Identification and characterization of CKIP-1, a novel pleckstrin homology domain-containing protein that interacts with protein kinase CK2.

The catalytic subunits of protein kinase CK2, CK2alpha and CK2alpha', are closely related to each other but exhibit functional specialization. To test the hypothesis that specific functions of CK2alpha and CK2alpha' are mediated by specific interaction partners, we used the yeast two-hybrid system to identify CK2alpha- or CK2alpha'-binding proteins. We report the identification and characterization of a novel CK2-interacting protein, designated CKIP-1, that interacts with CK2alpha, but not CK2alpha', in the yeast two-hybrid system. CKIP-1 also interacts with CK2alpha in vitro and is co-immunoprecipitated from cell extracts with epitope-tagged CK2alpha and an enhanced green fluorescent protein fusion protein encoding CKIP-1 (i.e. EGFP-CKIP-1) when they are co-expressed. CK2 activity is detected in anti-CKIP-1 immunoprecipitates performed with extracts from non-transfected cells indicating that CKIP-1 and CK2 interact under physiological conditions. The CKIP-1 cDNA is broadly expressed and encodes a protein with a predicted molecular weight of 46,000. EGFP-CKIP-1 is localized within the nucleus and at the plasma membrane. The plasma membrane localization is dependent on the presence of an amino-terminal pleckstrin homology domain. We postulate that CKIP-1 is a non-enzymatic regulator of one isoform of CK2 (i.e. CK2alpha) with a potential role in targeting CK2alpha to a particular cellular location.

ate a broad range of cellular proteins located in a variety of cellular compartments (mainly the nucleus and cytoplasm) and is involved in important cellular processes such as transcription, translation, morphogenesis, and cell cycle progression (reviewed in Refs. [1][2][3][4][5][6][7]. These observations support an important role for CK2 in a variety of cellular functions; however, its specific roles and mode of regulation in cells remain poorly understood. Moreover, these results suggest that CK2 is involved in a complex array of interactions with a wide selection of cellular proteins that are present in a broad variety of cellular locations. CK2 is a tetrameric protein comprised of two regulatory subunits (CK2␤) and two catalytic subunits (CK2␣ and/or CK2␣Ј). CK2␣ and CK2␣Ј are the products of separate genes, and their amino acid sequences are highly conserved between higher eukaryotes (reviewed in Ref. 7). In fact, in mammals and birds, CK2␣ and CK2␣Ј exhibit greater than 90% identity over their 330 amino-terminal amino acids (7). This aminoterminal sequence identity is in stark contrast to the unrelated carboxyl-terminal sequences of CK2␣ and CK2␣Ј that exhibit no obvious similarity (8 -10). This sequence divergence between the carboxyl-terminal domains of CK2␣ and CK2␣Ј suggests that important functional differences that exist between the two different catalytic isozymes result from these unique sequences (11).
Previous studies have failed to demonstrate significant catalytic differences between CK2␣ and CK2␣Ј in vitro (12), a result that likely reflects the high similarity exhibited by the catalytic domains of CK2␣ and CK2␣Ј (12). By comparison, there is mounting evidence in support of functional specialization of CK2␣ and CK2␣Ј in cells (11,13). For example, in HeLa cells, differences in the localization of CK2␣ and CK2␣Ј have been observed, with further indications that the localization of both CK2␣ and CK2␣Ј may be altered in a cell cycle-dependent manner (13). However, these differences in subcellular localization may be cell type-specific, since major differences in the localization of CK2␣ and CK2␣Ј have not been observed in all cells where they have been examined (14,15). It has also been demonstrated that CK2␣, but not CK2␣Ј, is phosphorylated in mitotic cells of mammalian and avian origin (16), suggesting that the functions of the two isoforms are independently regulated during cell division. The mitotic sites of phosphorylation of CK2␣ were identified as Thr 344 , Thr 360 , Ser 362 , and Ser 370 , all of which are located within the carboxyl-terminal domain of CK2␣ (17). There is also a "PXXP" motif adjacent to two of the phosphorylation sites. Interestingly, "PXXP" motifs have been implicated in protein-protein interactions, most notably with SH3 domains of a variety of regulatory proteins (18,19).
Biochemical studies have localized CK2 activity within the cytoplasm, the nucleus, and with other cellular structures including the plasma membrane (1,2,20,21). Immunofluorescence studies have confirmed that CK2 is localized in the nucleus and in the cytoplasm (13-15, 22, 23). The factors or mechanisms that regulate the subcellular distribution of CK2 remain poorly understood. In this regard, it may be notable that there is mounting evidence demonstrating that CK2 interacts with a variety of cellular proteins. For example, CK2␤ has been reported to interact with FGF-2 (24), A-Raf (25,26), Nopp140 (27), p53 (28 -30), p21 WAF1/CIP1 (31), c-MOS (32), and CD5 (33). CK2␣ and CK2␣Ј have also been observed to interact with specific protein partners. For instance, both CK2␣ and CK2␣Ј have been shown to interact with nucleolin (34), ATF1 (35,36), whereas only CK2␣ has been shown to interact with HSP90 (37,38) and with PP2A (39). Interestingly, the domain of CK2␣ implicated in the interaction with PP2A was localized to the sequence 166 HEHRKL (human and chicken CK2␣), which is one of the few regions of non-identity between CK2␣ and CK2␣Ј (HQQKKL). Collectively, these observations suggest that the subcellular localization of CK2, and presumably its ability to phosphorylate a number of its target proteins, may be regulated by interactions with specific protein partners. The resulting compartmentalization of CK2 could provide a mechanism for regulating CK2 activity in cells. In fact, this mode of regulation could reconcile the observations that although CK2 is a predominantly nuclear enzyme, it can also phosphorylate cytoplasmic proteins, as well as nuclear proteins.
The possible regulation of CK2 through its interactions with specific protein partners in cells together with the observed functional differences between CK2␣ and CK2␣Ј led us to hypothesize that the catalytic subunits of CK2 are involved in isozyme-specific protein-protein interactions and that these interactions may be mediated through the unique carboxyl-terminal domains of CK2␣ or CK2␣Ј. To examine this hypothesis, we used a yeast two-hybrid system to screen an Epstein-Barr virus-transformed human B-cell library using GAL4 DNAbinding domain fusions encoding full-length CK2␣ or CK2␣Ј, as well as their respective carboxyl-terminal domains, as bait. In this study, we report the isolation and identification of a novel protein, designated casein kinase interacting protein 1 (i.e. CKIP-1), that binds CK2␣ but not CK2␣Ј. In addition to providing further indications of functional specialization for CK2␣ and CK2␣Ј, the identification of this novel protein may provide new insights for understanding how the access of CK2 to some of its cellular targets is regulated. Examination of the predicted amino acid sequence of CKIP-1 reveals the presence of a pleckstrin homology domain that could mediate interactions between CKIP-1 and cellular membranes or proteins as well as motifs (including a leucine zipper and PXXP) that could be involved in protein-protein interactions. Overall, our results raise the possibility that CKIP-1 is a non-enzymatic regulator of CK2 that could play a role in controlling the access of CK2␣ to specific cellular targets through its ability to sequester CK2 or to recruit CK2 to specific cellular locations.

