Ki-1/57 interacts with RACK1 and is a substrate for the phosphorylation by phorbol 12-myristate 13-acetate-activated protein kinase C.

Ki-1/57, the 57-kDa human protein antigen recognized by the CD30 antibody Ki-1, is a cytoplasmic and nuclear protein that is phosphorylated on serine and threonine residues. When isolated from the Hodgkin's lymphoma analogous cell line L540 Ki-1/57 co-immunoprecipitated with a Thr/Ser protein kinase activity. It has been also found to interact with hyaluronic acid and has therefore been termed intracellular IHABP4 (hyaluronan-binding protein 4). Recent studies demonstrated, however, that Ki-1/57 engages in specific interaction with the chromo-helicase-DNA-binding domain protein 3, a nuclear protein involved in chromatin remodeling and transcription regulation. We used the yeast two-hybrid system to find proteins interacting with Ki-1/57 and identified the adaptor protein RACK1 (receptor of activated kinase 1). Next, we confirmed this interaction in vitro and in vivo, performed detailed mapping studies of the interaction sites of Ki-1/57 and RACK-1, and demonstrated that Ki-1/57 also co-precipitates with protein kinase C (PKC) when isolated from phorbol 12-myristate 13-acetate (PMA)-activated L540 tumor cells and is a substrate for PKC phosphorylation in vitro and in vivo. Interestingly, the interaction of Ki-1/57 with RACK1 is abolished upon activation of L540 cells with PMA, which results in the phosphorylation of Ki-1/57 and its exit from the nucleus. Taken together, our data suggest that Ki-1/57 forms a stable complex with RACK-1 in unstimulated cells and upon PMA stimulation gets phosphorylated on threonine residues located at its extreme C terminus. These events associate Ki-1/57 with the RACK1/PKC pathway and may be important for the regulation of its cellular functions.

The first monoclonal antibody that specifically detected the malignant Hodgkin's and Sternberg-Reed cells in Hodgkin's lymphoma was called Ki-1 and binds to the 120-kDa lymphocyte co-stimulatory molecule CD30 (Ki-1/120) on the surface of the Hodgkin's cells (1,2). It has however been noticed early on that this antibody also cross-reacts with an intracellular phosphoprotein antigen of 57 kDa termed Ki-1/57 (3,4). In vitro phosphorylation experiments performed with the Ki-1/57 antigen isolated from tumor cells demonstrated that it is associated with a serine/threonine protein kinase activity (5). Electron microscopic analysis showed that the Ki-1/57 antigen is located in the cytoplasm, at the nuclear pores, and in the nucleus, where it is frequently found in association with the nucleolus and other nuclear bodies (6). Tryptic digestion of the Ki-1/57 antigen resulted in the cloning of a partial cDNA encoding Ki-1/57 (7). The isolated contig 1 of 1380 bp length encodes the C-terminal 60% of the Ki-1/57 protein. Later, another group cloned the full-length Ki-1/57 cDNA (8). Huang et al. (8) found that Ki-1/57 has a hyaluronan binding activity and gave it the second name, intracellular hyaluronan-binding protein 4 (IHABP4). They also found that IHABP4/Ki-1/57 binds to other negatively charged glycosaminoglycans like chondroitin sulfate, heparane sulfate, and RNA, although with lower affinity. The functional meaning of Ki-1/57 interaction with these macromolecules remains open.
When we were searching the sequence data bank for Ki-1/57 related molecules, we found the human protein CGI-55, which amino acid sequence has 40.7% identity and 67.4% similarity with that of Ki-1/57 (9). This high degree of similarity suggests that both proteins might be paralogues and may have related functions. CGI-55 has also been described to bind to the 3Јregion of the mRNA encoding the plasminogen activator inhibitor (PAI) type 1 (10). Heaton et al. (10) have therefore termed CGI-55 as PAI RNA-binding protein 1 and suggested that it could be involved in the regulation of the stability of the PAI mRNA, although they do not provide experimental data to support this hypothesis.
We explored the yeast two-hybrid system to identify possible interacting proteins for both Ki-1/57 and CGI-55 and in this way obtain clues for the functional context of these proteins. Our analysis resulted in the identification of the human protein chromo-helicase-DNA-binding domain protein 3 (CHD3) as a partner for both proteins (9). The CHD proteins are members of the chromo domain family, a class of proteins that are involved in transcriptional regulation and chromatin remodeling (11)(12)(13)(14)(15)(16)(17). The binding of the proteins Ki-1/57 and CGI-55 to CHD3 might define them as a family of CHD3-binding proteins and suggested the possibility that they could be involved in nuclear functions associated with the remodeling of chromatin and the regulation of transcription. Whereas in the case of the CGI-55, 42% of the found interacting clones represented CHD3, only 4% of the clones interacting with Ki-1/57 represented CHD3 (9).
