Identification and Characterization of PS-GAP as a Novel Regulator of Caspase-activated PAK-2*

p21-activated protein kinase (PAK)-2 is a member of the PAK family of serine/threonine kinases. PAKs are activated by the p21 G-proteins Rac and Cdc42 in response to a variety of extracellular signals and act in pathways controlling cell growth, shape, motility, survival, and death. PAK-2 is unique among the PAK family members because it is also activated through proteolytic cleavage by caspase-3 or similar proteases to generate the constitutively active PAK-2p34 fragment. Activation of full-length PAK-2 by Rac or Cdc42 stimulates cell survival and protects cells from cell death, whereas caspase-activated PAK-2p34 induces a cell death response. Caspase-activated PAK-2p34 is rapidly degraded by the 26 S proteasome, but full-length PAK-2 is not. Stabilization of PAK-2p34 by preventing its polyubiquitination and degradation results in a dramatic stimulation of cell death. Although many proteins have been shown to interact with and regulate full-length PAK-2, little is known about the regulation of caspase-activated PAK-2p34. Here, we identify PS-GAP as a regulator of caspase-activated PAK-2p34. PS-GAP is a GTPase-activating protein for Cdc42 and RhoA that was originally identified by its interaction with the tyrosine kinase PYK-2. PS-GAP interacts specifically with caspase-activated PAK-2p34, but not active or inactive full-length PAK-2, through a region between the GAP and SH3 domains. The interaction with PS-GAP inhibits the protein kinase activity of PAK-2p34 and changes the localization of PAK-2p34 from the nucleus to the perinuclear region. Furthermore, PS-GAP decreases the stimulation of cell death induced by stabilization of PAK-2p34.

In multicellular organisms, cell metabolism needs to be tightly regulated by extracellular signals and intracellular signaling pathways. Because of the variety of signals cells receive, the precise balance and modulation of various signals are critical for normal function. Dysregulation of cell signaling pathways can result in cell death or malignant transformation. Protein kinases play a critical role in modulation of a wide variety of signals, and many protein kinases have been identified as oncogenes or tumor suppressor genes, demonstrating their critical role in cell signaling (1,2).
PAKs have been implicated in a variety of cellular functions, including regulation of cell shape and motility through effects on the actin cytoskeleton and integrin signaling pathways and regulation of cell survival and death (14 -19). PAKs appear to accomplish these different functions by interaction with a variety of other signaling molecules. The best known PAK interaction partners are the p21 monomeric G-proteins Cdc42 and Rac. Indeed, the PAK family was first identified in an overlay screen for proteins that interact with activated Rac (3). Binding of Cdc42 and Rac to the so-called PBD (p21-binding domain) or CRIB (Cdc42/Rac-interactive binding) domain results in activation of PAKs. Interestingly, sphingolipids also interact with the same region and activate PAKs (20). Additionally, the adaptor protein Nck binds to a proline-rich motif within the regulatory domain of PAK (residues 12-16 of PAK-1) and has been implicated in recruiting PAKs to activated growth factor receptor complexes (21)(22)(23). Pix/COOL proteins, which are guanine nucleotide exchange factors, have also been shown to modulate PAK activity through binding to a third, atypical proline-rich region within the regulatory domain of PAK (residues 187-196 of PAK-1) (24 -26). More recently, PAK-3 has been shown to interact with paxillin, which acts as a scaffolding adaptor protein in integrin signaling, through a region within the PAK regulatory domain that may include the Nck-binding site (27). Paxillin can compete with Nck for PAK binding; surprisingly, it also appears to compete with Pix, even though they are believed to interact with distinct regions within the regulatory domain of PAK.
PAK-1, PAK-2, and PAK-4 have been shown to suppress cell death and to promote cell survival through phosphorylation of the pro-apoptotic protein Bad (28 -31). Constitutive activation of PAKs appears to be involved in malignant transformation, cancer development, and cancer cell invasion. Expression of constitutively active PAK-4 results in anchorage-independent growth (32,33). Elevated protein or activity levels of PAK-1, PAK-2, and PAK-4 have been detected in various cancer cell lines, and elevated PAK activity has been shown to be required for proliferation of MDA-MB435 breast cancer cells (33)(34)(35). Dominant-negative PAK constructs reduce invasion of MDA-MB435 breast cancer cells (35).
PAK-2 is unique among the PAKs because of the existence of a cleavage site for caspase-3 or a caspase-3-like protease within the regulatory domain. Proteolytic cleavage C-terminal of Asp 212 removes most of the regulatory domain, generating a constitutively active PAK-2p34 catalytic fragment (36,37). Caspase-mediated generation of PAK-2p34 has been observed in response to a variety of apoptotic stimulants (31,36,38,39). Additionally, ectopic expression of PAK-2p34 stimulates cell death (40 -42). Therefore, PAK-2 appears to have dual and opposing functions in the regulation of cell survival and death. Activated full-length PAK-2 stimulates cell survival and suppresses cell death, whereas proteolytically activated PAK-2p34 induces a cell death response. We have shown recently that localization and protein levels of PAK-2p34 are tightly regulated (42). Subcellular localization of PAK-2 is regulated by nuclear export and nuclear localization signals. In full-length PAK-2, the nuclear export signal dominates over the nuclear localization signal, resulting in cytoplasmic localization. Caspase cleavage disrupts the nuclear export signal and results in nuclear accumulation of PAK-2p34. Protein levels of PAK-2p34 are regulated by ubiquitination and degradation by the 26 S proteasome. Caspase-activated PAK-2p34 is rapidly degraded by the 26 S proteasome, but full-length PAK-2 is not. Expression of epitope-tagged ubiquitin stabilizes PAK-2p34 by preventing its polyubiquitination and degradation, and stabilization of PAK-2p34 results in dramatic stimulation of programmed cell death (42).
Interestingly, cleavage of PAK-2 to the PAK-2p34 fragment removes interaction sites for Nck, Pix/COOL proteins, paxillin, and Cdc42/Rac/sphingolipids, thereby freeing the PAK-2p34 fragment from these known regulators of PAK signaling. Here, we report the identification of a novel PAK-2 regulator, PS-GAP. PS-GAP is a GTPase-activating protein (GAP) for Cdc42 and RhoA previously identified by interaction with the tyrosine kinase PYK-2 (43). PS-GAP interacts selectively with caspaseactivated PAK-2p34 both in vitro and in vivo, but does not interact with full-length PAK-2. The interaction with PS-GAP regulates the activity and subcellular localization of caspaseactivated PAK-2p34. PS-GAP inhibits the protein kinase activity of PAK-2p34 in vitro and changes the localization of PAK-2p34 from the nucleus to the perinuclear region. Furthermore, PS-GAP appears to regulate the ability of caspase-activated PAK-2p34 to induce programmed cell death. Expression of PS-GAP reduces levels of cell death induced by stabilization of PAK-2p34. PS-GAP is the first identified protein that specifically regulates pro-apoptotic caspase-activated PAK-2p34, but not anti-apoptotic full-length PAK-2, a critical step in elucidating the pro-apoptotic PAK-2p34 signaling pathway.

