The Pleckstrin homology domain of CK2 interacting protein-1 is required for interactions and recruitment of protein kinase CK2 to the plasma membrane.

CKIP-1 is a recently identified interaction partner of protein kinase CK2 with a number of protein-protein interaction motifs, including an N-terminal pleckstrin homology domain. To test the hypothesis that CKIP-1 has a role in targeting CK2 to specific locations, we examined the effects of CKIP-1 on the localization of CK2. These studies demonstrated that CKIP-1 can recruit CK2 to the plasma membrane. Furthermore, the pleckstrin homology domain of CKIP-1 was found to be required for interactions with CK2 and for the recruitment of CK2 to the plasma membrane. In this regard, point mutations in this domain abolish membrane localization and compromise interactions with CK2. In addition, replacement of the pleckstrin homology domain with a myristoylation signal was insufficient to elicit any interaction with CK2. An investigation of the lipid binding of CKIP-1 reveals that it has broad specificity. A comparison with other pleckstrin homology domains revealed that the pleckstrin homology domain of CKIP-1 is distinct from other defined classes of pleckstrin homology domains. Finally, examination of CK2alpha for a region that mediates interactions with CKIP-1 revealed a putative HIKE domain, a complex motif found exclusively in proteins that bind pleckstrin homology domains. However, mutations within this motif were not able to abolish CKIP-1-CK2 interactions suggesting that this motif by itself may not be sufficient to mediate interactions. Overall, these results provide novel insights into how CK2, a predominantly nuclear enzyme, is targeted to the plasma membrane, and perhaps more importantly how it may be regulated.

CK2 1 (formerly casein kinase II) is a ubiquitously expressed extraordinary conserved Ser/Thr kinase found in all eukaryotic cells (1)(2)(3). CK2 has been shown to be essential for viability in a variety of models ranging from yeast to mammalian cells (4 -6). An elevated CK2 activity has been detected in leukemic cells, healthy tissues with high mitotic index, and in a variety of human cancers (7)(8)(9). In addition, CK2 exhibits oncogenic activity when overexpressed in transgenic mice (10). The majority of CK2 is found in the nucleus of logarithmically growing cells (11)(12)(13); however, there have been indications that the nuclear/cytoplasmic distribution of CK2 is regulated in a cell cycle-dependent manner (14). Antibodies that interfere with the nuclear uptake of CK2 inhibit mitogenic stimulation upon micro-injection into cells (15), indicating the importance of the nature of CK2 localization for its biological function. To date, CK2 has been shown to phosphorylate and/or interact with a broad range of proteins located in a variety of cellular compartments, including nuclear proteins (16 -20), cytoplasmic proteins (21)(22)(23), and proteins localized at the plasma membrane (24 -26).
This tetrameric enzyme is composed of two catalytic (␣ and ␣Ј) and two regulatory subunits (␤). The two catalytic subunits are products of separate genes and show greater than 90% sequence identity over their N-terminal 330 amino acids (27). Despite the fact that there are no obvious catalytic differences between ␣ and ␣Ј, there is strong evidence to support functional specialization (28,29). For example, a knockout of CK2␣Ј illustrates the inability of CK2␣ to compensate during spermatogenesis (29). CK2␣ and ␣Ј also show differences in phosphorylation (30 -32), cell cycle-dependent differences in localization (14), and differences in protein partners. Both ␣ and ␣Ј interact with nucleolin (33), yet only ␣ interacts with HSP90 (34,35), Pin-1 (36), and PP2A (21). Interestingly enough, the interaction between PP2A and CK2␣ was mapped to one of the few areas of non-identity between ␣ and ␣Ј in the N-terminal 330 amino acids (21). These observations suggest that the subcellular localization of CK2 and its ability to phosphorylate a number of its target proteins may be regulated by interactions with specific protein partners (37). Clearly, this could explain why a predominantly nuclear enzyme has been reported to phosphorylate cytoplasmic and plasma membrane bound targets.
To test the hypothesis that specific functions of CK2␣ and CK2␣Ј are mediated by unique interaction partners, a yeast two-hybrid strategy was utilized to search for novel interacting partners. Using full-length CK2␣ and CK2␣Ј GAL4 binding domain fusions as bait, we isolated a novel CK2-interacting protein designated CKIP-1 (37), which interacted with CK2␣, but not ␣Ј in yeast two-hybrid assays. Interactions between CK2 were subsequently confirmed using GST pull-down assays and co-immunoprecipitation from transfected and non-transfected cells indicating that CKIP-1 and CK2 interact under physiological conditions (37).
CKIP-1 cDNA is broadly expressed and encodes a protein with a predicted molecular mass of 46 kDa. Although its overall biological functions remain poorly understood, CKIP-1 has recently been implicated in phosphatidylinositol 3-kinase-regulated muscle cell differentiation (38). Examination of the amino acid sequence of CKIP-1 revealed the presence of an N-terminal pleckstrin homology (PH) domain, which could mediate interactions between CKIP-1 and cellular membranes or proteins, and a putative C-terminal leucine zipper, which is thought to be involved in protein-protein interactions (37). In addition, this protein contains five PXXP motifs, two of which match the consensus sequence for phosphorylation by cyclindependent kinases and mitogen-activated kinases.
Based on the fact that CKIP-1 contains a number of proteinprotein interaction motifs, and its discrete cellular localization, we hypothesized that CKIP-1 is a non-enzymatic regulator of CK2. To test this hypothesis we examined the interactions between CK2 and CKIP-1 in cells and in vitro. Our data demonstrate that CKIP-1 is capable of redistributing CK2 to the plasma membrane and that the PH domain of CKIP-1 is required for this activity. This demonstration provides novel insight into a possible mechanism for how CK2, a predominantly nuclear enzyme, is capable of phosphorylating proteins present at the cellular membrane.

Materials
Human osteosarcoma Saos-2 cells and U2-OS cells were obtained from ATCC (Manassas, VA) and were maintained in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum (CanSera International, Inc.). Anti-CKIP-1 antibodies were raised in rabbits by BabCO (Richmond, CA) as previously described (37), and anti-GST monoclonal antibodies, anti-FLAG M2 monoclonal antibodies, and rhodamine-conjugated goat anti-rabbit antibodies were obtained from Sigma; horseradish peroxidase-conjugated goat antimouse antibodies were obtained from Bio-Rad. PIP-Strips were obtained from Echelon Research Laboratories Inc., and reagents for enhanced chemiluminescence (ECL) were obtained from Pierce. Plasmids encoding enhanced yellow fluorescent protein (EYFP-C1) and enhanced cyan fluorescent protein (ECFP-C1) were obtained from Clontech, Pfu DNA polymerase was obtained from Stratagene, and Fugene6 transfection reagent was obtained from Roche Applied Science. AirVol was from Air Products and Chemicals, Inc.; all other reagents were of reagent grade.

