pICln Binds to a Mammalian Homolog of a Yeast Protein Involved in Regulation of Cell Morphology*

Since its cloning and tentative identification as a chloride channel, the function of the pICln protein has been debated. Although there is no consensus regarding the specific function of pICln, it was suggested to play a role, directly or indirectly, in the function of a swelling-induced chloride conductance. Previously, the protein was shown to exist in several discrete protein complexes. To determine the function of the protein, we have begun the systematic identification of all proteins to which it binds. Here we show that four proteins firmly bind to pICln and identify the 72-kDa pICln-binding protein by affinity purification and peptide microsequencing. The interaction between this protein and pICln was verified several ways, including the extraction of several pICln clones from a cDNA library using the 72-kDa protein as a bait in a yeast two-hybrid screen. The protein is homologous to the yeast Skb1 protein. Skb1 interacts with Shk1, a homolog of the p21Cdc42/Rac-activated protein kinases (PAKs). The known involvement of PAKs in cytoskeletal rearrangement suggests that pICln may be linked to a system regulating cell morphology.

Since its cloning and tentative identification as a chloride channel, the function of the pICln protein has been debated. Although there is no consensus regarding the specific function of pICln, it was suggested to play a role, directly or indirectly, in the function of a swellinginduced chloride conductance. Previously, the protein was shown to exist in several discrete protein complexes. To determine the function of the protein, we have begun the systematic identification of all proteins to which it binds. Here we show that four proteins firmly bind to pICln and identify the 72-kDa pICln-binding protein by affinity purification and peptide microsequencing. The interaction between this protein and pICln was verified several ways, including the extraction of several pICln clones from a cDNA library using the 72-kDa protein as a bait in a yeast two-hybrid screen. The protein is homologous to the yeast Skb1 protein. Skb1 interacts with Shk1, a homolog of the p21 Cdc42/Racactivated protein kinases (PAKs). The known involvement of PAKs in cytoskeletal rearrangement suggests that pICln may be linked to a system regulating cell morphology.
Expression of the pICln cDNA in Xenopus laevis oocytes was correlated with the appearance of a nucleotide-sensitive chloride current (I ϭ current, Cl ϭ chloride, n ϭ nucleotide-sensitive) (1)(2)(3)(4). Although pICln was tentatively identified as an integral component of the chloride channel (1), several observations were inconsistent with the channel hypothesis for pICln. First, pICln lacks predicted hydrophobic membranespanning domains and structural homology to known channel proteins (1). Second, in mammalian cells and Xenopus oocytes, pICln was abundant and exhibited a predominantly cytoplasmic and nuclear localization, whereas a small fraction (Ͻ5%) was associated with the cytoskeleton (5). No pICln was de-tected in the plasma membrane.
The chloride conductance associated with expression of pICln was similar to an endogenous Xenopus oocyte chloride current elicited by hypotonic challenge (6). An anti-pICln antibody specifically ablated the swelling-induced chloride current in Xenopus oocytes (5), a finding supported by antisense experiments in mammalian cells (7). For the reasons stated above, we proposed that pICln was a cytosolic regulator of a swelling-induced chloride channel rather than a channel itself (5). In contrast, Paulmichl and co-workers (8) maintain that pICln is the swelling-induced chloride channel itself. Recently, data were presented suggesting that the chloride channel evoked by pICln expression has properties different from the swelling-induced chloride current, including a higher permeability to NO 3 Ϫ , stronger outward rectification, and voltage-dependent nucleotide block (3). The molecular identification of the swelling-induced chloride channel has proven difficult, and several proteins including P-glycoprotein, pICln, ClC-2, and ClC-3 have been proposed to constitute this channel (9,10). Although it seems unlikely that either pICln or P-glycoprotein are themselves chloride channels, both ClC-2 and ClC-3 are well established members of a family of chloride channel proteins. In contrast, pICln exhibits no significant homology to any known mammalian protein and contains no domains that suggest a specific function.
