiquitination in vitro.

We propose that intramolecular interactions may hinder the association of ATF2 with JNK or/and other polypeptides which bind the amino terminus of ATF2 and target its ubiquitination (Fig. 10). In this model, the events which disrupt intramolecular inhibition (such as ATF2 association with E1A or c-Jun) and lead to increased ATF2 dimerization would result in conformational changes of the ATF2 molecule favoring its association with targeting proteins and subsequent ubiquitination. Our data partially support this hypothesis, since: (i) background in vivo ubiquitination was primarily observed in 293T cells which express E1A and was not seen in three or four other experimental cell lines which do not express E1A; (ii) overexpression of c-Jun increases ATF2 ubiquitination and degradation; (iii) dimerization of ATF2 leads to changes in its conformation, as assessed by calpain cleavage; (iv) ubiquitination of the dimerization-deficient ATF2^L408P mutant is impaired even in the presence of E1A and c-Jun; (v) the ATF2^Δ150–248 mutant (which is constitutively active as a result of a lower degree of intramolecular inhibition) exhibits a distinctly different conformation, a higher level of basal ubiquitination, and a significantly shorter half-life; (vi) an increase in the level of endogenous ATF2 proteins in response to the proteasome inhibitor was primarily observed for the natural analogue ATF2^Δ150–248; and (vii) induction of c-jun expression in F9 teratocarcinoma cells coincides with the degradation of endogenous ATF2. One cannot exclude the possibility that additional regulatory mechanisms (i.e., ATF2 phosphorylation) also control the relationship between ATF2 dimerization and transactivation and ATF2 ubiquitination and degradation. We also found that ATF2 dimers are more efficiently digested in vitro by calcium-dependent calpain protease compared with the monomeric form of ATF2 (Fig. 7). Therefore, we cannot rule out the possibility that in addition to the ubiquitin-proteasome pathway, the calpain pathway may participate in the elimination of active ATF2 species in vivo.

FIG. 10.

Proposed model for the regulation of ATF2 ubiquitination. Conversion of ATF2 monomers into ATF2 homo- or heterodimers, while enabling the transactivation of ATF2, leads to conformational changes which favor the binding of E3 ubiquitin ligase and the ubiquitination of ATF2. The association of ATF2 with proteins which do not disrupt intramolecular inhibition (i.e., CREB [1]) does not increase the degree of ATF2 ubiquitination. While ATF2 homodimerization or heterodimerization with JunD or c-JunΔ31–57 suffices for ATF2 presentation to ubiquitin (Ub) ligase, a further increase in ATF2 ubiquitination is provided in trans via factors brought to the complex through docking sites, such as the δ domain on c-Jun, which further promotes targeting through JNK association.

An ATF2 molecule may form a homodimer with another ATF2 molecule, thereby exposing both to targeted ubiquitination. Similarly, members of the bZIP family which are capable of heterodimerization with ATF2 may contribute to the targeting of ATF2 ubiquitination as well (Fig. 10). Consistent with this idea, the effect of different ATF2 partners on the susceptibility of ATF2 to ubiquitination may vary depending on the specific conformation of dimerized ATF2. For instance, depletion of CREB did not affect WCE targeting activity (Fig. 3A). As CREB association with ATF2 does not disrupt ATF2 intramolecular inhibition (1), this result suggests that a dimerization-dependent conformational change is important for ubiquitination. Conversely, the heterodimer with c-Jun is susceptible to ubiquitination in vitro and in vivo (Fig. 3A and 4B). c-Jun LZM lacking the leucine zipper does not promote ATF2 ubiquitination. Moreover, expression of this mutant decreases ATF2 ubiquitination (Fig. 4C), probably due to titration of targeting molecules (i.e., JNK). A similar sequestering effect of c-Jun has been shown for the alteration of p53 degradation (8). c-Jun increases ATF2 ubiquitination more efficiently than JunD, which does not bind JNK (12, 15), or c-JunΔ31–57, which lacks the JNK binding domain (Fig. 4). These results imply that the presence of a JNK docking site may elicit trans-ubiquitination by facilitating the presentation of targeting molecules for the ubiquitination of heterodimerized ATF2 (Fig. 10).