Plasmid Constructs
pEYFP Construct-Full-length CK2␣ was amplified out of pGBT9-CK2␣ using the following forward 5Ј-TTG ACT GTA TCG CCG AAT TCC CCG GGG ATC CTG-3Ј and reverse 5Ј-CTT GGC TGC AGG GTC GAC GAT CCC CCG GGC-3Ј primers. The amplified product was gelpurified and ligated into the pCR-Blunt vector (Invitrogen) then digested with EcoRI and SalI. The respective EcoRI/SalI fragment was subcloned into the EcoRI/SalI sites of pEYFP-C1.
pECFP Constructs-Full-length CKIP-1 was amplified out of pGBT9-CKIP-1 using the following forward 5Ј-AC TGT ATC GCC GAA TTC CCC ATG ATG AAG A-3Ј and reverse 5Ј-CTT GGC TGC AGG GTC GAC CCA CCC TGC CCT-3Ј primers. The PCR product was gel-purified, ligated into the pCR-Blunt vector, digested with EcoRI and SalI, and subcloned into EcoRI/SalI-digested pECFP-C1 (Clontech). A monomeric form of ECFP-C1 was constructed by incorporating an A206K mutation as described previously (39) using a sequential PCR strategy.
pEGFP Constructs-Full-length cDNA encoding CKIP-1 was subcloned into the EcoRI and SalI sites of the pEGFP-C3 vector. To examine the residues important for interactions with CK2, key residues within the PH domain (Lys-42, Arg-44, and Trp-123) of CKIP-1 were mutated. Full-length cDNA encoding CKIP-1, as described previously (37), incorporated into a pEGFP vector (pRSP9), was utilized as the template in a PCR reaction to amplify the DNA and incorporate specific point mutations. To begin with, W123A was incorporated into the full-length cDNA using sequential PCR. The first two PCR products for the W123A mutation were generated using the CKIP-1 forward primer  5Ј-TGA TCA GAA TTC CCA TGA TGA AGA AGA ACA ATT CCG CCA  AG-3Ј and the W123A reverse primer with 5Ј-AAT GAT CCA CGA TTC  CTT CTC TTC-3Ј, and the W123A forward primer 5Ј-GAA GAG AAG  GAA TCG GCG ATC AAT GCC CTC-3Ј with the CKIP-1 reverse primer  5Ј-GAA TTC GTC GAC CCC ACC CTG CCC TCA CAT CAG G-3Ј. These products were gel-purified and added to a second PCR reaction mixture along with the CKIP-1 forward and reverse primers. The PCR product was gel-purified and ligated into pCR-Blunt. The clone was then digested with EcoRI/SalI, and the respective fragment was subcloned into the EcoRI/SalI-digested pEGFP-C3 vector to create a construct with an N-terminal GFP tag. The sequence of this clone, pMO7, was confirmed by sequencing using the pEGFP-C primer. The K42C mutant, R44C mutant, and K42C/R44C double mutant were made in a similar fashion, using K42C forward 5Ј-TTC AGG GAG ATT TGG TGT AAC CGC TAT GTG-3Ј and K42C reverse 5Ј-GCG GTT ACA CCA AAT CTC CCT GAA-3Ј primers, R44C forward 5Ј-GGG AGA TTT GGA AAA ACT GCT ATG TGG TGC TG-3Ј and R44C reverse 5Ј-CAC ATA GCA GTT TTT CCA AAT CTC CC-3Ј primers, and K42C/R44C forward 5Ј-TTC AGG GAG ATT TGG TGT AAC TGC TAT GTG-3Ј and K42C/R44C reverse 5Ј-GCA GTT ACA CCA AAT CTC CCT GAA-3Ј primers, respectively, along with the CKIP-1 primer set. The resulting clones, pMO9, pMO55, and pCZ12, respectively, were confirmed by sequencing using the pEGFP-C primer. pMO7 was used as the template to construct the double mutants containing K42C and W123A, and R44C and W123A, and the triple mutant containing K42C, R44C, and W123A. PCR was used to create point mutations utilizing either K42C primers or R44C primers for the double mutants, and the K42C/R44C primers for the triple mutant. The resulting clones, pMO12, pMO54, and pMO13, respectively, were verified by sequencing using the pEGFP-C primer.
Myc-CK2␤-CAAX Construct-To construct a membrane localized form of CK2␤, we incorporated the Ras membrane-targeting signal as described by Clark et al. (40) at the immediate C-terminal of Myc-CK2␤. The Ras membrane-targeting signal is a 19 amino acid peptide ending in CAAX, where C ϭ cysteine, A ϭ aliphatic amino acid, and X ϭ any amino acid (40). The resulting construct was designated Myc-CK2␤-CAAX.
pACT2 Constructs-Full-length cDNA encoding CKIP-1 was amplified using the CKIP-1 forward primer and the CKIP-1 reverse primer. Primers were designed to create the number of CKIP-1 deletion mutants. CKIP-PH was constructed using the CKIP-1 forward primer and PH reverse primer (5Ј-GAA TTC GTC GAC TCA GGT GAT GGC AGA GTT GAG GGC A-3Ј), which incorporates a stop site after the PH domain; CKIP-LZ was constructed using the LZ forward primer (5Ј-TGA TCA GAA TTC CGG ATT CTG AGT CAG AGC AGC T-3Ј) with the CKIP-1 reverse primer; CKIP-⌬PH was constructed using the ⌬PH forward primer (5Ј-TGA TCA GAA TTC CCC GAG CCA AGA ACC GTA TCT TGG-3Ј) with the CKIP-1 reverse primer; CKIP-⌬LZ was constructed using CKIP-1 forward primer with the ⌬LZ reverse primer (5Ј-GAA TTC GTC GAC TCA AGA ATC CGG CGG AGA CCG AGG G-3Ј), which incorporated a stop codon immediately before the LZ; and finally, CKIP-⌬PH⌬LZ was constructed using the ⌬PH forward primer with the ⌬LZ reverse primer. Each PCR product was gel-purified, ligated into pCR-Blunt, and digested with EcoRI, and the respective fragments were subcloned into EcoRI-digested pACT2 with the exception of CKIP-⌬PH⌬LZ, which was digested from the pCR-Blunt construct with EcoRI/SalI and inserted into EcoRI/XhoI-digested pACT2. Constructs were verified for orientation using restriction digests and sequencing.
