Molecular Cloning and Sequencing of the Cytostatic G Protein-activated Protein Kinase PAK I*

The serine/threonine protein kinase PAK I (p21-acti- vated protein kinase), a ubiquitous multipotential protein kinase of 58–60 kDa, has been shown to have cyto- static properties. Data from our laboratory show that PAK I is highly active in oocytes and quiescent and serum-starved cells, and injection of active PAK I into one blastomere of two-cell frog embryos inhibits cleav- age of the injected blastomere. To clone the cDNA encoding PAK I, purified peptides from rabbit PAK I were sequenced, degenerate oligonucleotides were used to isolate PAK I clones from a rabbit spleen library, and the 5 (cid:42) -terminus was obtained by polymerase chain reaction. The entire cDNA sequence extends over 4471 nucleo- tides, with an open reading frame for a protein of 524 residues and a 3 (cid:42) -noncoding region of 2826 nucleotides. Clones with the same open reading frame but with 3 (cid:42) noncoding regions of 1055 and 2478 nucleotides were isolated, suggesting the generation of different transcripts by alternative termination of transcription. The amino acid sequence of PAK I shows high homology to the p21-activated protein kinases from human placenta and rat brain and to yeast STE20. PAK I is activated by Cdc42(GTP). The PAK enzymes have been proposed to regulate the stress-activated protein kinase (also known as the Jun kinase) signaling pathway Cell 81, 1137–1146;

PAK I activity is elevated in quiescent and serum-starved cells, as opposed to actively dividing cells. PAK I activity is also high in frog oocytes, but is rapidly diminished following fertilization. 2 Injection of endogenously active or proteolytically activated PAK I into two-cell frog embryos leads to cleavage arrest (11). Other protein kinases, including casein kinase II, protein kinase C, and the catalytic subunit of cAMP-dependent protein kinase, have no effect on cleavage.
The cloning and sequencing of the cDNA encoding PAK I, described herein, has revealed a mode of regulation of PAK I activity. The N-terminal regulatory domain contains a G protein-binding region similar to that of the G protein-activated protein kinases including rat brain PAK65 (12), human placenta PAK65 (13), and yeast STE20 (14,15). Binding of the small G protein Cdc42 stimulates autophosphorylation and protein kinase activity, as shown by phosphorylation of H4.

EXPERIMENTAL PROCEDURES
Materials-32 P-Labeled and 35 S-labeled nucleotides as well as Hybond-N membranes were purchased from Amersham Corp. Immobilon-P was obtained from Millipore Corp., and ProBlott was from Applied Biosystems Inc. Endopeptidase Asp-N was purchased from Boehringer Mannheim. The Vydac reverse-phase C 18 column was from The Separations Group. Pyrostase was obtained from Molecular Genetic Resources, and the nested deletion kit was from Pharmacia Biotech Inc. The messenger RNA isolation kit, the rabbit spleen Uni-ZAP library, Stratascript RNase H Ϫ reverse transcriptase, and cloned Pfu DNA polymerase were purchased from Stratagene. T4 polynucleotide kinase, T4 DNA ligase, Klenow DNA polymerase, and restriction enzymes were from New England Biolabs Inc. The 5Ј-Amplifinder TM RACE kit was obtained from CLONTECH, and the Sequenase Version 2.0 DNA sequencing kit was from U. S. Biochemical Corp. The Qiaprep spin kit and the plasmid midi kit were from QIAGEN Inc.; the Geneclean kit was from BIO 101, Inc. Expression clones for Rac1, Cdc42Hs, and RhoA as GST fusion proteins were generously provided by Dr. Channing Der (University of North Carolina, Chapel Hill, NC).
Purification and Analysis of PAK I Peptides-The inactive PAK I holoenzyme was purified to apparent homogeneity from rabbit reticulocytes by chromatography on DEAE-cellulose, SP-Sepharose, and protamine-agarose and by fast protein liquid chromatography on Mono Q and Mono S as described elsewhere. 1 Peptides of PAK I were obtained by chemical cleavage using CNBr, partial proteolysis with trypsin, or a combination of CNBr and endopeptidase Asp-N.