Materials
Human osteosarcoma Saos-2 cells and SV40 large T-transformed green monkey kidney COS-7 cells were obtained from ATCC (Manassas, VA) and were maintained in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum (Life Technologies, Inc.). Anti-HA antibodies (i.e. 12CA5 monoclonal antibodies) were obtained from Babco; biotinylated rat anti-HA antibodies (clone 3F10) were obtained from Roche Molecular Biochemicals; horseradish peroxidase-conjugated mouse monoclonal anti-biotin antibodies were obtained from Jackson ImmunoResearch Laboratories Inc.; and anti-GFP antibodies were obtained from CLONTECH or were a generous gift from Dr. L. Berthiaume (University of Alberta). Rabbit anti-bodies against CK2␤ were described previously (40). Nitrocellulose and polyvinylidene difluoride for immunoblots as well as reagents for the colorimetric development of immunoblots were obtained from Bio-Rad. Reagents for enhanced chemiluminescence (ECL) were obtained from Amersham Pharmacia Biotech. All other reagents were of reagent grade.
To generate CK2␣-(332-391)(4D), the nucleotides coding for the amino acid residues representing the four mitotic sites of phosphorylation at amino acid residues (Thr 344 , Thr 360 , Ser 362 , and Ser 370 ) within CK2␣ (17) were mutated to replace each of the phosphorylation sites with aspartic acid. These mutations were engineered in order to mimic the charges brought on by phosphorylation of the four residues. By using two rounds of PCR amplification, we produced the CK2␣-(332-391)(4D) mutant. In the first round of PCR, two products were made using the following primers: 5Ј CCT GAG CTA CTT GTA GAC TA 3Ј (sense primer for the 1st product), 5Ј TGG CAC TGA AGA AAT CCC TGA CAT CAT ATT GGC GCT GCT GGG ATC ACT GCC CCC 3Ј (antisense primer for the 1st product that introduces the Thr 344 /Asp mutation), 5Ј GGG ATT TCT TCA GTG CCA GAT CCT GAT CCC CTT GGA CCT GCA GGC GAT CCA GTG 3Ј (sense primer for the 2nd product that introduces Thr 360 /Asp, Ser 362 /Asp, and Ser 370 /Asp mutations), and 5Ј TCC CCC ACC TCT GCT CAG 3Ј (antisense primer for the 2nd product). After amplification, the DNA fragments of interest were gel-purified, denatured, allowed to anneal, and then subjected to a second round of PCR using the sense primer from the 1st product in the first round of PCR and the antisense primer from the 2nd product in the first round of PCR as primers. The PCR products from the second round of amplification were directly cloned into the pCRII vector and sequenced. An EcoRI/SphI fragment from the resultant plasmid encoding the 4D variant of the carboxyl-terminal domain of CK2␣ was then used to replace a similar fragment in pGEX-CK2␣ (17). By using this construct as a template, EcoRI and BamHI sites were generated with PCR using the following primers: 5Ј CCG GAA TTC GCT CGA ATG GGT TCA TCT 3Ј (sense primer with EcoRI site) and 5Ј CGG ATC CAC CTC TGC TCA GGC ATC 3Ј (antisense primer with BamHI site). The plasmid pGBT9-CK2␣-(332-391) (4D) construct was derived from this construct using the strategy described above for pGBT9-CK2␣-(332-391) and pGBT9-CK2␣Ј-(333-350). The constructs were sequenced to confirm identity and to ensure that no mutations were introduced during amplification.
The two deletion constructs of CK2 (i.e. CK2␣-(2-330) and CK2␣-(2-350)) were generated using PCR. Briefly, both constructs were amplified using the same sense primer (5Ј GCC ATG GGA AAA AGT TGT TGT TAA AAT TCT C 3Ј) and the following antisense primers: 5Ј CCT CTA GAG TCG ACT CAC TGG TCC TTC ACA ACA GTG 3Ј which introduces a stop codon after residue 330 followed by a SalI site for CK␣-(2-330), or 5Ј GGT CTA GAG TCG ACT CAA TTG GCG CTG CTG ACG GG 3Ј which introduces a stop codon after residue 350 followed by a SalI site for CK2␣-(2-350). Amplified products were subcloned into pCR-Blunt, sequenced, and then digested with NcoI and SalI. The respective NcoI/SalI fragments were subcloned into SalI/NcoI-digested pGBT9-CK2␣ to replace the 3Ј-coding region of CK2␣. These constructs were transformed into yeast as described below to ensure that they did not autonomously activate transcription and to verify that the truncated CK2␣ hybrid proteins still exhibited interactions with pACTII-CK2␤. The latter test ensures that the deleted proteins maintain the overall integrity of CK2␣ that is required to interact with CK2␤.
Other Constructs-A strategy similar to that described in the preceding section was used to subclone the 1.5-kb BglII fragment and the 300-bp BamHI/BglII fragment into pGEX-3X to generate glutathione S-transferase (GST) fusion proteins encoding the entire open reading frame (designated GST-CKIP-1) and a GST fusion protein encoding the carboxyl-terminal 102 residues of CKIP-1 (designated GST-CKIP-1-(308 -409)). A plasmid encoding a GST fusion protein of CK2␣ was constructed by subcloning the BamHI fragment encoding full-length CK2␣ into the BamHI site of pGEX-3X. Constructs encoding CK2 subunits with HA epitope tags in pRc/CMV (Invitrogen) were previously described (14). Briefly, HA-CK2␣ and HA-CK2␣Ј encode amino-terminally tagged CK2␣ and CK2␣Ј, respectively, whereas CK2␣-HA encodes CK2␣ with a carboxyl-terminal HA tag. The construct designated HA-CK2␣Ј/␣ encodes an amino-terminally tagged chimera where the carboxyl-terminal domain of CK2␣ has been attached to the amino-terminal domain of CK2␣Ј as described (14).

Anti-CKIP-1 Antibodies
Anti-CKIP-1 antibodies were raised in rabbits by Babco (Richmond, CA) using a GST fusion protein encoding the carboxyl-terminal 102 residues of CKIP (i.e. GST-CKIP-(308 -409)) as the antigen. To characterize these antibodies, lysates were prepared from human osteosarcoma Saos-2 cells using Laemmli sample buffer and were analyzed on Western blots that were developed with pre-or post-immune serum from either of the two immunized rabbits in the presence or absence of GST-CKIP-(308 -409) or GST. Antiserum from either of the two immunized rabbits, but not the pre-immune serum, recognized a band of approximately 50 kDa in extracts of Saos-2 cells that was blocked by the inclusion of GST-CKIP-(308 -409), but not by GST, during the incubation with antibody. Affinity-purified anti-GST antibodies were obtained from this rabbit serum by passing the serum through a GST-affinity column that was made by coupling purified GST to Affi-Gel 10 (Bio-Rad) according to the manufacturer's recommendations. Once this serum had been depleted of anti-GST antibodies, it was passed through a GST-CKIP-1-(308 -409) affinity column that was similarly prepared to isolate affinity-purified anti-CKIP-1 antibodies. Antibodies were recovered from affinity columns as described (45).

Screening of Yeast Two-hybrid Positives for CK2␤
Since we had previously demonstrated that interactions between CK2␣ (or CK2␣Ј) and CK2␤ can be detected using the two-hybrid system (42), we utilized a PCR strategy to examine positive colonies from the screens performed using CK2␣ or CK2␣Ј as the bait for the presence of the cDNA encoding CK2␤. For this analysis, yeast were grown overnight at 30°C in liquid culture. Yeast were disrupted by vortexing with glass beads, and plasmid DNA was isolated as described previously (43). The primers used for PCR were 5ЈGCG GGG ATC CTG AGC AGC TCA GAG GAG 3Ј (sense primer located within the domain of CK2␤ that interacts with CK2␣) and 5ЈCTA CCA GAA TTC GGC ATG CCG GTA GAG GTG TGG TCA 3Ј (antisense primer located within the ADH-terminator of pACT) (46). Colonies that were positive for CK2␤ were not further analyzed.