Myristoylation Signal Constructs-A myristoylation sequence from v-Src (41), MGSSKSKPKDPSQR, was used to replace the PH domain at the N-terminal of CKIP-1 using the following myristoylation primer, 5Ј-AAG GAA TTC ATG GGT AGC AGC AAG AGC AAG CCC AAG GAT CCC AGC CAG CGG CGA GCC AAG AAC CGT ATC TTG GAT GAG GTC AC-3Ј. To subclone myristoylated CKIP-1 into FLAG-pRC/CMV, EGFP-CKIP-1 was used as a template for PCR using the myristoylation primer and FLAG-pRC/CMV reverse primer. The PCR product was gel-purified, ligated into pCR Blunt, digested with EcoRI/NotI, ligated into EcoRI/NotI-digested FLAG-pRC/CMV vector and sequence verified, resulting in clone pMO47. Similarly, a construct lacking the FLAG tag on the N-terminal was made by digesting a pCR Blunt clone with HindIII/ApaI, and then subcloning into HindIII/ApaI-digested pRC/ CMV. The resulting clone, pMO48, was verified by sequencing.

Transfection of Saos-2 and U2-OS Cells
U2-OS and Saos-2 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 1% penicillin/ streptomycin (Invitrogen). Cells were grown to 80% confluence, harvested using Trypsin-EDTA, and counted using a hemocytometer. For visualization studies, cells were seeded the day before transfection at ϳ1.6 ϫ 10 5 cells into 6-well, flat bottom tissue culture plates containing coverslips. The following day, 2-5 h prior to transfection, the medium was changed. Transfections were carried out using Fugene6 (Roche Applied Science) as described by the manufacturer. Plasmids encoding the various constructs were added at a concentration of 1 g per 35-ϫ 20-mm well of a 6-well plate, or in the case of co-transfections, 0.5 g of each construct was used. For pull-down experiments, U2-OS cells were seeded at 2.7 ϫ 10 6 cells per 150-mm plate and transfected by calcium phosphate transfection. Precipitates were left on the cells for 16 -18 h following transfection, after which the precipitates were washed off using PBS, pH 7.4. Medium was replaced, and the cells were allowed to grow another 24 h. After the 24 h, cells were washed well with PBS, and cell extracts were prepared by scraping cells from plates into interaction buffer (20 mM HEPES, pH 7.4, 150 mM NaCl, 10% glycerol, 0.05% Nonidet P-40, 1% aprotinin, 2 g/ml leupeptin, 10 g/ml para-methylsulfonyl fluoride), followed by sonication on ice (3 ϫ 10-s bursts). The resulting lysates were ultracentrifuged at 55,000 rpm for 20 min at 4°C. The supernatant was collected and used immediately for GST pull-down assays or alternatively stored at Ϫ80°C.

Visualization of Fluorescent Fusion Proteins in Transfected Cells
To examine the co-localization of CK2␣ with CKIP-1, transfected U2-OS or Saos-2 cells were washed twice with PBS, then fixed for 30 min at 37°C using 3.7% paraformaldehyde in PBS, and finally washed in PBS and mounted using AirVol. Where indicated, a 10-min incuba-tion using digitonin in PBS at a concentration of 25 g/ml was performed before fixation for some slides. To visualize FLAG-CKIP-1 and Myr-CKIP localization, Saos-2 cells were subjected to indirect immunofluorescence using anti-CKIP-1 antibodies and rhodamine-labeled goat anti-rabbit antibodies. For indirect immunofluorescence, cells on coverslips were washed twice with PBS and were then fixed for 30 min with 3.7% paraformaldehyde in PBS prior to saturation of the cells for 30 min with 0.1 M glycine and permeabilization with 0.1% Triton X-100 in PBS. The fixed and permeabilized cells were incubated for 1 h with anti-CKIP-1 antibodies (1/250 dilution), washed for 3 ϫ 5 min with PBS, and then incubated for 1 h with rhodamine-conjugated goat antirabbit antibodies (1/1000 dilution). Finally coverslips were washed well with PBS, and mounted in AirVol. For non-confocal images, EYFP and ECFP fluorescence was visualized using an Axiovert inverted fluorescence microscope fitted with a YFP/CFP filter set. The images were pseudocolored using Adobe Photoshop. For confocal images, EGFP, EYFP, and rhodamine fluorescence were visualized with a Zeiss LSM 410 or Zeiss LSM 510 Meta inverted confocal microscope. The images were pseudocolored using built-in Zeiss LSM software, and transferred into Adobe Photoshop. In all cases, controls were used to ensure that the settings used for microscopy eliminated bleed-through between channels.

GST Pull-downs and in Vitro Binding Assays
Glutathione S-transferase (GST) fusion proteins encoding full-length CKIP-1, CKIP-1 deletion mutants (LZ, ⌬PH, etc.), CK2␣ or GST itself were expressed in BL21(DE3) bacteria and purified using glutathioneagarose as described previously (31) and left on the beads for use in GST pull-down assays or in vitro binding studies. GST pull-down assays were performed by incubating the purified proteins with whole cell lysates from CK2␣-HA-transfected U2-OS cells for 1 h at 4°C. Alternatively, in vitro binding assays were performed by incubating 11 l of GST or GST-CK2␣ beads in 100 l of Nonidet P-40 buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 1% Nonidet P-40) with 35 S-labeled FLAG-CKIP-1, and FLAG-CKIP-1 PH domain mutants produced using a TNT kit (Promega) with T7 polymerase according to the manufacturer's recommendations and incubating for 1 h at 4°C. Following incubation, glutathione-agarose was collected by centrifugation, supernatant was removed, and the beads were washed four times with interaction buffer or Nonidet P-40 buffer, respectively. Proteins remaining on the glutathione-agarose were eluted by the addition of Laemmli sample buffer and were subjected to SDS-polyacrylamide gel electrophoresis. Proteins from GST pull-down assays were transferred to a polyvinylidene difluoride membrane, and the presence of CK2␣-HA was confirmed by Western blot with anti-HA antibodies. In the case of the in vitro binding assays, radiolabeled proteins from three independent experiments were visualized using a PhosphorImager (Amersham Biosciences), and quantitated using ImageQuaNT software (Amersham Biosciences) to determine the percentage of total radiolabeled product that was bound to the GST fusion proteins. The average percentage of total radiolabeled protein that was bound to GST or GST-CK2␣ Ϯ S.D. is displayed in a graphical format.