Here we report that the vast majority of clones (54%) found to interact with Ki-1/57 represent the scaffold and regulatory protein RACK-1 (receptor of activated kinase 1), a protein that we did not identify in the interaction screen of the putative Ki-1/57 paralogue CGI-55. RACK1 has a molecular mass of 36 kDa and is composed of seven WD repeats (18,19). Its overall structure resembles that of the ␤-subunit of G proteins (20,21). RACK1 has been reported to interact with PKC␤ (22)(23)(24); Src (25); ␤-integrins (26); PDE4D5 (27); the ␤-subunit of the granulocyte-macrophage colony-stimulating factor, interleukin 3, and interleukin 5 receptors (28); type 1 interferon receptor (29); STAT1 (30); and a number of viral proteins (31)(32)(33). RACK1 is up-regulated in human carcinomas and during tissue regeneration after ischemic renal injury (34,35). Furthermore, RACK1 has been functionally implicated in the development of cardiac hypertrophy (36), the regulation of cell adhesion (37), the increase of focal adhesion (38), and the protection from viral, E1A protein-induced apoptosis (32). On a molecular level the interaction of RACK1 with Src has been described to result in an inhibition of the kinase activity of Src (25). The activity of PDE4E5 on the other hand was unaffected by the binding of RACK1 (27). Although RACK1 has been reported to have a stimulatory effect on the substrate phosphorylation by PKC (22,39), others found that RACK1 does not influence the kinase activity of serine/threonine kinases such as PKC, cAMP-dependent protein kinase, and casein kinase II (25), indicating that the RACK1 activity on PKC may be influenced by the kind of substrate involved.
Here we show that RACK1 interacts with Ki-1/57, confirm this interaction in vitro and in vivo, and map the interaction sites of Ki-1/57 and RACK-1 in detail. Furthermore, we found that Ki-1/57 is a substrate for PKC and that its interaction with RACK1 is abolished in the course of the PMA activation of the cells. Our data suggest that Ki-1/57 is involved in specific protein-protein interactions and provide a plausible explanation for the long known fact that Ki-1/57, which does not contain a kinase domain, in fact co-precipitates with kinase activity. The co-precipitated kinase activity appears to be PKC. This could be confirmed by the co-immunoprecipitation of Ki-1/57 with PKC, which associates with Ki-1/57 after PMA stimulation of the cells. Our results further suggest that the cellular functions of Ki-1/57 may be subject to regulation via a PKC/ RACK1 pathway.

EXPERIMENTAL PROCEDURES
Plasmid Constructions-Several sets of oligonucleotides were designed to allow subcloning of the complete Ki-1/57 coding region in different expression vectors. Insertion into pGEX-2TK (Amersham Biosciences) allowed expression of Ki-1/57(1-413) as a C-terminal fusion to GST (GST-Ki-1/57). The cDNAs encoding full-length Ki-1/57, Ki-1/ 57(122-413), and the eight other indicated deletion constructs (numbered in the same way) were inserted into the yeast two-hybrid expression vector pBTM-116 (40 -43). Other deletions were also subcloned into bacterial expression vectors peT28a, pProEx, or pGEX to allow their expression as His 6 -tagged or GST-tagged fusion proteins. In a similar fashion the cDNAs encoding the indicated deletions of the protein RACK-1 were amplified and inserted into the yeast two-hybrid vector pGAD424 (Clontech), and full-length RACK1 was inserted into the bacterial expression vector pET-28a (Novagen) to allow expression of the His 6 -RACK1 fusion protein.
Yeast Two-hybrid Screening and Interaction Analysis-The pBTM116-Ki-1/57(122-413) (40) vector was used to express a fragment spanning 60% of the C terminus of the protein Ki-1/57 linked to the C terminus of LexA DNA-binding domain peptide in Saccharomyces cerevisiae strain L40. A human fetal brain cDNA library (Clontech) expressing GAL4 activation domain fusion proteins was co-transformed with the recombinant pBTM116-Ki-1/57 vector. Selection of transfor-mants, the ␤-galactosidase activity test, plasmid DNA extraction, and sequencing were performed as described previously (9).
In Vitro Binding Assay, Western Blot Analysis, Antibodies, and Cell Culture-GST or GST-Ki-1/57 fusion proteins were coupled to glutathione-Sepharose beads. After washing, the beads were incubated with His 6 -RACK-1 for 2 h at 4°C and then washed with buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl). The proteins bound to the beads were separated by SDS-PAGE, transferred to a polyvinylidene difluoride membrane, and visualized by immunochemiluminescence using a mouse anti-GST antibody (for control of equal loading of beads) or anti-His 5 monoclonal antibody (Qiagen) and secondary anti-mouse IgGhorseradish peroxidase conjugate. The anti-RACK-1 monoclonal antibody was from Transduction Laboratories. The specific anti-Ki-1/57 monoclonal antibodies A26, E203 (7), and Ki-1 (1) have been described previously. Anti-Ki-1/67 control antibody had been provided by Prof. Dr. Hilmar Lemke (45). An anti-phospho-PKC antibody sampler kit was purchased from Cell Signaling Technology. L540 Hodgkin's analogous cells (46) were cultivated in RPMI 1640 medium, supplemented with 20% fetal calf serum, 2 mM L-glutamine, penicillin (100 units/ml), and streptomycin (100 g/ml) at 37°C and 5% CO 2 (L540 standard medium). HeLa cells were cultivated in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM L-glutamine, penicillin (100 units/ml), and streptomycin (100 g/ml) under equal conditions.
