14-3-3 Proteins Associate with A20 in an Isoform-specific Manner and Function Both as Chaperone and Adapter Molecules

A20, a novel zinc finger protein, is an inhibitor of tumor necrosis factor-induced apoptosis. The mechanism by which A20 exerts its protective effect is currently unknown. Several isoforms of the 14-3-3 proteins were found to interact with A20 in a yeast two-hybrid screen. A20 bound several 14-3-3 isoforms in vitro. Moreover, transfected A20 was found to preferentially bind the endogenous η14-3-3 isoform, whereas the β/ζ isoforms co-immunoprecipitated much less efficiently, and ϵ14-3-3 had an intermediate affinity. Importantly, c-Raf, a previously described 14-3-3-interacting protein, also preferentially bound the η isoform. The cellular localization and subcellular fractionation of A20 was dramatically altered by co-transfected 14-3-3, providing the first experimental evidence for the notion that 14-3-3 can function as a chaperone. Furthermore, c-Raf and A20 co-immunoprecipitated in a 14-3-3-dependent manner, suggesting that 14-3-3 can function as a bridging or adapter molecule.

Tumor necrosis factor-␣ (TNF) 1 is a catabolic pro-inflammatory cytokine that is capable of inducing apoptotic death in a number of tumor cell lines. The systemic toxicity of TNF and the generation of TNF resistant clones have in large part contributed to the therapeutic failure of TNF as an anti-cancer agent. To understand the molecular mechanism of TNF and the basis of sensitivity and resistance to TNF killing, we cloned TNF-inducible primary response genes by differential hybridization (1). One of the genes, designated A20, conferred resistance to TNF killing when transfected into sensitive NIH3T3, WEHI 164, and MCF7 cells (2). Further, breast carcinoma cell lines that were resistant to TNF cytotoxicity expressed higher levels of A20 than corresponding sensitive lines (2). Interestingly, the expression of A20 is subverted by a number of gene products including the LMP-1 gene product of the Epstein-Barr virus, a known inhibitor of apoptosis (3). It is plausible that virally mediated induction of A20 and subsequent resistance to apoptosis and cytokine killing contributes to host immune response attenuation and viral persistence.
Analysis of the full-length A20 cDNA revealed an open reading frame coding for a protein of 790 amino acid residues, which showed no significant homology to previously described proteins (4). However, within A20 a distinct repeated element was found that included seven novel zinc finger motifs of the form CX 4 CX 11 CX 2 C. The repression of its own promoter activity, coupled with the presence of zinc fingers, initially suggested that A20 might be a transcriptional factor (5). However, immunolocalization studies revealed A20 to be a cytoplasmic protein, 2 consistent with the finding that other zinc finger proteins, including protein kinase C and c-Raf, are non-nuclear.
To derive a biochemical understanding of how A20 functions, a yeast two-hybrid screen was utilized to identify A20-interacting proteins. The 14-3-3 family of proteins was found to specifically bind A20 and could function as adapter molecules mediating the interaction of A20 with c-Raf, a Ser/Thr kinase of pivotal importance to a number of signaling pathways.
14-3-3 proteins are highly conserved and are ubiquitous in the animal and plant kingdoms. At least seven mammalian isoforms have been identified. Previously, 14-3-3 proteins had been shown to function as regulators of protein kinase C, tryptophan, and tyrosine hydroxylases and to be essential for the stimulation of exocytosis from chromaffin cells (see Ref. 6 and references therein). However, interest in 14-3-3 has recently been stimulated by the discovery that oncogene products, including Raf-1(7-13), Bcr-Abl, Bcr (14,15), polyoma middle T antigen (16), and cell cycle control proteins such as Cdc25 phosphatases (17) associate with 14-3-3. Moreover, the Schizosaccharomyces pombe 14-3-3 homologues, rad24 and rad25, are required for DNA damage checkpoint control and, therefore, the timing of mitosis (18). The importance of 14-3-3 in signal transduction is further corroborated by reports that 14-3-3 interacts with phosphatidylinositol 3 kinase (19) and glycoprotein Ib-IX (20).