Phospholipid Binding Assay
GST fusion proteins encoding full-length CKIP-1, as well as each of the CKIP-1 PH domain mutants, and GST alone were expressed in BL21(DE3) bacteria and purified using glutathione-agarose as described previously (31). Purified GST fusion proteins were used in an overlay assay with a nitrocellulose membrane spotted with various phospholipids (44, 45) (PIP-Strip, Echelon Research Laboratories Inc.) with the protein at a final concentration of 0.5 g/ml as directed by the manufacturer. Anti-GST monoclonal antibodies and horseradish peroxidase-conjugated goat anti-mouse antibodies were used along with ECL to determine the spectrum of phospholipid binding. Tris-buffered saline with 3% fatty acid-free bovine serum albumin (Sigma) was used to block the PIP-Strip and was also used in the antibody mixtures. (37,46), we hypothesized that CKIP-1 could be a non-enzymatic regulator of CK2 involved in recruitment of CK2 to specific cellular locations. To test this hypothesis, we examined the ability of CKIP-1 to alter the distribution of CK2 within cells using CK2␣ as an EYFP fusion and CKIP-1 as an ECFP fusion. Constructs encoding each of these fusions were transfected into human osteosarcoma Saos-2 cells and analyzed by fluorescence microscopy (Fig. 1). As expected, EYFP-CK2␣ displays mainly nuclear localization with some cytoplasmic staining (Fig. 1, top left panel). By comparison, ECFP-CKIP-1 is localized to the plasma membrane and/or membrane ruffles with additional cytoplasmic staining (Fig. 1, top right  panel). When EYFP-CK2␣ was expressed in cells together with ECFP-CKIP-1, a notable alteration in the distribution of EYFP-CK2␣ is observed with an apparent increase in the amount of EYFP-CK2␣ in the cytoplasm and some of the EYFP-CK2␣ localized on the membrane (Fig. 1, bottom left  panel). Examination of the ECFP-CKIP-1 in the same cells (Fig. 1, bottom right panel) revealed that the re-localized CK2␣ present on the plasma membrane was co-localized with ECFP-CKIP-1. A similar redistribution of EYFP-CK2␣ in cells expressing ECFP-CKIP-1 was observed in human osteosarcoma U2-OS cells examined by confocal microscopy (data not shown). Furthermore, the redistribution of EYFP-CK2␣ to the plasma membrane is unaffected when a dimerization-defective mutant of ECFP (39) is used for the expression of ECFP-CKIP-1 indicating that this redistribution does not result from dimerization of EYFP and ECFP fusion. Overall, the differences observed in CK2␣ localization in the presence and absence of CKIP-1 indicates that CKIP-1 possesses the ability to re-localize CK2␣ in vivo.

CKIP-1 Recruits CK2 to the Plasma Membrane-Based on previous results
Comparison of the Re-localization of CK2 to the Membrane by CKIP-1 and CK2␤-CAAX-Results illustrated in Fig. 1 demonstrate re-localization of CK2␣ to the plasma membrane by CKIP-1. However, it is evident that a major proportion of CK2␣ was still localized within the nucleus. Accordingly, to examine the full extent to which CK2␣ can be re-localized to the plasma membrane, we devised a strategy to examine the ability of a membrane-targeted version of CK2␤, an integral component of tetrameric CK2 complexes, to recruit CK2␣ to the membrane. In this respect, we constructed a mutant of CK2␤ designated Myc-CK2␤-CAAX with a C-terminal CAAX motif that is isoprenylated in cells to direct membrane localization and an Nterminal Myc-tag to permit detection (40). Initial immunoprecipitation studies demonstrated that Myc-CK2␤-CAAX is not impaired in its ability to interact with the catalytic CK2 subunits (data not shown). Indirect immunofluorescence indicates that Myc-CK2␤-CAAX is localized primarily to the plasma membrane ( Fig. 2A) with some cytoplasmic staining. Occasionally, this mutant exhibited nuclear staining in some cells (not shown). To compare the localization of CKIP-1 to that of Myc-CK2␤-CAAX using indirect immunofluorescence, we constructed a variant of CKIP-1 with an N-terminal FLAG epitope. Indirect immunofluorescence ( Fig. 2A) demonstrates that FLAG-CKIP-1 localizes to the plasma membrane and/or membrane ruffles with some cytoplasmic staining ( Fig. 2A) in a manner identical to that seen with ECFP-CKIP-1. Staining with either anti-Myc or anti-FLAG was absent in non-transfected cells ( Fig. 2A, top panels). To compare the abilities of Myc-CK2␤-CAAX and CKIP-1 to re-localize CK2 to the plasma membrane, U2-OS cells were transfected with EYFP-CK2␣ in the presence or absence of either Myc-CK2␤-CAAX or FLAG-CKIP-1 and EYFP fluorescence visualized by confocal microscopy. When EYFP-CK2␣ was co-transfected with Myc-CK2␤-CAAX or with FLAG-CKIP-1, a proportion of the EYFP-CK2␣ could be visualized at the plasma membrane (Fig. 2B, middle and right panels, respectively), which was not evident when EYFP-CK2␣ was expressed by itself (Fig. 2B, left panel). It is also evident with Myc-CK2␤-CAAX (Fig. 2B, middle panel), as is the case with FLAG-CKIP-1 (Fig. 2B, right panel) and with ECFP-CKIP-1 (Fig. 1), that a major proportion of CK2 is retained within the nucleus. The latter observation indicates that, even with an integral component of CK2 complexes, it is only possible to re-localize a fraction of the total cellular CK2, an observation that likely reflects the complex regulation of CK2 and its interactions with a large number of cellular proteins. In relation to Myc-CK2␤-CAAX, it is therefore apparent that CKIP-1 is nearly as effective at eliciting the re-localization of CK2␣ to the plasma membrane.
To confirm that the fluorescent proteins were located at the membrane, cells were treated with digitonin to permeabilize the cells and release their soluble cytoplasmic contents without releasing membrane-associated proteins. U2-OS cells were transfected with EYFP-CK2␣ in the presence or absence of either Myc-CK2␤-CAAX or FLAG-CKIP-1 and incubated with, or without digitonin, prior to visualization of EYFP-CK2␣ by fluorescence microscopy (Fig. 2C). Digitonin treatment (Fig.  2C, bottom left panel) results in the complete loss of EYFP-CK2␣ indicating that EYFP-CK2␣ by itself is not membraneassociated. By comparison, a significant proportion of the EYFP-CK2␣ was retained at the cell membrane following digitonin treatment of cells that were co-expressing EYFP-CK2␣ with either Myc-CK2␤-CAAX or FLAG-CKIP (Fig. 2C, bottom  middle and bottom right panels). These results validate the membrane localization observed without digitonin treatment and reinforce the conclusion that CKIP-1 re-localizes CK2.