EXPERIMENTAL PROCEDURES
Materials-Yeast strain PJ69-4a was a generous gift from Dr. P. James (University of Wisconsin, Madison, WI) (44). Yeast strain Y190, two-hybrid vectors pAS2-1 and pACT2, the Advantage 2 PCR kit, mouse heart Marathon-Ready cDNA, the monoclonal anti-green fluorescent protein (EGFP) Living Colors antibody, and expression vector pRevTRE were obtained from Clontech. The Frozen-EZ Yeast Transformation II kit was from Zymo Research. Restriction enzymes and T4 DNA ligase were obtained from New England Biolabs Inc. The QIAprep spin miniprep kit and plasmid midi kit and the pDRIVE PCR cloning kit were purchased from QIAGEN Inc. The Geneclean III kit was from BIO 101, Inc. Biomax MR and Biomax MS autoradiography films were from Eastman Kodak. The Thermoscript RT-PCR kit, Platinum Taq polymerase, plasmid pcDNA3.1, Dulbecco's modified Eagle's medium, Dulbecco's phosphate-buffered saline (D-PBS), trypsin/EDTA, Express Five SFM medium, Cellfectin, anti-Myc monoclonal antibody, and customized oligonucleotide primers were obtained from Invitrogen. Fetal bovine serum was from Hyclone Laboratories. The baculovirus expression vector pAcG2T and BaculoGold baculovirus helper DNA were from Pharmingen. Genejammer transfection reagent and Escherichia coli XL2-Blue were from Stratagene. TransIT-LT1 transfection reagent was from Mirus. Tris-buffered saline and SuperSignal chemiluminescent reagent were from Pierce. Bacterial expression vector pGEX2-T and reduced glutathione-Sepharose were obtained from Amersham Biosciences. Immuno-Fluore mounting medium and [␥-32 P]GTP were from ICN Biomedicals, Inc. [␥-32 P]ATP was obtained from PerkinElmer Life Sciences. Mouse RNA from various tissues was a gift from Dr. S. Duncan (Medical College of Wisconsin). Rabbit anti-PS-GAP polyclonal antibody was a gift from Dr. W.-C. Xiong (University of Alabama at Birmingham, Birmingham, AL). Anti-FLAG monoclonal antibody, agarose-conjugated anti-FLAG monoclonal antibody, and anti-hemagglutinin (HA) monoclonal antibody were obtained from Sigma. Agaroseconjugated anti-HA polyclonal antibody was obtained from Santa Cruz Biotechnology. Mammalian expression vectors pExpress/HA and pRet-roIRES/GFP were generated previously (42,45). Mammalian expression clones for FLAG-tagged PAK-2p34 and PAK-2p34-K278R in pRet-roIRES/GFP (42) and for FLAG-tagged PAK-2, PAK-2-L106F, and PAK-2-K278R in pRevTRE 2 were generated previously. Plasmid pMT107 encoding His-tagged ubiquitin (46) was provided by Dr. D. Bohmann (University of Rochester, Rochester, NY). The TNT T7 coupled reticulocyte lysate system for in vitro transcription/translation was from Promega.
Yeast Two-hybrid Library Screening and Analysis-The kinase-deficient PAK-2p34-K278R mutant was subcloned into the Gal4 DNAbinding domain vector pAS2-1 in-frame with the Gal4 DNA-binding domain. The PAK-2p34-K278R bait was used to screen a mouse embryonic fibroblast cDNA library in vector pACT2 (47). Yeast PJ69-4a cells were cotransformed with 0.2 mg of both bait and library plasmids using the Frozen-EZ yeast transformation II kit and plated onto leucine-, tryptophan-, and histidine-deficient medium. Yeast cells were allowed to grow and form colonies on this medium for 2 weeks before replating onto adenine-deficient medium. Adenine-deficient medium allows for stringent growth selection of yeast cells that are capable of transcribing the ADE2 reporter gene, which is under the control of the GAL4 promoter in strain PJ69-4a (44). All colonies capable of growth on adeninedeficient medium were considered potential positive clones. Plasmid DNA was isolated from these clones by digestion with lyticase using QIAGEN spin column miniprep kits. Plasmid DNA was propagated in E. coli XL2-Blue and analyzed by restriction digestion and DNA sequencing. Unique library clones were transformed into yeast strain Y190 with various PAK-2 constructs in pAS2-1 or in a pAS2-1 empty vector control. Filter-lift ␤-galactosidase assays were performed to assay for interactions and to eliminate false positives.
Glutathione S-Transferase (GST) Pull-down Assays-GST fusion proteins of wild-type and mutant PAK-2 were expressed in E. coli or baculovirus-infected insect cells and purified as described previously (37,48). p/K18 cDNA was subcloned into pcDNA3.1, and protein was synthesized and labeled with [ 35 S]methionine by a coupled in vitro transcription/translation reaction. An aliquot of 10 g of GST fusion protein was mixed with 10% of the in vitro transcription/translation reaction in 200 l of pull-down buffer (50 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM EGTA, 1.5 mM MgCl 2 , and 1% Triton X-100) plus 0.8% bovine serum albumin and incubated on ice for 1 h. An aliquot of 20 l of glutathione-Sepharose beads was added, and the mixture was incubated for 1 h at 4°C. Beads were washed with pull-down buffer and analyzed by SDS-PAGE and autoradiography.