In Vivo Binding Assay-5.0 ϫ 10 7 L540 cells were or were not stimulated with PMA (100 ng/ml) for 4 h (33). The cells were lysed in 1 ml of buffer NaCl/Tris (25 mM Tris, pH 7.5, 137 mM NaCl, 2.7 mM KCl, 1% Triton X-100, protease inhibitors). The lysates were treated with DNase (Promega) and cleared at 14.000 ϫ g for 30 min. Next 20 l of protein A-Sepharose beads (Amersham Biosciences) were loaded with the indicated antibodies overnight at 4°C, washed in buffer NaCl/Tris, and incubated with the L540 lysate overnight at 4°C. After further three washes with the buffer Tris/EDTA (10 mM Tris, pH 7.5, 1 mM EDTA, 0.5 M NaCl), the beads were resuspended in SDS-PAGE loading buffer, boiled, and analyzed by SDS-PAGE and Western blot using the indicated antibodies. Western blots were developed by chemiluminescence as described previously (44). Loading controls consisted either of protein detection by SDS-PAGE or of Western blot development with control antibodies as indicated in the figures.
In Vitro Phosphorylation Assay and Phosphoamino Acid Analysis-5 ϫ 10 7 L540 cells were treated or not with PMA, collected, and lysed as described above. Endogenous PKC was immunoprecipitated from these lysates with anti-phospho-PKC-Pan antibody (Cell Signaling) coupled to protein A-Sepharose beads. Next these beads were incubated with purified GST-Ki-1/57, His 6 -RACK-1, both, or GST in kinase buffer (25 mM Tris, pH 7.5, 1.32 mM CaCl 2 , 5 mM MgCl 2 , 1 mM EDTA, 1.25 EGTA, 1 mM dithiothreitol) containing 10 nM PMA, 5 M ATP, and 0.5 Ci of [␥-32 P]ATP, in a total volume of 25 l for 30 min at 30°C. Phosphorylated proteins were run out by SDS-PAGE. The gel was stained, dried, and exposed to x-ray film. In other experiments purified GST, GST-Ki-1/57, His 6 -RACK1, and deletion constructs of Ki-1/57 were phosphorylated in complete kinase buffer in a final volume of 50 l at 30°C with purified PKC-Pan, PKC, or PKC for 15 min. The PKC or PKC are human recombinant His-tagged and affinity-purified proteins (Promega). PKC-Pan was purified from rat brain and consists predominantly of the PKC isoforms ␣, ␤, and ␥ (Promega). Radioactively labeled proteins were visualized as described above.
Phosphoamino acid analysis was basically performed as described in Machado et al. (47). Briefly, the 32 P-radiolabeled phosphorylated proteins were hydrolyzed with 6 N HCl for 60 min at 90°C. The hydrolysate was lyophilized, dissolved in water, and spotted onto Sigma cell type 100 cellulose thin layer chromatography plates (Sigma). The solvent system was isobutyric acid, 0.5 M ammonium hydroxide (5:3). Phosphoserine, phosphothreonine, and phosphotyrosine standards (2 g) (Sigma) were mixed with the radiolabeled protein hydrolysate and spotted together on the TLC plates. Amino acids were visualized with 0.2% ninhydrin in ethanol, and radiolabeled residues were detected by autoradiography (AR). Theoretical phosphorylation site prediction was performed by the software NetPhos 2.0 Prediction server available at the web site of the Center for Biological Sequence Analysis (www.cbs.dtu.dk/services/NetPhos).
Metabolic Labeling, in Vivo Phosphorylation Assay, and Kinase Inhibitors-5 ϫ 10 6 L540 cells were preincubated or not for 1 h with protein kinase inhibitors: Ro-32-0432 (28 nM), and staurosporine (0.7 nM) (Calbiochem). This inhibitor incubation was performed with phosphate-free L540 standard medium (the fetal calf serum in this medium had been dialyzed against a 150 mM NaCl solution). Next the cells were activated or not by the addition of 100 ng/ml of PMA for a second hour. In parallel to the PMA treatment, the cells were metabolically labeled by the addition of 0.4 mCi of radioactive 32 P-labeled inorganic phosphate (Amersham Biosciences). After lysis Ki-1/57 was immunoprecipitated from the lysates of the metabolically labeled L540 cells with anti-Ki-1 antibody A26 coupled to protein A-Sepharose beads and analyzed by autoradiography and SDS-PAGE.
Preparation of Cytoplasmic and Nuclear Cell Fractions-L540 cells were harvested and incubated with 300 l of hypotonic buffer A (10 mM Tris, pH 7.5, 10 mM KCl, 0.1 mM EDTA, 1.5 mM MgCl 2 , 0.5 mM dithiothreitol, and a mixture of protease inhibitors) for 30 min on ice (33). The nuclei were recovered by centrifugation at 14,000 rpm for 10 min. The supernatant represents the cytoplasmic fraction. To obtain the nuclear fraction, the crude nuclear pellet was resuspended in 200 l of hypertonic buffer B (20 mM Tris, pH 8.0, 0.4 M NaCl, 0.1 mM EDTA, 1.5 mM MgCl 2 , 0.5 mM dithiothreitol, 25% v/v glycerol) followed by incubation on ice for 30 min. After centrifugation, the fractions were incubated with the antibodies at 4°C overnight. On the next day 20 l of protein A-Sepharose were added for 2 h.
Immunofluorescence Analysis-HeLa cells grown on glass coverslips were stimulated or not with PMA for 4 h at 37°C. The cells were fixed with 100% methanol and immunostained with primary antibody monoclonal mouse Ki-1, mouse anti-RACK1, or rabbit anti-Phospho-PKC, and secondary antibody fluorescein anti-mouse or rhodamine anti-rabbit antibody. The cells were examined with a Nikon microscope. DAPI staining was used to show the positions of the nuclei. The cells were examined with Nikon fluorescence microscope. Immunolabeled proteins were presented with the respective color. Superimposing the two colors (merge) results in a yellow/orange signal.