Isolation of the CKIP-1 cDNA
After excluding all of the positives that encoded CK2␤, one positive from the screen with CK2␣ was further characterized. Plasmid DNA was obtained from yeast by glass bead preparation and electroporated into the KC8 strain of Escherichia coli. To select for those bacteria that harbored the pACT plasmid, bacteria were selected on ϪLeu plates. Plasmid DNA was then isolated and used to re-test for interactions with Gal4 DNA binding domain fusions of CK2␣ or CK2␣Ј in yeast, as described above. Transformations were also performed with a number of control constructs (as indicated). Plasmid DNA was also used as a template for sequencing manually using T7 polymerase or by automated sequencer (Perkin-Elmer).

Northern Blot Analysis
A Human Multiple Tissue Northern blot (CLONTECH) was probed as recommended in the accompanying manual. Probes that were labeled with 32 P were as follows: an XhoI fragment of the CKIP-1 cDNA, or the full-length cDNAs encoding CK2␣, CK2␣Ј or CK2␤.

In Vitro Binding Assay
A GST fusion protein encoding full-length CK2␣ or GST itself were expressed in bacteria and purified using glutathione-agarose as described previously (17). Purified GST-CK2␣ and GST were then coupled to Affi-Gel 10 (Bio-Rad) at a concentration of 4 mg/ml according to the manufacturer's recommendations. Radiolabeled ( 35 S-labeled) CKIP-1 was produced by in vitro transcription and translation using a TЈNЈT kit (Promega) with T7 polymerase according to manufacturer's recommendations. For this procedure, a PCR product encoding CKIP-1 was generated using a heat-stable DNA polymerase (Pfu obtained from Stratagene) from pACT-38 using the following primers: 5Ј TAA TAC GAC TCA CTA TAG GGA GAC CAC ATG GAT GAT GTA TAT AAC TAT CTA TTC 3Ј (sense primer) and 5Ј CTA CCA GAA TTC GGC ATG CCG GTA GAG GTG TGG TCA 3Ј (antisense primer). The former primer anneals within the activation domain of Gal4 that is encoded by pACT and introduces a T7 promoter to be used for in vitro transcription, whereas the latter primer anneals to a sequence within the ADH terminator of pACT. Radiolabeled CKIP-1 (5 l of the reticulocyte lysate reaction) was incubated for 1 h at 4°C with 5 l of Affi-Gel 10-GST beads in a total volume of 50 l in interaction buffer (20 mM Hepes, pH 7.4, 150 mM NaCl, 10% glycerol, 0.05% Nonidet P-40) supplemented with protease inhibitors (0.1 mM phenylmethylsulfonyl fluoride, 1% aprotinin). After the 1-h incubation, the Affi-Gel beads were collected by centrifugation, and the supernatant was removed, and the beads were washed three times with the interaction buffer. Proteins retained on the Affi-Gel beads were eluted by the addition of Laemmli sample buffer and were subjected to SDS-polyacrylamide gel electrophoresis. Radiolabeled proteins were visualized using a PhosphorImager (Molecular Dynamics).

Transfection of COS-7 and Saos-2 Cells
Cells were grown to 80% confluence and then split 1:8 into 100 ϫ 20-mm culture plates or 35 ϫ 20-mm wells containing coverslips. The following day, the media were changed 2-5 h prior to transfection. Transfections were carried using calcium phosphate as described (44) or using Fugene 6 (Roche Molecular Biologicals) as described by the manufacturer. Plasmids encoding various pEGFP constructs were added at concentrations of 5 g per 100 ϫ 20-mm culture plates or 1-2 g per 35 ϫ 20-mm wells of a 6-well flat bottom tissue culture plate. The various pRc/CMV vectors encoding CK2 constructs were added at concentrations of 15-25 g per 100 ϫ 20-mm culture plate or 5-10 g per 35 ϫ 20-mm well of a 6-well flat bottom tissue culture plate. For calcium phosphate precipitations, precipitates of DNA were left on cells for 16 -18 h after which time the cells were washed thoroughly with PBS, pH 7.4 -7.5, to remove the precipitate. Media was added to cells which were incubated a further 24 h. Cells were then washed thoroughly with PBS, and the liquid was removed. At this time, cells grown on coverslips were mounted onto glass slides for fluorescence microscopy, whereas cells on the plates were frozen at Ϫ80°C until needed.

Immunoprecipitation
Immunoprecipitations were carried out in a 1:1 mixture comprised of transfected cell extract (described above) and binding buffer (200 mM NaCl, 20% glycerol, 1% Triton X-100). Anti-HA immunoprecipitations were performed using clone 12CA5 anti-HA antibodies (1:250), and anti-CKIP-1 immunoprecipitations were performed using anti-CKIP-1 (1/250). To minimize interference on immunoblots from the presence of the immunoglobulin heavy chain, these antibodies were initially crosslinked to protein A-Sepharose using dimethylpimelimidate as described (45). Samples were rocked on ice for 1 h. Sepharose beads were then washed with 3 ϫ 500 l of a 1:1 mixture of lysis buffer/binding buffer and then resuspended in Laemmli sample buffer. Samples were analyzed by SDS-polyacrylamide gel electrophoresis followed by Western blot analysis. Relevant proteins were detected using either anti-GFP antibody (1/2000), anti-CKIP-1 antibody (1/1000), or biotinylated anti-HA (1/500) as primary antibody and then appropriate secondary antibodies conjugated to horseradish peroxidase followed by detection with enhanced chemiluminescence. Since CK2␣-HA has a molecular weight of approximately 48,000 that is similar to that of the immunoglobulin heavy chain, the biotinylated anti-HA antibodies (clone 3F10) were used for immunoblots to detect CK2␣-HA without any interference from the immunoglobulin heavy chain.
Immunoprecipitations of endogenous CKIP-1 from non-transfected Saos-2 cells were performed by extracting Saos-2 cells with RIPA buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS) supplemented with protease inhibitors (10 g/ml leupeptin, 0.1 mM phenylmethylsulfonyl fluoride, 1% aprotinin) (45). These extracts were clarified by centrifugation as described above and incubated with anti-CKIP-1 antibodies (antiserum at 1:250 or 5 g of affinity purified antibody) for 2 h at 4°C. Since it was apparent that the anti-CKIP-1 antibodies recognized a protein in cell extracts of approximately 50 kDa, antibodies were initially cross-linked to protein A-Sepharose using dimethylpimelimidate as described (45) to minimize interference from the immunoglobulin heavy chain. Immunoprecipitates were washed 3 times with RIPA buffer and once with 20 mM Tris, pH 7.5, before analysis on Western blots using anti-CKIP-1 antibodies. For these experiments, blots were developed colorimetrically using 5-bromo-4-chloro-3-indolyl phosphate and nitro blue tetrazolium as the substrates for alkaline phosphatase-conjugated secondary goat antirabbit antibodies.