Molecular Cloning-To obtain full-length PS-GAP cDNA, mouse heart cDNA was amplified by PCR (49) using the forward primer PS-GAP-5Ј (5Ј-TTGTGTTCATATGGGGCTGCAGCCCCTGGAGTTTA), corresponding to the start codon region of PS-GAP (43), and the reverse primer PS-GAP-3Ј (5Ј-ACATTCTAGACTACAGGAGCTTGACATAATT-CTGTGGA), corresponding to the stop codon region of clone p/K18 and elongated using the Advantage 2 PCR kit according to the manufacturer's instructions. PCR products were purified by agarose gel electro-phoresis and extraction using the Geneclean III kit and ligated into the pDRIVE vector. Clones were identified through plasmid isolation using QIAprep spin miniprep kits, followed by restriction digestion and agarose gel electrophoresis. Selected clones were analyzed by DNA sequencing. Sequences were assembled and analyzed using Vector NTI Suite 7.1 (Informax). BLAST (NCBI Protein Database) was used for homology searches. To facilitate subcloning of PS-GAP, an internal BamHI site in PS-GAP-a was disrupted by site-directed mutagenesis according to the megaprimer PCR method (50,51) without changing the amino acid sequence. The megaprimer was amplified with PS-GAP-5Ј and ⌬BamHI-3Ј (TCCTCTTCGAGCTCCGCTTCGTGCGCGCGCGTAT-CCTCTCCCGGAACCA) using full-length PS-GAP-a as a template. Then, the megaprimer was used together with PS-GAP-3Ј to amplify full-length PS-GAP-a. The cDNA encoding PS-GAP-a with the disrupted BamHI site was subcloned into pExpress/HA and pExpress/Myc for expression in mammalian cells.
Cell Culture and Transfection-Human embryonic kidney 293T cells (American Tissue Culture Collection) were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and penicillin/streptomycin and grown at 37°C in a humidified atmosphere of 5% CO 2 . For transfection, cells were seeded at densities that will allow them to reach 50% confluency within 16 -24 h. Plasmid DNAs for epitope-tagged PS-GAP-a and PAK-2 constructs were transfected into cells using Genejammer or TransIT-LT1 transfection reagent. At 48 h after transfection, cells were harvested and lysed. To stabilize recombinant PAK-2p34, cells were cotransfected with pMT107 encoding Histagged ubiquitin (42).
Immunoprecipitation and Western blotting-293T cells transfected with epitope-tagged PS-GAP-a and PAK-2 constructs were lysed in modified radioimmune precipitation assay buffer (50 mM HEPES (pH 7.5), 150 mM NaCl, 1% Triton X-100, 0.25% deoxycholate, 10 g/ml aprotinin, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, and 0.2 mM sodium orthovanadate). Protein concentrations were determined by the Bradford assay using bovine ␥-globulin as a protein standard. 500 g of lysate protein was diluted with PBS to a final concentration of ϳ5 g/l and incubated overnight with 20 l of agarose-conjugated anti-FLAG or anti-HA antibody at 4°C. Immunocomplexes were washed with pull-down buffer (50 mM HEPES (pH 7.5), 150 mM NaCl, 1 mM EGTA, 1.5 mM MgCl 2 , and 1% Triton X-100) and analyzed by Western blotting. Western blotting was performed using cell lysates (30 g of protein) or immunoprecipitates (from 150 -250 g of cell lysate) by SDS-PAGE, followed by semidry transfer to polyvinylidene membranes. Chemiluminescence detection was performed using SuperSignal reagent and horseradish peroxidaseconjugated secondary antibodies.
Rho-GAP Assays-RhoA, Rac1, and Cdc42 were expressed in E. coli XL2-Blue as GST fusion proteins using pGEX2-T, whereas PS-GAP-a was expressed as a GST fusion protein in TN-5B1-4 cells using pAcG2T. Recombinant GST fusion proteins were absorbed to glutathione-Sepharose as described above and eluted with 10 mM reduced glutathione. Purified GST-RhoA, GST-Rac1, and GST-Cdc42 were dialyzed overnight at 4°C in 40 mM Tris-HCl (pH 7.5), 5 mM EDTA, 1% ␤-mercaptoethanol, and 10% glycerol. For Rho-GAP assays, G-proteins were preloaded at 30°C for 5 min using 50 mM Tris-HCl (pH 7.6), 2 mM EDTA, 100 mM NH 4 Cl, 0.5 mg/ml bovine serum albumin, 1 mM dithiothreitol, and 0.1 mM [␥-32 P]GTP (800 cpm/pmol) and placed on ice. MgCl 2 and GTP were added to 12 and 2 mM, respectively. GTPase activity was monitored by incubating preloaded Rho GTPase for 5 min at room temperature in the absence or presence of PS-GAP and spotting on nitrocellulose filters. Filters were washed with 10 ml of wash buffer (50 mM Tris-HCl (pH 7.6), 100 mM NH 4 Cl, 1 mM MgCl 2 , and 7 mM ␤-mercaptoethanol) and dried, and the remaining [␥-32 P]GTP was analyzed by scintillation counting. The effects of PAK-2p34 on GTPase activity were measured by preincubation of PS-GAP-a with 10 mM MgCl 2 and 200 M ATP for 15 min at 30°C in the presence or absence of GST-PAK-2p34 prior to performing the GAP assay as outlined above.
Kinase assays were performed in the presence and absence of purified GST fusion proteins of PS-GAP containing the region between the GAP and SH3 domains, the SH3 domain, or the region between the GAP and SH3 domains plus the SH3 domain. Autophosphorylation of PAK-2p34 was analyzed by SDS-PAGE on 12% gels and autoradiography. 32 P incorporation into myelin basic protein or histone H4 was analyzed by SDS-PAGE on 12 or 15% gels, respectively, followed by scintillation counting of excised protein bands.
Immunofluorescence-293T cells plated on polylysine-coated chamber slides were transfected with pRevTRE-HA-PS-GAP-a and pRe-vTRE-EGFP-PAK-2p34. At 48 h after transfection, cells were washed twice with D-PBS and fixed with 4% paraformaldehyde in D-PBS for 15 min at room temperature. Cells were washed twice with PBS and permeabilized using D-PBS containing 0.2% Triton X-100 for 10 min. Cells were washed for 5 min with D-PBS containing 0.1% Tween 20 and blocked in D-PBS containing 10% goat serum, 100 mM ethanolamine, and 0.1% Tween 20. Cells were washed with D-PBS plus 0.1% Tween 10 and incubated overnight at 4°C in D-PBS containing 5% goat serum, 0.1% Tween 20, and 10 g/l Alexa Fluor 595-conjugated anti-HA antibody. Cells were washed three times with D-PBS containing 0.2% Tween 20 and once with D-PBS. Nuclei were stained by incubation in 0.05 g/ml 4Ј,6-diamidino-2-phenylindole in D-PBS for 5 min. Cells were then washed once with D-PBS, and slides were mounted using Immuno-Fluore mounting medium. Immunofluorescence was analyzed by fluorescence microscopy.