For in situ CNBr cleavage, PAK I was subjected to electrophoresis on a 12.5% SDS-polyacrylamide gel (16) and transferred to an Immobilon-P membrane. PAK I was cleaved on the membrane with 4 mg/ml CNBr for 15 h at room temperature, and the peptides were eluted with 100 l of 70% isopropyl alcohol and 1% trifluoroacetic acid for 3 h at room temperature, followed by 100 l of 40% acetonitrile for 1 h at 37°C. The extracts and the original cleavage solution were dried in a Speed Vac concentrator and separated by electrophoresis on a 16.5% SDS-polyacrylamide gel (16). The peptides were transferred to a ProB-* This work was supported by United States Public Health Service Grant GM26738. 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 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EMBL Data Bank with accession number(s) U46915.
‡ To whom correspondence should be addressed. lott membrane and subjected to amino acid sequence analysis. The N-terminal sequence of CNBr peptide 1 (20 kDa) was PEQWARLL-GTSNXTKLEQKK, and that of CNBr peptide 2 (8 kDa) was XXXXFSTGGKDPLSANHXL.
To obtain the amino-terminal sequence of the p37 peptide, PAK I was partially digested with 4 g/ml trypsin on ice for 1 min as described by Tahara and Traugh (1), with the omission of bovine serum albumin. Peptides were separated by electrophoresis on a 12% SDS-polyacrylamide gel (16) and transferred to Immobilon-P, and the sequence was determined to be SVIDPIPAPVGDSHV.
For CNBr/endopeptidase Asp-N cleavage, PAK I was cleaved with 10 mg/ml CNBr for 15 h at room temperature, and the peptides were denatured, reduced, and carboxymethylated as described (17). The CNBr peptides were digested with endopeptidase Asp-N, and the digest was acidified with glacial acetic acid and separated by reverse-phase high pressure liquid chromatography on a Vydac C 18 column with a acetonitrile gradient from 0 to 100%. Amino acid sequence analysis and mass spectrometry were performed at the Biotechnology Instrumenta- cDNA Cloning by Reverse Transcription and Polymerase Chain Reaction-Poly(A) ϩ RNA was isolated from rabbit liver, spleen, kidney, and brain using the messenger RNA isolation kit. One g of poly(A) ϩ RNA was reverse-transcribed with Stratascript RNase H Ϫ reverse transcriptase using an oligo(dT) primer. The resulting cDNA was amplified over 35 cycles in a polymerase chain reaction (18) using Pyrostase, the 48-fold degenerate sense oligonucleotide primer (5Ј-ATGGAYGARCAR-CARATHGC-3Ј) (1 M), and the 512-fold degenerate antisense oligonucleotide primer (5Ј-CCNCKYTTYTCNACRTCCAT-3Ј) (5 M). A PCR product of 431 bp was subcloned into pBluescript SK ϩ .
cDNA Library Screening-For high density screening in a rabbit spleen Uni-ZAP library, ϳ5 ϫ 10 5 plaque-forming units were plated with Escherichia coli XL-1Blue. Two replicas were prepared from each plate on Hybond-N membranes. The 431-bp PAK I cDNA was labeled by the multipriming procedure (19) with [␣-32 P]dCTP. Hybridization at 30°C and washes at 55°C were carried out as described elsewhere (20). After autoradiography, positive clones were purified over two more rounds of screening, excised in vivo from the Uni-ZAP vector as pBluescript SK ϩ clones, and analyzed by Southern blot hybridization (20,21).