RESULTS
Yeast Two-hybrid Screen-To identify Ki-1/57 interacting proteins, the yeast two-hybrid system (40 -43) was employed, utilizing a human fetal brain cDNA library (Clontech). In a first screen we used a fragment of the Ki-1/57 cDNA that encodes its C-terminal 60% as a bait. 2.0 ϫ 10 6 screened cotransformants yielded 250 clones positive for both His3 and LacZ reporter constructs. Library plasmids DNA of 80 clones were sequenced. 54% of the sequenced clones all encoded the full-length protein RACK1 (48). Another protein identified was CHD3, which had already been described previously elsewhere (9) and represented 4% of the interacting clones. Other nuclear proteins involved in the regulation of transcription have also been identified but will be described elsewhere.
Furthermore, we mapped the RACK1 regions that are required for the interaction with Ki-1/57. N-and C-terminal deletion constructs of the RACK1 protein were fused to the Gal4 activation domain (vector pACT2; Fig. 1C) and tested for their ability to bind to full-length Ki-1/57 (Fig. 1D). None of the four different deletion constructs of RACK1 interacted with Ki-1/57. This shows that full-length RACK1 is required for an interaction with Ki-1/57.

In Vitro Confirmation of the Ki-1/57-RACK1 Interaction with
Purified Fusion Proteins-To confirm the interaction between Ki-1/57 and RACK1 in vitro, we next performed in vitro pulldown assays with purified recombinant proteins that had been expressed in E. coli (GST, GST-Ki-1/57, and His 6 -RACK1) and purified by affinity chromatography. GST-Ki-1/57 bound specifically to RACK1, whereas the control protein GST did not ( Fig. 2A, left panel). We controlled the equal loading of the glutathione beads with GST or GST-Ki-1/57 fusion protein by developing the same membrane with an anti-GST monoclonal antibody ( Fig. 2A, right panel).
Co-precipitation of Ki-1/57 with His 6 -RACK1 from a Lysate of L540 Cells-When a lysate of L540 cells was incubated with His 6 -RACK1-loaded Ni-NTA-Sepharose beads, Ki-1/57 could be specifically pulled down (Fig. 2B, left panel). On the other hand, when Ni-NTA-Sepharose beads were loaded with control proteins such as the nonrelated protein His 6 -FEZ1 (49) or the Ki-1/57 homologue protein CGI-55 (9), no co-precipitated Ki-1/57 band could be detected in the Western blot. We obtained a corresponding result when we used glutathione-Sepharose control beads loaded with GST-Ki-1/57 (Fig. 2C, left panel). RACK1 was only pulled down from the lysate of L540 cells with the GST-Ki-1/57 but not with the GST-loaded beads. Both Western blot experiments were checked by detecting the input proteins. These were run out on SDS-PAGE, transferred to a polyvinylidene difluoride membrane, and visualized by anti-His 5 monoclonal antibody or anti-GST antibody, depending on the nature of the fusion part (Fig. 2, right panels).
In Vitro Phosphorylation of Ki-1/57 by PKC Isolated from PMA-stimulated L540 Cells-Early experiments with the Ki-1/57 antigen had demonstrated that the Ki-1 antibody immunoprecipitated an serine/threonine protein kinase activity that had been initially attributed to the Ki-1/57 antigen (5). The cloning of the cDNA encoding Ki-1/57 revealed, however, that the Ki-1/57 sequence does not encode a kinase domain (7,8). Further experiments with the full-length recombinant Ki-1/57 protein also did not reveal a kinase activity of the Ki-1/57 protein toward itself or other proteins (data not shown). Our finding that Ki-1/57 strongly interacts with the protein RACK1 in the yeast two-hybrid system immediately suggested a hypothesis for an alternative explanation for the co-immunoprecipitation of the kinase activity with Ki-1/57. RACK1 is an receptor for activated protein kinase C and interacts in vitro and in vivo with activated PKC. Therefore, the observed kinase activity in the Ki-1 immunoprecipitate might be PKC associated to RACK1 and/or to Ki-1/57. To test this hypothesis, we first analyzed whether Ki-1/57 can be a substrate for the phosphorylation by PKC immunoprecipitated from the lysate of unstimulated and PMA-stimulated L540 cells (Fig. 3A, left  panel). The GST-Ki-1/57 fusion protein used in this assay was only weakly phosphorylated by the PKC that had been immunoprecipitated from the lysate of nonactivated cells by an anti-Phospho-Pan-PKC antibody (Fig. 3A, left panel, second lane). It was, however, strongly phosphorylated by the PKC isolated from the lysate of PMA-activated L540 cells (Fig. 3A, fifth lane). RACK1 itself did not suffer phosphorylation by PKC under these conditions (Fig. 3A, left panel, first and fourth lanes), nor did its presence influence the extend of phosphorylation of Ki-1/57 (Fig. 3A, left panel, third and sixth lanes). The right panel of Fig. 3A demonstrates the equal protein loading of the different lanes. GST control protein was not phosphorylated by PKC (Fig. 3B, upper panel, first lane).