Certain evidence indicates that ATF2 homodimers can be bound to DNA target sequences before transactivation (37). Our studies did not address the possible role of ATF2 dimers that bind to specific target motifs in the regulation of ubiquitination and degradation. Nevertheless, the addition of oligonucleotides bearing the jun2 target sequence to an in vitro reaction did affect the degree of ATF2 ubiquitination (data not shown). It has also been suggested that heterodimers of ATF2 with newly synthesized c-Jun replace less transcriptionally potent ATF2 homodimers on the jun2 promoter, thus forming a positive-feedback loop (36). Our data showing that the expression of exogenous c-Jun in NIH 3T3 and 293T cells or the upregulation of endogenous c-Jun in F9 cells potentiates ATF2 ubiquitination and degradation provide a clue for understanding how the very same heterodimerization could eventually restrict the duration of transcriptional activity.

It appears from the data presented here and previously published findings that ATF2 transcriptional activity is very tightly regulated. Stable ATF2 protein species are transcriptionally inactive (18), and active ATF2 dimers are unstable (this study). These characteristics of ATF2 are expected to play important roles in limiting the response of cells to viral aggression, stress stimuli, or inflammatory cytokines, as well as in regulating the antigen receptor-mediated stimulation of T and B lymphocytes (17, 35). We recently obtained evidence that ATF2 plays an important role in the radiation resistance of human melanoma cells (31) as well as in the UV-induced apoptosis of human melanoma cells (14a). Indeed, while the introduction of ATF2 in 293T cells resulted in an increased frequency of apoptotic cells, transcriptionally active ATF2^Δ150–248 was twice as active in the induction of apoptosis as the wild-type protein (data not shown).

Dimerization-dependent ubiquitination and degradation may constitute a general mechanism for limiting the transactivation of other bZIP family members. First, structural similarities between ATF2 and ATFa transcription factors make the latter a candidate for such regulation. ATFa is capable of dimerization and has been shown to bind JNK (3). We cannot exclude the possibility that ATFa also is degraded upon upregulation of c-Jun expression by retinoic acid in F9 cells (Fig. 9B). Second, some bZIP family members which cannot form homodimers are expected to be regulated through heterodimerization. Indeed, evidence indicates that the ubiquitination and degradation of c-Fos are dependent on its heterodimerization with c-Jun (27, 33). Heterodimerization would apparently be important for regulation of the ubiquitination and degradation of proteins which may not directly or efficiently associate with the ubiquitination-targeting proteins. For instance, JunD cannot associate directly with JNK (12, 15). Nevertheless, JunD, as part of the JunD–c-Jun heterodimer, can be still phosphorylated by activated JNK which is presented to the phosphoacceptor site of JunD by the JNK docking site on heterodimerized c-Jun (15). Therefore, it can be also assumed that in nonstressed cells, inactive JNK presented by c-Jun may target the trans-ubiquitination and degradation of JunD. Interestingly, an inverse correlation between the levels of expression of c-Jun and JunD proteins in mouse fibroblasts has been demonstrated (28). In addition, the dimerized conformation may favor the ubiquitination and degradation of bZIP transcription factors which can directly bind targeting molecules (ATF2 and c-Jun). Since the binding of JNK to c-Jun is a prerequisite for c-Jun phosphorylation and efficient c-Jun phosphorylation by JNK in vivo was shown to require dimerization (15), the targeting of c-Jun for ubiquitination by JNK (6) may also require c-Jun dimerization.

Negative regulation of signal transduction pathways via the ubiquitin-proteasome system has been documented so far for the JAK-STAT pathway, protein kinase C, and c-Kit (9). The preferential ubiquitination and degradation of transcriptionally active species of ATF2 and, perhaps, of some other bZIP transcription factors provide the underlying mechanism for regulating the duration and magnitude of transcriptional output.