Cloning of the 5Ј-cDNA Region by RACE-PCR-The 5Ј-end of the PAK I cDNA was obtained by RACE-PCR using the 5Ј-Amplifinder TM RACE kit. Total RNA from rabbit spleen was isolated by the guanidinium thiocyanate method (22), and poly(A) ϩ RNA was isolated by two rounds of oligo(dT)-cellulose chromatography (23). For the cDNA synthesis, poly(A) ϩ RNA (2 g) was extended with reverse transcriptase using the oligonucleotide primer P1 (5Ј-GCCTGTAAACACTCTCTGCA-CACAG-3Ј). The synthesized cDNA was purified and ligated to a synthetic anchor. PCR was carried out over 35 cycles with cloned Pfu DNA polymerase using the oligonucleotide primer P2 (5Ј-CTTCATCCATG-CAGGTTTCTGTTAC-3Ј) and a primer corresponding to the anchor sequence. A PCR product of ϳ1150 bp was subcloned into pBluescript SK ϩ .
Sequence Analysis of PAK I cDNA Clones-Plasmid DNA was isolated with the plasmid midi kit; subclones were constructed using internal EcoRI and PstI sites, and nested deletions were created with the double-stranded nested deletion kit. DNA sequencing was carried out by the dideoxy chain termination method of Sanger et al. (24) using ␣-35 S-dATP and T7 DNA polymerase. Nucleotide sequences were aligned to form the full-length cDNA sequence, and the sequence was analyzed with the University of Wisconsin Genetics Computer Group Package Version 8 (25).
Stimulation of PAK I Autophosphorylation and Activity by G Proteins-GST fusion proteins of Rac1, Cdc42Hs, and RhoA were expressed individually in E. coli DH5␣; the bacteria were sonicated in the presence of the protease inhibitors aprotinin, leupeptin, pepstatin (40 g/ ml), and phenylmethylsulfonyl fluoride (0.5 mM), and the fusion proteins were purified on glutathione-Sepharose beads. GST-Rac1, GST-Cdc42Hs, or GST-RhoA (1 g) was preloaded with 0.18 mM GTP␥S or 0.18 mM GDP in 20 mM Tris-HCl, pH 7.5, and 50 mM NaCl for 10 min at 30°C in a volume of 20 l. Then, the G proteins were incubated with PAK I (0.1 g) in 70 l of phosphorylation buffer (50 mM Tris-HCl, pH 7.4, 10 mM MgCl 2 , 30 mM 2-mercaptoethanol) with 0.2 mM [␥-32 P]ATP (4000 dpm/pmol) at 30°C. Aliquots were removed at 5, 10, and 15 min; autophosphorylation was analyzed by electrophoresis on a 12.5% SDSpolyacrylamide gel followed by autoradiography.
The effects of autophosphorylation on activity were measured by preincubation of G protein with PAK I for 10 min as described above. A 20-l sample was assayed in a final volume of 70 l containing phosphorylation buffer, 4.0 g of H4, and 0.2 mM [␥-32 P]ATP (2000 dpm/ pmol) by incubation for 15 min at 30°C. Phosphorylation of H4 was analyzed by electrophoresis on a 15% SDS-polyacrylamide gel followed by autoradiography (26) and quantified by scintillation counting of the excised H4 band.

RESULTS
Preparation of a PAK I-specific Hybridization Probe-Peptides from the inactive PAK I holoenzyme purified from rabbit reticulocytes were obtained by cleavage with CNBr, CNBr/ endopeptidase Asp-N, or trypsin. The peptides were separated, purified, and partially sequenced as described under "Experimental Procedures." Two CNBr peptides of 20 and 8 kDa and six CNBr/endopeptidase Asp-B peptides ranging from 5497 to 1438 Da were isolated. A peptide of 37 kDa, isolated following limited proteolysis with trypsin, contained the active catalytic domain. 1 All nine peptides were subjected to partial sequence analysis as described under "Experimental Procedures." The N-terminal sequences of CNBr/Asp-N peptides 2-5 could be located within the alignment of the catalytic domains of other protein kinases as described by Hanks et al. (27,28).