In a similar approach we next wanted to know whether PKC isoforms have a differential phosphorylation activity toward Ki-1/57 and tested therefore a panel of monoclonal antibodies against different PKC subtypes isolated from PMA-activated L540 cells (Fig. 3B, upper panel). The PKCs had been immunoprecipitated by anti-phospho-PKC antibodies or as indicated in Fig. 3B. We found a strong phosphorylation of Ki-1/57 by PKC␣␤, PKC␦, PKC/, and especially by PKC, however not by PKC. These data show that Ki-1/57 can serve in principal as a substrate for a wide variety of different PKC isoforms but also that its phosphorylation is strongest with PKC. The equal loading of the different lanes is demonstrated by the control SDS-PAGE shown at the bottom panel of Fig. 3B.
We also performed phosphorylation experiments with commercial purified PKC-Pan and found a strong phosphorylation of Ki-1/57, which was neither promoted nor inhibited by the presence of His 6 -RACK1 protein in equal amounts or slight excess (Fig. 3C).
Ki-1/57 Interaction with RACK1 Is Abrogated by Its Phosphorylation or by the PMA Activation of the Cell in Vivo-When we performed the pull-down experiments of RACK1 with a GST-Ki-1/57 protein that we had previously submitted to in vitro phosphorylation with PKC-Pan, we observed a complete abrogation of the interaction (Fig. 4A, left panel, lane 3). Such a down-regulation of the interaction by the phosphorylation of Ki-1/57 might be functionally relevant and could serve to downregulate the adaptor functions of RACK1 once that PKC has phosphorylated Ki-1/57 in vivo. Therefore we tested whether this effect can also be observed in vivo. We found that His 6 -RACK1 fusion protein coupled to Ni-NTA-Sepharose beads cosediments an approximately three times smaller quantity of Ki-1/57 from lysates of PMA stimulated then from unstimulated L540 cells (not shown). This indicates that the phosphorylation of Ki-1/57 in vivo also diminishes its capacity to interact with external recombinant RACK1.
Ki-1/57 Is Only Phosphorylated on Its Extreme C Terminus (Residues 346 -413)-We now expressed several deletion constructs of the Ki-1/57 protein and submitted the purified recombinant proteins to in vitro phosphorylation experiments to determine the regions of Ki-1/57 that are a target for the phosphorylation by PKC (Fig. 4B). From these studies it became clear that neither the N-terminal region (1-150) nor the middle region of Ki-1/57(151-263) but only its C-terminal region (both 264 -413 and 122-413) are phosphorylated by PKC-Pan (Fig. 4B, left panel). Recombinant PKC and PKC gave the same results as purified PKC-Pan (not shown). The loading of the in vitro phosphorylation reaction with equal amounts of recombinant protein fragments is shown in Fig. 4B (right  panel). These results suggested performance of a more detailed deletional analysis of the C terminus of Ki-1/57, because this region contains 15 Ser/Thr residues (Fig. 5A), all of which could be target residues of phosphorylation by PKC. Therefore, we generated the indicated subdeletions (Fig. 5A) of the fragment Ki-1/57(264 -413) and expressed them in bacteria as GST fusions proteins for in vitro studies and in the yeast as LexA fusion proteins to be able to assess their capacity to still interact with RACK1. Two of the constructs, Ki-1/57(294 -413) and Ki-1/57(346 -413) were still able to interact with RACK1 (Fig.  5B). Most interestingly, these same two constructs, when expressed in fusion with GST and used as substrates in the in vitro phosphorylation assays with PKC-Pan, were the only two of the five tested subdeletions that could be phosphorylated (Fig. 5C, left panel). In vitro phosphorylations of these five fragments with PKC and PKC gave the same result (not shown). Fig. 5C (right panel) shows equal loading of the reactions with proteins or control protein GST, which was not phosphorylated.
Ki-1/57 Phosphorylation Can Be Blocked by Protein Kinase Inhibitors in Vitro and in Vivo-To gather further evidence that the kinase that phosphorylates Ki-1/57 is PKC, we tested a series of protein kinase and PKC inhibitors for their potential to block Ki-1/57 phosphorylation in vitro (Fig. 4C) and in vivo (Fig. 4D). We found that the general kinase inhibitor staurosporine and the PKC-specific inhibitor Ro-32-0432 were the most effective inhibitors of the phosphorylation of His 6 -Ki-1/57 (264 -413) by PKC-Pan (Fig. 4C, lanes 7 and 8), PKC, and PKC (not shown) in vitro. There was no difference in the inhibition profile for the three PKCs tested. We then tested the best two inhibitors in vivo and found that only Ro-32-0432 but not staurosporine (at the tested relative low concentration) can inhibit the phosphorylation of Ki-1/57 in vivo (Fig. 4D, lanes 3  and 4). These data support the hypothesis that Ki-1/57 is also a substrate for PKC phosphorylation in vivo. A comparison of lanes 1 and 2 of Fig. 4D demonstrates the increased phosphorylation of Ki-1/57 after the stimulation of the L540 cells with PMA and in the absence of inhibitor. The equal loading of the lanes with immunoprecipitated Ki-1/57 is shown in the lower panel of Fig. 4D.