Elucidation of Domains on CKIP-1 Responsible for Interactions with CK2-To understand the mechanistic basis for interactions between CK2 and CKIP-1 we performed yeast two-hybrid assays to identify the domain(s) on CKIP-1 responsible for interactions with CK2. Constructs encoding full-length CKIP-1 and different CKIP-1 deletion mutants fused to the GAL4 activation domain in the pACT2 plasmid (Fig. 3A) were each co-transformed into yeast with GAL4 binding domain fusions of either CKIP-1 or CK2␣ in the pGBT9 plasmid. Each of the co-transformations into the PJ69-4a yeast strain was plated on both Trp-Leu-, to monitor transformation efficiency, and His-Ade-Leu-Trp-, to examine interactions. Positive inter- actions between pGBT9-CK2␣ and pACT2-CKIP-1 and its deletion mutants were only detected when CK2 was co-transformed with pACT2-CKIP-1 and pACT2-CKIP-⌬LZ (Fig. 3A). These results suggest that the PH domain of CKIP-1 is required for interactions with CK2␣ but, on its own, not sufficient, because pACT2-CKIP-PH did not show positive interactions on -HALT. Positive interactions between pGBT9-CKIP-1 and pACT2-CKIP-1 and its deletion mutants were only detected when CKIP-1 was co-transformed with pACT2-CKIP-1, pACT2-CKIP-LZ, and pACT2-CKIP-⌬PH (Fig. 3A) indicating possible dimerization of CKIP-1 through its putative leucine zipper. To confirm the results of the yeast two-hybrid studies, we analyzed the interactions between CKIP-1 and CK2 using GST pull-down assays. As illustrated in Fig. 3B, CK2 interacts with GST-CKIP-1 and GST-CKIP-⌬LZ, and very slightly with GST-CKIP-PH. Overall, the results concur with yeast twohybrid studies, indicating that the PH domain of CKIP-1 is necessary but, on its own, is not sufficient for interactions with CK2.
The PH Domain of CKIP-1, Not Its Membrane Localization, Is Required for CK2 Re-localization-Because PH domains are involved in membrane localization, we decided to investigate the possibility that the membrane localization itself may be important for interactions with CK2␣ in vivo, rather than the PH domain. Therefore, the PH domain of CKIP-1 was replaced with a myristoylation signal (MGSSKSKPKDPSQR) to target CKIP-1 to the cell membrane in the absence of its the PH domain (41). Confocal microscopy (Fig. 4A) demonstrates that FLAG-CKIP-1 and Myr-CKIP exhibit identical distribution when expressed in Saos-2 displaying primary localization on the cell membrane/ruffles together with some cytoplasmic staining.
The ability of FLAG-CKIP-1 or FLAG-Myr-CKIP to interact directly with CK2␣ was then tested in vitro employing pulldown assays using GST or GST-CK2␣ as affinity matrices with radiolabeled FLAG-CKIP-1 (50 kDa) or FLAG-Myr-CKIP (42 kDa) that were obtained by in vitro transcription and translation (Fig. 4B). Approximately 50% of FLAG-CKIP-1 was retained on GST-CK2␣ with negligible amounts (Ͻ0.5%) retained on GST alone (Fig. 4B, left panel). Negligible amounts (i.e. Ͻ0.5%) of FLAG-Myr-CKIP were retained on either GST-CK2␣ or GST beads (Fig. 4B). These results indicate that the PH domain of CKIP-1 is required for its interactions with CK2. To extend these findings, we examined the ability of CKIP-1 to interact with CK2␣ in vivo (Fig. 4C). Saos-2 cells were cotransfected with EYFP-CK2␣ and either ECFP-CKIP-1 or Myr-CKIP, and the subcellular localization of EYFP-CK2␣ was examined by confocal microscopy. As with the experiments shown elsewhere (Figs. 1 and 2) expression of CKIP-1 resulted in re-localization of some CK2␣ to the cell membrane (indicated by arrows, Fig. 4C). However, when EYFP-CK2␣ was expressed with Myr-CKIP, CK2␣ was found solely in the nucleus

FIG. 3. The PH domain of CKIP-1 is required for interactions with CK2.
A, to examine the interactions between CK2 and CKIP-1, and CKIP-1 with itself, cDNA encoding full-length CKIP-1 was expressed in yeast as a fusion with the DNA activation domain (AD) of GAL4, using the plasmid pACT2 as described under "Experimental Procedures." Constructs encoding GAL4 DNA activation domain fusions encoding deletion mutants of CKIP-1 (LZ, PH, ⌬LZ, ⌬PH, and ⌬PH ⌬LZ, as indicated) were also prepared. The pleckstrin homology (PH) from residues 21-132 of CKIP-1 is indicated as is the putative leucine zipper (LZ) from residues 347-373. Plasmids encoding CKIP-1 or CK2␣ as GAL4 DNA binding domain fusions were co-transformed into yeast along with pACT2 constructs encoding wild type CKIP-1 or its various deletion mutants. Positive interactions were indicated by the ability of transformants to grow on synthetic complete media deficient in His, Ade, Leu, and Trp (designated -HALT). B, GST pull-down assays were performed on whole cell lysates from CK2␣-HA-transfected U2-OS cells. GST, GST-CKIP-1, GST-CKIP-LZ, GST-CKIP-PH, GST-CKIP-⌬LZ, GST-CKIP-⌬PH⌬LZ, and GST-CKIP-⌬PH, were used in pull-down assays on the whole cell lysates. Proteins bound to GST-fusion proteins were eluted from beads using Laemmli sample buffer. CK2␣-HA was detected on immunoblots using anti-HA antibodies and enhanced chemiluminescence. of the cells. Overall, these results demonstrate that, although FLAG-CKIP-1 and Myr-CKIP exhibit the same subcellular distribution, the presence of the PH domain itself is required for interactions with CK2␣.
Evaluation of the Role of a Potential HIKE Domain in CK2␣-As reported previously, the unique C-terminal of CK2␣ is not responsible for interactions between CK2␣ and CKIP-1 but, rather, a portion of the N-terminal that is highly related to CK2␣Ј (37). It is conceivable that the interaction between the two lies in a region of the N-terminal of CK2␣ exhibiting non-identity to CK2␣Ј. Closer examination of amino acid sequences of the N termini of the CK2 catalytic subunits revealed a potential HIKE domain present in CK2␣ (Fig. 5A). HIKE is a highly conserved sequence motif, as shown in Fig. 5A, that selectively occurs in proteins that bind PH domains (47). This motif was originally identified by sequence homology analysis of the proteins considered the strongest PH binding candidates, such as PKC and Akt (48). A conserved ␤ strand-loop-␤ strand structure is exhibited by HIKE, despite the fact that the proteins in which they are found have widely different threedimensional structures (48). There is a highly conserved acidic residue 8 amino acids downstream from the most C-terminal conserved Lysine. It is this conserved acidic residue, present in CK2␣ but not CK2␣Ј (Fig. 5A), at which mutagenesis was directed to evaluate the domain. The resulting constructs were co-transformed along with pACT2-CKIP-1 into the PJ69-4a yeast strain and subjected to yeast two-hybrid analysis. It was found that the E167Q mutation in the HIKE domain of CK2␣ was not sufficient to abolish interactions between CKIP-1 and CK2␣, nor was the Q168E mutation in CK2␣Ј sufficient to promote interactions with CKIP-1 (Fig. 5B). These results suggest that residues other than the mutated residue contribute to the interaction between protein kinase CK2 and CKIP-1, reflecting the complexity of the HIKE domain.