Following reverse transcription of mRNA from rabbit spleen and brain, degenerate oligonucleotides corresponding to CNBr/ Asp-N peptides 4 and 5 were used to amplify a PAK I-specific polynucleotide. After 35 cycles of amplification, PCR products of 431 bp were subcloned and analyzed by DNA sequencing. The clones from spleen and brain were identical and contained the complete amino acid sequences of CNBr/Asp-N peptides 2, 4, and 5.
Cloning of the cDNA Encoding PAK I-The partial 431-bp cDNA from rabbit spleen was used to screen a rabbit spleen cDNA library, and nine clones were detected within 5 ϫ 10 5 plaques. Restriction analysis and partial cDNA sequencing showed that all nine clones contained the same open reading frame, but were incomplete at the 5Ј-end. Differences in the length of the 3Ј-untranslated region were observed. PAKI-4.1 was the clone that contained the longest open reading frame and had the longest 3Ј-untranslated region (Fig. 1). The 5Ј-end of the cDNA for PAK I was obtained by RACE-PCR from rabbit spleen mRNA. Six positive clones with an insert of ϳ1150 bp were characterized by restriction analysis and partial DNA sequencing. All clones overlapped with the 5Ј-end of PAKI-4.1 and contained the same open reading frame starting with ATG; only the lengths of the 5Ј-untranslated region were different. The longest clone isolated by RACE-PCR (clone 1150-4) contained the longest 5Ј-untranslated region.
Complete Sequence Analysis of PAK I cDNA-Both strands of the library clone PAKI-4.1 and the RACE-PCR clone 1150-4 were sequenced completely. For the sequence analysis of PAKI-4.1, subclones E1100 and P2900 were constructed using internal EcoRI and PstI sites (Fig. 1). Overlapping nested deletions were created to sequence 2665 nucleotides from the 3Ј-end of the noncoding strand of PAKI-4.1, the entire coding strand of subclone P2900, and both entire strands of subclone E1100 and of RACE-PCR clone 1150-4.
The complete cDNA sequence of PAK I consists of 4471 bp, including a poly(A) tail of 18 nucleotides (Fig. 2). The cDNA sequence contains an open reading frame from nucleotides 74 to 1645, 73 nucleotides of 5Ј-untranslated region, and 2826 nucleotides of 3Ј-untranslated region. The 3Ј-untranslated regions of other clones isolated by screening of the rabbit spleen library were shorter than that of PAKI-4.1, but could be aligned The deduced amino acid sequence of the open reading frame encodes a polypeptide of 524 amino acid residues with a calculated molecular mass of 58,027 Da. All nine partial amino acid sequences, obtained by microsequence analysis of purified peptides of PAK I, could be aligned within the deduced amino acid sequence (Fig. 2). The peptide beginning with SVID at residue 197 indicates the N terminus of the p37 peptide generated by limited trypsin digestion, which contains the catalytic domain. The beginning of the catalytic domain at residue 247 was determined by comparison with the sequences of other protein kinases using the alignment of Hanks et al. (27,28). Peptide p37 contains part of the regulatory domain (from residues 197 to 246) and all 11 conserved subdomains characteristic of the catalytic domain of protein kinases. The GXGXXG motif involved in nucleotide binding begins at residue 256; residue 278 is the invariant lysine that contacts the ␣and ␤-phosphates of the ATP; residue 368 is the aspartate that acts as a general base for the removal of a proton from the hydroxyl group of the protein substrate; residue 370 is the lysine that binds to the ␥-phosphate of the ATP; and residue 386 is the aspartate that is part of the conserved DFG motif and chelates the Mg 2ϩ bound to the ATP. A putative binding site for Rho-like G proteins was identified in the regulatory domain between residues 73 and 107 by sequence comparison with PAK65 from rat brain and human placenta and yeast STE20 (12)(13)(14)(15).