The function of 14-3-3 in these cases has not been elucidated in detail. The interaction of Raf with 14-3-3 leads to Raf activation in several in vivo systems (7,9,10), while no effect is observed in vitro (13). Overexpression of 14-3-3 in Jurkat cells leads to decreased phosphatidylinositol 3-kinase stimulation by anti-CD3 (19). One potential mechanism by which 14-3-3 could exert a variety of functions is by its ability to form dimers. It could act as an adapter between two molecules and thereby modulate their activity (21).
In this report we demonstrate that 14-3-3 can mediate the interaction of A20 and Raf by functioning as a bridging adapter molecule. Additionally, 14-3-3 proteins can serve as chaperones, promoting the transition of A20 from insoluble punctate cytoplasmic structures to the soluble cytoplasmic compartment.

MATERIALS AND METHODS
Yeast Two-hybrid Screen-A GAL4 DNA binding domain-A20 fusion construct in the pAS1CYH2 vector was used as bait (22) to screen a human B-cell cDNA library fused to the GAL4 activation domain in the pACT prey plasmid as described previously (23). Transformants were selected on ϪTrp, ϪLeu, ϪHis ϩ 30 mM Triazole and potential positives identified by assaying for ␤-galactosidase activity using 5-bromo-4chloro-3-indolyl-␤-D-galactoside as a chromogenic substrate. Positive clones were cured of the bait plasmid by growth in ϪLeu medium and the library plasmid rescued by transformation of XL1Ϫ Blue competent Escherichia coli cells. Recovered plasmids were initially characterized by partial DNA sequencing. To demonstrate specificity, nonidentical clones were co-transformed into yeast with full-length A20 bait, A20 NH 2 -terminal domain (amino acid residues 45-366), A20 COOH-terminal zinc finger domain (379 -749), or a battery of heterologous proteins expressed as GAL4-DNA binding domain fusions in the pAS1CYH2 vector.
In Vitro Binding-A20 containing an NH 2 -terminal FLAG epitope tag was expressed by transient transfection in 293T cells. Following lysis in RIPA buffer (150 mM NaCl, 10 mM Tris, pH 7.4, 10% glycerol, 5 mM EDTA, 1% Triton X-100, 1% deoxycholate, and protease inhibitors; 5 g/ml leupeptin, 5 g/ml aprotinin, 5 g/ml pepstatin, 50 g/ml soybean trypsin inhibitor, and 0.5 mM phenylmethylsulfonyl fluoride), the lysates were incubated overnight at 4°C with anti-FLAG antibodycoupled agarose (Eastman Kodak Co.) and the beads subsequently washed in 50 mM NaCl, 10 mM Tris, pH 8.0, 1% Triton X-100. As a negative control, an unrelated protein, B94 (26) similarly FLAG epitope-tagged, was also expressed and bound to anti-FLAG beads. Library-encoded cDNAs to be tested for A20 binding were amplified by polymerase chain reaction and then subjected to coupled in vitro transcription/translation in the presence of [ 35 S]methionine according to the manufacturer's instructions (Promega, Madison, WI). Five l of the radiolabeled translation product was mixed with anti-FLAG beads containing either FLAG-A20 or FLAG-B94 and 30 l of a 50% slurry of bovine serum albumin agarose beads (Sigma) as carrier. Binding reactions were incubated for 4 h at 4°C in 200 l of binding buffer (50 mM NaCl, 10 mM Tris pH, 7.6, 1% Triton X-100, 100 M CaCl 2 , 13.5 mM MgCl 2 , 1 mM dithiothreitol, and protease inhibitors), followed by three 500-l washes with binding buffer, boiling in sample buffer, and analysis of the eluted proteins by SDS-PAGE and fluorography.
Purified tagged 14-3-3 isoforms were obtained by immunoprecipitation of the in vitro transcription/translation mixture with tag-specific antibodies. Phosphate-buffered saline washes were followed by elution in 0.1 M glycine, pH 3.0. The pH of the eluted fractions was adjusted to 8.0 by 1 M Tris base.