Immune Complex Kinase Assays
Immunoprecipitations were performed as described above with the exception that cells were lysed with 1% Nonidet P-40 in PBS instead of RIPA buffer. Immunoprecipitates were washed twice with 1% Nonidet P-40 in PBS and then twice with 50 mM Tris, 150 mM NaCl. Assays to measure CK2 activity were performed for 10 -20 min at 30°C with gentle shaking in a 30-l reaction containing 50 mM Tris-Cl, pH 7.5, 150 mM NaCl, 10 mM MgCl 2 , 1 mM dithiothreitol, 20 M ATP (specific activity approximately 0.01 Ci/pmol) using the synthetic peptide RRRDDDSDDD (0.1 mM) as a specific substrate for CK2 (47). Reactions were initiated by the addition of immune complex and were terminated by the addition of EDTA to a concentration of 25 mM. Reactions were centrifuged briefly to pellet immune complexes, and half of the supernatant was spotted on P81 paper to measure incorporation of 32 P into synthetic peptide as described previously (47). All samples were assayed in the presence or absence of synthetic peptide to control for background phosphorylation of proteins in the immune complexes. Heparin (50 g/ml) or GTP (100 M) were added as indicated.

Visualization of EGFP Fusion Proteins in Transfected Cells
To examine co-distribution of EGFP-CKIP-1 with the plasma membrane, transfected COS-7 cells were stained with rhodamine-labeled concanavalin A (Sigma), a lectin that binds to glycoproteins on the surface of cells. To examine co-distribution of EGFP-CKIP-1 with endogenous CK2, transfected Saos-2 cells were subjected to indirect immunofluorescence using anti-CK2␣ antibodies and rhodamine-labeled goat anti-rabbit antibodies as described (15). For indirect immunofluorescence, cells on coverslips were washed twice with PBS and were then fixed for 20 min with 3% paraformaldehyde in PBS prior to permeabilization for 5 min with 1% Triton X-100 in PBS. The fixed and permeabilized cells were incubated for 30 min with anti-CK2␣ antibodies and then with rhodamine-conjugated goat anti-rabbit antibodies as described (15). EGFP fluorescence and rhodamine fluorescence were visualized with a Zeiss LSM 410 inverted confocal microscope. The images were pseudocolored using built-in LSM software and overlaid.

RESULTS
Yeast Two-hybrid Screen-In order to identify protein-binding partners of CK2␣ or CK2␣Ј, we performed independent yeast two-hybrid screens of a human B-cell cDNA library using constructs encoding full-length CK2␣ and CK2␣Ј as well as the carboxyl-terminal domains of the respective proteins as bait (Fig. 1A). The carboxyl-terminal domains of CK2␣ or CK2␣Ј were each utilized in screens in order to examine the possibility that these unique carboxyl-terminal domains were sufficient for interactions with binding partners. Screens with each of these constructs were performed simultaneously. Neither of the carboxyl-terminal domain constructs yielded any positive clones (data not shown). However, using the full-length CK2␣ and CK2␣Ј, we obtained 243 and 196 positives, respectively. We expected that a number of the positives could be CK2␤, the regulatory subunit of CK2 that is known to interact strongly with both CK2␣ and CK2␣Ј in yeast two-hybrid assays (42,46). Therefore, we used a PCR screening procedure to test positive colonies for the presence of the CK2␤ cDNA being expressed as a fusion with the transcriptional activation domain of GAL4. The results from the PCR analysis demonstrated that the majority of the positives contained CK2␤, a result that validates the utility of the yeast two-hybrid system for the identification of binding partners for the catalytic subunits of CK2. One colony (designated 38) derived from the screen with CK2␣ did not show any evidence of CK2␤. Consequently, the pACT library plasmid from colony 38 was isolated, as described under "Experimental Procedures," and subjected to further examination.
Specificity of Interaction of pACT-38 with CK2␣-In order to determine whether the protein encoded by the pACT plasmid in colony 38 (designated pACT-38) encoded a protein able to interact specifically with CK2␣, we re-examined its ability to interact with CK2␣ as well as a variety of other proteins in a yeast two-hybrid assay (Fig. 1B). Positive interactions involving pACT-38 were only detected when it was co-expressed with CK2␣. Despite the high degree of similarity between CK2␣ and CK2␣Ј, we did not observe any interaction with CK2␣Ј. This latter result is consistent with the absence of this cDNA arising from the screen that was performed with CK2␣Ј. In these reconstruction assays, the integrity of the CK2␣ and CK2␣Ј constructs was confirmed by testing them successfully for interactions with pACT-CK2␤ (data not shown). The interaction of the pACT-38-encoded protein with CK2␣, but not with CK2␣Ј, suggests that differences existing between CK2␣ and CK2␣Ј are sufficient to confer interaction specificity. Moreover, Novel PH Domain-containing CK2-interacting Protein the observation that the pACT-38-encoded protein was not able to interact with either the CK2␣-(332-391) or CK2␣Ј-(333-350) in the yeast two-hybrid system indicates that the carboxylterminal domains by themselves are not sufficient to mediate the interaction between CK2␣ and the protein encoded by pACT-38, which we subsequently designated CKIP-1 (i.e. CK2 interacting protein-1).
To extend further the analysis of interactions between CK2␣ and the protein encoded by pACT-38, we were interested in determining whether the unique carboxyl-terminal domain of CK2␣ is required for interactions. Consequently, we con-structed two deletions of CK2␣, CK2␣-(2-330) and CK␣-(2-350). As illustrated in Fig. 1A, both of these deletion constructs contain all of the subdomains that are found in all protein kinase family members. Moreover, to residue 330 of CK2␣, the two isoforms of CK2 exhibit approximately 90% identity, and there is no similarity between CK2␣ and CK2␣Ј beyond that point. The former construct (i.e. CK2␣-(2-330)) completely removes all of the unrelated residues, whereas the latter construct (i.e. CK2␣-(2-350)) generates a form of CK2␣ that is the same length as CK2␣Ј. In preliminary tests, these deletion constructs failed to activate transcription autonomously indicating that they are suitable for use in two-hybrid assays (data not shown). Furthermore, both constructs exhibited interactions with CK2␤ in two-hybrid assays suggesting that the overall structural integrity of CK2␣ had been maintained (data not shown). When these constructs were tested for interactions with pACT-38, their capacity for interactions is similar to that exhibited by CK2␣ (Fig. 1C). This result suggests that although the major difference between CK2␣ and CK2␣Ј lies within their carboxyl-terminal domain, the carboxyl-terminal domain of CK2␣ is not required for interactions with CKIP-1. It is noteworthy that protein phosphatase 2A also exhibits interactions with CK2␣ and not with CK2␣Ј and that the residues important for this interaction are within the amino-terminal 330 residues of CK2 and not within the carboxyl-terminal domain (39). More detailed future analysis of CK2␣ will be required to identify the specific residues of CK2␣ that are important for the selective interactions with CKIP-1.