Amino acid sequence alignment of the total proteins revealed 95% identity between rabbit PAK I and human PAK65 and 78% identity between rabbit PAK I and rat PAK65 (Fig. 3). Within the catalytic domain, rabbit PAK I has 99% identity with human PAK65, 92% with rat PAK65, and 65% with yeast STE20. In the regulatory domain, PAK I has significantly less homology with STE20 and rat PAK65, except for the G proteinbinding region. Compared with rat PAK65, PAK I was 20 residues shorter because of five gaps in the regulatory domain ranging from 1 to 10 residues. Human PAK65 was 18 residues shorter than rabbit PAK I because of differences at the N terminus over the first 30 amino acid residues; the rest of the amino acid sequence was very similar to that of rabbit PAK I.
G Protein Activation of PAK I-Previously, a number of compounds known to modulate protein kinase activity have been examined as possible physiological regulators of PAK I; none of these had any effect on PAK I activity. To examine the effects of the G proteins Rac1, Cdc42Hs, and RhoA on autophosphorylation and activation of purified inactive PAK I, GST fusion proteins expressed in E. coli were purified on glutathione-Sepharose beads and preloaded with GTP␥S. A low basal level of autophosphorylation of inactive PAK I was detected in the absence of G protein (Fig. 4). Upon addition of GST-Rac1, a slight stimulation of autophosphorylation over the basal level was observed over a 15-min period of incubation, while GST-RhoA had little effect. Stimulation of autophosphorylation upon addition of GST-Cdc42Hs was observed within 5 min. The rate of autophosphorylation was linear up to 10 min and began to level off by 15 min.
To examine the effects of autophosphorylation on activation of PAK I, protein kinase activity was measured by phosphorylation of the substrate H4 (Fig. 5). PAK I was incubated under autophosphorylation conditions for 10 min, either alone or with GST-Rac1, GST-Cdc42Hs, or GST-RhoA preloaded with GTP␥S or GDP, and then assayed with H4. Phosphorylation of H4 was dependent on the level of autophosphorylation of PAK I. Without addition of G protein, a low level of autophosphorylation and PAK I activity with H4 was observed, which was subtracted from the data obtained in the presence of G protein.
Addition of GST-Cdc42Hs(GTP␥S) resulted in an increased autophosphorylation of PAK I of 0.97 mol/mol and a stimulation of the rate of phosphorylation of H4 of 18,261 pmol/min/ mg. With GST-Cdc42Hs(GDP), stimulation of autophosphorylation was 0.12 mol/mol, and the activity was 2154 pmol/min/ mg. These values were ϳ8-fold lower than those observed with Cdc42Hs(GTP␥S). GST-Rac1(GTP␥S) resulted in only a modest stimulation, ϳ10-fold lower than that observed with Cdc42Hs-(GTP␥S); autophosphorylation and activity were 0.09 mol/mol and 1515 pmol/min/mg, respectively. GST-RhoA(GTP␥S) had no stimulatory effect on either autophosphorylation or activity. DISCUSSION The cDNA isolated by screening of a rabbit spleen cDNA library and by 5Ј-RACE-PCR using rabbit spleen mRNA contains the information for the complete sequence of PAK I. The deduced amino acid sequence contains the catalytic domain and an N-terminal regulatory domain, including a region shown previously to bind Rho-like G proteins. All of the partial amino acid sequences determined by microsequencing of nine peptides from PAK I could be aligned with the deduced amino acid sequence. The calculated molecular mass of 58,027 Da is in accordance with the apparent molecular mass of 58 -60 kDa displayed by native PAK I on SDS-polyacrylamide gels (5,6).
The cDNA for PAK I from rabbit spleen contains a long 3Ј-untranslated region of 2826 nucleotides. Long 3Ј-untranslated regions are characteristic of highly regulated genes and are involved in the stability and translational efficiency of the mRNA (31, 32). The isolation of PAK I clones from rabbit spleen with shorter 3Ј-untranslated regions of 1055 and 2478 nucleotides, including the poly(A) tails, suggests the existence of different transcripts for PAK I generated by alternative transcription termination and polyadenylation sites. Transcripts of PAK I with 3Ј-untranslated regions of different lengths could have different stability and/or translation efficiency. Therefore, the amount of PAK I protein could be regulated in part by the expression of transcripts with 3Ј-untranslated regions of different lengths.