Analysis of Programmed Cell Death-293T cells were cotransfected with constructs for EGFP-PAK-2p34, His-tagged ubiquitin (His-Ub), and/or Myc-PS-GAP-a. Cells were stained with 10 g/ml Hoechst 33342 for 10 min and analyzed by fluorescence microscopy. To determine the levels of programmed cell death, 500 cells expressing EGFP-PAK-2p34 were counted and analyzed for apoptotic chromatin condensation.

Identification of a Novel PAK-2-interacting
Protein-To identify novel proteins that interact with PAK-2, we performed a yeast two-hybrid screen in a mouse embryonic fibroblast cDNA library. So far, no substrates or regulators have been identified for caspase-activated PAK-2p34. Therefore, we used a construct containing residues 213-524, which correspond to the caspase-cleaved PAK-2p34 fragment, as a bait. Transformation of yeast with PAK-2p34 results in an ϳ10-fold lower number of yeast colonies and delayed growth compared with the kinasedeficient PAK-2p34-K278R mutant. PAK-2p34 is also toxic in E. coli 2 and induces cell death in mammalian cells (40 -42). Therefore, we used the kinase-deficient PAK-2p34-K278R mutant for the yeast two-hybrid screening. As a result of the library screening, we isolated 10 clones that selectively induced the reporter genes when cotransfected with PAK-2p34-K278R, but not with an empty vector control. Partial DNA sequencing revealed that clones p/K1 and p/K6 and clones p/K4, p/K11, p/K14, and p/K17 were duplicates, whereas clones p/K5, p/K15, p/K16, and p/K18 were unique. Sequence analysis and a BLAST search of the NCBI Database revealed that clone p/K18, an ϳ600-bp cDNA, has homology to the C terminus of Rho-GAP proteins such as Graf and oligophrenin (52)(53)(54). A subsequent BLAST search several months later revealed that clone p/K18 encodes a 191-amino acid C-terminal fragment of PS-GAP, a Rho-GAP recently identified in a yeast two-hybrid screen using the tyrosine kinase PYK-2 as a bait (43).
The interaction between p/K18 and PAK-2 was further examined by yeast two-hybrid analyses and GST pull-down assays (Fig. 1). Clone p/K18 was cotransformed into yeast with PAK-2p34-K278R, wild-type PAK-2, kinase-deficient PAK-2-K278R, constitutively active PAK-2-T402E, the regulatory domain of PAK-2, or an empty vector as a negative control. Interaction was examined by filter-lift assay for ␤-galactosidase activity (Fig. 1A). Cotransformation of clone p/K18 with PAK-2p34-K278R resulted in strong induction of ␤-galactosid-ase activity, whereas cotransformation with the empty vector as a negative control showed no induction of ␤-galactosidase activity, indicating that induction requires interaction with PAK-2. However, cotransformation of full-length wild-type, kinase-deficient, or constitutively active PAK-2 or the regulatory domain fragment of PAK-2 with clone p/K18 did not result in induction of ␤-galactosidase activity. Therefore, clone p/K18 interacts specifically with PAK-2p34, but not full-length PAK-2, in yeast.
To verify the results obtained by yeast two-hybrid analysis, the interactions between clone p/K18 and the various PAK-2 constructs were examined by GST pull-down assays. p/K18 protein was synthesized and labeled with [ 35 S]methionine by in vitro transcription/translation and incubated with GST fusion proteins of PAK-2p34-K278R, wild-type PAK-2, kinase-deficient PAK-2-K278R, constitutively active PAK-2-T402E, the regulatory domain of PAK-2, or GST alone as a negative control. Interaction was detected by GST pull-down with glutathione-Sepharose, followed by SDS-PAGE and autoradiography (Fig. 1B). p/K18 protein was pulled down by GST-PAK-2p34-K278R, but not by GST alone, indicating specificity of the interaction. However, no p/K18 protein was pulled down by GST fusions of full-length wild-type, kinase-deficient, or constitutively active PAK-2 or by the regulatory domain alone. Therefore, clone p/K18 also interacts specifically with PAK-2p34, but not full-length PAK-2, in vitro.
Molecular Cloning of Three Variants of PS-GAP-To isolate a cDNA clone encoding full-length PS-GAP, we performed PCR with mouse brain cDNA using a 5Ј-primer derived from the published sequence for PS-GAP (43) and a 3Ј-primer derived from p/K18. Restriction digest analysis indicated that we obtained three unique transcripts, each of which had been amplified by at least two independent PCRs. Sequencing revealed that we isolated three variants of PS-GAP that were probably generated by alternative splicing (Fig. 2). PS-GAP-a, the long-est splice variant, encodes a 786-amino acid protein that contains a pleckstrin homology (PH) domain from residues 276 to 375, a Rho-GAP domain from residues 398 to 550, and a Cterminal SH3 domain from residues 732 to 786. Compared with PS-GAP-a, splice variant PS-GAP-b contains a gap of 66 nucleotides in the region between the PH and GAP domains and the 5Ј-region of the Rho-GAP domain. This sequence gap preserves the reading frame and results in PS-GAP-b, a protein of 767 amino acids that lacks 10 residues in the region between the PH and GAP domains as well as the 12 N-terminal residues of the GAP domain. Compared with PS-GAP-a, splice variant PS-GAP-c contains a gap of 153 nucleotides within the region between the GAP and SH3 domains. This nucleotide sequence gap preserves the reading frame and results in PS-GAP-c, a 735-amino acid protein that lacks 51 residues within the region between the GAP and SH3 domains. The additional splice variant PS-GAP-s, which lacks the 103 N-terminal residues and starts at Met 104 , was described previously (43).
PS-GAP-a, PS-GAP-b, and PS-GAP-c differ from the previously described PS-GAP-m and PS-GAP-s in a stretch of 4 consecutive amino acids within the region between the PH and GAP domains. In addition, there are several single nucleotide polymorphisms, some of which result in amino acid polymorphisms. These polymorphisms were observed in several clones that were amplified by independent PCRs, suggesting that they correspond to genetic diversity between different animals. The cDNA used in the PCR amplifications was from a pool of 200 mice. Seven amino acid polymorphisms exist between PS-GAP-a, PS-GAP-b, and PS-GAP-c. Four more exist between these variants and the previously described PS-GAP-m and PS-GAP-s variants.