Interactions between CKIP-1 and CK2␣ in Vitro-The ability of CKIP-1 to interact directly with CK2␣ was tested in vitro using purified GST fusion proteins and the products of in vitro transcription and translation. In Fig. 2A, we detected a major radiolabeled translation product of approximately 50 kDa in the lysate containing the pACT-38 DNA (lane designated as 38). This band is not present in the control lysate which did not contain any pACT-38 DNA (lane C). By using GST-CK2␣ and GST proteins as affinity matrices in pull-down assays, it was evident that the CKIP-1 product interacts with GST-CK2␣ but FIG. 1. Isolation of CK2-interacting proteins using the yeast two-hybrid system. A, to isolate cDNAs encoding CK2-interacting proteins, cDNAs encoding full-length CK2␣ or CK2␣Ј were expressed in yeast as fusions with the DNA binding domain (BD) of GAL4 using the plasmid pGBT9 as described under "Experimental Procedures." Constructs encoding GAL4 DNA binding domain fusions encoding truncated forms of CK2␣ (i.e. CK2␣-(2-350) and CK2␣-(2-330)) as well as fusions encoding the unique carboxyl-terminal domains of CK2␣ (i.e. CK2␣-(332-391)) and CK2␣Ј (i.e. CK2␣Ј-(333-350)) were also prepared. The closely related amino-terminal domains of CK2␣ and CK2␣Ј that contain all of the subdomains found in protein kinase family members are indicated in solid gray. The unique carboxyl-terminal domains of CK2␣ and CK2␣Ј are denoted with solid black shading or with hatched bars, respectively. B, the pACT plasmid from one positive that did not contain CK2␤ (designated pACT-38) was transformed into yeast to test for interactions with a variety of independent GAL4-DNA binding domain fusions that were expressed using pGBT9 or pAS1 including pGBT9-huntingtin (82), pGBT9-fragmentin (83), and pAS1-RAD7 (84). Positive interactions were indicated by the ability of transformants to grown on media deficient in Trp, Leu, His, and Ade (designated ϪHALT). C, the plasmid designated pACT-38 was transformed into yeast to test for interactions with pGBT9 constructs encoding fulllength CK2␣ or CK2␣Ј as well as a construct encoding a deletion of CK2␣ that shortens CK2␣ to the same length as CK2␣Ј (i.e. CK2␣-(2-350)) and a deletion of CK2 that completely removes its unique carboxyl-terminal domain (i.e. CK2␣-(2-330)). As in B, positive interactions were indicated by growth on selective media.

FIG. 2. In vitro interactions between CK2␣ and CKIP-1.
In vitro transcription/translation reactions were performed using a reticulocyte lysate-based system with a PCR product derived from pACT-38 (designated 38) as template or a control reaction (designated C) lacking this PCR product. A, 35 S-labeled reaction products from the in vitro transcription/translation reactions were subjected to electrophoresis on 12% SDS-polyacrylamide gels and visualized using a PhosphorImager. The major translation product of approximately 50,000 that was obtained using pACT-38 DNA as template is indicated by an arrow. This band is absent in the control reaction (designated C). The positions of molecular weight markers are also indicated. In vitro translation products were incubated with GST-CK2␣ or GST coupled to Affi-Gel beads as described under "Experimental Procedures." B, radiolabeled proteins that were retained on either of the respective Affi-Gel beads were subjected to SDS-polyacrylamide gel electrophoresis and visualized as in A. C, approximately 25% of each supernatant was similarly analyzed. not with GST (Fig. 2B). Quantitation of the amount of radioactivity that was recovered on Affi-Gel beads (Fig. 2B) or remained in the supernatant (Fig. 2C) indicated that approximately 50% of the CKIP-1 was retained on the GST-CK2␣, whereas the amount of CKIP-1 that was retained on GST was negligible. A lower band (i.e. below the 47.5K marker) that is most evident in the lysate ( Fig. 2A) and in the GST-CK2 pull down (Fig. 2B) could either be an alternative translation product of CKIP-1 resulting perhaps from a different start codon or the band could be a proteolytic degradation product of CKIP-1. However, the precise identity of this band is not known. As was the case with GST, in vitro translated CKIP-1 exhibited negligible interactions with GST-CK2␤ (data not shown). Taken together, the results from the yeast two-hybrid assays and the in vitro interaction studies suggest that the CKIP-1 protein is capable of interacting specifically and directly with CK2␣. We also demonstrated that in vitro translated CK2␣ could interact with GST-CKIP-1 (data not shown), further confirming the in vitro interactions between the two proteins.
Nucleotide and Deduced Amino Acid Sequences of CKIP-1-The nucleotide sequence of CKIP-1 was determined by sequencing the 1469-bp CKIP-1 insert located within the BglII sites of pACT. We discovered a large open reading frame (ORF), encompassing nucleotides 1-1391, which was in frame with the GAL-4 activation domain of pACT. The ORF was followed by a TGA stop codon at nucleotides 1392-1394. Near the 5Ј end, we also observed two successive putative ATG start codons starting at nucleotide 165, in which the second ATG more closely resembles a Kozak consensus sequence (48). The deduced amino acid sequence of the ORF (bp 165-1391) (Fig. 3) encodes a protein of 409 amino acids with a theoretical molecular mass of 46,236.93 Da. Notably, this predicted size is consistent with that observed for the in vitro translation product observed previously (Fig. 2).
Screening of the nucleotide sequence of CKIP-1 with the "BLAST" programs identified one significant, partial match. The carboxyl-terminal domain of CKIP-1 (encompassing amino acids 338 -409 of the predicted amino acid sequence) displayed 95% identity with the predicted amino acid sequence from a partial cDNA encoding an ORF of 72 amino acids that was identified as a putative c-Jun leucine zipper interactive mouse protein (cDNA JZA-20/pir B46132) (50). This ORF does contain a sequence that resembles a leucine zipper (shown in Fig. 3) that may explain why it exhibited the capacity to interact with the leucine zipper of c-Jun. Isolation of a full-length cDNA and further characterization of the protein encoded by JZA-20 was not performed (50). Consequently, the significance of interactions between JZA-20 and c-Jun remains undefined. Analysis of the deduced amino acid sequence of CKIP-1 by "ProfileScan" and "Pfam" demonstrated the presence of a pleckstrin homology (PH) domain between residues 21 and 132 of the deduced protein sequence (Fig. 3).
Northern Blot Analysis-To determine whether we had isolated a cDNA encoding the full-length CKIP-1 protein and to examine its expression patterns, we probed a human multiple tissue northern blot (CLONTECH) using a fragment of the CKIP-1 cDNA. This probe detected a major band of approximately 1.5 kb as well as an additional band of significantly lower intensity at 4.4 kb. The overall intensity signal of the 1.5-kb band of CKIP-1 was highest in the skeletal muscle and heart lanes. Intermediate signals were observed in brain, placenta, and lung lanes, whereas weaker signals were seen in the liver, kidney, and pancreas lanes. We also probed the same Northern with CK2␣, CK2␣Ј, and then CK2␤ (Fig. 4, B-D,  respectively), in order to compare expression levels of the different mRNAs throughout the various human tissues. Al-though the expression patterns between CKIP-1, CK2␣, and CK2␣Ј were quite similar, there were a few noticeable differences. The signal for CK2␣ (Fig. 4B) was weaker in the lung lane but stronger in both the kidney and pancreas lanes as compared with the CKIP-1 blot (Fig. 4A). The signal for CK2␣Ј (Fig. 4C) was much stronger in the pancreas lane and stronger in the placenta lane. The significance of the differences is unclear; however, it is clear is that we could easily detect mRNA of CKIP-1 as well as CK2␣ and CK2␣Ј in a variety of tissues. CK2␤ mRNA was readily detected in a fairly uniform fashion in all tissues, except for an apparently lower signal detected in lung. This Northern blot analysis suggested that the size of the CKIP-1 mRNA is about 1.5 kb, which is consistent with the size of the CKIP-1 cDNA isolated from pACT.