Multiple Sequence Alignment of the PH Domain of CKIP-1 with Other PH Domains-PH domains are found in over 70 proteins involved in signal transduction and cytoskeletal structures (49) and are thought to have a role in either membrane localization and/or protein-protein interactions (37,50,51). Multiple sequence alignments of PH domains indicate that the domain itself is a series of rather poorly conserved peptides, ϳ120 amino acids in length, containing a nearly invariant Trp residue in the C-terminal 15 amino acids (51), and a fairly conserved "ϩXϩ" motif (52). A multiple sequence alignment (MSA) was performed on the PH domain of CKIP-1, along with representative PH domains from established PH domain classes. The alignment illustrated in Fig. 6A shows that the domain itself is not highly conserved. A phylogenetic tree was then drawn from a MSA containing nine different PH domains. The branch lengths serve as a basis of the evolutionary conservation between items on the tree. As shown in Fig. 6B, there is not a high degree of conservation between PH domains, and CKIP-1 does not show high conservation to any PH domain in the given alignment. Overall, the results demonstrate the poor sequence identity and large amount of diversity exhibited by PH domains and make it evident that the PH domain of CKIP-1 is distinct from the known classes of PH domains. Only two residues were identical between the five PH domains aligned, and very few residues were even considered conserved or semiconserved. The first conserved residue in the alignment (lysine 42) conforms to a known "ϩXϩ" motif, because it is followed by another positive residue (arginine 44) two amino acids C-terminal. The second conserved residue (tryptophan 123) was identical in all five domains aligned and is known to be critical for PH domain integrity.
Systematic Mutation of the PH Domain of CKIP-1-Based on the evidence that the PH domain is required for interactions with CK2, we were interested in generating mutants of CKIP-1 to disrupt key residues within the domain without affecting its overall structure. As noted in the previous section, Lys-42 and Arg-44 of CKIP-1 conform to a "ϩXϩ" motif so were mutated to cysteine (K42C and R44C). The mutations conferring a change from lysine or arginine to cysteine were based on a mutation in Bruton's tyrosine kinase, which was identified as the defective gene in X-linked immunodeficiency mice. This mutation in Bruton's tyrosine kinase caused B-cells to respond abnormally to activating signals (53) but did not alter Bruton's tryosine kinase expression or in vitro kinase activity in these cells (54). Comparable mutations made in PKD promoted an activated state of PKD (55), whereas mutations in Akt resulted in abnormal regulation of the enzyme (53,56). We also generated W123A mutants of CKIP-1, because this residue is conserved in all known PH domains and is the only invariant amino acid in the C-terminal ␣-helix. A similar mutation in PKD (55) was found to be as effective as a partial or complete deletion of the PH domain in promoting an active state of PKD. In addition to the above point mutations, double and triple mutants of CKIP-1 were also constructed by combining the K42C, R44C, and W123A substitutions.
Each of the CKIP-1 PH domain mutants were transfected into Saos-2 cells and examined by confocal microscopy (Fig. 7A) to determine their subcellular location. In accordance with experiments shown elsewhere (Figs. 1, 2, and 4), CKIP-1 itself is primarily localized to the plasma membrane. The K42C, R44C, and double mutant incorporating these two mutations had no dramatic effect on localization. However, in the case of the W123A mutant and the double and triple mutants incorporating this mutation, membrane localization was disrupted. The exact cellular localization of these mutations is unknown at this time and requires further examination.

FIG. 5. Evaluation of a potential HIKE domain in CK2␣.
A, alignment of a portion of the amino acid sequence of CK2␣ and CK2␣Ј showing residues conserved between the two catalytic subunits. The consensus sequence for the HIKE domain, which is found in a number of proteins known to bind PH domains (47) is also shown. HIKE residues in CK2␣ and CK2␣Ј are underlined and indicated in bold. B, mutations were made in the potential HIKE domain of CK2␣ and/or CK2␣Ј as described under "Experimental Procedures." A yeast two-hybrid assay was performed using constructs in which the potential HIKE domain was mutated and expressed in yeast as a fusion with the DNA binding domain (BD) of GAL4, using the plasmid pGBT9 as described under "Experimental Procedures." CKIP-1 was co-transformed as a fusion with the GAL4 DNA activation domain. Positive interactions were indicated by the ability of transformants to grow on synthetic complete media lacking His, Ade, Leu, and Trp (-HALT).