In Vivo Binding-FLAG-A20 was expressed in 293T cells by transient transfection. Forty-eight hours following transfection, cells were washed in phosphate-buffered saline and lysed in 3 ml of lysis buffer (10 mM Tris pH 7.6, 50 mM NaCl, 5 mM EDTA, 1% Triton X-100, 25 mM NaF, 1 mM Na 3 VO 4 , 1 mM dithiothreitol, protease inhibitors). Lysates were cleared by microcentrifugation, diluted 3-fold in lysis buffer, anti-FLAG beads added (25 l of a 50% slurry), and incubated at 4°C for the indicated times. Beads were washed three times in lysis buffer, boiled in sample buffer, and the eluted proteins resolved by SDS-PAGE, transferred to Immobilon P membrane (Millipore, Bedford, MA), and probed with isoform-specific 14-3-3 antibodies. Raf/14-3-3 co-precipitations were similarly carried out using 293T cells transfected with a c-Raf expression construct and immunoprecipitated with anti c-Raf antibodies (Transduction Laboratories, Lexington, KY).
For Flag-A20/Raf co-precipitations, cells were lysed in a low stringency Nonidet P-40 lysis buffer (20 mM Tris, pH 8.0, 137 mM NaCl, 10% glycerol, 2 mM EDTA, 5 mM Na 3 VO 4 , 1% Nonidet P-40, protease inhibitors) and the lysates cleared by centrifugation. For high stringency immunoprecipitations, deoxycholate and SDS were added to a final concentration of 0.5% and 0.1%, respectively. The immunoprecipitates were washed four times with the respective immunoprecipitation buffers prior to analysis by SDS-PAGE.
Immunostaining-Immunostaining was performed on transiently transfected 293 cells grown on gelatin-coated glass coverslips as described (28).

RESULTS AND DISCUSSION
Two-hybrid Screen-A total of 1.3 ϫ 10 6 primary transformants were screened following transformation of a Y190 yeast strain expressing A20 bait with a B-cell library. Sixty-two colonies were ␤-galactosidase-positive. Forty-three of the positive clones were sequenced and 18 clones identified as fusions between 14-3-3 proteins and the library vector encoded Gal4 activation domain (Table I). Specificity of the interaction was corroborated by co-transformation with heterologous baits.
Intriguingly, all 14-3-3 sequences contained an untranslated segment fused in-frame to the Gal4 activation domain. To assess the significance of the 5Ј-untranslated sequence, ␤14-3-3 constructs with only the coding region fused to the Gal4 DNA binding domain were tested for two-hybrid interaction with A20 (Table II). The ␤14-3-3 without the untranslated segment interacted with A20 in the prey vector; therefore, the noncoding region was not necessary for binding A20. Similar untranslated 14-3-3 sequences were also obtained during a twohybrid screen with Raf as bait (9). Moreover, both the aminoand carboxyl-terminal halves of A20 were found to interact with 14-3-3, indicating that binding was not restricted to a distinctly defined single domain. Full-length A20 also interacted with the amino-and carboxyl-terminal halves of A20 (Table II) signifying that A20 also bound itself through multiple domains.
In Vitro Interactions-To confirm the interaction between 14-3-3 and A20, in vitro binding experiments were performed using recombinant A20 expressed in 293T cells and in vitro transcription/translated 14-3-3 (Fig. 1). All isoforms specifically bound A20 but not an unrelated protein B94. Quantitative comparison of the affinities of 14-3-3 isoforms for A20 was difficult due to the presence of endogenous 14-3-3 proteins in the reticulocyte lysates. Endogenous proteins also lowered the binding capacity of the FLAG-A20 beads for 14-3-3 so that significant depletion of the supernatant was only achieved with an excess of A20 beads (data not shown). Therefore, in vitro binding experiments were repeated by using tagged and purified in vitro transcription/translated ␤ and 14-3-3. Using the phosphoimager, it was determined that 32% (Ϯ 13) of the

14-3-3 Associates with A20
14-3-3 bound to A20, while only 11% (Ϯ 4) of the ␤14-3-3 isoform bound to the same amount of A20 beads under identical conditions. The isoform has therefore an affinity for A20 that is 2.8-fold higher than the ␤ isoform. The same experiments were performed with c-Raf expressed in 293 cells, and the ratio of /␤ affinity for Raf binding was 1.6-fold. These experiments demonstrate that A20 and 14-3-3 proteins interact directly and in an isoform-specific manner.