Sequence comparison shows high sequence homology to the yeast protein kinase STE20 (14,15) and PAK65 from rat brain and human placenta (12,13), suggesting that they belong to a family of related protein kinases. Rat PAK65 appears to be a brain-specific isoform; a protein with the corresponding molecular mass was detected only in brain extracts, but not in other tissues, by overlay assays with radiolabeled Rac1(GTP) and Cdc42Hs(GTP) and by a specific antibody raised against the N-terminal region of rat brain PAK65 (12,13). Unlike brain-specific rat PAK65, inactive PAK I was purified from a number of tissue and cell types, including rabbit reticulocytes, liver, and skeletal muscle; chicken gizzard, liver, and brain; bovine liver; mouse 3T3-L1 cells (1-6); and frog oocytes and embryos. 2 The protein reacted with antibodies raised against PAK I from rabbit reticulocytes. PAK I from rabbit spleen and PAK65 from human placenta have a high degree of sequence homology, but the N-terminal 30 amino acid residues are completely different. Human PAK65 lacks 18 amino acid residues, and the remaining 12 amino acid residues cannot be aligned with either rabbit PAK I or rat PAK65. Therefore, it is unlikely that the two protein kinases are homologous enzymes, but they appear to be closely related isoforms.
PAK65 from rat brain and human placenta as well as two additional PAK enzymes detected in human neutrophils were shown to be autophosphorylated upon binding of the G proteins Rac1 and Cdc42, but not RhoA; autophosphorylation resulted in activation as shown by stimulation of phosphorylation of the universal substrate, myelin basic protein (12,13,29,30). Binding of the GTP-bound form of GST-Cdc42Hs, but not GST-Rac1 or GST-RhoA, greatly induced autophosphorylation of PAK I from rabbit reticulocytes, and autophosphorylation resulted in activation of the protein kinase. STE20 in budding yeast plays a key role in signal transduction pathways connecting membrane receptors with the MAP kinase cascade (33). In the mating differentiation pathway, a pheromone initiates differentiation by activation of STE20, which activates the MAP kinase cascade. STE20 and MAP kinases are also involved in other signal transduction pathways in yeast, including the invasive growth response and pseudohyphal development pathways initiated by nutritional starvation. Different isoforms of STE20 and MAP kinases could have alternative functions in regulating different signal transduction pathways (33). PAK I and other PAK enzymes, as the mammalian counterparts of yeast STE20 protein kinases, also appear to be involved in the coordination of signal transduction pathways. It has recently been hypothesized (34,35) that the Rac1/Cdc42-regulated PAK enzymes are a key point regulator of the stress-activated protein kinase or Jun kinase signaling pathway, which regulates c-Jun activity. Biochemical evidence from our laboratory suggests that PAK I is involved in regulation of cytostasis as shown by the high level of activity in quiescent and serum-starved cells and lower levels of PAK I activity in actively dividing cells. 2 PAK I activity is also high in mature frog oocytes, but is greatly diminished following fertilization. 2 Injection of active PAK I into two-cell embryos has been shown to inhibit cell division in the injected blastomere (11). Thus, PAK I appears to be a candidate for regulation of the signaling cascade that is activated in response to stress. FIG. 5. Stimulation of PAK I activity by autophosphorylation. PAK I (0.1 g) was autophosphorylated for 10 min in the presence of the GST fusion protein of Cdc42Hs, Rac1, or RhoA preloaded with GTP␥S or GDP. Following autophosphorylation, assays with H4 were carried out and analyzed on a 15% SDS-polyacrylamide gel, followed by autoradiography and scintillation counting of the excised H4 bands. Phosphorylation of H4 in the absence of G protein was used as background and subtracted from the data.