Although work from several laboratories supports a link between pICln expression and activation of a chloride current, the nature of this link is not clear. Currently there are no data to suggest that pICln directly regulates a chloride channel. Indeed, pICln may act far upstream from any plasma membrane-associated event and participate in such diverse functions as transcriptional or translational regulation, cytoskeletal rearrangement, or any one of several signal transduction cascades. pICln was shown previously to exist in several discrete complexes with other cytosolic proteins (5). We reasoned that the identification of proteins interacting with pICln might reveal functional connections to signaling pathways or known cellular mechanisms. Here we report the identification of one such pICln-interacting protein, a 72-kDa protein that appears to be the human homolog of Skb1. Skb1 is a yeast protein that interacts with Shk1, a homolog of the p21 Cdc42/Rac -activated protein kinases (PAKs). 1 Although the function of PAKs are only beginning to be understood, they appear to affect cell morphology through interactions with the cytoskeleton (11).

EXPERIMENTAL PROCEDURES
IBP72 Affinity Purification-Rat pICln coding sequence was subcloned into the pGEX-2T plasmid (Amersham Pharmacia Biotech). The GST-pICln fusion protein was expressed in BL-21 bacteria and purified over glutathione-Sepharose according to the manufacturer's protocols. GST-pICln was immobilized using ActiGel ALD (Sterogene) at 2 mg of protein/ml of gel. Bovine ventricular tissue was minced and homogenized by Polytron (setting 7) for 3 ϫ 30 s in MB buffer (10 mM Na-HEPES (pH 7.5), 20 mM KCl, 1 mM EGTA, 3 mM MgCl 2 , 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and 2 mg/ml each of aprotinin, leupeptin, and pepstatin). Following centrifugation at 100,000 ϫ g, the supernatant (2.6 g of protein) was loaded onto a 2 ϫ 25 cm DEAE-Sephacel (Amersham Pharmacia Biotech) column and * This work was supported by the Howard Hughes Medical Institute and National Institutes of Health training grants (to W. P. and K. W.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  washed with MB containing 100 mM NaCl. pICln-containing complexes (as detected by Western blotting) were eluted with MB ϩ 400 mM NaCl. The eluate was supplemented with Triton X-100 (1% final concentration) and rotated overnight with 100 l of GST-pICln resin. After washing beads with MB, 400 mM NaCl, 1% Triton X-100, bound proteins were solubilized in SDS loading buffer, separated by SDS-polyacrylamide gel electrophoresis, transferred to polyvinylidene fluoride film, and visualized with Coomassie staining. The 72-kDa protein band was excised, digested with trypsin and cyanogen bromide, and microsequenced (Mayo Foundation).
Constructs, Northern Blot, Yeast Two-hybrid Analysis, and Cell Transfection-IBP72 coding sequence was subcloned into pGEX-2T, and the GST-IBP72 fusion protein was produced and purified as described above. The full-length IBP72 clone was subcloned into pCDNA 3.1(Ϫ) (Invitrogen) and translated in vitro using the TNT system (Promega). pICln deletions were generated by polymerase chain reaction (PCR) subcloning of truncated versions of human pICln coding sequence into pCDNA3.1(ϩ) or (Ϫ); all constructs were "tagged" with an aminoterminal FLAG epitope (DYKDDDDK). 10 g of each construct was used for calcium phosphate transfection of 50 -70% confluent HEK293 cells plated in 10-cm culture dishes. 48 h after transfection, proteins were in vivo labeled for 5-12 h using 50 Ci/ml [ 35 S]methionine (Amersham). IBP72 coding sequence was subcloned into pBTM-116KN vector (18) using restriction sites introduced by PCR. The resultant construct was verified by DNA sequencing and used as the bait for twohybrid screening of a human heart Matchmaker library (CLONTECH) in the yeast strain L40 (18). Full-length, FLAG-tagged pICln coding sequence was cloned into pGAD424 (CLONTECH). Yeast his3 expression was assayed by growth on dropout plates lacking histidine, tryptophan, and leucine and supplemented with 5 mM 3-aminotriazole. ␤-Galactosidase activity was measured using o-nitrophenyl-␤-D-galactopyranoside (Sigma) as substrate. The multi-tissue human Northern blot (CLONTECH) was probed according to the manufacturer's specifications with a [ 32 P]dCTP random-labeled fragment (Stratagene) consisting of the entire human IBP72 human coding sequence.