Isoform Specificity of in Vivo Interactions-The overall amino acid sequence of the 14-3-3 proteins is highly conserved (6). The different isoforms are even more conserved between species, which suggest distinct conserved functions within each isoform. To assess the affinity of each isoform for A20, we analyzed the in vivo interaction between FLAG-A20 and endogenous 14-3-3 isoforms in 293T cells. The endogenous ␥, ⑀, , and ␤/ 14-3-3 isoforms can be detected with the available antibodies by immunoblotting 293T cell lysates ( Fig. 2A, total lysate). No signal was obtained in the lysate nor immunoprecipitate with antibodies against and isoforms. The relative affinity of A20 for the different isoforms was estimated by comparison of the ECL signal of the total lysate to that obtained with the same antibody in the immunoprecipitate. Reproducibly, the largest relative signal in the anti-FLAG-A20 immune complex compared to total cell lysate was obtained with antiantibodies. The immunoprecipitate to lysate ratio for the ⑀ isoform was also significantly higher, while much lower ratios were obtained with anti ␤/ and ␥ antibodies. A20 therefore preferentially bound the isoform, followed by the ⑀ isoform. The ␤/ and ␥ isoforms bound least well. Probing the blot with PAN 14-3-3 antibodies revealed an additional band between the slow migrating ⑀ band and the 30-kDa band that represents the bulk of the isoforms ( Fig. 2A, arrow). The same pattern was obtained using a different PAN-14-3-3 antibody raised against the peptide KSELVQKAKLAEQAERYDD (S.C. . As this band was highly enriched in the A20 immunoprecipitate, A20 binds strongly a 14-3-3 protein that is not recognized by the available isoform-specific antibodies. This protein could be a new 14-3-3 isoform or a post-translational modification that modifies the epitopes recognized by the polyclonal isoform-specific antibodies. Postranslational modifications of the ␤ and isoform have been identified as the ␣ and ␦ isoform, respectively (29). These post-translational modifications are still recognized by the antibodies used in our study (27), making it more likely that the A20 associated 14-3-3 protein is a new isoform.
Co-precipitations with c-Raf were performed to determine if Raf also interacted preferentially with 14-3-3 isoforms. c-Raf was found to bind 14-3-3 ( Fig. 2B) with an isoform preference similar to that of A20 (i.e. bound more strongly than ␤/ with ⑀ showing intermediate binding). The affinity of c-Raf for the 14-3-3 isoforms was not identical to A20, as most clearly demonstrated by the absence of the novel isoform in the Raf immunoprecipitates.

Bait
Prey  2). Precipitated 14-3-3 isoforms were resolved on a 15% SDS-PAGE gel, transferred to a membrane, and probed with isoformspecific antibodies. Membranes were stripped and reprobed. The X indicates a reference point and the arrow an unknown 14-3-3 isoform that preferentially binds A20. The panel is representative of three independent experiments. S.C. PAN 14-3-3, gel probed with PAN 14-3-3 antibodies from Santa Cruz Biotechnology. B, 293T cells were transfected with c-Raf. Lanes 1 and 3 were cotransfected with v-Src. Anti-c-Raf immunoprecipitates (lanes 3 and 4) were compared to total lysates (lanes 1 and 2; 1/100 of total) by immunoblotting for 14-3-3 isoforms as in A.
Also of note is the fact that the difference in affinities of the ␤ and isoforms for A20, albeit with the same preferences, are much larger in vivo than in vitro. This suggests that posttranslational modifications of 14-3-3 also contribute to the binding.
Interestingly, a recent peptide binding study defined the 14-3-3 binding sequence as RSXphosphoSXP (30). No isoform preference was found in these studies, indicating that additional motifs are likely responsible for the isoform specificity observed with native proteins.