To confirm that we indeed possessed a cDNA encoding for the full-length CKIP-1 protein, we submitted the CKIP-1 nucleotide sequence to a BLAST against tentative human consensus sequences. This enabled us to identify an expressed sequence tag (EST) with a reported insert size of 1.7 kb that we obtained from ATCC (410320/H14297). Results of the nucleotide sequencing of this clone (not shown) demonstrated that a large central region of the EST, encompassing the entire 409 amino acid ORF, was completely identical to the nucleotide sequence of the CKIP-1. The major differences between CKIP-1 and EST H14297 are the existence of sequence extensions at both the 5Ј and 3Ј ends of the EST sequence. Importantly, at the 5Ј end of EST H14297, we observed an in-frame TGA stop codon, with no intervening ATG codons between this stop codon and the pu- The predicted amino acid sequence of CKIP-1 was determined by sequencing the insert of pACT-38 and was confirmed by sequencing a cDNA that was represented as an expressed sequence tag (EST) as described in the text. PXXP motifs are underlined, and serine or threonine residues that are within PX(S/T)P motifs that conform to the consensus for phosphorylation by mitogen-activated protein kinases or cyclin-dependent kinases are marked with asterisks. B, schematic representation of CKIP-1 illustrates the presence of an amino-terminal PH domain and also shows the region of similarity to that of the deduced amino acid sequence obtained from a partial cDNA previously identified as a c-Jun leucine zipper interacting protein (50). Residues that form a putative leucine zipper are underlined.
tative start codon(s) of the ORF that we had identified in CKIP-1. At the 3Ј end of EST H14297, we identified an AT-rich region that was not observed in the CKIP-1 nucleotide sequence. Taken together, these results support the conclusion that a cDNA that encoded the full-length CKIP-1 protein had been isolated.
Detection of a 50-kDa Protein with Anti-CKIP-1 Antiserum-In order to study the CKIP-1 protein, polyclonal rabbit antibodies directed against a GST fusion protein encoding the carboxyl-terminal 102 amino acids of CKIP-1 (i.e. GST-CKIP-1-(308 -409)) were raised as described under "Experimental Procedures." By immunoblot analysis, anti-CKIP-1 antibodies detect a protein of approximately 50 kDa in lysates and in anti-CKIP-1 immunoprecipitates (Fig. 5A) that is not detected in anti-GST immunoprecipitates. Importantly, the 50-kDa band was not evident when immunoprecipitations were performed with pre-immune sera from either of the rabbits (data not shown). Moreover, the 50-kDa band is effectively competed away when Western blots with anti-CKIP-1 sera are performed in the presence of GST-CKIP-1-(308 -409) (Fig. 5B) but not when these blots are performed in the presence of GST (not shown). Overall, these results support the conclusion that the anti-CKIP-1 antibodies specifically recognize a protein of 50 kDa. Since the apparent molecular weight of this band is very similar to that predicted for the 409-amino acid ORF of CKIP-1, we believe that the 50-kDa protein is endogenous CKIP-1.
Measurement of CK2 Activity in Anti-CKIP-1 Immunoprecipitates-Immunoprecipitates of CKIP-1 were subsequently analyzed for kinase activity toward the specific CK2 substrate peptide RRRDDDSDDD (64) to determine whether there are interactions between CK2 and CKIP-1 in mammalian cells. As illustrated in Fig. 6, immunoprecipitates performed with anti-CKIP-1 antiserum derived from either of two immunized rabbits displayed CK2 activity that was approximately 10-fold above the background activity that was observed in immunoprecipitates performed with pre-immune serum derived from the same rabbits. Kinase activity toward the CK2 substrate peptide was nearly completely abolished by heparin, a known inhibitor of CK2 (7). Furthermore, inclusion of GTP in the kinase activity markedly diminished 32 P incorporation into the substrate peptide. Since CK2 is one of the few protein kinases capable of utilizing GTP as substrate, the latter observation is consistent with the conclusion that the kinase being measured in anti-CKIP-1 immunoprecipitates is CK2. Collectively, these results provide strong evidence for interactions between CK2 and CKIP-1 in mammalian cells. However, we cannot rigorously exclude the possibility that kinases distinct from CK2, but with similar properties to CK2, may have contributed to the activity measured in anti-CKIP-1 immunoprecipitates. The CK2 activity that was measured in anti-CKIP-1 immunoprecipitates represents a relatively low percentage (less than 4%) of the activity that was measured in anti-CK2␤ immunoprecipitates. It is important to note that this estimate is based on the kinase activities that were measured in immune complex kinase assays performed using antibodies (anti-CKIP-1 or anti-CK2␤) that may exhibit differences in immunoprecipitation efficiency. Nevertheless, these results suggest that CKIP-1⅐CK2 complexes represent a rather small fraction of total cellular CK2.
Expression of EGFP Fusion Proteins Encoding CKIP-1-Based on the evidence from co-immunoprecipitation assays that CKIP-1 and CK2 do indeed interact in mammalian cells, we were interested in examining the subcellular distribution of FIG. 5. Immunoblot analysis of anti-CKIP-1 immunoprecipitates. Extracts were prepared from human osteosarcoma Saos-2 cells using RIPA buffer as described under "Experimental Procedures" and were subjected to immunoprecipitation with affinity-purified anti-CKIP-1 antibodies or with affinity-purified anti-GST antibodies. Immunoprecipitates and the cell lysate were subjected to electrophoresis on 12% SDS-polyacrylamide gels and analyzed by immunoblotting using anti-CKIP-1 antiserum (A) or using anti-CKIP-1 antiserum that had been preincubated with the immunogen GST-CKIP-1-(308 -409) (B). A band of approximately 50 kDa that is present in the lysate and in the anti-CKIP-1 immunoprecipitate is marked with an arrow in A. This band is not detected in B in the presence of the competing antigen. The immunoglobulin heavy chain is detected as a diffuse band in all of the immunoprecipitates and is marked (H). Molecular weight markers are also indicated. Immune complexes were detected by colorimetric detection as described under "Experimental Procedures." CKIP-1 since we speculated that CKIP-1 could function as a protein that recruits CK2 to a particular cellular location. To achieve this objective, we prepared an expression construct encoding CKIP-1 as an EGFP fusion protein using pEGFP-C2, a plasmid that allows for the expression of fusion proteins at the carboxyl terminus of EGFP. An EGFP fusion protein encoding the carboxyl-terminal 102 amino acids of CKIP-1 was also constructed using an internal BamHI site within the CKIP-1 cDNA. The latter fusion protein lacks the PH domain that is located near the amino terminus of CKIP-1 but retains the putative leucine zipper. To effect a more precise deletion of the PH domain of CKIP-1, we also generated a construct encoding EGFP-CKIP-1-(133-409) as described under "Experimental Procedures." As a prelude to examining the localization of these protein within cells, we examined their expression by Western blot analysis of transfected cell lysates using anti-CKIP-1 antiserum or anti-GFP antiserum (Fig. 7). As expected, anti-GFP antibodies (Fig. 7) detected EGFP and each of the fusion proteins encoding CKIP-1 or deletions of CKIP-1 (lanes 2-5). The anti-CKIP-1 antibodies detect bands in lanes 2-4 that appear to be identical to the bands that are detectable with anti-GFP but do not detect GFP (lane 5). The low levels of endogenous CKIP-1 that are detected in COS-7 cells are not visible at the levels of detection that are represented in this figure.