FIG. 2. In vitro binding assays of RACK1 and Ki-1/57. A, glutathione-
Sepharose beads were loaded with purified GST or GST-Ki-1/57 proteins. The beads were then washed and incubated with purified His 6 -RACK1, washed three times, separated by SDS-PAGE, transferred to a polyvinylidene difluoride membrane, and probed with anti-His 5 antibody (left panel) or anti-GST monoclonal antibody (right panel). B, Ni-NTA-Sepharose beads were loaded with His 6 -RACK1 or control proteins His 6 -CGI-55 or His 6 -FEZ1. Loaded beads were than incubated with the total cell lysate of 1 ϫ 10 7 L540 cells and washed three times. The bound proteins were separated by SDS-PAGE and transferred to a polyvinylidene difluoride membrane and probed with Ki-1/57 antibody (A26). C, glutathione-Sepharose beads were loaded with GST or GST-Ki-1/57 proteins and incubated with lysate of L540 cells for 3 h at 4°C. After washes and processing as above in B, RACK1 was probed with specific antibody. Purified GST, GST-Ki-1/57, and His 6 -RACK1 proteins were used to identify the precipitated bands on the blot. The arrows indicate the detected proteins. A and B, the panels shown on the right represent membranes that were developed with control antibodies against the fusion part GST or His 6 of the used recombinant proteins to demonstrate equal loading. WB, Western blot.
The extreme C-terminal fragment Ki-1/57(346 -413) in fusion with GST is only phosphorylated on threonine using PKC-Pan. PKC and PKC also phosphorylated this fragment only on threonine, but the degree of phosphorylation was lower (not shown). This suggests that the two threonine residues present in this fragment (Thr 354 and Thr 375 ) might be the main target residues for phosphorylation by PKC in vitro. This also demonstrates that the phosphorylation of Ki-1/57 by PKC is highly specific, considering that there are 34 Ser/Thr residues in the whole amino acid sequence of Ki-1/57, and apparently only the two most C-terminal threonines are targets of phosphorylation in vitro. The reaction was controlled with free GST protein, which itself does not suffer phosphorylation by the three PKCs tested (not shown; see also Fig. 5C).
Ki-1/57 Interacts with RACK1 in Vivo Only before PMA Activation and with PKC Afterward-Next we wanted to test whether Ki-1/57 engages also in interaction with RACK1 and PKC in human cells. To assess the subcellular localization of the co-immunoprecipitated proteins, we analyzed the cytoplasmic and nuclear compartments of the L540 cells separately. When we immunoprecipitated RACK1 from the lysate of L540 cells, we detected the 57-kDa band of co-immunoprecipitated Ki-1/57 in the Western blot developed with A26 antibody only in the nuclear compartment of the cell but both before and after cell stimulation with PMA (Fig. 6, lanes 25 and 26). There was a slight decrease in the co-precipitated amount of Ki-1/57 after the addition of PMA.
When we immunoprecipitated Ki-1/57 we also co-immunoprecipitated RACK1 but only from the nucleus and in the absence of PMA (Fig. 6, lane 16). These results suggest that Ki-1/57 and RACK1 form a stable complex in human L540 cells, until the cells are activated by PMA. Interestingly, we also detected co-immunoprecipitated Ki-1/57 when we used antiphospho-PKC (and anti-phospho-PKC, not shown) antibody in the immunoprecipitation, in the nucleus and a little less in the cytoplasm, but only after PMA activation (Fig. 6, lanes 30  and 32). Anti-phospho-PKC␣␤II did not co-immunoprecipitate Ki-1/57, neither with our without PMA treatment (not shown).
Ki-1/57 Exits the Nucleus upon PMA Activation-Ki-1/57 is located in the cytoplasm and the nucleus of cells (6) and interacts with nuclear proteins involved in the regulation of transcription and the remodeling of chromatin such as CHD3 (9). Therefore, we were interested to know whether its localization to the nucleus or that of its interacting proteins RACK1 and PKC is affected by the cell stimulation with the PKC activator PMA and whether the interaction of RACK1 and Ki-1/57 is affected by PMA. First, we stimulated L540 cells with PMA or not and then fractionated the cellular lysates into cytoplasmic and nuclear fractions, from which we then immunoprecipitated Ki-1/57, RACK1, and Phospho-PKC (Fig. 6). When immunoprecipitated by antibody A26, the nuclear fraction of Ki-1/57 is no longer detectable after 4 h of PMA stimulation of the cells (Fig.  6, lanes 35 and 36), whereas there was no alteration in the amounts of cytoplasmic Ki-1/57. When immunoprecipitated with antibody E203, this decrease of the amount of nuclear Ki-1/57 could also be noticed, albeit to a lesser degree in comparison with A26 (Fig. 6, lanes 39 and 40). There is slightly more RACK1 in the cytoplasmic fraction after PMA stimulation (Fig. 6, lanes 1 and 2), whereas the nuclear fraction is unaffected by the PMA stimulation (Fig. 6, lanes 3 and 4). The amount of RACK1 co-immunoprecipitated with phospho-PKC-Pan is increased both in the cytoplasm as well as in the nucleus, whereas RACK1 that co-immunoprecipitated with PKC␣␤II was only detectable in the cytoplasm after PMA stimulation (Fig. 6, lane 10).
The fact that Ki-1/57 is only associated with RACK1 in the nucleus and in the absence of cell stimulator PMA (Fig. 6, lanes  15 and 16) suggests that the interaction of Ki-1/57 with RACK1 is only stable under unstimulated conditions but is abrogated after the activation with PMA. These experiments were con-trolled with an antibody against the nonrelated protein Ki-67 (46), which co-immunoprecipitates neither RACK1 (Fig. 6,  lanes 17-20) nor Ki-1/57 (Fig. 6, lanes 41-44).