Raf immunoprecipitates from 293T cells transfected with c-Raf alone or c-Raf and v-Src (a potent activator of c-Raf activity) display identical 14-3-3 isoform binding (Fig. 2B, lanes  3 and 4). Therefore, the 14-3-3/c-Raf interaction and isoform preference was not influenced by the state of activation of Raf.
Whereas this is the first report on Raf isoform specificity, the association of Raf with 14-3-3 has been reported previously to be independent of Raf activation in baculovirus overexpression systems (31,32). In contrast to our work is the finding that activated c-Raf from serum-activated NIH3T3 does not associate with 14-3-3 (12).
14-3-3 Mediates A20/Raf Complex Formation-The dimerization domain of 14-3-3 protein has been localized recently to the amino terminus (21). Potentially, 14-3-3 could serve to bridge two 14-3-3-interacting molecules. To explore this, anti-FLAG immunoprecipitations were performed either at high stringency (0.1% SDS, 0.5% deoxycholate, 1% Nonidet P-40) or at low stringency (1% Nonidet P-40), from cell lysates containing FLAG-A20 and c-Raf. Immunoblotting of the A20-FLAG immune complex with c-Raf antibody revealed that the molecules co-precipitated at low stringency (Fig. 4, upper panel). 14-3-3 was also co-precipitated, suggesting the formation of a trimolecular complex (lane 3, lower panel). At high stringency, in the absence of co-precipitating 14-3-3 (lane 4, lower panel), no complex formation could be detected; the FLAG-A20 precipitates contained almost no co-precipitating Raf (lane 4, upper panel). Because c-Raf and A20 do not interact directly (Table  II), these results suggest that 14-3-3 functions as an adapter, allowing A20 to bind c-Raf. An adapter function for 14-3-3 proteins has been suggested (21), and while this manuscript was in preparation, such a function was shown for the Raf/Bcr interaction (15). The adapter function for 14-3-3 may also be responsible for the interaction of Cdc25 (a phosphatase for cyclin-dependent kinases) with Raf in baculovirus-infected Sf9 lysates (17). In vitro binding studies with purified components will be required to reconstitute the interaction and determine whether additional components participate in complex formation.
In summary, the data presented here adds A20 to a growing list of molecules capable of interacting with the 14-3-3 proteins (17,19,32). Raf, phosphatidylinositol 3-kinase, Bcr-Abl, platelet glycoprotein GpIb-IX, and Cdc25 all interact with 14-3-3 proteins, and all are involved in signal transduction. Unfortunately, the functional consequence of these interactions is poorly understood (32). The ability of A20 to interact with c-Raf through 14-3-3 (Fig. 4) and the dramatic effect of 14-3-3 on A20 localization (Fig. 3) strongly suggest both an adapter and chaperone function for 14-3-3. An adapter role for 14-3-3 was also reported for the Raf-BCR interaction (15). Most recently, char-acterization of the 14-3-3/Raf interaction revealed phosphoserine to be the critical residue responsible for mediating protein/protein interaction (30). This suggests that 14-3-3 is a phosphoserine-regulated adapter molecule akin to SH2-containing proteins like the p85 subunit of phosphatidylinositol 3-kinase, which are phosphotyrosine-regulated adapter proteins (33). A20 contains one perfect match for the consensus 14-3-3 binding sequence (RSXSXP) and several related ones in the COOH-terminal zinc finger region. However, the isoform specificity observed in the binding of 14-3-3 to A20 and to c-Raf suggests that additional protein contacts may also play a role. A recent report identifying the binding domain of 14-3-3 for glycoprotein Ib␣ (34) suggests hat additional motifs can mediate 14-3-3 binding.
A20 not only inhibits TNF-induced apoptosis but also downregulates the activity of its own promoter (5), as well as TNF-induced transcription of several NF-B containing genes 3 and other TNF responses (35) (36) These observations and the association of A20 with proteins known to be involved in signaling implicate A20 in early events of TNF signal transduction.