Expression of PS-GAP Variants-The mouse brain cDNA used to amplify full-length PS-GAP was derived from a pool of 200 mice. Therefore, differences between the three isolated PS-GAP variants could be due to differences between individual animals. To determine whether the sequence gaps observed in PS-GAP-b and PS-GAP-c correspond to polymorphisms in different animals or represent splice variants within the same animal, we performed RT-PCR using mRNA from brain, heart, kidney, lung, and testes of a single mouse as well as mRNA from BALB/3T3 mouse fibroblasts (Fig. 3). Primer sets flanking the gaps observed in PS-GAP-b and PS-GAP-c were used in RT-PCR experiments to amplify PS-GAP cDNA. PS-GAP-a, PS-GAP-b, and PS-GAP-c cDNA clones were amplified as positive controls and markers. Amplification of PS-GAP-a and PS-GAP-c with primers flanking the gap at the start of the GAP domain (P1/P2) produced an ϳ400-bp PCR product, whereas amplification of PS-GAP-b, which lacks a sequence encoding 22 amino acids within this region, produced a PCR product of ϳ340 bp. Amplification of PS-GAP-a and PS-GAP-b using primers flanking the gap in the region between the GAP and SH3 domains (P3/P4) produced an ϳ480-bp PCR product, whereas amplification of PS-GAP-c, which lacks a sequence encoding 51 amino acids within this region, produced a PCR product of ϳ330 bp.
When primer pair P1/P2 was used to amplify transcripts from mouse tissues and BALB/3T3 fibroblasts, PCR products of 400 and 340 bp corresponding to the isolated variants were generated. Both PCR products were generated in all tissues tested, but relative amounts varied between tissues. For example, the 340-bp PCR product was present at the lowest levels in testes and BALB/3T3 fibroblasts. In addition, products Ͼ400 bp and Ͻ340 bp were also generated, suggesting that there may be additional variants within this region. When primer pair P3/P4 was used to amplify cDNA from mouse tissues and BALB/3T3 fibroblasts, PCR products of 480 and 330 bp corre- sponding to the isolated variants were generated. Both products were generated in all tissues tested, but relative levels varied between tissues. For example, the 480-bp product was present at the lowest levels in testes, whereas the 330-bp product was detected at the lowest levels in brain and heart. The data indicate that the frame-preserving sequence gaps observed within PS-GAP-b and PS-GAP-c represent splice variants of PS-GAP-a.
PS-GAP protein levels were examined in various mouse tissues as well as in BALB/3T3 mouse fibroblasts and human embryonic kidney 293T cells. Tissue or cell lysates were analyzed by Western blotting using the anti-PS-GAP antibody (Fig. 4). Immunoreactive bands were present at ϳ140, 95, and 85 kDa, which were previously shown to immunoreact with the anti-PS-GAP antibody, but not with preimmune serum, and probably represent endogenous PS-GAP (43). The relative amounts of the PS-GAP variants varied between different tissues and cell lines. Heart, liver, spleen, and 293T cells showed extremely low levels of the 95-kDa band, whereas the 85-kDa band was low in brain and extremely low in BALB/3T3 fibroblasts. Recombinant PS-GAP-a comigrated with the band at 95 kDa, suggesting that this immunoreactive protein corresponds to the PS-GAP-a variant. The 85-kDa band could represent PS-GAP-b and/or PS-GAP-c, as their calculated molecular masses are 86 and 83 kDa, respectively. So far, no clone that could represent the 140-kDa band has been described.
PS-GAP Interacts with PAK-2 in Mammalian Cells-We have identified PS-GAP as a novel PAK-2-interacting protein.
In yeast two-hybrid analyses and GST pull-down assays, clone p/K18, the 191-amino acid C-terminal fragment of PS-GAP-a, interacted specifically with PAK-2p34-K278R, but not with full-length PAK-2. However, the interaction with full-length PAK-2 could require more N-terminal amino acid sequences of PS-GAP-a, which are not present within clone p/K18. Therefore, we used full-length wild-type PAK-2, kinase-deficient PAK-2-K278R, and constitutively active PAK-2-L106F in addition to PAK-2p34 and PAK-2p34-K278R to analyze the interaction with full-length PS-GAP-a. 293T cells were cotransfected with PS-GAP-a containing an N-terminal HA tag (HA-PS-GAP-a) and PAK-2 constructs containing an N-terminal FLAG tag (FLAG-PAK-2, FLAG-PAK-2-K278R, FLAG-PAK-2-L106F, FLAG-PAK-2p34, and FLAG-PAK-2p34-K278R). PAK-2p34 and kinase-deficient PAK-2p34-K278R are rapidly degraded by the 26 S proteasome; however, degradation of PAK-2p34 and PAK-2p34-K278R can be inhibited by coexpression of epitopetagged ubiquitin (42). Therefore, cells were cotransfected with a construct encoding His-Ub to stabilize FLAG-PAK-2p34 and FLAG-PAK-2p34-K378R. Cell lysates were used in reciprocal co-immunoprecipitation experiments with agarose-conjugated anti-FLAG or anti-HA antibody. Lysates from untransfected cells were used as negative controls.
Anti-FLAG immunoprecipitates and cell lysates were analyzed by Western blotting with anti-PS-GAP and anti-FLAG antibodies. The anti-PS-GAP antibody detected HA-PS-GAP-a in lysates of all cells transfected with HA-PS-GAP-a, but not in untransfected control cells. Additional immunoreactive bands probably correspond to endogenous forms of PS-GAP other than PS-GAP-a. HA-PS-GAP-a co-immunoprecipitated only with FLAG-PAK-2p34, but not with FLAG-PAK-2 or FLAG-PAK-2-K278R (Fig. 5A). No immunoreactive bands corresponding to endogenous forms of PS-GAP other than PS-GAP-a co-immunoprecipitated with FLAG-PAK-2p34, indicating that they may not interact with PAK-2p34. Western blotting with the anti-FLAG antibody was performed to verify immunoprecipitation of FLAGtagged proteins. As the levels of immunoprecipitated FLAG-PAK-2 and FLAG-PAK-2-K278R were approximately equal to those observed for FLAG-PAK-2p34, we conclude that HA-PS-GAP-a selectively precipitates with FLAG-tagged constructs of PAK-2p34, but not full-length PAK-2 or kinase-deficient PAK-2-K278R. Untransfected cells used as negative controls did not show immunoreactive bands corresponding to HA-PS-GAP-a and FLAG-tagged PAK-2 proteins.