We next tested whether the observed disappearance of Ki-1/57 from the nucleus (Fig. 6, lanes 36 and 40) can be seen by immunofluorescence localization studies in human HeLa cells (Fig. 7). We observed that both RACK1 (Fig. 7C) and Ki-1/57 (Fig 7, A and B) exit the nucleus upon PMA activation. The exit of Ki-1/57 from the nucleus is accompanied by that of phospho-PKC␣␤II and phospho-PKC/ (Fig. 7), and that of RACK1 is accompanied by the exit of phospho-PKC␣␤II. These experiments were repeated with L540 cells and essentially gave the same results (not shown).

DISCUSSION
To find a functional context for the protein Ki-1/57 we set out to perform a yeast two-hybrid screen to identify possible interacting protein partners. Screens of a human fetal brain cDNA library with both Ki-1/57 and with its homologue protein CGI-55 previously identified the chromatin remodeling factor CHD3 (9). This was the first report that described a specific protein-protein interaction for both CGI-55 and Ki-1/57 and could define them as a new family of CHD3 interacting proteins. The majority (54%) of clones found to interact with Ki-1/57, however, represent the signaling adapter molecule RACK1.
The interactions between Ki-1/57 and RACK1 were con-firmed in vitro and in vivo by co-precipitation assays from L540 Hodgkin's disease analogous cells. Because RACK1 has been described previously to be an adapter protein for activated protein kinases C and helps to maintain PKC in an activated

FIG. 5. Mapping of the phosphorylation and interaction characteristics of Ki-1/57 C-terminal region (264 -413) and phosphoaminoacid analysis.
A, schematic representation of the C-terminal region of Ki-1/57 with indication of the present Ser and Thr residues. Underlined bold type indicates residues with a predicted probability of Ͼ78% to be phosphorylated, whereas residues without underlining have a probability of Ͻ45% (prediction made by using the NetPhos 2.0 Prediction Server). B, interaction of the indicated Ki-1/57 fragments with RACK1 in the yeast two-hybrid system. The ability of the co-transformant cells to grow on minimal medium,-Trp,-Leu,-His was tested in the presence of 10 mM 3-amino-1,2,4-triazole inhibitor for suppression of unspecific reporter activation. Presence of bait and prey plasmids in the co-transformed cells was controlled by growth on minimal medium-Trp,-Leu (data not shown). state, we tested whether Ki-1/57 also interacts with cellular PKC and whether it represents a target molecule for PKC phosphorylation. Our phosphorylation assays with PKC show that Ki-1/57 is a substrate for PKC isolated from PMA-activated but not from nonactivated control L540 cells. Immunoprecipitation of PKC from PMA-stimulated but not from unstimulated L540 cells showed co-immunoprecipitated Ki-1/57 protein, thereby demonstrating that PKC interacts with Ki-1/57 after cell activation. The interaction of Ki-1/57 with RACK1, however, was abolished after PMA stimulation, suggesting that this interaction is regulated. Together these results might suggest a hypothesis for a sequential mode of interactions between the three molecules Ki-1/57, RACK1, and PKC: (i) before PMA activation Ki-1/57 is firmly attached to RACK1, this interaction occurs mainly in the nucleus; (ii) after PMA activation Ki-1/57 gets phosphorylated, this results in the abrogation of the interaction with RACK1; and (iii) the newly created phosphoamino acid groups in the C terminus of Ki-1/57 could now serve as docking sites for the interaction of kinases or other proteins with Ki-1/57.
It had been shown previously that different proteins that interact with RACK1 interact with different docking sites involving one or more of the seven individual blades of the putative propeller structure of RACK (18). In case of the protein Src, the smallest unit of RACK1 that was capable of an interaction consists of only a single WD repeat blade (50). The binding of the interferon receptor on the other hand involves five of the seven blades (30). Therefore, we set out to map the regions of RACK1 involved in the interaction with Ki-1/57 and found that none of our constructed deletions but only full-length RACK1 was able to engage in protein-protein interaction with Ki-1/57. This is corroborated by the fact that all of the clones we identified in the yeast two-hybrid screen contained the full-length RACK1 coding region. Ki-1/57 seems therefore to be the first protein found that only interacts with full-length RACK1. The mapping of the interaction site of Ki-1/57 on the other hand demonstrated that its extreme C terminus (amino acids 346 -413) is fully capable to interact with RACK1. This suggests that Ki-1/57 might be a multi-domain protein with its C terminus containing a docking domain/motif for RACK1.
It was shown before that although RACK1 engages in protein interaction with several structurally and functionally different proteins (22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33), its interaction among proteins of the same family is highly specific (50,27). Yarwood et al. (27) have for instance shown that RACK1 interacts with PDE4D5 but with none of the other PDE4 isoforms tested. This high degree of selectivity of the interaction holds true in the opposite direction, too. PDE4 does not interact with any of a series of WD repeat-containing proteins other than RACK1 (27). Our results confirm this trend, because none of the other proteins identified so far in our yeast two-hybrid analysis of the Ki-1/57 protein belongs to the WD repeat-containing family of proteins (not shown). RACK1 interacts neither in the yeast two-hybrid sys-tem nor in vitro with the Ki-1/57 homologue protein CGI-55 (not shown).