FIG. 6. Multiple sequence alignment of the PH domain of CKIP-1 with other well known PH domains. A, a multiple sequence alignment was done using ClustalW (version 1.82, available at www.ebi.ac.uk/clustalw/). The PH domain of CKIP-1 and representative PH domains from the known classes of PH domains were aligned using default parameters. The representative proteins from each grouping of PH domains were as follows: Bruton's tyrosine kinase (group 1), the N-terminal PH domain of Pleckstrin (group 2), Akt-2 (group 3), and Dynamin-1 (group 4). Acidic residues are shown in blue, basic residues in magenta, small hydrophobic residues in red, and residues that contain hydroxyls and amines and are basic are shown in green. The consensus symbols are as follows: '*' indicates that the residue is identical in all sequences in the alignment, ':' indicates that conserved residues have been observed according to color designations, and '.' indicates that semi-conserved substitutions are observed. Overall, the alignment of the PH domains shows a very poorly conserved domain. B, a phylogenetic tree was drawn for To determine whether membrane localization exhibited any correlation with phospholipid binding, we examined the effects of these CKIP-1 PH domain mutations on binding to various phosphoinositols using overlay assays performed with PIP strips (44,45). Accordingly, GST fusion proteins of CKIP-1 and each of the CKIP-1 PH domain mutants were expressed in bacteria. Each of the GST fusion proteins was expressed to comparable degrees (Fig. 7B). As illustrated in Fig. 7C with representative data from the PIP strip assays, it is evident that GST-CKIP-1 exhibits broad spectrum of binding. The GST-CKIP-1 K42C mutant shared the same broad spectrum of binding; however, the K42C mutant bound the phosphoinositols to a lesser degree (Fig. 7C) than was observed for GST-CKIP-1. The R44C and K42C/R44C double mutant showed similar binding to that observed with the K42C mutant (data not shown), again to a weaker extent than GST-CKIP-1. All mutants with a variety of proteins containing PH domains, including the proteins shown in the MSA shown in panel A, using NJ Plot (www.pbil.univ-lyoul.fr/ njplot.hml). Also included in the phylogenetic tree are ␤-ARK-1 (group 2), the C-terminal PH domain of pleckstrin (unclassed), and the N-terminal (group 1) and C-terminal PH domains (group 4) of TIAM1. Phylograms are branching tree diagrams assumed to be an estimate of phylogeny, with branch lengths being proportional to the amount of inferred evolutionary change. The phylogram shown here demonstrates the uniqueness of the PH domain CKIP-1 from other known PH domains. the W123A substitution, including the K42C/R44C/W123A mutant (Fig. 7C) and W123A, K42C/W123A, and R44C/W123A mutants (data not shown) exhibited a complete loss of phospholipid binding (Fig. 8C). These observations concur with the cellular localization observed by confocal microscopy. Overall, these results suggest that the highly conserved tryptophan residue is important for the integrity of the entire domain. Mutation of this residue to alanine causes the protein to no longer be associated with the plasma membrane, with a corresponding loss of phospholipid binding.
Examination of Critical Residues in CKIP-1 Important for Interactions with CK2-Based on the differences in subcellular localization and phospholipid binding of the PH domain mutants of CKIP-1 we were interested in examining their effect on the ability of CKIP-1 to interact with CK2␣ in vivo. As a means of achieving this objective, representative FLAG-tagged CKIP-1 mutants (K42C, W123A, K42C/W123A, and K42C/ R44C/W123A) were co-transfected into U2-OS cells with EYFP-CK2␣. As illustrated in Fig. 8A (middle left panel), FLAG-CKIP-K42C caused re-localization of a proportion of CK2␣ to a degree similar to that observed with wild-type FLAG-CKIP-1 (upper right panel). By comparison, CK2␣ was not observed at  Fig. 7, individual mutants are designated according to the individual residue(s) (i.e. K42C, R44C, and/or W123A) that were mutated. EYFP fluorescence was visualized using confocal microscopy. Arrows indicate plasma membrane-localized EYFP-CK2␣. B, in vitro transcription/translation reactions were performed using a reticulocyte lysate-based system with plasmids encoding FLAG-tagged CKIP-1 PH domain mutants (as indicated) under the control of a T7 promoter. The 35 S-labeled in vitro translation products were incubated with either GST or GST-CK2␣ on glutathione-agarose. The 35 S-protein that was bound to the beads as well as ϳ20% of each unbound supernatant was subjected to SDS-polyacrylamide gel electrophoresis and visualized by phosphorimaging analysis. C, graphical representation of results from pull-down assays. Assays were performed in three independent experiments with the percentage of total radiolabeled protein bound to GST or GST-CK2␣ determined by phosphorimaging analysis using ImageQuaNT software. % Bound represents the percentage of the total protein that was bound to GST-CK2␣ or GST (average Ϯ S.D. for data derived from the three independent experiments). the plasma membrane when co-transfected with the W123A, K42C/W123, or K42C/R44C/W123A mutants, an observation consistent with the failure of these mutants to localize to the plasma membrane.
To complement the localization studies, we examined the ability of the PH domain mutants of CKIP-1 to interact directly with CK2␣ using GST pull-down assays with radiolabeled translation products of each of the PH domain mutations prepared by in vitro transcription and translation. Representative autoradiograms of these pull-down assays are shown in Fig. 8B with graphical representations illustrated in Fig. 8C. The K42C and K44C mutants both retained the ability to interact directly with CK2␣ in vitro (Fig. 8, B and C) to the same extent as wild-type CKIP-1. By comparison, all CKIP-1 mutants harboring a W123A substitution, including W123A, K42C/W123A, R44C/W123A, and K42C/R44C/W123A displayed a marked reduction of at least 70% in binding to CK2␣ as compared with wild-type CKIP-1.
Taking into account the results demonstrated in Figs. 7 and 8 (localization studies, phospholipid binding, and in vitro and in vivo interactions with CK2␣), it is apparent that the CKIP-1 PH domain mutations can be grouped into two classes. First, there are the mutations (K42C and/or R44C) that are still localized to the plasma membrane, retain phospholipid binding, and retain interactions with CK2␣ both in vitro and in vivo to the same extent as is seen with wild type CKIP-1. Second, there are the mutations (W123A by itself or in combination with K42C and/or R44C) that are no longer localized to the plasma membrane, have lost phospholipid binding, and have diminished interactions with CK2␣ both in vitro and in vivo. Collectively, these results demonstrate the importance of the integrity of the PH domain of CKIP-1 not only for the plasma membrane localization of CKIP-1, but also for interactions with CK2␣ and recruitment of CK2 to the plasma membrane.

DISCUSSION
In this study, we report the investigation of the interaction between protein kinase CK2 and CKIP-1 together with a functional analysis of CKIP-1. As mentioned previously, CKIP-1 has a number of protein-protein interaction motifs. Based on the presence of these motifs and the discrete cellular localization of CKIP-1, we hypothesized that CKIP-1 is a non-enzymatic regulator of CK2. In this respect, one can envisage the possibility that CKIP-1 functions in a manner analogous to the AKAP family of proteins. Protein kinase A anchoring proteins (AKAPs) bind the regulatory subunit of cAMP-dependent protein kinase (protein kinase A) to direct the kinase to discrete intracellular locations (57)(58)(59)(60). By binding to additional signaling molecules, AKAPs function to coordinate multiple components of signal transduction pathways. CK2, like protein kinase A, has multiple substrates in a variety of cellular locations, and to date, it is unclear how subpopulations of CK2 are targeted to specific cellular locations (1,61). CKIP-1 has domains such as a PH domain and putative LZ that make it ideal as a potential targeting molecule. The possibility also exists that CKIP-1 could act as a scaffold/adaptor protein, similar to the JIPs (JNK-interacting proteins), that coordinates signal transduction events involving CK2␣ and distinct signaling pathways. JIPs were previously identified as scaffolding proteins in the JNK (c-Jun NH 2 -terminal kinase) signaling pathway (62, 63) but were recently found to bind directly to kinesin (64). JIPs act as a linker between kinesin and its cargo, allowing kinesin to transport many different cargoes and to concentrate and respond to signaling pathways at certain sites within the cell (64 -66).