Cell Lysis, Immunoprecipitation, and Immunoblotting-Total cell lysate for immunoprecipitation was obtained as 100,000 ϫ g supernatant after cell lysis in buffered solution containing 1% Triton X-100 and 350 mM NaCl. Soluble proteins were isolated by Dounce homogenization followed by pelleting of microsomal fractions at 40,000 rpm in an SW-55 rotor for 30 min at 4°C. The supernatant was supplemented with Triton X-100 to a final concentration of 1% and with NaCl to 350 mM. For immunoprecipitation, samples were precleaned with 40 l of Protein A/G-Sepharose (Amersham Pharmacia Biotech) for 1.5 h at 4°C. pICln was immunoprecipitated using the aFP antibody as described (5). GST-pICln (1 g) was used for precipitation of in vitro translated IBP72. Precipitation of the deletion constructs and interacting proteins was achieved using Protein G-Sepharose and anti-FLAG M2 monoclonal antibody (Kodak). The rabbit anti-IBP72 antibody was generated against the GST-IBP72 fusion protein and affinity-purified.

RESULTS
pICln was immunoprecipitated from [ 35 S]methionine-labeled Madin-Darby canine kidney (MDCK) total cell lysates using a polyclonal antibody generated to a GST-pICln fusion protein. Several proteins consistently co-immunoprecipitated with pICln (pICln-binding proteins (IBP)) with electrophoretic mobilities corresponding to molecular masses of 72, 43, 29, and 17 kDa (Fig. 1, lane 1). Since the same set of associated proteins was co-immunoprecipitated from the water-soluble cell fraction, we concluded that the IBPs are not membrane-associated proteins. Their association with pICln was judged to be specific because the same set of proteins was co-immunoprecipitated with a different anti-pICln antibody (data not shown).
IBP72 was purified by affinity to pICln. Cytosolic extracts from bovine heart were used as the source of IBP72 since initial experiments indicated that it was relatively abundant in this tissue. IBP72 was enriched significantly in eluates from the GST-pICln resin (Fig. 1, lane 2). Immobilized GST did not bind this protein, indicating that the 72 kDa protein interacted with GST-pICln specifically (Fig. 1, lane 3). The purified 72-kDa protein was digested with trypsin and cyanogen bromide, and five different peptides were sequenced. The peptide sequences obtained from the 72-kDa protein were used to screen the expressed sequence tag (EST) data base. Several overlapping clones were identified that predicted a single open reading frame (ORF) of 1911 base pairs and whose translation contained sequences identical with the 72-kDa protein-derived peptides. Subsequently, two EST clones containing an identical 2.4-kb insert were identified (GenBank TM accession numbers R13970 and AA099674) that spanned the ORF. Using 5Ј-rapid amplification of cDNA ends with a human fetal brain library, an additional 60 bases of 5Ј-untranslated sequence were identified.
The ORF predicts a protein of 637 amino acids with a molecular mass of 72.6 kDa ( Fig. 2A). Consistent with this prediction, in vitro translation of the cDNA yielded a protein with an apparent molecular mass of 72 kDa (Fig. 2B). Only two residues are not conserved between the bovine and human proteins within the sequence specified by the five fragments. A 2.4-kb transcript for IBP72 was identified in a wide range of human tissues including skeletal muscle, brain, heart, placenta, kidney, pancreas, lung, and liver (Fig. 2C). Although the cloned 72-kDa protein has no significant homology with other cloned mammalian proteins and contains no consensus structural motifs, it does exhibit moderate homology to putative proteins encoded in the Caenorhabditis elegans and Saccharomyces cerevisiae genomes, as well as significant homology to the skb1 gene product (12) from Schizosaccharomyces pombe (52% homology; Fig. 2A). Recently, a human cDNA identical with our IBP72 was cloned by homology to Skb1 and submitted to GenBank TM (accession number AF015913).
The cloned 72-kDa protein appears to be IBP72 as indicated by several approaches. First, the in vitro translated protein exhibited specific binding to the GST-pICln fusion protein but not to GST alone (Fig. 3, left panel). Second, an affinity-purified polyclonal antibody raised against the recombinant 72-kDa protein recognized IBP72 co-immunoprecipitated with pICln (Fig. 3, right panel). Third, a LexA-IBP72 fusion protein specifically interacts with a GAL4 activation domain-ICln fusion protein in the yeast two-hybrid system (Table I). Moreover, when a human heart cDNA library was screened in the yeast two-hybrid system using the LexA-IBP72 fusion protein as bait, six independent ICln clones were obtained. These results argue strongly that our cloned protein is IBP72.