Co-immunoprecipitation of EGFP-CKIP-1 and CK2␣-HA-To characterize further the EGFP-CKIP-1 fusion protein, we examined its ability to interact with CK2 in mammalian cells using co-immunoprecipitation assays. To achieve this objective, we co-transfected cells with a construct encoding EGFP-CKIP-1 and a construct encoding epitope-tagged CK2␣ (i.e. CK2␣-HA with a carboxyl-terminal HA tag) or with an empty vector (i.e. pRc/CMV). Anti-HA immunoprecipitates were performed with lysates obtained from these cells and examined on Western blots with anti-CKIP-1 antiserum (Fig.  8A). Alternatively, immunoprecipitates performed with anti-CKIP-1 serum were analyzed with biotinylated anti-HA antibodies on Western blots (Fig. 8B). In Fig. 8A, the EGFP-CKIP-1 fusion protein can be visualized in a transfected cell lysates (marked lysate). As in Fig. 7, this band was not observed in a lysate from non-transfected cells (not shown). The EGFP-CKIP-1 protein is also present in anti-HA immunoprecipitates (Fig. 8A) derived from cells co-transfected with EGFP-CKIP-1 and CK2␣-HA but not in immunoprecipitates from cells transfected with EGFP-CKIP-1 and the empty vector (marked pRc/ CMV). In Fig. 8B, a band corresponding to the CK2␣-HA is detected exclusively in lysates of cells transfected with CK2␣-HA and in anti-CKIP-1 immunoprecipitates. Biotinylated anti-HA antibodies were utilized to eliminate the appearance of the immunoglobulin heavy chain that confounded the detection of CK2␣-HA on immunoblots in a number of preliminary experiments. With these biotinylated antibodies, nonspecific bands are observed in all lysates, but since their migration on polyacrylamide gels is significantly different than that of CK2␣-HA, they do not interfere with its detection. Moreover, these nonspecific bands are not observed in immunoprecipitates. Overall, the results of the co-immunoprecipitation assays suggest that the EGFP-CKIP-1 protein retains the capacity to interact with CK2␣ in mammalian cells.
Subcellular Localization of EGFP-CKIP-1 Constructs by Fluorescence Microscopy-To examine the subcellular localization of CKIP-1 in mammalian cells, we examined COS-7 cells that had been transfected with constructs encoding EGFP, EGFP-CKIP-1, EGFP-CKIP-1-(133-409), or EGFP-CKIP-1-(308 -409) by fluorescence microscopy. With EGFP, we observed general cellular fluorescence that included prominent nuclear fluorescence (Fig. 9D). By comparison, EGFP-CKIP-1 does not exhibit the general cellular fluorescence and appears to localize chiefly to discrete areas of the cell consistent with plasma membrane and/or membrane ruffles (Fig. 9A). The deletion mutant, EGFP-CKIP-1-(308 -409), displays a general cytosolic distribution similar to that observed with EGFP ( Fig. 9G) but FIG. 6. Measurement of CK2 activity in anti-CKIP-1 immunoprecipitates. A, extracts were prepared from human osteosarcoma Saos-2 cells using Nonidet P-40 lysis buffer and were subjected to immunoprecipitation with antiserum from either of 2 rabbits that were immunized with GST-CKIP-1-(308 -409) (designated 1 and 2) or with pre-immune serum obtained from either of the same rabbits (designated as 1(pre) and 2(pre)). Immunoprecipitates were utilized in kinase assays using the specific CK2 substrate peptide RRRDDDSDDD to measure CK2 activity. Incorporation of [ 32 P]phosphate into synthetic peptide from [ 32 P]ATP was determined by P81 filter paper assay as does not appear to exhibit any membrane fluorescence. Similar observations are made with EGFP-CKIP-1-(133-409), a mutant that represents a precise deletion of the PH domain of CKIP-1 (data not shown). Overall, these results suggest that CKIP-1 contains information that is able to affect the cellular distribution of EGFP. It also appears that the amino-terminal domain of CKIP-1 that contains the PH domain is required for the plasma membrane staining that is observed. To characterize further the putative membrane localization of EGFP-CKIP-1, transfected cells were also stained with rhodamineconjugated concanavalin A, a lectin known to interact with cell surface glycoproteins. EGFP fluorescence and rhodamine fluorescence were detected by confocal microscopy. Examination of the fluorescence patterns for EGFP-CKIP-1 (Fig. 9A) and rhodamine concanavalin A (Fig. 9B) reveals a similar pattern of staining that is clearly evident when optical slices of these images are overlaid (Fig. 9C). Prominent yellow staining that is observed in the overlay demonstrates that a large proportion of the EGFP-CKIP-1 is co-distributed with the rhodamine concanavalin A, a result that is consistent with the conclusion that some, but not all, of the EGFP-CKIP-1 is localized to the plasma membrane. In comparison to EGFP-CKIP-1, the vast majority of the EGFP and EGFP-CKIP-1-(308 -409) are not co-distributed with the concanavalin A.
We also examined the localization of EGFP-CKIP-1 fusion proteins in human osteosarcoma Saos-2 cells because these cells exhibit higher level of endogenous CKIP-1 than do COS-7 cells. In Saos-2 cells, EGFP-CKIP-1 exhibits prominent staining of the cell periphery as is the case in COS-7 cells. Moreover, a proportion, but not all, of the transfected Saos-2 cells also exhibited prominent nuclear localization of CKIP-1 (Fig. 10B). Based on the prediction that CKIP-1 could be a non-enzymatic regulator of CK2 that controls the cellular activity of CK2 through its ability to sequester CK2 or to target CK2 to specific cellular locations, we were also interested in determining where CK2 and CKIP-1 exhibit co-distribution within cells. For these experiments, endogenous CK2 was examined in these cells by indirect immunofluorescence. Fluorescent proteins were detected by confocal microscopy. As expected, in Saos-2 cells (Fig. 10A) and also in COS-7 cells (not shown), CK2 exhibits a predominantly nuclear localization with some non-nuclear staining. As noted above, in Saos-2 cells, EGFP-CKIP-1 was detected at the cell periphery and in the nucleus (Fig. 10B). Co-distribution between EGFP-CKIP-1 and CK2 in the nucleus is clearly illustrated by the presence of the yellow color that is seen in the overlay (Fig. 10C). However, with the resolution of these techniques, we could not conclusively determine whether or not a small proportion of the CK2 exhibited co-distribution with the EGFP-CKIP-1 that is at the cell periphery. Similarly, in COS-7 cells where EGFP-CKIP-1 is predominantly localized to the cell periphery, we could not conclusively determine whether EGFP-CKIP-1 and CK2 are co-distributed. In light of the predominant nuclear localization of CK2 and the undoubted complexity of interactions between CK2 and its many cellular targets (including substrates), these latter observations are perhaps not surprising. DISCUSSION In this study, we report the identification of a novel protein of unknown function that we have designated CKIP-1 on the basis that we initially identified it as a CK2-interacting protein. This protein was originally isolated as a CK2-interacting protein using the yeast two-hybrid system, but interactions between CK2 and CKIP-1 were also observed in vitro using GST pull-down assays and using co-immunoprecipitation assays from both transfected and non-transfected cells. Furthermore, using the yeast two-hybrid system we have observed that CKIP-1 is able to interact with the CK2␣, but not the CK2␣Ј, isoform of protein kinase CK2 suggesting that CKIP-1 exhibits isoform specificity. Based on this observation, it is tantalizing to speculate that the isoform specificity of CKIP-1 may in part contribute to the functional specialization that has been observed for CK2␣ and CK2␣Ј (11,13,51).
Our results from yeast two-hybrid assays also suggest that, although the major difference between CK2␣ and CK2␣Ј lies within their unrelated carboxyl-terminal domains, this domain does not appear to be required for interactions between CK2␣ and CKIP-1. It is noteworthy that protein phosphatase 2A also exhibits isoform specificity and interacts with CK2␣ but not CK2␣Ј (39). Interestingly, the residues that are critical for interactions between protein phosphatase 2A and CK2␣ reside within the amino-terminal region of CK2␣ and do not involve its unique carboxyl-terminal domain. Future investigation will be required to identify the precise residues within the aminoterminal 330 amino acids of CK2␣ that are involved in interactions with CKIP-1.