The recruitment of CK2␣ to the plasma membrane by CKIP-1 may bring the CK2 tetramer or CK2␣ itself into com-plexes requiring kinase activity, thereby leading to specific phosphorylation events. In this respect, CK2 phosphorylates a broad range of cellular proteins (61), which includes substrates located at the plasma membrane such as spectrin (24), the insulin receptor (67), caveolin (26), IGF-II receptor (68), dynamin (25), and IRS-1 (69). Although it remains to be determined how the phosphorylation of each of these specific proteins by CK2 is regulated, this present study with CKIP-1 may provide a novel paradigm for understanding how CK2 can phosphorylate targets that are localized at specific sites within cells.
Our results from co-localization studies on CKIP-1 and CK2␣ indicate that CKIP-1 is capable of re-localizing a proportion of the cellular CK2 to the plasma membrane in both U2-OS and Saos-2 cells (Figs. 1 and 2). Although it was apparent that CKIP-1 could recruit CK2 to the plasma membrane, the majority of the CK2 was retained in the nucleus. Consequently, we performed studies to examine the amount of CK2␣ that can be re-localized to the membrane by CK2␤, an integral CK2 subunit to which a membrane localization signal had been fused. These studies demonstrated that, as was the case with CKIP-1, Myc-CK2␤-CAAX is capable of re-localizing only a small fraction of the CK2 to the plasma membrane. In fact, in relation to Myc-CK2␤-CAAX, it is apparent that CKIP-1 is nearly as effective at eliciting the re-localization of CK2␣ to the plasma membrane. This inability of even an integral component of the CK2 complex to re-localize the entire population of CK2␣ may be due in part to the complex regulation of CK2 as well as the diverse and dynamic nature of its interactions within cells (70,71).
It is currently unknown whether the interaction between CKIP-1 involves only the catalytic CK2␣ subunit, or the entire CK2 tetramer. In addition, it has become clear that CKIP-1 interacts with CK2␣ through a region of the N-terminal 330 amino acids, and not the C-terminal, which is unique from CK2␣Ј. It should be noted that protein phosphatase 2A also interacts with CK2␣ but not CK2␣Ј. Similarly, in this case the residues critical for interactions between protein phosphatase 2A and CK2␣ also lie within the N-terminal of CK2␣ (21), where there is a region of non-similarity between the two catalytic subunits.
In an effort to characterize interactions between CK2 and CKIP-1, we focused our attention on a putative HIKE domain in CK2␣. This potential HIKE domain was identified in the C-terminal of CK2␣ in an area where there is non-identity between ␣ and ␣Ј (Fig. 5A). HIKE is a highly conserved sequence motif that selectively occurs in proteins that bind pleckstrin homology domains (47,48). HIKE was identified in strong PH-binding candidates such as G ␤ , protein kinase C, and Akt. HIKE also encompasses a consensus sequence for phosphoinositide binding, specifically charged residues and their spacing appear highly conserved between HIKE and PI binding sequences (48). The HIKE motif in different proteins shares a ␤ strand-loop-␤ strand structure despite the widely different three-dimensional structures displayed by these proteins, with the angle between the two ␤ strands varying considerably in amplitude, suggesting that this may contribute to PH domainbinding specificity (47,48). In keeping with these features, the potential HIKE domain identified in CK2␣ is in a region of a ␤ strand-loop-␤ strand structure. Initial studies on the HIKE domain of CK2␣ were performed by swapping a residue that fits the HIKE motif with the residue found in CK2␣Ј, and vice versa. These constructs were utilized in yeast two-hybrid studies (Fig. 5B), however this residue alone was found not to be responsible for the differences seen between CK2␣ and ␣Ј and their interaction with CKIP-1. These results suggest that this particular residue of the putative HIKE domain is not solely responsible for the interaction between protein kinase CK2 and CKIP-1 and that residues other than this one are likely to be involved in the interaction between CK2 and CKIP-1. Overall, the observed results reflect the complexity of the HIKE domain.
Studies on the PH domain of CKIP-1 have revealed that the domain itself is unique from other PH domains classified on the basis of phosphoinositide binding. From the experiments shown in Fig. 7, we conclude that CKIP-1 exhibits a very broad spectrum of binding to phosphoinositols (Fig. 7C). CKIP-1 also shows uniqueness from other PH domain in the MSA and phylogenetic tree (Fig. 6). By comparison, a recent report, implicating CKIP-1 in muscle cell differentiation (38), showed a more specific phospholipid binding spectrum for CKIP-1 and suggested that its membrane localization was dependent on cell stimulation. We do not have a precise explanation for the apparent discrepancies. However, it is noteworthy that we find that CKIP-1 is consistently localized to the plasma membrane even in the absence of stimulation when introduced into cells. This latter observation is consistent with its broad spectrum phosphoinositol binding (Fig. 7). Polyphosphoinositide binding is a property shared by most, if not all PH domains, however, in only very few cases is phosphoinositide binding by PH domains of high affinity and specificity (72)(73)(74). By far, the majority of PH domain/phosphoinositide interactions reported thus far are of low affinity and display little or no stereospecificity (72). Although "high affinity" PH domains can function independently as signal-related membrane-targeting modules, the functions of "low affinity" PH domains is less clear, although evidence exists supporting the importance of these PH domains and a role for their weak interactions (72). It has also been suggested by Lemmon et al. (72) that the weak phosphoinositide binding by a PH domain may cooperate with interactions mediated by an entirely separate domain within the same protein, as is seen with Tiam-1 (75). The domains that cooperate in membrane association may be in different proteins, and the regulation of oligomer formation could control the membrane targeting, as suggested for dynamin (72,76).
In conclusion, we have demonstrated that CKIP-1 is capable of redistributing a proportion of CK2␣ to the plasma membrane. As noted earlier, the PH domain of CKIP-1 was found to be responsible for the interaction between CK2␣ and CKIP-1, but on its own is not sufficient. Replacement of the PH domain with a myristoylation signal results in the inability of CKIP-1 to recruit CK2␣ to the plasma membrane and causes a loss of interaction between the two in vitro. Based on these findings we speculate that CKIP-1 is acting as a non-enzymatic regulator of CK2, with the ability of sequestering a fraction of the CK2 to the plasma membrane. CKIP-1 may also function as an adaptor protein that integrates signals from CK2 to other signaling molecules or possibly to mediate interactions between CK2␣ and other proteins that could be targets of CK2. Overall, this study has provided a means of how the predominantly nuclear CK2 can have targets at the plasma membrane and provided novel insights into the complex regulation of CK2.