To identify the domain(s) of pICln critical for interaction with IBP72, we generated several epitope-tagged (FLAG) human pICln deletion constructs and examined their ability to bind native IBP72. The full-length FLAG-pICln protein and endogenous pICln interacted with the same set of proteins in human embryonic kidney (HEK293) cells (data not shown). All deletion constructs were expressed at levels equivalent to or higher than the full-length FLAG-pICln protein, as assessed by immunocytochemical analysis and immunoprecipitation of [ 35 S]methionine-labeled proteins (data not shown). Based on this approach, we conclude that the extreme carboxyl terminus of human pICln, specifically the last 37 amino acids, is critical for interaction with IBP72 (Fig. 4). Consistent with this result, one of the pICln clones identified by yeast two-hybrid selection

TABLE I
Interaction of IBP72 with pICln in the yeast two-hybrid system Yeast-harboring plasmids expressing the indicated proteins were assayed for growth on selective plates containing 5 mM 3-aminotriazole (AT) and for ␤-galactosidase activity. ␤-Galactosidase activities are given in ␤-galactosidase units (19) and represent the average Ϯ S.D. of three independent measurements. GAD is the Gal4 activation domain (amino acids 768 -881). GAD-pICln-DN103 is a pICln clone that lacks the amino-terminal 103 amino acids.
pICln Interacts with Mammalian Homolog of Skb1 has the amino-terminal 103 residues deleted but retains full ability to interact with IBP72 (Table I).

DISCUSSION
In an effort to identify the functional role of pICln, we are characterizing IBPs. Four major IBPs consistently co-purify with pICln in several tissues. We cloned the human cDNA for IBP72 based on microsequence data obtained from affinitypurified bovine IBP72. The interaction between the cloned human IBP72 and pICln was confirmed by several lines of evidence, including the extraction of pICln from a cDNA library using the full-length coding sequence for IBP72 as the bait in a yeast two-hybrid screen. IBP72 is ubiquitously expressed and has no identified mammalian homologs or recognizable structural motifs that would suggest a specific function. Currently we are cloning the other IBPs and will use similar approaches to verify the specificity of their interaction with pICln.
Although IBP72 has no known human homologs, sequence similarity suggests that IBP72 represents a human homolog of the Skb1 protein from S. pombe (12). Skb1 was identified by a yeast two-hybrid screen using the S. pombe protein kinase Shk1 as bait. The Shk1 kinase is linked to Ras-and Cdc42-dependent signaling cascades regulating cell viability, morphology, and mitogen-activated protein kinase-mediated pheromone responses (13). S. pombe lacking Skb1 are less elongated than wild-type yeast, emphasizing the role of this protein in the regulation of cell morphology. Shk1 is a homolog of the mammalian PAK, of which there are three cloned isoforms (14). PAK kinases are activated by GTP-bound forms of the small GTPbinding proteins Rho, Rac, and Cdc42 and have been implicated in control of cytoskeletal rearrangement and cell morphology (11,14).
One consistent conclusion from previous studies of pICln function is that its overexpression induces, either directly or indirectly, the appearance of a chloride conductance (1,2,4). Given the biochemical characteristics of pICln, we favor the hypothesis that pICln is not a channel itself but rather part of a pathway either closely or remotely connected to a chloride current, possibly through cytoskeletal rearrangement. Indeed, actin co-immunoprecipitated with pICln, and a fraction of pICln associated with insoluble cytoskeletal elements (5). The protein identified in this report, IBP72, may provide a link between pICln and cytoskeletal rearrangement. Regulation of swelling-induced chloride channels is likely to involve cytoskeletal rearrangement (15,16). Also, recent evidence links p21 Rhodependent cytoskeletal reorganization to a swelling-induced chloride conductance (17). Whether pICln and IBP72 are linked to a swelling-induced chloride current (5-7), a volume-insensitive chloride conductance (2, 3), or both will be determined only by understanding all elements of the pathway. pICln Interacts with Mammalian Homolog of Skb1