In a reciprocal immunoprecipitation, anti-HA immunoprecipitates and cell lysates were analyzed by Western blotting with anti-PS-GAP and anti-FLAG antibodies. The anti-FLAG antibody detected FLAG-PAK-2, FLAG-PAK-2-K278R, and FLAG-PAK-2p34 in lysates of cells transfected with the corresponding constructs, but not in untransfected control cells. FLAG-PAK-2p34, but not FLAG-PAK-2 or FLAG-PAK-2-K278R, co-immunoprecipitated with HA-PS-GAP-a (Fig. 5B). Western blotting with the anti-PS-GAP antibody was performed to verify immunoprecipitation of HA-PS-GAP-a. As the levels of immunoprecipitated HA-PS-GAP-a were approximately equal, we conclude that HA-PS-GAP-a selectively precipitates with FLAG-tagged constructs of PAK-2p34, but not full-length PAK-2 or kinase-deficient PAK-2-K278R. Untransfected cells used as negative controls did not show immunoreactive bands corresponding to HA-PS-GAP-a and FLAG-tagged PAK-2 proteins.
To verify that the interaction between PS-GAP-a and PAK-2p34 is specific, 293T cells were cotransfected with PS-GAP-a containing an N-terminal Myc tag (Myc-PS-GAP-a) and FLAG-PAK-2p34, FLAG-PAK-2p34-K278R, or FLAG-PAK-2-L106F. 293T cells transfected with Myc-PS-GAP-a alone were used as negative controls. Cell lysates were used in co-immunoprecipitation experiments with agarose-conjugated anti-FLAG antibody. Anti-FLAG immunoprecipitates and cell lysates were analyzed by Western blotting with anti-Myc and anti-FLAG antibodies (Fig. 5C). The anti-Myc antibody detected Myc-PS-GAP-a in all cell lysates. Myc-PS-GAP-a co-immunoprecipitated with FLAG-PAK-2p34 and FLAG-PAK-2p34-K278R, whereas with FLAG-PAK-2-L106F, the levels of Myc-PS-GAP-a were low and indistinguishable compared with the negative control. Western blotting with the anti-FLAG antibody was performed to verify immunoprecipitation of FLAG-tagged proteins. As the levels of immunoprecipitated FLAG-PAK-2 constructs were approximately equal, it appears that Myc-PS-GAP-a precipitates with FLAG-tagged constructs of PAK-2p34 and kinase-deficient PAK-2p34-K278R, but not constitutively active PAK-2-L106F. The results of the immunoprecipitation experiments indicate that PS-GAP-a specifically interacts with caspase-activated PAK-2p34, but not with active or inactive forms of full-length PAK-2, in mammalian cells.
PAK-2p34 Does Not Affect the GAP Activity of PS-GAP-To examine the physiological significance of PAK-2p34/PS-GAP-a interactions, we examined the activity of a GST fusion protein of full-length PS-GAP as a GAP for the p21 monomeric G-proteins RhoA, Rac1, and Cdc42 in the absence and presence of PAK-2p34 (Fig. 6). Purified recombinant RhoA, Rac1, and Cdc42 were preloaded with [␥-32 P]GTP, and GTPase activity was analyzed in the absence and presence of purified full-length PS-GAP-a. Fulllength PS-GAP stimulated the intrinsic GTPase activity of Cdc42 and, to a much lower extent, RhoA in a dose-dependent manner, but had no significant effect on the GTPase activity of Rac1 (Fig.  6A). To examine whether the interaction with PAK-2p34 affects the GAP activity of PS-GAP, we determined the stimulation of the intrinsic GTPase activity of RhoA, Rac1, and Cdc42 in the presence of an equimolar amount or a 5-fold excess of purified PAK-2p34. Because PS-GAP contains a putative PAK-2 phosphorylation site (KRAS 264 ) directly N-terminal to the PH domain, the GAP activity of PS-GAP could be modified through phosphorylation of PS-GAP by PAK-2p34. Therefore, PS-GAP-a was preincubated with GST-PAK-2p34 in the presence of ATP and MgCl 2 prior to GAP assays. As a control, PS-GAP-a was preincubated with ATP and MgCl 2 alone. To compensate for different levels of GAP activity, we used 20 nM PS-GAP-a in experiments with RhoA and Rac1 and 5 nM PS-GAP-a in experiments with Cdc42. Preincubation in the presence of PAK-2p34 had no significant effect on the GAP activity of PS-GAP-a for RhoA, Rac1, or Cdc42 (Fig. 6B). Similar results were obtained when PAK-2p34 was added directly to the GAP assay without a preincubation period with ATP and MgCl 2 (data not shown). The data indicate that the interaction with PAK-2p34 does not affect the GAP

FIG. 5. PS-GAP-a interacts with PAK-2p34 in vivo.
For co-immunoprecipitation, 293T cells were cotransfected with epitope-tagged PS-GAP-a and PAK-2 constructs. His-Ub was cotransfected to inhibit degradation of PAK-2p34 and PAK-2p34-K278R. A, 293T cells were cotransfected using constructs for expression of HA-PS-GAP-a and FLAG-PAK-2, FLAG-PAK-2p34, or FLAG-PAK-2-K278R. As a negative control, untransfected cells were used. Lysates were immunoprecipitated using anti-FLAG antibody, and lysates (Lys) and immunoprecipitates (IP) were analyzed by Western blotting with the anti-PS-GAP antibody to detect coimmunoprecipitated HA-PS-GAP-a (upper panel). Western blotting with the anti-FLAG antibody was performed to verify immunoprecipitation of FLAG-PAK-2 constructs (lower panel). B, lysates were also immunoprecipitated using the anti-HA antibody, and immunoprecipitates were analyzed by Western blotting with the anti-FLAG antibody to detect coimmunoprecipitated FLAG-PAK-2 constructs (upper panel). Western blotting with the anti-PS-GAP antibody was performed to verify immunoprecipitation of HA-PS-GAP-a (lower panel). C, 293T cells were cotransfected using constructs for expression of Myc-PS-GAP-a and FLAG-PAK-2p34, FLAG-PAK-2p34-K278R, or FLAG-PAK-2-L106F. As a negative control, cells were transfected with Myc-PS-GAP-a alone. Lysates were immunoprecipitated using the anti-FLAG antibody, and lysates and immunoprecipitates were analyzed by Western blotting with the anti-Myc antibody to detect co-immunopre-  (n ϭ 6). B, intrinsic GTPase activity of p21 GTPases (Intr.) and GTPase activity in the presence of PS-GAP-a (control (con)) and PS-GAP-a preincubated with a 1-fold (1ϫ) or 5-fold (5ϫ) molar ratio of PAK-2p34 over PS-GAP-a were determined over 5 min. Assays with RhoA and Rac1 were performed with 20 nM PS-GAP-a, whereas assays with Cdc42 were performed with 5 nM PS-GAP-a. Results are shown as the means Ϯ S.D. (n ϭ 4). significant phosphorylation of PS-GAP-a was observed, suggesting that PS-GAP-a is not a substrate for PAK-2p34 (data not shown). However, a fragment of PS-GAP that contains the region between the GAP and SH3 domains inhibited autophosphorylation of PAK-2p34 (Fig. 7A). To examine whether PS-GAP binding inhibits PAK-2p34 kinase activity, we performed kinase assays using myelin basic protein as a PAK substrate in the presence of purified GST-PS-GAP-a constructs (Fig. 7B). GST fusion proteins of full-length PS-GAP-a, the region between the GAP and SH3 domains, and the region between the GAP and SH3 domains plus the SH3 domain resulted in decreased PAK-2p34 activity for myelin basic protein. A GST fusion protein of the SH3 domain of PS-GAP showed no inhibition of PAK-2p34 activity for myelin basic protein. Therefore, the region between the GAP and SH3 domains appears to be the PAK-2p34-binding domain of PS-GAP-a. The GST fusion protein of the PAK-2p34-binding domain plus the SH3 domain had no effect on the activity of full-length PAK-2-T402E, which is consistent with the results showing that full-length PAK-2 does not interact with PS-GAP.