RACK1 has been described to have an activating influence FIG. 7. Subcellular localization of endogenous RACK1, Ki-1/57, phospho-PKC␣␤ II , and phospho-PKC/before and after treatment of cells with PMA. HeLa cells grown on glass coverslips were stimulated or not with PMA for 4 h at 37°C. The cells were fixed with 100% methanol, and the proteins were immunodetected with the following primary antibodies: Ki-1 monoclonal mouse antibody or mouse anti-RACK1 and the indicated rabbit anti-phospho-PKC antibodies. Fluorescein-coupled anti-mouse (green) or rhodamine-conjugated anti-rabbit antibodies (red) served as secondary antibodies. DAPI staining (blue) was used to show the positions of the nuclei. The cells were examined with a Nikon fluorescence microscope. Superimposing the two colors (merge) results in a yellow/orange signal. on PKC activity and even a small peptide derived of amino acids 234 -241 of WD blade 6 of RACK1 bound to PKC and activated it in vitro and in vivo (22,46). However, in contradiction to this finding another report showed no influence of RACK1 on the kinase activity of PKC, cAMP-dependent protein kinase, or casein kinase II toward peptide substrates but an inhibitory effect of RACK1 on the autophosphorylation activity of Src and Yes and on the peptide phosphorylation activity of Src and Lck (25). These results demonstrate that it is not yet possible to conclude whether RACK1 has an overall stimulatory or inhibitory role on the kinase activity of different kinases but rather suggest that not only the type of kinase but most likely also the kind of substrate involved might be of importance. Our results with the phosphorylation of the PKC substrate Ki-1/57 in the presence of RACK1 did not show any influence of RACK1 on the outcome of the kinase reaction (Fig. 3C).
Ki-1/57 has been also described previously as an intracellular hyaluronan-binding protein (IHABP4), because of its capacity to interact with a series of negatively charged macromolecules, including hyaluronan, heparan sulfate, chondroitin sulfate, and RNA (8). According to Huang et al. (8), the binding of IHABP4/Ki-1/57 to hyaluronan depends on the presence of so called hyaluronan binding motifs of the structure (R/K)X 7 (R/K). However, the majority of nuclear proteins are overproportionally rich in the positively charged amino acids Lys and Arg. Our analysis of several randomly selected nuclear proteins revealed that all of them contained several of such putative hyaluronanbinding motifs. However, some of them contained many of such putative hyaluronan-binding motifs: CHD3 (accession number NM_001272.1) contains 49; Topors (accession number AF098300) contains 36; human polycomb2 (accession number AF013956) contains 7; p53 (accession number AAH03596) contains 3; and c-Fos (accession number K00650) contains 2. This would suggest that the majority of Arg/Lys-rich nuclear proteins have the potential to interact with hyaluronan. Even Huang et al. (8) state that it remains open whether hyaluronan is indeed a natural ligand for IHABP4/Ki-1/57 (8). They speculate that because both hyaluronat and Ki-1/57 have been found in the nucleus and cytoplasm, Ki-1/57 might be involved in the regulation of hyaluronan functions (8).
Our recent studies point to other possible nuclear functions of Ki-1/57 as well as its homologue CGI-55 (9). We found that CGI-55 and Ki-1/57 interact with CHD3, a nuclear protein involved in the remodeling of chromatin and the regulation of transcription (9). Furthermore, both Ki-1/57 (6) and CGI-55 are localized in the nucleus, nucleolus, and other small nuclear bodies, and CGI-55 has been shown to co-localize to p80 coilinpositive nuclear coiled bodies, which have been functionally implicated in the regulation of transcription and the processing of RNA. 2 Furthermore, the other proteins identified to interact with CGI-55 in the yeast two-hybrid screen are in their majority nuclear proteins, and several of them, including the proteins Daxx (a Fas-binding protein), Topors (a topoisomerasebinding protein), and hPc2 (human polycomb 2), are like CHD3 involved in the regulation of transcription. 2 In fact we also had identified both Daxx and Topors as interacting partners of Ki-1/57 in our yeast two-hybrid screen (data not shown). These findings demonstrate that Ki-1/57 and CGI-55 have common interacting nuclear protein partners (CHD3, Daxx, and Topors) as well as specific interaction partners like RACK1 for Ki-1/57 and hPc2 for CGI-55. They further point to the possibility that both Ki-1/57 and CGI-55 might be involved in nuclear functions such as the remodeling of chromatin and the regulation of transcription, like several of its interacting nuclear protein partners.
In this context our observation of the nuclear exit of Ki-1/57 after stimulation of the cells with PMA may be of functional relevance. It has been shown recently that the activity of the chromatin-remodeling factor HDAC7 is regulated by its PMAinduced export from the nucleus (51,52). The combined PMA/ ionomycin treatment mimics the T cell receptor activation, and the PMA-induced nuclear export of HDAC7 was accompanied by a drop in a HDAC7-dependent Nur77 promotor activity, which controls a luciferase reporter gene. This demonstrates how the regulated nuclear export of a protein can affect the transcriptional regulation of genes. Because Ki-1/57 has been shown to interact with CHD3, another factor involved in chromatin remodeling and transcriptional regulation, it is tempting to speculate that the PMA-dependent nuclear export of Ki-1/57 could have functional consequences for CHD3s activity. While this manuscript was in the review process we became aware of a recent publication by Ozaki et al. (53). This group had found that RACK1 interacts with the C terminus of the p53 homologue protein p73. Most interestingly, Ozaki et al. were able to demonstrate that RACK1 inhibits both p73-mediated transcription from a test promoter as well as p73-mediated apoptosis. Future experiments will address whether and how Ki-1/57 and CGI-55 are involved in the regulation of transcription and what are the exact functions of these interesting novel proteins.