On the basis that our two-hybrid screens to identify proteins that interact with the catalytic subunits of CK2 yielded predominantly the regulatory ␤ subunit of CK2, our screens must be classified as successful. Furthermore, since the CKIP-1 cDNA was isolated from the screen using the same criteria that yielded CK2␤, it is not surprising that CKIP-1 appears to be a physiologically relevant CK2-interacting protein. In utilizing the yeast two-hybrid system to identify proteins that interact with CK2, we anticipated that we could isolate novel substrates or regulators of CK2. Regulators of CK2 could be classified as direct regulators (i.e. activators or inhibitors of CK2) or indirect regulators that could have the potential to control the cellular activity of CK2 by controlling its localization in cells or recruiting the enzyme to its substrates. In this regard, we have examined the possibility that CKIP-1 is a substrate for CK2, but we have been unable to observe any significant phosphorylation of either GST-CKIP-1 or GST-CKIP-1-(308 -409) by purified CK2 in vitro under conditions that yield extensive phosphorylation of casein (data not shown). Similarly, neither GST-CKIP-1 nor GST-CKIP-1-(308 -409) alters the activity of purified CK2 in vitro (data not shown). These results suggest that CKIP-1 is neither a substrate nor direct regulator of CK2 activity. Consequently, the notion that CKIP-1 may be an indirect regulator of CK2 may be reasonable.
The presence of a putative PH domain suggested that CKIP-1 may interact with the plasma membrane since the PH domains of other proteins have been implicated in localizing these proteins to the plasma membrane (52)(53)(54)(55). This prediction is supported by the observations made with EGFP-CKIP-1 in COS-7 cells (Fig. 9) and in Saos-2 cells (Fig. 10). In fact, the location of EGFP-CKIP-1 in COS-7 cells appears to be remarkably similar to that reported for the amino-terminal PH domain of pleckstrin (53). In that study, Ma et al. (53), demonstrated that the amino-terminal PH domain expressed in COS-1 cells associated with peripheral membrane ruffles and dorsal membrane projections, and that the presence of the amino-terminal PH domain of pleckstrin is critical for targeting pleckstrin to the plasma membrane.
The prominent nuclear localization of EGFP-CKIP-1 that is observed in a significant proportion of the Saos-2 cells suggests that the regulation and/or functions of CKIP-1 may be rather complex. In particular, since not all of the Saos-2 cells exhibited the nuclear localization of EGFP-CKIP-1, we speculate that its nuclear distribution is perhaps regulated in a cell cycle-dependent manner or in response to particular stimuli. We have also observed nuclear localization of EGFP-CKIP-1 in MC-3T3 cells and in Rat-1 fibroblasts (data not shown) but not in the nucleus of COS-7 cells (Fig. 9). As yet, we do not have any precise insights into why CKIP-1 is observed in the nucleus of some cells but not in others. Based on the observation that we observe EGFP-CKIP-1 in the nucleus of Saos-2 cells, we cannot exclude the possibility that CKIP-1 has a role in facilitating interactions between CK2 and its many nuclear targets. Moreover, the observation that EGFP-CKIP-1 resides in the nucleus of some, but not all, Saos-2 cells raises the intriguing possibility that CKIP-1 has a role in regulating the import or perhaps the export of CK2 from the nucleus. Clearly, a thorough understanding of CKIP-1 and its possible role(s) in regulating the cellular functions of CK2 will require a better understanding of how CKIP-1 and its localization or interactions with CK2 are regulated. Given that CK2 is an enzyme with a complex series of potential targets involved in a myriad of cellular functions, it is perhaps not surprising that CKIP-1 appears to also have a complex regulation.
It is intriguing to speculate that CKIP-1 may act to target or sequester CK2 to specific cellular locations and that CKIP-1 indirectly regulates the cellular functions of CK2 by controlling access to its cellular targets. This speculation is supported by our evidence that CK2 and CKIP-1 exhibit interactions in a number of assays including the yeast two-hybrid system, in Human osteosarcoma Saos-2 were transfected with pEGFP-CKIP-1, and then indirect immunofluorescence was carried out to visualize endogenous CK2 using rhodamine-conjugated secondary antibodies as described under "Experimental Procedures." Rhodamine fluorescence (A) and EGFP fluorescence (B) were detected by confocal microscopy, and optical slices of these cells were overlaid (C). The nuclei of two transfected cells that were observed in the field are marked with arrows.
co-immunoprecipitations, and that they exhibit co-distribution in the nucleus of Saos-2 cells. Although a precise understanding of how CKIP-1 could be involved in the regulation of CK2 must await a more thorough understanding of how the subcellular distribution of CKIP-1 is regulated, it is not inconceivable that CKIP-1 could recruit CK2 to specific cellular locations. For example, this suggestion would be consistent with a recent report demonstrating that CK2 is associated with the plasma membrane (21,33), along with reports of plasma membranelocalized CK2 substrates (including spectrin (56), insulin receptor (57), caveolin (58), IGF-II (59), dynamin (60), IRS-1 (61), and most recently CD5 (33)). These findings could implicate both CKIP-1 and CK2␣ in membrane-associated cellular processes such as receptor-mediated signaling, vesicle transport, and/or cytoskeleton organization. In this regard, CK2 has been previously implicated in insulin signaling (62)(63)(64), although the nature of its involvement is not clear and somewhat controversial (7,65,66). IRS-1 is a key intracellular regulatory protein that transduces the insulin receptor signal to a variety of proteins (67). It is interesting to note that CK2, the insulin receptor, IRS-1, and dynamin have all been co-localized to caveolae (58, 68 -71) and that, like CKIP-1, IRS-1 and dynamin both possess PH domains. In addition to the possibility that CKIP-1 targets CK2 to the plasma membrane, it is also possible that, through its PH domain, CKIP-1 mediates interactions between CK2 and proteins involved in vesicle transport or cytoskeletal organization. Interestingly, likely physiological targets for CK2 include a number of proteins that regulate intracellular trafficking such as p65 (synaptotagmin) (72), furin (73,74), and dynein (75). The observations that CK2␣ is implicated in the polarized growth of both fission and budding yeast cells (76,77), as well as neuritogenesis in mouse neuroblastoma (N2A) cells (78), also suggests a role for CK2 in the organization of the actin cytoskeleton. In fact, many proteins that interact with the cytoskeletal regulating GTPases (Rac, Rho, and cdc42) possess PH domains, leucine zippers, prolinerich regions, and/or kinase domains (79). With its PH domain, putative leucine zipper and proline-rich motifs, CKIP-1 could have the capacity to recruit CK2␣ which would bring kinase activity to this complex. Whether the CK2␣-CKIP-1 interaction plays a role in cytoskeletal signaling remains entirely speculative but is worthy of further investigation.
In conclusion, we have identified a protein, designated CKIP-1, that is able to interact with CK2␣. As noted earlier, the selective interactions between CKIP-1 and CK2␣ provide one further indication of functional specialization for the isoforms of CK2. Based on the subcellular localization of an EGFP-CKIP-1 fusion protein and the presence of a number of potential protein-protein interaction motifs, we speculate that CKIP-1 is a non-enzymatic regulator of CK2 that may act to sequester a fraction of the total cellular CK2 to specific subcellular locations. Alternatively, CKIP-1 may function as an adaptor that integrates CK2-mediated signals with signals from other signaling molecules or that mediates interactions between CK2␣ and other cellular proteins that could be targets or regulators of CK2. In this regard, CKIP-1 may regulate CK2 activity in a similar fashion to that of the many modular protein complexes that regulate the activity of protein kinases through protein-protein interactions (reviewed in Refs. 18, 55, 80, and 81).