To verify that inhibition of PAK-2p34 kinase activity by PS-GAP-a is not limited to phosphorylation of myelin basic protein, kinase assays were also performed using histone H4 as a substrate (Fig. 7C). The GST fusion protein of the PS-GAP-a fragment containing the PAK-2p34-binding domain plus the SH3 domain also inhibited PAK-2p34 kinase activity for histone H4, indicating that inhibition of PAK-2p34 kinase activity by binding to PS-GAP is a general phenomenon and is not limited to phosphorylation of myelin basic protein. To characterize the inhibition of PAK-2p34 activity by binding of PS-GAP, we performed a dose-response curve using different ratios of the GST fusion protein of the PS-GAP-a fragment containing the PAK-2p34-binding domain plus the SH3 domain (Fig. 7D). Maximal inhibition was observed at an ϳ20-fold molar excess of the PS-GAP-a fragment. These data suggest that PS-GAP down-regulates PAK-2p34 kinase activity and that this effect is mediated through the direct interaction with the PAK-2p34-binding domain of PS-GAP.
PS-GAP Prevents Nuclear Accumulation of PAK-2p34 -To examine the effects of PS-GAP on PAK-2p34 in vivo, we used transient expression of recombinant PS-GAP-a and PAK-2p34 in 293T cells. Expression of EGFP fusion constructs was used to study the localization of PS-GAP-a and PAK-2p34 in live cells (Fig. 8A). EGFP-PS-GAP-a was detected in the cytoplasm, but not in cell nuclei. EGFP-PAK-2p34 was detected primarily in nuclei. This is consistent with previous results showing that caspase-activated PAK-2p34 is targeted to the nucleus, whereas full-length PAK-2 is localized in the cytoplasm (42). Surprisingly, coexpression of EGFP-PAK-2p34 and HA-PS-GAP-a changed the localization of EGFP-PAK-2p34 from the nucleus to the cytoplasm.
To further characterize the change in PAK-2p34 localization by PS-GAP-a, we performed a combination of EGFP fluorescence and direct immunofluorescence experiments using recombinant expression of EGFP-PAK-2p34 and HA-PS-GAP-a (Fig. 8B). EGFP-PAK-2p34 and HA-PS-GAP-a were expressed in 293T cells and detected in paraformaldehyde-fixed cells using EGFP fluorescence and direct immunofluorescence with Alexa Fluor 594-conjugated anti-HA antibody. In the absence of HA-PS-GAP-a, EGFP-PAK-2p34 was detected primarily in the nucleus. HA-PS-GAP-a was detected in the cytoplasm at the perinuclear region and the cell periphery. However, coexpression of both EGFP-PAK-2p34 and HA-PS-GAP-a resulted in co-localization of both proteins to the perinuclear region within the cytoplasm. Similar results were obtained when FLAG-PAK-2p34 was used and detected by direct immunofluorescence with a fluorescein isothiocyanate-conjugated anti-FLAG antibody. Therefore, the interaction with PS-GAP-a appears to prevent nuclear accumulation of PAK-2p34.
PS-GAP Reduces the Ability of PAK-2p34 to Induce Cell Death-To examine whether the interaction with PS-GAP pro-  (Fig. 9). Coexpression with Myc-PS-GAP-a slightly stabilized the levels of EGFP-PAK-2p34, whereas coexpression with His-Ub resulted a dramatic stabilization of EGFP-PAK-2p34 levels. In addition, coexpression with Myc-PS-GAP-a and PS-GAP did not stabilize EGFP-PAK-2p34 to higher levels compared with coexpression with His-Ub alone. Western blotting with the anti-Myc antibody was performed to examine the expression levels of Myc-PS-GAP-a. Although the interaction with PS-GAP appears to slightly reduce degradation of PAK-2p34, it does not completely protect PAK-2p34 from degradation.
To examine the effects of PS-GAP on the ability of PAK-2p34 to induce cell death, we analyzed apoptotic chromatin condensation of 293T cells that were cotransfected either with EGFP-PAK-2p34 and His-Ub or with EGFP-PAK-2, His-Ub, and Myc-PS-GAP-a. Expression and localization of EGFP-PAK-2p34 were analyzed by EGFP fluorescence experiments; nuclear morphology was analyzed by staining with Hoechst 33342 (Fig. 10). Coexpression of Myc-PS-GAP-a resulted in the perinuclear localization of EGFP-PAK-2p34 and decreased levels of cell death. A quantitative analysis showed that 86% of cells cotransfected with EGFP-PAK-2p34 and His-Ub displayed apoptotic chromatin condensation, indicating that stabilization of PAK-2p34 results in stimulation of cell death. In contrast, only 37% of cells that were cotransfected with EGFP-PAK-2p34, His-Ub, and Myc-PS-GAP displayed apoptotic chromatin condensation. Therefore, PS-GAP appears to reduce the ability of PAK-2p34 to induce programmed cell death. DISCUSSION PAK-2 appears to be a bifunctional modulator of both cell survival and cell death pathways (42,55). Full-length PAK-2 stimulates cell survival, whereas caspase-activated PAK-2p34 stimulates programmed cell death. Therefore, it is critical to understand the regulation of anti-apoptotic PAK-2 and proapoptotic PAK-2p34 signaling pathways. Although full-length PAK-2 is known to be regulated through interactions with a number of other signaling molecules, little is known about the regulation of PAK-2p34 signaling. In this study, we have identified PS-GAP as a novel regulator of caspase-activated PAK-2p34. PS-GAP interacts specifically with PAK-2p34, but not with full-length PAK-2. This is surprising since PAK-2p34 is a proteolytic product of PAK-2, which is identical to residues 213-524 of PAK-2. Therefore, the interaction site with PS-GAP appears to be masked in full-length PAK-2. The crystal structure of PAK-1 predicts that inactive PAKs exist as antiparallel dimers that dissociate or unfold upon activation by p21 Gproteins (12). Therefore, the interaction site with PS-GAP could be masked by interactions between the PAK monomers. However, this does not explain why PS-GAP does not interact with PAK-2-T402E and PAK-2-L106F, two constitutively active mutants of PAK-2 that should be in an open conformation. Therefore, the interaction site with PS-GAP appears to be masked by intramolecular interactions within full-length PAK-2 that are released by removing the N-terminal 212 residues through proteolytic cleavage.
PS-GAP was originally detected in a yeast two-hybrid screen using a fragment of PYK-2 as a bait protein, and the SH3 domain of PS-GAP was shown to interact with PYK-2 (43). The region between the GAP and SH3 domains of PS-GAP-a was sufficient to inhibit the protein kinase activity of PAK-2p34 and was named the PAK-2p34-binding domain. Interestingly, the splice variant PS-GAP-c lacks 51 amino acids within this region immediately N-terminal to the SH3 domain, and interactions between PAK-2p34 and PS-GAP could be modulated by alternative splicing. Splice variant PS-GAP-b differs from PS-GAP-a in the region between the PH and GAP domains, including the N terminus of the GAP domain. Because PS-GAP-b lacks the N-terminal 12 amino acids of the GAP domain, it is unclear if it can function as a Rho-GAP. Two additional PS-GAP variants, PS-GAP-m and PS-GAP-s, were described previously (43). PS-GAP-m could correspond to PS-GAP-a, but there are some differences in the sequence, viz. in 4 consecutive amino acids within the region between the PH and GAP domains. Because this appears to be a region where splicing occurs, these differences could also be due to alternative splicing. The results of our RT-PCR with primers flanking the sequence gap observed in PS-GAP-b suggest that two more splice variants exist within this region. Therefore, PS-GAP appears to be regulated by multiple alternative splicing events, and the various splice variants may serve different functions within the cell.
To analyze the physiological function of the interaction between PS-GAP and PAK-2p34, we measured the GAP activity of full-length PS-GAP-a for RhoA, Rac1, and Cdc42, and examined whether PAK-2p34 affects the GAP activity of PS-GAP. Full-length PS-GAP-a stimulated the GTPase activity of Cdc42 and, to a lesser extent, RhoA, but not Rac1. In previous studies, the isolated GAP domain of PS-GAP was shown to stimulate the GTPase activity of RhoA and, to a lesser degree, Cdc42, but not Rac1 or Ran; however, recombinant expression of fulllength PS-GAP in 293T cells decreased the levels of active GTP-bound Cdc42 more than those of RhoA (43). Therefore, it appears that full-length PS-GAP and the isolated GAP domain differ in their affinities for Cdc42 and RhoA and that Cdc42 is the preferred target of full-length PS-GAP. The interaction between PS-GAP and PAK-2p34 could affect the Rho-GAP activity of PS-GAP, and this could involve phosphorylation of PS-GAP by PAK-2p34. PS-GAP has a consensus site for phosphorylation by PAK-2 (KRX(S/T)) at Ser 264 . However, PAK-2p34 (or full-length PAK-2) did not phosphorylate PS-GAP-a to a significant level. Therefore, it appears that PS-GAP-a is not a substrate for PAK-2p34. In addition, PAK-2p34 did not affect the GAP activity of PS-GAP for Cdc42, RhoA, or Rac1. Therefore, the interaction with PAK-2p34 does not regulate the GAP activity of PS-GAP by phosphorylation or allosteric interaction.
Although PS-GAP is not a substrate or target of PAK-2p34, it appears to be a specific regulator of PAK-2p34. In protein kinase assays, full-length PS-GAP-a and fragments containing the PAK-2p34-binding domain inhibited PAK-2p34 kinase activity toward itself and toward exogenous substrates such as myelin basic protein and histone H4. Although inhibition of PAK-2p34 kinase activity was pronounced, PS-GAP did not decrease the kinase activity of the constitutively active fulllength PAK-2-T402E mutant. Therefore, the interaction with PS-GAP specifically inhibits the protein kinase activity of caspase-activated PAK-2p34, but not full-length PAK-2. Maximal inhibition of PAK-2p34 activity required a 20-fold molar excess of PS-GAP, which is quite high for a physiologically relevant event. However, physiological inhibition of PAK-2p34 by PS-GAP could require a post-translational modification or a cofactor of PS-GAP that was not present in the in vitro protein kinase assays.
In addition to inhibition of PAK-2p34 activity, PS-GAP affects the subcellular localization of PAK-2p34. Coexpression of PS-GAP changes the localization of recombinant PAK-2p34 from the nucleus to the cytoplasm. We have shown recently that full-length PAK-2 is localized in the cytoplasm, whereas caspase-activated PAK-2p34 accumulates in the nucleus (42). PS-GAP is a cytoplasmic protein that localizes to the perinuclear region and the cell periphery (43). Because coexpressed PAK-2p34 and PS-GAP co-localize to the perinuclear region, it appears that PS-GAP prevents nuclear accumulation of caspase-activated PAK-2p34.
Caspase-activated PAK-2p34 is rapidly degraded by the 26 S proteasome, but full-length PAK-2 is not. Expression of epitope-tagged ubiquitin stabilizes PAK-2p34 by inhibiting its polyubiquitination and degradation. Stabilization of PAK-2p34 results in a dramatic stimulation of cell death (42). Although expression of PS-GAP slightly increases PAK-2p34 levels, it does not result in significant stabilization of PAK-2p34 or in stimulation of cell death. However, expression of PS-GAP with PAK-2p34 and epitope-tagged ubiquitin reduces the levels of cell death, indicating that PS-GAP counteracts cell death induced by caspase-activated PAK-2p34. Therefore, PS-GAP appears to protect cells from cell death by inhibiting the kinase activity and/or by preventing nuclear accumulation of proapoptotic caspase-activated PAK-2p34.